Existence Theory for Generalized

Newtonian Fluids 1st Edition Breit -

eBook PDF
Go to download the full and correct content document:
https://ebooksecure.com/download/existence-theory-for-generalized-newtonian-fluids-
ebook-pdf/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Hydrodynamic Lubrication of Non-Newtonian Fluids 1st

Edition - eBook PDF

https://ebooksecure.com/download/hydrodynamic-lubrication-of-non-
newtonian-fluids-ebook-pdf/

The Physics of Fluids and Plasmas: An Introduction for

Astrophysicists (eBook PDF)

http://ebooksecure.com/product/the-physics-of-fluids-and-plasmas-
an-introduction-for-astrophysicists-ebook-pdf/

(eBook PDF) Mechanics of Fluids, SI Edition 5th Edition

http://ebooksecure.com/product/ebook-pdf-mechanics-of-fluids-si-
edition-5th-edition/

(eBook PDF) Riddles of Existence: A Guided Tour of

Metaphysics: New Edition 2nd Edition,

http://ebooksecure.com/product/ebook-pdf-riddles-of-existence-a-
guided-tour-of-metaphysics-new-edition-2nd-edition/
The Generalized Triangle Inequalities in Symmetric
Spaces and Buildings with Applications to Algebra

http://ebooksecure.com/product/the-generalized-triangle-
inequalities-in-symmetric-spaces-and-buildings-with-applications-
to-algebra/

(eBook PDF) Fluids & Electrolytes Made Incredibly Easy!

(Incredibly Easy! Series) 6th Edition

http://ebooksecure.com/product/ebook-pdf-fluids-electrolytes-
made-incredibly-easy-incredibly-easy-series-6th-edition/

Malware Diffusion Models for Wireless Complex Networks.

Theory and Applications 1st Edition Karyotis - eBook
PDF

https://ebooksecure.com/download/malware-diffusion-models-for-
wireless-complex-networks-theory-and-applications-ebook-pdf/

Solutions for Biot's Poroelastic Theory in Key

Engineering Fields. Theory and Applications 1st Edition
Edition Yuanqiang Cai And Honglei Sun (Auth.) - eBook
PDF
https://ebooksecure.com/download/solutions-for-biots-poroelastic-
theory-in-key-engineering-fields-theory-and-applications-ebook-
pdf/

(eBook PDF) Practical Risk Theory for Actuaries

http://ebooksecure.com/product/ebook-pdf-practical-risk-theory-
for-actuaries/
EXISTENCE
THEORY FOR
GENERALIZED
NEWTONIAN
FLUIDS
Mathematics in Science and
Engineering

EXISTENCE
THEORY FOR
GENERALIZED
NEWTONIAN
FLUIDS

Dominic Breit
Mathematical & Computer Sciences; Mathematics
Heriot-Watt University
Edinburgh, Midlothian, United Kingdom

Series Editor
Goong Chen

AMSTERDAM • BOSTON • HEIDELBERG • LONDON

NEW YORK • OXFORD • PARIS • SAN DIEGO
SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO
Academic Press is an imprint of Elsevier
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1800, San Diego, CA 92101-4495, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
Copyright © 2017 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and
retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publisher’s permissions policies and our arrangements
with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency,
can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating
and using any information, methods, compounds, or experiments described herein. In using such
information or methods they should be mindful of their own safety and the safety of others,
including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-811044-7

For information on all Academic Press publications

visit our website at https://www.elsevier.com

Publisher: Nikki Levy

Acquisition Editor: Graham Nisbet
Editorial Project Manager: Susan Ikeda
Production Project Manager: Paul Prasad Chandramohan
Designer: Maria Ines Cruz
Typeset by VTeX
Dedicated
to
Janine
PREFACE

