100% found this document useful (1 vote)
175 views24 pages

Book Embed PDF

This document outlines the contents of a book on fluid systems and dynamics. Chapter 1 covers preliminaries including notation, measures and distributions, function spaces, and Fourier analysis. Chapter 2 discusses dynamical systems concepts like attractors and Lyapunov exponents applied to fluid systems. Subsequent chapters cover viscous fluid systems, analysis of complete fluid models, large time behavior, and simplified fluid models.

Uploaded by

Vasi Uta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (1 vote)
175 views24 pages

Book Embed PDF

This document outlines the contents of a book on fluid systems and dynamics. Chapter 1 covers preliminaries including notation, measures and distributions, function spaces, and Fourier analysis. Chapter 2 discusses dynamical systems concepts like attractors and Lyapunov exponents applied to fluid systems. Subsequent chapters cover viscous fluid systems, analysis of complete fluid models, large time behavior, and simplified fluid models.

Uploaded by

Vasi Uta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

Contents

1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Euclidean space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.3 Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.4 Differential operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Measures and distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Continuously differentiable functions . . . . . . . . . . . . . . . . . 4
1.2.2 Compactness of sets of continuous functions . . . . . . . . . . 5
1.2.3 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.4 Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Bochner and Sobolev spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Lebesgue (Bochner) spaces . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2 Sobolev spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.3 Dual Sobolev spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.4 Embeddings for dual Sobolev spaces . . . . . . . . . . . . . . . . . 12
1.3.5 Traces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.6 Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Fourier analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.1 Fourier transform of tempered distributions . . . . . . . . . . . 15
1.4.2 Boundedness of Fourier multipliers . . . . . . . . . . . . . . . . . . 16
1.5 Compactness of families of functions . . . . . . . . . . . . . . . . . . . . . . . 17
1.5.1 Weak compactness of integrable functions . . . . . . . . . . . . 17
1.5.2 Young measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Gronwalls lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2 Dynamical systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1 Dynamical systems and attractors . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 Fractal dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 Exponential attractor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Method of trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
X Contents

2.5 Method of Lyapunov exponents . . . . . . . . . . . . . . . . . . . . . . . . . . . 43


2.5.1 Coverings of ellipsoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.5.2 Lyapunov exponents and solution operators . . . . . . . . . . . 48
2.6 Extension to the case without uniqueness . . . . . . . . . . . . . . . . . . . 53
2.7 Bibliographical remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3 Viscous fluid systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59


3.1 Fields and balance laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.1.1 Balance law - integral formulation . . . . . . . . . . . . . . . . . . . 61
3.1.2 Balance law - formal interpretation . . . . . . . . . . . . . . . . . . 64
3.1.3 Weak sequential stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2 Kinematics of mass transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2.1 Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3 Momentum equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.4 Total energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.1 Kinetic and internal energy balance . . . . . . . . . . . . . . . . . . 72
3.4.2 Conservative boundary conditions and the total
energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.5.1 Entropy production as a non-negative Radon measure . . 74
3.6 Fluid description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.6.1 General constitutive relations . . . . . . . . . . . . . . . . . . . . . . . 76
3.6.2 Navier-Stokes-Fourier system . . . . . . . . . . . . . . . . . . . . . . . 77
3.7 Constitutive equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.7.1 Hypothesis of thermodynamic stability . . . . . . . . . . . . . . . 82
3.7.2 Monoatomic gas state equation . . . . . . . . . . . . . . . . . . . . . . 82
3.7.3 Third law of thermodynamics, entropy . . . . . . . . . . . . . . . 83
3.7.4 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.7.5 Transport coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

4 Complete fluid systems, analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 87


4.1 Equilibrium states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.1.1 Total mass conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.1.2 Total dissipation balance . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.1.3 Static states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.2 Weak sequential stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2.1 Stability of solutions under a uniform energy bound . . . 95
4.3 Global-in-time solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.3.1 Existence theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.4 Long-time behavior of conservative systems . . . . . . . . . . . . . . . . . 120

5 Large time behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125


5.1 Large time behavior revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.1.1 Uniform decay of density oscillations . . . . . . . . . . . . . . . . . 129
5.2 Energy blow up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Contents XI

5.2.1 A general result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138


5.2.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.3 Rapidly oscillating driving forces . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.3.1 Energy estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.3.2 Uniform bounds via iteration . . . . . . . . . . . . . . . . . . . . . . . 151
5.3.3 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.3.4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

6 Complete and simplified fluid systems . . . . . . . . . . . . . . . . . . . . . 155


6.1 Dynamical systems approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.1.1 Dissipativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.1.2 Global trajectories and attractors . . . . . . . . . . . . . . . . . . . 158
6.2 Dynamics of reduced systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.2.1 Barotropic compressible fluids . . . . . . . . . . . . . . . . . . . . . . . 161
6.2.2 Incompressible fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.3 Bibliographical comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