In continuum mechanics, the motion of an incompressible, homo-

geneous fluid is described by the velocity field and the hydrodynamical
pressure. The time evolution of the fluid is governed by the Navier–
Stokes system of partial differential equations which describes the balance
of mass and momentum. In the classical formulation – which goes back
to C.-L. Navier and G.G. Stokes – the relation between the viscous stress
tensor and the symmetric gradient of the velocity field is linear (i.e. we
have a Newtonian fluid). This system is already quite challenging from a
mathematical point of view and has fascinated many mathematicians. How-
ever, it can only model fluids with a very simple molecular structure such
as water, air and several oils. In order to study more complex fluids one has
to deal with a generalized Navier–Stokes system. In this model for gen-
eralized Newtonian fluids, the viscosity ν is assumed to be a function of
the shear-rate γ̇ . Very popular among rheologists is the power-law model
in which the generalized viscosity function is of power-type ν ∼ γ̇ p−2 with
p > 1. For a specific fluid physicists can identify this power by experiments.
Instead of the Laplacian, as in the classical Navier–Stokes system, the main
part of the mathematical model is a power-type nonlinear second order dif-
ferential operator. So, in addition to the convective term, a second highly
nonlinear term appears. The only framework which is available today is the
concept of weak solutions. These solutions belong to appropriate Sobolev
spaces: Derivatives have to be understood in the sense of distributions and
singularities may occur.
The mathematical observation of power-law fluids began in the late
sixties in the pioneering work of J.-L. Lions and O.A. Ladyzhenskaya. The
cases they could handle were restrictive for real world situations. Never-
theless, it was a breakthrough in the theory of partial differential equations.
Since then, there has been huge progress in the mathematical theory of
generalized Newtonian fluids. The first systematic study was initiated by
the group around J. Nečas in the 1990s. Situations which are realistic for
physical and industrial applications could finally be handled. In the 2000s
the Lipschitz truncation (the approximation of a Sobolev function by a Lip-
schitz continuous one done in such a way that both are equal outside of a
small set whose size can be controlled) was shown to be a powerful tool
in the analysis of generalized Newtonian fluids. A very wide range of non-
Newtonian fluids could finally be included in the mathematical theory. The

ix
x Preface

best known bound p > 65 for three dimensional flows of power-law fluids
was achieved.
A drawback of the Lipschitz truncation is its nonlinear and nonlo-
cal character. In fact, the property of a function to be solenoidal (that
is, divergence-free) is lost by truncating it. So, one has to introduce the
pressure function in the weak formulation which results in additional tech-
nical difficulties. An advanced pressure decomposition via singular integral
operators is necessary in the non-stationary case. An improved version, the
“solenoidal Lipschitz truncation”, was developed only recently. It allows the
existence of weak solutions to the generalized Navier–Stokes system to be
shown without the appearance of the pressure function and therefore highly
simplifies the proofs. Moreover, it allows the Prandtl–Eyring fluid model to
be studied which was out of reach before. In this model the power-growth
is replaced by some logarithmic function: the law ν ∼ ln(1 + γ̇ )/γ̇ was in-
troduced in 1936 based on a molecular theory. This leads to a limit case in
the functional analytical setting of generalized Newtonian fluids. It is not
possible to introduce the pressure in the expected function space. Neither
can the divergence be corrected.
The aim of this book is to present a complete and rigorous mathe-
matical existence theory for generalized Newtonian fluids – for stationary,
non-stationary and stochastic models. The balance laws are formulated in
all situations via a generalized Navier–Stokes system. The proofs are pre-
sented as self-contained as possible and require from the reader only basic
knowledge of nonlinear partial differential equations.
The heart of this book is the construction of the “solenoidal Lipschitz
truncation”. It has numerous applications and is of interest for future re-
search beyond the scope of this monograph. The stationary truncation is
presented in Chapter 3 and the non-stationary version in Chapter 6. Based
on the “solenoidal Lipschitz truncation” the existence of weak solutions to
generalized Navier–Stokes equations is shown. The existence proof itself is
only slightly more complicated than the classical monotone operator theory
and easy to follow.
In Chapter 4 we study the stationary Prandtl–Eyring model in two di-
mensions. Here, several important tools like the Bogovskiı̆ operator and
Korn’s inequality loose some of their continuity properties. Optimal re-
sults, which might allow for a loss of integrability, can be achieved in the
framework of Orlicz spaces. They are flexible enough to study fine prop-
erties of measurable functions which are required in the Prandtl–Eyring
fluid model. We present optimal versions of the Bogovskiı̆ operator, Nečas’
 Preface xi