7 Incompressible model of Ladyzhenskaya . . . . . . . . . . . . . . . . . . . 169


7.1 Statement of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
7.1.1 Assumptions on the stress tensor . . . . . . . . . . . . . . . . . . . . 171
7.1.2 Mathematical formulation of the problem . . . . . . . . . . . . . 173
7.1.3 Existence and uniqueness results . . . . . . . . . . . . . . . . . . . . 175
7.2 2D Navier-Stokes equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
7.2.1 Dissipativity and absorbing set . . . . . . . . . . . . . . . . . . . . . . 181
7.2.2 Differentiability of the solution operator . . . . . . . . . . . . . . 183
7.2.3 Estimates of the trace. Attractor dimension . . . . . . . . . . . 184
7.3 2D Ladyzhenskaya model; case r 2 . . . . . . . . . . . . . . . . . . . . . . . 187
7.3.1 Dissipativity and absorbing set . . . . . . . . . . . . . . . . . . . . . . 188
7.3.2 Differentiability of the solution semigroup . . . . . . . . . . . . 190
7.3.3 Estimates of the trace. Attractor dimension. . . . . . . . . . . 192
7.4 3D Ladyzhenskaya model; case r > 11/5 . . . . . . . . . . . . . . . . . . . 194
7.4.1 Existence of exponential attractor . . . . . . . . . . . . . . . . . . . 195
7.4.2 Explicit dimension estimates for r > 12/5 . . . . . . . . . . . . 203

8 2D Model of Ladyzhenskaya for r < 2 . . . . . . . . . . . . . . . . . . . . . 217


8.1 2D Ladyzhenskaya model; case r (1, 2) . . . . . . . . . . . . . . . . . . . 218
8.1.1 Basic mathematical theory . . . . . . . . . . . . . . . . . . . . . . . . . 218
8.1.2 Dissipativity and the dynamics of trajectories . . . . . . . . . 224
8.1.3 Exponential attractor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
8.1.4 Improved estimates for r > 4/3 . . . . . . . . . . . . . . . . . . . . . . 234
8.2 Fluids with pressure-dependent viscosities . . . . . . . . . . . . . . . . . . 238
8.2.1 Basic mathematical theory . . . . . . . . . . . . . . . . . . . . . . . . . 241
8.2.2 Dissipativity and absorbing set . . . . . . . . . . . . . . . . . . . . . . 246
8.2.3 Existence of exponential attractor . . . . . . . . . . . . . . . . . . . 251
8.3 Bibliographical remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
XII Contents

9 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
9.1 Embeddings and interpolations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
9.2 Korn and Poincare type inequalities . . . . . . . . . . . . . . . . . . . . . . . 262
9.3 Div-Curl lemma and related topics . . . . . . . . . . . . . . . . . . . . . . . . . 264
9.4 A result of P.-L. Lions on weak continuity . . . . . . . . . . . . . . . . . . 267
9.5 Coverings of compacts in Bochner spaces. . . . . . . . . . . . . . . . . . . . 272
9.5.1 Bochner spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
9.5.2 Asymptotics of the sequences . . . . . . . . . . . . . . . . . . . . . . . 275
9.5.3 Aubin-Lions lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
9.6 Lieb-Thirring inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
9.7 Properties of the stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
1
Notation and basic definitions

1.1 Notation

If not indicated otherwise, the symbols throughout the book are used in the
following manner:

1.1.1 Symbols

The symbols const, c, ci denote generic real constants. Their concrete value
may change at different parts of the text.
The symbols Z and C denote the sets of integers, and complex numbers,
respectively. The symbol R denotes the set of real numbers, RN is the N -
dimensional Euclidean space.

1.1.2 Euclidean space

The symbol RN stands for a domain - an open connected subset of


the Euclidean space RN . The closure of a set Q RN is denoted by Q, its
boundary is Q. By the symbol 1Q we denote the characteristic function of
the set Q. The outer normal vector to Q is usually denoted by n . The symbol
B(a; r) denotes an (open) ball in RN of center a RN and radius r > 0.
We also write BX (a; r) to indicate a ball with respect to a more general
norm X.
Vectors and vector valued functions ranging in an Euclidean space are
denoted by symbols beginning by a boldface minuscule, for example u , v .
Matrices (tensors) and matrix valued functions are represented by special
Roman characters as S, T, in particular, the identity matrix is denoted by
I = {i,j }N
i,j=1 . The symbol I may also be used to denote the identity operator
in a general setting.
2 1 Preliminaries
T
The transpose of a square matrix A = {ai,j }N N
i,j=1 is A = {aj,i }i,j=1 . The
P N
trace of a square matrix A = {ai,j }N
i,j=1 is trace[A] = i=1 ai,i .