negative norm theorem and Korn’s inequality in this framework in Chap-

ter 2. The background about Orlicz- and Orlicz–Sobolev spaces is revised
in Chapter 1.
In Chapter 7 we deal with non-stationary flows of power law fluids.
A first step is to approximate the equations by a system whose solutions are
known to exist. In order to pass to the limit in the regularization parame-
ter, one has to apply compactness of the velocity, the “solenoidal Lipschitz
truncation” and arguments from monotone operator theory.
In the last part of the book we study stochastic partial differential equa-
tions in fluid mechanics. Probabilistic models have become more and more
important for applications and earned a strong interest amongst mathemati-
cians. They can, for instance, take into account physical uncertainties and
model turbulence in the fluid motion. We present first existence theorems
for generalized Navier–Stokes equations under random perturbations. The
results and methods build a basis for future research on stochastic partial
differential equations in the analysis of generalized Newtonian fluids. All
probabilistic tools are presented as well. The Chapter can be viewed as an
introduction to stochastic partial differential equations from an analytical
point of view. Thus, the proofs of the main result given in Chapter 10 are
accessible to analysts without prior knowledge in stochastic analysis.
ACKNOWLEDGMENT

This monograph is based on my habilitation thesis [26]. It contains re-

sults from the papers [27,28,30,31,33,34]. Note that the results from [31]
are an improvement on [32] and are not yet contained in [26]. Thanks goes
to J. Frehse, J. Málek and E. Süli for being referees of [26].
Special thanks go to all the people without whom the writing of this
book would have been impossible: to M. Fuchs for introducing me to
mathematical fluid mechanics, to L. Diening for being a constant source of
inspiration, for much valuable advise and his unique view on the Lipschitz
truncation method, to A. Cianchi for sharing his deep knowledge of Or-
licz spaces and to M. Hofmanová for opening the door for me to stochastic
partial differential equations. I would like to extend my sincere appreciation
to all the people with whom I have discussed research related to this mono-
graph, in particular: M. Bildhauer, J. Frehse, S. Schwarzacher. Moreover,
thanks goes to all people who carefully read parts of it: L. Diening, M. Hof-
manová, M.V. Lawson, M. Ottobre, S. Paulus, O. Penrose, B.P. Rynne,
Z.B. Wyatt.

Dominic Breit
Edinburgh
August 2016

xiii
NOTATION

Formula Meaning
C Continuous functions
Cα α -Hölder continuous functions
Ck k-times continuously differentiable functions
C∞ ∞-times continuously differentiable functions
spt(ϕ) Support of ϕ
C0∞ C ∞ -functions with compact support
C0k C k -functions with compact support
D Dual of C0∞
C0∞,div C0∞ -functions with vanishing divergence

Ddiv Dual of C0∞,div
C0∞,⊥ C0∞ -functions with vanishing mean value
X Dual space of X
 · X Norm on X
·, ·X Inner product on X
Ln n-dimensional Lebesgue measure
Hs s-dimensional Hausdorff measure
B Borelian σ -algebra
Lp Lebesgue-space of p-integrable functions
p
Lloc Lebesgue-space of locally p-integrable functions
p
Ldiv L p -functions with vanishing divergence
L⊥p L p -functions with vanishing mean
 
p Dual exponent of p: p = p/ p − 1
L A /A(L ) Orlicz-space generated by A
L⊥A /A(L )⊥ L A -functions with vanishing mean