The scalar product of vectors a = [a1 , . . . , aN ], b = [b1 , . . . , bN ] is denoted


by
N
X
a b = ai b i ,
i=1

the scalar product of tensors A = {Ai,j }N N


i,j=1 , B = {Bi,j }i,j=1 reads

N
X
A:B= Ai,j Bj,i .
i,j=1

The symbol a b denotes the tensor product of vectors a , b , specifically,

a b }i,j = ai bj .
a b = {a

The vector product a b is the antisymmetric part of a b. If N = 3, the


vector product of vectors a = (a1 , a2 , a3 ), b = (b1 , b2 , b3 ) is identified with a
vector
a b = (a2 b3 a3 b2 , a3 b1 a1 b3 , a1 b2 a2 b1 ).

The product of a matrix A with a vector b is a vector Abb whose components


are
N
X
[Abb]i = Ai,j bj for i = 1, . . . , N,
j=1

while the product of a matrix A = {Ai,j }N,M M,S


i,j=1 and a matrix B = {Bi,j }i,j=1
is a matrix AB with components
M
X
[AB]i,j = Ai,k Bk,j .
k=1

1.1.3 Norms

The norm in a Banach space X is denoted by the symbol kkX . The Euclidean
norm of a vector a RN is denoted by
v
uN
uX
a| = a a = t
|a a2i .
i=1
1.1 Notation 3

The distance of a vector a to a set K RN is denoted as

a, K] = inf{|a
dist[a a k | | k K},

the semidistance of sets A, B is

dist[A, B] = sup{dist[x, B] | x A}.

We occasionally use the symbol distX [A, B] to indicate the metric used.
The diameter of K is

diam[K] = sup |x y|.


(x,y)K 2

The closure of K with respect to the topology X is denoted by clX (K), the
Lebesgue measure of a set Q is |Q| or vol Q.

1.1.4 Differential operators

The symbol
g
yi g(y) (y), i = 1, . . . , N,
yi
denotes the partial derivative of a function g = g(y), y = [y1 , . . . , yN ], with
respect to the variable yi evaluated at a point y RN . The same notation is
used for the generalized (distributional) derivatives introduced below. In most
cases, we consider functions g = g(t, x) of the time variable t (0, T ) and
the spatial coordinate x = [x1 , x2 , x3 ] R3 . We use italics rather than
boldface minuscules to denote the independent variables although they may
be vectors as the case may be.
Occasionally, we write d/dt to indicate the derivative with respect to the
variable t; it is also often used if u = u(t, x) is considered as a Bochner
function, namely a function t 7 u(t, ) with values in suitable function space,
thus suppressing the variable x.
The gradient of a scalar function g = g(t, x) with respect to the spatial
variable x is a vector

x g(t, x) = [x1 g(t, x), x2 g(t, x), x3 g(t, x)].

The gradient of a vector function v = [v1 (t, x), v2 (t, x), v3 (t, x)] with re-
spect to the space variables x is the matrix

xv (t, x) = {xj vi (t, x)}3i,j=1 ;


4 1 Preliminaries

The divergence of a vector valued function depending on the spatial and


time variables v = [v1 (t, x), v2 (t, x), v3 (t, x)] with respect to the space variable
x is a scalar
X3
divxv (t, x) = xi vi (t, x).
i=1

The divergence of a tensor (matrix-valued) function B = {Bi,j (t, x)}3i,j=1 with


respect to the space variable x is a vector
3
X
[divB]i = [divx B(t, x)]i = xj Bi,j (t, x), i = 1, . . . , 3.
j=1

The vorticity of a vectorial function v = [v1 (t, x), . . . , v3 (t, x)] is an antisym-
metric matrix
n o3
curlxv = xv Tx v = xj vi xi vj .
i,j=1

The vorticity operator in R3 may be interpreted as a vector curl v = x v .

The symbol = x denotes the Laplace operator,

x = divx x .

1.2 Measures and distributions


If not otherwise stated, all function spaces considered in what follows are real.
The duality pairing between an abstract vector space X and its dual X is
denoted as h; iX ;X , or simply h; i. In particular, if X is a Hilbert space, the
symbol h; i denotes the scalar product in X.

1.2.1 Continuously differentiable functions

For Q RN , the symbol C(Q) denotes the set of all continuous functions on
Q. In Q is bounded, the symbol C(Q) denotes the Banach space of functions
continuous on the closure Q endowed with norm

kgkC(Q) = sup |g(y)|.


yQ

Similarly, C(Q; X) denotes the Banach space of vectorial functions continuous


in Q and ranging in a Banach space X with norm
1.2 Measures and distributions 5

kgkC(Q) = sup kg(y)kX .


yQ

The symbol Cweak (Q; X) denotes the space of all vector-valued functions on
Q ranging in a Banach space X continuous with respect to the weak topology.
More specifically, g Cweak (Q; X) if the mapping y 7 kg(y)kX is bounded
and
y 7 hf ; g(y)iX ;X
is continuous on Q for any linear form f belonging to the dual space X .
We say that gn g in Cweak (Q; X) if

hf ; gn iX ;X hf ; giX ;X in C(Q) for all f X .

The symbol C k (Q), Q RN , where k is a non-negative integer, denotes


the space of functions on Q that are restrictions of k-times continuously dif-
ferentiable functions on RN . C k, (Q), (0, 1) is the subspace of C k (Q) of
functions having their k-th derivatives -Holder continuous in Q. C k,1 (Q) is a
subspace of C k (Q) of functions whose k-th derivatives are Lipschitz on Q. For
a bounded domain Q, the spaces C k (Q) and C k, (Q), (0, 1] are Banach
spaces with norms
kukC k (Q) = max sup | u(x)|
||k xQ

and
| u(x) u(y)|
kukC k, (Q) = kukC k (Q) + max sup ,
||=k (x,y)Q2 , x6=y |x y|

where u stands for the partial derivative x11 . . . xNN u of order || =


PN
i=1 i . The spaces C
k,
(Q; RM ) are defined in a similar way. Finally, we
k
set C = k=0 C .

1.2.2 Compactness of sets of continuous functions

Arzel`
a-Ascoli Theorem:
Theorem 1.1 Let Q RM be a compact set and X a compact topological
metric space endowed with a metric dX . Let {vn }n=1 be a sequence of func-
tions in C(Q; X) that is equi-continuous, meaning, for any > 0 there is
> 0 such that
h i
dX vn (y), vn (z) provided |y z| < independently of n = 1, 2, ...
6 1 Preliminaries

Then {vn }n=1 is precompact in C(Q; X), that is, there exists a subse-
quence (not relabeled) and a function v C(Q; X) such that
h i
sup dX vn (y), v(y) 0 as n .
yQ

Proof: See Kelley [96, Chapter 7, Theorem 17].




1.2.3 Measures

The symbol Cc (Q) denotes the space of continuous functions with compact
support in a locally compact Hausdorff metric space Q. The symbol M(Q)
stands for the space of signed Borel measures on Q. The symbol M+ (Q)
denotes the cone of non-negative Borel measures on Q. A measure M+ (Q)
such that [Q] = 1 is called probability measure. Lebesgue measure of a set
A RN is denoted |A| or vol A.

Riesz Representation Theorem:

Theorem 1.2 Let Q be a locally compact Hausdorff metric space. Let f be


a non-negative linear functional defined on the space Cc (Q).
Then there exist a algebra of measurable sets containing all Borel sets
and a unique non-negative measure on f M+ (Q) such that

hf ; gi = g df for any g Cc (Q). (1.1)
Q

Moreover, the measure f enjoys the following properties:


f [K] < for any compact K Q.

f [E] = sup {f [K] | K E, K compact}


for any open set E Q.

f [V ] = inf {(E) | V E, E open}


for any Borel set V .
If E is f measurable, f (E) = 0, and A E, then A is f measurable.

Proof: See Rudin [140, Chapter 2, Theorem 2.14].



1.3 Bochner and Sobolev spaces 7

Corollary 1.1 Assume that f : Cc (Q) R is a linear and non-negative


functional, where Q is a domain in RN .
Then there exists a measure f enjoying the same properties as in The-
orem 1.2 such that f is represented through (1.1).

Proof: See Evans and Gariepy [61, Chapter 1.8, Corollary 1].


1.2.4 Distributions

The symbol Cck (Q; RM ), k {0, 1, . . . , } denotes the vector space of func-
tions belonging to C k (Q; RM ) and having compact support in Q. The space
Cc (Q; RM ) may be endowed with the topology induced by the convergence:

n Cc (Q)

if
supp[n ] K, K Q a compact set,
(1.2)
n in C k (K) for any k = 0, 1, . . .
We write Cc (Q) instead of Cc (Q; R).
The dual space [Cc (Q; RM )] is the space of distributions on with values
in RM . Continuity of a linear form belonging to this space is understood with
respect to the convergence introduced in (1.2).
A differential operator of order || can be identified with a distribution

||
h v; i = (1) v dy
Q

for any locally integrable function v.

1.3 Bochner and Sobolev spaces

1.3.1 Lebesgue (Bochner) spaces

The Lebesgue spaces Lp (Q; X) are spaces of (Bochner) measurable functions


v ranging in a Banach space X such that the norm

kvkpLp (Q;X) = kvkpX dy is finite, 1 p < .
Q

Similarly, v L (Q; X) if v is (Bochner) measurable and


8 1 Preliminaries

kvkL (Q;X) = ess sup kv(y)kX < .


yQ

The symbol Lploc (Q; X) denotes the vector space of locally Lp -integrable
functions, meaning

v Lploc (Q; X) if v Lp (K; X) for any compact set K in Q.