A Young conjugate of A
u∗ Decreasing rearrangement of u
u∗v Convolution of u and v
Mf Hardy-Littlewood maximal operator applied to f
Br Ball with radius r
κ Br Ball with same center as Br and radius κ r
k Averaged Taylor polynomial of order k
W k ,p Sobolev functions with differentiability k and
integrability p

xv
xvi Notation

Formula Meaning
Wlock,p Sobolev functions with differentiability k and local
integrability p
k ,p
Wdiv W k,p -functions with vanishing divergence
W0k,p W k,p -functions with vanishing trace
W −k,p Dual space of W0k,p
k ,p
Wdiv W k,p -functions with vanishing divergence
W0k,div
,p
W0k,p -functions with vanishing divergence
−k,p
Wdiv Dual space of W0k,div
,p

W 1,A Orlicz-Sobolev space generated by A

W01,A W 1,A -functions with vanishing trace
×d
Rdsym Symmetric d × d matrices
Symmetric tensor product
Ep L p -functions with symmetric gradient in L p
E0p Ep -functions with vanishing trace
E0p,div E0p -functions with vanishing divergence
EA L A -functions with symmetric gradient in L A
E0A EA -functions with vanishing trace
E0A,div E0A -functions with vanishing divergence
BD Functions of bounded deformation
M Bounded signed measures
H A /H A(L) L A -functions with divergence in L A
H0A /H0A(L) H A -functions with vanishing normal trace
Bog Bogovskiı̆-operator
−1 Solution operator to the Laplace equation
−2 Solution operator to the bi-Laplace equation
curl−1 Solution operator to the curl equation
Qr Parabolic cube with radius r
κ Qr Cube with same center as Qr and radius κ r
Mα α -parabolic maximal function
Mασ α -parabolic maximal function with power σ
L p (0, T ; V ) Bochner measurable functions with values in V and
integrability p
C ([0, T ]; V ) Continuous functions with values in V
C α ([0, T ]; V ) α -Hölder continuous functions with values in V
Cw ([0, T ]; V ) Weakly continuous functions with values in V
 Notation xvii

Formula Meaning
W k,p (0, T ; V ) k-times weakly differentiable functions with values in V
and integrability p
(, F , P) Probability space with sample space , σ -algebra F and
probability measure P
(Ft )t≥0 filtration
L p (, F , P; V ) Random variable over (, F , P) with values in V and
moments of order p
M2 Quadratically integrable martingales
M2c Quadratically integrable continuous martingales
M2c,loc Locally in time quadratically integrable continuous
martingales
L2 (H1 , H2 ) Hilbert-Schmidt operators from H1 → H2
 
X t t≥0 Quadratic variation process of (Xt )t≥0
 
X , Y t t≥0 Covariation process of (Xt )t≥0 and (Yt )t≥0
Aq Stokes operator on L q
Aq A-Stokes operator on L q
D Domain of an operator
CHAPTER 1

Preliminaries
Contents
1.1 Lebesgue & Sobolev spaces 3
1.2 Orlicz spaces 8
1.3 Basics on Lipschitz truncation 11
1.4 Existence results for power law fluids 22

Abstract
In this chapter we present some preliminary material which will be needed in order to
study stationary models for generalized Newtonian fluids. We begin with the functional
analytic framework. In particular, we define Lebesgue-, Sobolev- and Orlicz-spaces and
describe their basic properties. After this we present the Lipschitz truncation method
in its classical framework and present two applications. Finally, we discuss some mod-
elling aspects concerning power law fluids and provide a historical overview on the
mathematical theory of weak solutions for stationary flows.

1.1 LEBESGUE & SOBOLEV SPACES

In this section we define various function spaces. For proofs, further details
and references we refer to [5].
Definition 1.1.1 (Classical function spaces). Let G ⊂ Rd be open and
k ∈ N. We define
C (G) := {u : G → R : u is continuous} ,
 
C k (G) := u : G → R : ∇ i u is continuous for i = 0, ..., k ,
 
C ∞ (G) := u : G → R : ∇ i u is continuous for all i ∈ N0 .
Remark 1.1.1. In Definition 1.1.1 if we replace G by the closed set G
we are considering functions whose derivatives are continuous up to the
boundary of G.
Definition 1.1.2. Let G ⊂ Rd be open and α ∈ (0, 1]. We define
   
u(x) − u(y)
C (G) := u : G → R : sup 
α
 <∞ ,
x=y x − yα
 
C α (G) := u : G → R : u ∈ C α (K ) for all K  G
as the set of (locally) α -Hölder continuous functions.