We write Lp (Q) for Lp (Q; R).


Let f L1loc (Q) where Q is an open set. A Lebesgue point a Q of f in
Q is characterized by the property

1
lim f (x)dx = f (a). (1.3)
r0+ |B(a, r)| B(a;r)

For f L1 (Q) the set of all Lebesgue points is of full measure, meaning
its complement in Q is of zero Lebesgue measure. A similar statement holds
for vector valued functions f L1 (Q; X), where X is a Banach space (see
Brezis [16]).
If f C(Q), then identity (1.3) holds for all points a in Q.

Aubin-Lions lemma:

Theorem 1.3 Let p, q [1, ), let X, Y and Z be Banach spaces, such


that Y X Z, the embedding Y Z being compact, and let moreover Y ,
Z be reflexive.
Then the space
 d
u Lp (0, T ; Y ) | u Lq (0, T ; Z)
dt
is compactly embedded into the space Lp (0, T ; X).

Proof: See Simon [149] for a general case; a constructive proof for certain
special cases is also given in Section 9.5 of the Appendix.


Linear Functionals on Lp (Q; X):

Theorem 1.4 Let Q RN be a measurable set, X a Banach space that is


reflexive and separable, 1 p < .
Then any continuous linear form [Lp (Q; X)] admits a unique rep-

resentation w Lp (Q; X ),

h; viLp (Q,X );Lp (Q;X)


1.3 Bochner and Sobolev spaces 9

= hw (y); v(y)iX ;X dy for all v Lp (Q; X),
Q

where
1 1
+ = 1.
p p
Moreover the norm on the dual space is given as

kk[Lp(Q;X)] = kw kLp (Q;X ) .

Accordingly, the spaces Lp (Q; X) are reflexive for 1 < p < as soon as X
is reflexive and separable.

Proof: See Gajewski et al. [74, Chapter IV, Theorem 1.14, Remark 1.9].


Identifying with w , we write



[Lp (Q; RN )] = Lp (Q; RN ), kk[Lp (Q;RN )] = kkLp (Q;RN ) , 1 p < .

This formula is known as Riesz representation theorem.


If the Banach space X in Theorem 1.4 is merely separable, we have

[Lp (Q; X)] = Lpweak() (Q; X ) for 1 p < ,

where
Lpweak() (Q; X )
n

:= : Q X y Q 7 h(y); viX ;X measurable for any fixed v X,

o
y 7 k(y)kX Lp (Q)

(see Edwards [56], Pedregal [128, Chapter 6, Theorem 6.14]).


H
olders inequality reads
1 1 1
kuvkLr (Q) kukLp(Q) kvkLq (Q) , = +
r p q

for any u Lp (Q), v Lq (Q), Q RN (see Adams and Fournier [1, Chapter
2]).
Interpolation inequality for Lp spaces reads

(1) 1 1
kvkLr (Q) kvkLp(Q) kvkLq (Q) , = + , p < r < q, (0, 1)
r p q

for any v Lp Lq (Q), Q RN (see Adams and Fournier [1, Chapter 2]).
10 1 Preliminaries

1.3.2 Sobolev spaces

The Sobolev spaces W k,p (Q; RM ), 1 p , k a positive integer, are


the spaces of functions having all distributional derivatives up to order k
in Lp (Q; RM ). The norm in W k,p (Q; RM ) is defined as
 1/p
PM P p
i=1 ||k k v k
i Lp (Q) if 1 p <
kvv kW k,p (Q;RM ) = ,




max1iM, ||k {k vi kL (Q) } if p =

where the symbol stands for any partial derivative of order ||.
If Q is a bounded domain with boundary of class C k1,1 , then there exists
a continuous linear operator which maps W k,p (Q) to W k,p (RN ); it is called
extension operator. If, in addition, 1 p < , then W k,p (Q) is separable and
the space C k (Q) is its dense subspace.
By a regular domain, we will understand a bounded domain with boundary
of class C 3,1 ; that is to say, the boundary can be covered by a finite number
of graphs of C 3 functions.
The space W 1, (Q), where Q is a bounded domain, is isometrically iso-
morphic to the space C 0,1 (Q) of Lipschitz functions on Q.
For basic properties of Sobolev functions, see Adams and Fournier [1,
Chapter 2] or Ziemer [163].

1.3.3 Dual Sobolev spaces

The symbol W0k,p (Q; RM ) denotes the completion of Cc (Q; RM ) with respect
to the norm k kW k,p (Q;RM ) . In what follows, we identify W 0,p (; RN ) =
W00,p (; RN ) with Lp (; RN ). 
1,p (Q) = W 1,p (Q)
We denote L p (Q) = {u Lp (Q) | Q u dy = 0} and W
Lp (Q). If Q R is a bounded domain, then Lp (Q) and W
N 1,p (Q) can be
p 1,p
viewed as closed subspaces of L (Q) and W (Q), respectively.

Dual Sobolev Spaces:

Theorem 1.5 Let RN be a domain, and let 1 p < . Then the dual
space [W0k,p ()] is a proper subspace of the space of distributions d ().
Moreover, any linear form f [W0k,p ()] admits a representation
X 
hf ; vi[W k,p ()] ;W k,p () = (1)|| w v dx, (1.4)
0 0
||k

1 1
where w Lp (), + = 1.
p p
1.3 Bochner and Sobolev spaces 11

The norm of f in the dual space is given as


n P 1/p o

p

inf ||k kw k
Lp ()
w satisfy (1.4)






for 1 < p < ;
kf k[W k,p()] =
0
n o



inf max ||k {kw k L (Q) } w satisfy (1.4)





if p = 1.

The infimum is attained in both cases.

Proof: See Adams and Fournier [1, Theorem 3.12], Mazya [121, Section
1.1.14].


The dual space to the Sobolev space W0k,p () is denoted as W k,p ().

The dual to the Sobolev space W k,p () admits formally the same representa-
tion formula as (1.4). However it cannot be identified as a space of distributions
on . A typical example is the linear form

hf ; vi = w f x v dx, with divxw f = 0

that vanishes on d() but generates a non-zero linear form when applied to
v W 1,p ().

Rellich-Kondrachov Embedding Theorem:

Theorem 1.6 Let RN be a bounded Lipschitz domain.


(i) Then, if kp < N and p 1, the space W k,p () is continuously embedded
in Lq () for any
Np
1 q p = .
N kp
Moreover, the embedding is compact if k > 0 and q < p .
(ii) If kp = N , the space W k,p () is compactly embedded in Lq () for any
q [1, ).
(iii) If kp > N then W k,p () is continuously embedded in C k[N/p]1, (),
where [ ] denotes the integer part and
12 1 Preliminaries

[ Np ] + 1 N
p if N
p
/ Z,
=
N
arbitrary positive number in (0, 1) if p Z.

Moreover, the embedding is compact if 0 < < [ Np ] + 1 N


p.

Proof: See Ziemer [163, Theorem 2.5.1, Remark 2.5.2].




1.3.4 Embeddings for dual Sobolev spaces

The following result may be regarded as a direct consequence of Theorem 1.6.

Embedding Theorem for Dual Sobolev Spaces:

Theorem 1.7 Let RN be a bounded domain. Let k > 0 and q <


satisfy
p Np
q> , where p = if kp < N,
p 1 N kp
q > 1 for kp = N,
or
q 1 if kp > N.

Then the space L () is compactly embedded into the space W k,p (),
q

1/p + 1/p = 1.

1.3.5 Traces

The Sobolev-Slobodeckii spaces W k+,p (Q), 1 p < , 0 < < 1, k =


0, 1, . . ., where Q is a domain in RL , L a positive integer, are Banach spaces
of functions with finite norm
p1
X   p
| v(y) v(z)|
W k+,p (Q) = kvkpW k,p (Q) + dy dz ,
Q Q |y z|L+p
||=k

see Necas[126, Section 2.3.8] for example.


Let RN be a bounded Lipschitz domain. Referring to the notation in-
troduced in (i), we say that f W k+,p () if (f )(I , ) W k+,p (RN 1 )
for any = B with B belonging to the covering B of and the
1.3 Bochner and Sobolev spaces 13

corresponding term in the partition of unity F. The space W k+,p () is a


Banach space endowed with a norm
M
X
kvkpW k+,p () = k(vi ) (I , )kpW k+,p (RN 1 ) .
i=1

In the above formulas (I , ) : RN 1 RN maps y to (y , (y )). For more


details see for example Necas [126, Section 3.8].


In the situation when RN is a bounded Lipschitz domain, the Sobolev-
Slobodeckii spaces admit similar embeddings as classical Sobolev spaces. The
embeddings
W k+,p () Lq () and W k+,p () C s ()
Np
are compact provided 0 < (k + )p < N , 1 q < N (k+)p , and s =
0, 1, . . . , k, (k s + )p > N , respectively. The former embedding remains
Np
continuous (but not compact) at the border case q = N (k+)p . See Adams
and Fournier [1, Chapter 6].

Trace Theorem for Sobolev Spaces and Greens formula:

Theorem 1.8 Let RN be a bounded Lipschitz domain.