Existence Theory for Generalized Newtonian Fluids.

DOI: http://dx.doi.org/10.1016/B978-0-12-811044-7.00002-1
Copyright © 2017 Elsevier Inc. All rights reserved. 3
4 Existence Theory for Generalized Newtonian Fluids

Remark 1.1.2. In the case α = 1 we obtain the set of Lipschitz continuous

functions.
Definition 1.1.3. Let G ⊂ Rd be open and u : G → R. We define the
support of u by

spt u := {x ∈ G : u(x) = 0}.

By C00 (G) and C0∞ (G) we denote subclasses of C 0 (G) and C ∞ (G) whose
elements satisfy spt u  G.
Definition 1.1.4 (Lebesgue spaces). Let (X, , μ) be a measure space. We
define
 
L (X, , μ) := u : X → R : u is μ-measurable,
p
|u|p dμ < ∞ , 1 ≤ p < ∞,
 X

L ∞ (X, , μ) := u : X → R : u is μ-measurable, inf sup |u(x)| < ∞ .

μ(N )=0 x∈X\N

The elements of L p (X, , μ) are equivalence classes. Two functions u and

v belong to the same equivalence class if u = v μ-almost everywhere in X,
i.e. if
μ {x ∈ X : u(x) = v(x)} = 0.

Remark 1.1.3. • L p (X, , μ) is a Banach space together with the natu-

ral norm
 1
p
u p := u L p (X) := |u| dμ
p
for 1 ≤ p < ∞,
X
u ∞ := u L ∞ (X) := inf sup |u(x)| .
μ(N )=0 x∈X\N

• We set p = p−p 1 for 1 < p < ∞, p = ∞ for p = 1 and p = 1 for p = ∞.

Let 1 ≤ p < ∞ and v ∈ L p (X, , μ). Then the mapping

L p (X, , μ) u → u v dμ
G
belongs to the dual space L p (X, , μ) . On the other hand each element
of L p (X, , μ) can be represented via a function v ∈ L p (X, , μ). In
particular, the mapping

L p (X, , μ) v → L p (X, , μ) u → u v dμ ∈ L p (X, , μ)
X

is an isomorphism.
 Preliminaries 5

• The space L p (X, , μ) is reflexive if and only if 1 < p < ∞.

• If μ(X) < ∞ we have that L q (X, , μ) ⊂ L p (X, , μ) for 1 ≤ p ≤ q ≤ ∞.
In particular, the following holds
u L p (X) ≤ c (d, p, q, μ(X)) u L q (X)

for every u ∈ L q (X, , μ).

• We can define vector- or matrix-valued L p -spaces component-wise.
For simplicity we do not mention this in the notation. However, when
considering functions with values in an infinite dimensional spaces X
we denote this by writing L p (X, , μ; X ).

The most important special case is (X, , μ) = (G, B(Rd ), Ld ) where

G ⊂ Rd , B(Rd ) is the Borel σ -algebra on Rd and Ld the d-dimensional
Lebesgue measure. In this case we abbreviate the notation by setting
L p (G) := L p (G, B(Rd ), Ld ). If 1 ≤ p < ∞ then C0∞ (G) is dense in L p (G). In
particular, for every u ∈ L p (G) there is (um ) ⊂ C0∞ (G) such that um → u in
L p (G).

Definition 1.1.5 (Local Lebesgue spaces). Let G ⊂ Rd be open. We define

p
Lloc (G) := {u : G → R : u ∈ L p (K ) for all K  G} for 1 ≤ p ≤ ∞.

Definition 1.1.6 (Weak derivative). Let G ⊂ Rd be open and u ∈ Lloc 1 (G).

• u is called weakly differentiable with respect to the i-th variable if there

is a function vi ∈ Lloc
1 (G) such that

 
u ∂i ϕ dz = − vi ϕ dz for all ϕ ∈ C0∞ (G).
G G

We call vi the weak derivative of u with respect to the i-th variable.