Then there exists a linear operator 0 with the following properties:

[0 (v)](x) = v(x) for x provided v C (),

k0 (v)k 1 1 ,p
p
ckvkW 1,p () for all v W 1,p (),
W ()

ker[0 ] = W01,p ()
provided 1 < p < .
Conversely, there exists a continuous linear operator
1
: W 1 p ,p () W 1,p ()

such that 1
0 ((v)) = v for all v W 1 p ,p ()
provided 1 < p < .
In addition, the following formula holds:
  
xi uv dx + uxi v dx = 0 (u)0 (v)ni dSx , i = 1, . . . , N,


for any u W 1,p (), v W 1,p (), where n is the outer normal vector to
the boundary .
14 1 Preliminaries

Proof: See Necas [126, Theorems 5.5, 5.7].



1 1
The dual [W 1 p ,p ()] to the Sobolev-Slobodeckii space W 1 p ,p ()) =
1
,p 1 ,p
W p () is denoted simply by W p ().

1.3.6 Interpolation

If RN is a bounded Lipschitz domain, then we have the interpolation


inequality

kvkW ,r () ckvkW ,p () kvk1


W ,q () , 0 1, (1.5)

for
1 1
0 , , 1, 1 < p, q, r < , = + (1 ), = +
r p q
(see Sections 2.3.1, 2.4.1, 4.3.2 in Triebel [155]). We also refer to Appendix for
proofs of several other interpolation, as for example Ladyzhenskaya inequality.

1.4 Fourier analysis


Let v = v(x) be a complex valued function integrable on RN . The Fourier
transform of v is a complex valued function Fx [v] of the variable RN
defined as
 N/2 
1
Fx [v]() = v(x) exp(i x) dx. (1.6)
2 RN

Therefore, the Fourier transform Fx can be viewed as a continuous linear


mapping defined on L1 (RN ) with values in L (RN ).
For u, v complex valued square integrable functions on RN , we have Par-
sevals identity:
 
u(x)v(x) dx = Fx [u]()Fx [v]() d,
RN RN

where bar denotes the complex conjugate. Parsevals identity implies that
Fx can be extended as a continuous linear mapping defined on L2 (RN )
with values in L2 (RN ). Its inverse reads
 1 N/2 
1
Fx [f ] = f ()exp(ix )d. (1.7)
2 RN
1.4 Fourier analysis 15

1.4.1 Fourier transform of tempered distributions

The symbol S(RN ) denotes the space of smooth rapidly decreasing (complex
valued) functions, specifically, S(RN ) consists of functions u such that

sup sup (1 + |x|2 )s | u| <


||m xRN

for all s, m = 0, 1, . . . . We say that un u in S(RN ) if

sup sup (1 + |x|2 )s | (un u)| 0, s, m = 0, 1, . . . . (1.8)


||m xRN

The space of tempered distributions is identified as the dual S (RN ). Con-


tinuity of a linear form belonging to S (RN ) is understood with respect to
convergence introduced in (1.8).
The Fourier transform of a tempered distribution f S (RN ) is defined
as
hFx [f ]; gi = hf ; Fx [g]ifor any g S(RN ). (1.9)
It is a continuous linear operator defined on S (RN ) onto S (RN ) with the
1
inverse Fx ,
1
hFx 1
[f ], gi = hf, Fx [g]i, f S (RN ), g S(RN ). (1.10)

We recall formulas

k Fx [f ] = Fx [ixk f ], Fx [xk f ] = ik Fx [f ], (1.11)

where f S (RN ), and


   
Fx [f g] = Fx [f ] Fx [g] , (1.12)

where f S(RN ), g S (RN ) and denotes denotes convolution.


A differential operator D of order m,
X
D= a ,
||m

can be defined by means of a Fourier multiplier in the form


X
=
D a (i) , = 11 . . . N
N
,
||m

in the sense that



X
1
D[v](x) = Fx a (i) Fx [v]() , v S(RN ).
||m
16 1 Preliminaries

The operators defined through the right-hand side of the above expression are
called pseudodifferential operators.
Various pseudodifferential operators used in the book are identified through
their Fourier symbols:
Riesz transform:
ij
Rj , j = 1, . . . , N.
||
Inverse Laplacian:
1
()1 .
||2
The double Riesz transform:
i j
{R}N
i,j=1 , R =
1
x x , Ri,j , i, j = 1, . . . , N.
||2
Inverse divergence:
ij
Aj = xj 1 , j = 1, . . . , N.
||2

We denote
3
X 3
X
A:R Ai,j Ri,j , R[vv ]i Ri,j [vj ], i = 1, 2, 3.
i,j=1 j=1

1.4.2 Boundedness of Fourier multipliers

H
ormander-Mikhlin Theorem:

Theorem 1.9 Consider an operator L defined by means of a Fourier mul-


tiplier m = m(),
1
L[v](x) = Fx [m()Fx [v]()] ,

where m L (RN ) has classical derivatives up to order [N/2]+1 in RN \{0}


and satisfies
| m()| c |||| , 6= 0,
for any multiindex such that || [N/2] + 1, where [ ] denotes the integer
part.
Then L is a bounded linear operator on Lp (RN ) for any 1 < p < .

Proof: See Stein [150, Chapter 4, Theorem 3].



1.5 Compactness of families of functions 17

1.5 Compactness of families of functions


Let X be a Banach space, BX the (closed) unit ball in X, and BX the (closed)
unit ball in the dual space X .
Here are some known facts concerning weak compactness:

(1) BX is weakly compact only if X is reflexive. This is stated in Kakutanis


theorem, see Theorem III.6 in Brezis [17].
(2) BX is weakly-(*) compact. This is Banach-Alaoglu-Bourbaki theorem,
see Theorem III.15 in Brezis [17].
(3) If X is separable, then BX is sequentially weakly-(*) compact, see The-
orem III.25 in Brezis [17].
(4) A non-empty subset of a Banach space X is weakly relatively com-
pact only if it is sequentially weakly relatively compact. This is stated
in Eberlein-Shmuliyan-Grothendieck theorem, see Kothe [97], Paragraph
24, 3.(8); 7.

In view of these facts:


(1) Any bounded sequence in Lp (Q), where 1 < p < and Q RN is a
domain, is weakly (relatively) compact.
(2) Any bounded sequence in L (Q), where Q RN is a domain, is weakly-
(*) (relatively) compact.

1.5.1 Weak compactness of integrable functions

Since L1 is neither reflexive nor dual of a Banach space, the uniformly bounded
sequences in L1 are in general not weakly relatively compact in L1 . On the
other hand, the property of weak compactness is equivalent to the property
of sequential weak compactness.

Weak Compactness in the Space L1 :

Theorem 1.10 Let V L1 (Q), where Q RM is a bounded measurable


set.
Then the following statements are equivalent:
(i) any sequence {vn }
n=1 V contains a subsequence weakly converging in
L1 (Q);
(ii) for any > 0 there exists k > 0 such that

|v(y)| dy for all v V;
{|v|k}
18 1 Preliminaries

(iii) for any > 0 there exists > 0 such for all v V

|v(y)|dy <
M

for any measurable set M Q such that

|M | < ;

(iv) there exists a non-negative function C([0, )),

(z)
lim = ,
z z
such that 
sup (|v(y)|) dy < .
vV Q

Proof: See Ekeland and Temam [58, Chapter 8, Theorem 1.3], or Pedregal
[128, Lemma 6.4].


Condition (iii) is termed equi-integrability of a given set of integrable


functions and the equivalence of (i) is Dunford-Pettis theorem (cf. Diestel
[47, p.93]). Condition (iv) is called De la Valle-Poussin criterion, see Pedre-
gal [128, Lemma 6.4]. The statement there exists a non-negative function
C([0, )) in condition (iv) can be replaced by there exists a non-
negative convex function on [0, ).

1.5.2 Young measures

Let Q RN be a domain. We say that : Q RM is a Caratheodory function


on Q RM if

for a. a. x Q, the function 7 (x, ) is continuous on RM ;
(1.13)

for all RM , the function x 7 (x, ) is measurable on Q.

We say that {x }xQ is a family of parametrized measures if x is a prob-


ability measure for a.a. x Q, and if

the function x RM () dx () := hx , i is measurable on Q

for all : RM R, C(RM ) L (RM )).
(1.14)
1.6 Gronwalls lemma 19

Fundamental Theorem of the Theory of


Parameterized (Young) Measures:

N
Theorem 1.11 Let {vv n } n=1 , v n : Q R RM be a sequence of func-
tions bounded in L (Q; R ), where Q is a domain in RN .
1 M

Then there exist a subsequence (not relabeled) and a parameterized fam-


ily {y }yQ of probability measures on RM depending measurably on y Q
with the following property:
For any Caratheodory function = (y, z), y Q, z RM such that

(, v n ) weakly in L1 (Q),

we have 
(y) = (y, z) dy (z) for a.a. y Q.
RM

Proof: See Pedregal [128, Chapter 6, Theorem 6.2].




The family of measures {y }yQ associated to a sequence {vv n }


n=1 ,

v n v weakly in L1 (Q; RM ),

is termed Young measure. We shall systematically denote by the symbol (, v )


the weak limit associated to {(, v n )}
n=1 via the corresponding Young mea-
sure constructed in Theorem 1.11. Note that Young measure need not be
unique for a given sequence.

1.6 Gronwalls lemma

Gronwalls lemma:

Theorem 1.12 Let be a nonnegative, continuous function, let a be a non-


negative integrable function, and let b 0 be a real constant. Assume that
 t
(t) b + a(s)(s) ds for all t [0, T ].
0

Then  
t
(t) b exp a(s) ds for all t [0, T ].
0

In particular, 0 on [0, T ] if b = 0.
20 1 Preliminaries

Proof: See for example Chicone [31, Chapter 2, Theorem 2.1].




You might also like