If weak derivatives with respect to all variables exist, we call u weakly
differentiable and denote by ∇ u = (v1 , ..., vn ) the weak gradient of u.
• Let α ∈ N0d and |α| := α1 + ... + αn . For a smooth function u we denote
|α|
by Dα u := ∂ α1 x∂1 ...∂uαn xn its α -th classical derivative. A function u ∈ Lloc
1 (G)

is called k-times weakly differentiable if for all α ∈ N0d with |α| ≤ k there
is a function vα ∈ Lloc 1 (G) such that

 
u Dα ϕ dz = (−1)|α| vα ϕ dz for all ϕ ∈ C0∞ (G).
G G

We call Dα u = vα the α -th weak derivative.

6 Existence Theory for Generalized Newtonian Fluids

Definition 1.1.7 (Sobolev spaces). For k ∈ N and 1 ≤ p ≤ ∞ we define


W k,p (G) := u : G → R, Dα u ∈ L p (G), for all α ∈ N0d , |α| ≤ k ,
W k,p (G)
W0k,p (G) := C0∞ (G) .

Remark 1.1.4. • W k,p (G) (1 ≤ p ≤ ∞) is a Banach space together with

the norm
 1
p
u k ,p := u W k,p (G) := |D u| dz
α p
for 1 ≤ p < ∞,
|α|≤k G

u k,∞ := u W k,∞ (G) := inf sup |Dα u(x)| .

Ld (N )=0 x∈G\N
|α|≤k

• The space W k,p (G) is reflexive iff 1 < p < ∞.

• A sequence (uk ) ⊂ W k,p (G) (1 < p < ∞) converges weakly to u in
W k,p (G) if
 
Dα uk v dz −→ Dα u v dz, k → ∞,
G G

for all v ∈ L p (G, RN ) (p := p−p 1 ) and for all α ∈ N0d with |α| ≤ k.
• The dual space of W0k,p (G) will be denoted by W −k,p (G).
Theorem 1.1.1 (Smooth approximation). Let G ⊂ Rd be open and bounded
with Lipschitz boundary (i.e. ∂ can be locally parametrized by Lipschitz con-
tinuous functions of d − 1 variables) and 1 ≤ p < ∞. Then C ∞ (G) is dense in
W k,p (G). In particular, for every u ∈ W k,p (G) there is (um ) ⊂ C0∞ (G) such that
um → u in W k,p (G).
Theorem 1.1.2 (Sobolev). Let G ⊂ Rd be open.
a) The embeddings
dp
W01,p (G) → L d−p (G) if p < d,
d
W01,p (G) → C 1− p (G) if p > d, Ld (G) < ∞,
are continuous.
b) Let G ⊂ Rd be open and bounded with Lipschitz-boundary. The embeddings
dp
W 1,p (G) → L d−p (G) if p < d,
1− dp
W 1,p (G) → C (G) if p > d,
are continuous.
 Preliminaries 7

Theorem 1.1.3 (Kondrachov). Let G ⊂ Rd be open and bounded. The em-

bedding
pd
W01,p (G) → L q (G), q< ,
d−p
is compact for all p < d.
In Definition 1.1.7 we have interpreted the boundary values of a
Sobolev function as follows: u = 0 on ∂ G iff u ∈ W01,p (G), where W01,p (G)
for p < ∞ denotes the closure of C0∞ (G) in W 1,p (G). We will develop a
rigorous definition of boundary values and show that it coincides with the
former one. In order to do so we need functions which are integrable over
the boundary of G. For G ⊂ Rd open let L p (∂ G) be equal to the set of all
Hd−1 -measurable functions with
 1
p
u L p (∂) := |u|p dHd−1 < ∞.
∂G

Here Hd−1 denotes the (d − 1)-dimensional Hausdorff measure. For

C 1 -functions we define the operator tr : C 1 (G) → L p (∂ G) by tr u := u|∂ G .
Lemma 1.1.1. Let u ∈ C 1 (G) with G ⊂ Rd open and bounded with Lipschitz–
boundary. Then we have
tr u L p (∂ G) ≤ c (d, p, G) u W 1,p (G) .

For u ∈ W 1,p (G) we consider the approximation sequence um ∈

C ∞ (G) with um − u 1,p → 0. Its existence follows from Theorem 1.1.1.
Lemma 1.1.1 shows that (tr um ) is a Cauchy-sequence in L p (∂ G). We define
its limit (which does not depend on the special choice of the sequence) as
the trace of u. The result is a linear operator tr : W 1,p (G) → L p (∂ G) which
coincides with the classical trace operator on C 1 (G).
Theorem 1.1.4. Let  ⊂ Rd be open and bounded with Lipschitz boundary.
Then we have
 
W01,p (G) = u ∈ W 1,p (G); tr u = 0
for 1 ≤ p < ∞.
All results of this section generalize in a straightforward manner to spaces
of vector-valued functions. In order to keep the notation simple we do not
use target spaces in the notation for our function spaces. It will follow from
the context (and the bold-symbol) when we are dealing with these. In fluid
8 Existence Theory for Generalized Newtonian Fluids

mechanics the velocity field is a function from Rd ⊃ G → Rd . In this setting

we need function spaces of solenoidal (that is, divergence-free) functions.
We will use the following notation for 1 ≤ p < ∞
1,p
Wdiv (G) := {ψ ∈ W 1,p (G) : div ψ = 0},
C0∞,div () := {ψ ∈ C0∞ (G) : div ψ = 0},
p L 2 (G) W 1,p (G)
Ldiv (G) := C0∞,div (G) , W01,div
,p
(G) := C0∞,div (G) .
−1,p 1,p
Finally we write Wdiv (G) for the dual of W0,div (G).

1.2 ORLICZ SPACES

In this section we present some important properties of Orlicz spaces
(see [125] and [5]).
A function A : [0, ∞) → [0, ∞] is called a Young function if it is con-
vex, left-continuous, vanishing at 0, and neither identically equal to 0 nor
to ∞. Thus, with any such function, it is uniquely associated a (non-trivial)
non-decreasing left-continuous function a : [0, ∞) → [0, ∞] such that
 s
A(s) = a(r ) dr for s ≥ 0. (1.2.1)
0

The Young conjugate A of A is the Young function defined by

A(s) = sup{rs − A(r ) : r ≥ 0} for s ≥ 0 .
For A we have the representation formula
 s
A(s) = a−1 (r ) dr for s ≥ 0, (1.2.2)
0
where a−1 denotes the (generalized) left-continuous inverse of a. Moreover,
for every Young function A,
r ≤ A−1 (r )A−1 (r ) ≤ 2r for r≥0 (1.2.3)
as well as
A = A. (1.2.4)
Let A be a Young function of the form (1.2.1). Then the convexity of A
and A(0) = 0 imply
A(λs) ≤ λA(s) for all λ ∈ [0, 1] (1.2.5)
and all s ≥ 0. If λ ≥ 1, then
λA(s) ≤ A(λs) for s ≥ 0. (1.2.6)
 Preliminaries 9

As a consequence, if λ ≥ 1, then
A−1 (λs) ≤ λA−1 (s) for s ≥ 0, (1.2.7)
where A−1 denotes the (generalized) right-continuous inverse of A.
A Young function A is said to satisfy the 2 -condition if there exists a
positive constant K such that
A(2s) ≤ KA(s) for s ≥ 0. (1.2.8)
If (1.2.8) just holds for s ≥ s0 for some s0 > 0, then A is said to satisfy the
2 -condition near infinity. We say that A satisfies the ∇2 -condition [near
infinity] if à satisfies the 2 -condition [near infinity].
A Young function A is said to dominate another Young function B near
infinity if there exist positive constants c and s0 such that
B(s) ≤ A(cs) for s ≥ s0 . (1.2.9)
The functions A and B are called equivalent near infinity if they dominate
each other near infinity.
Let G be a measurable subset of Rd , and let u : G → R be a measur-
able function. Given a Young function A, the Luxemburg norm associated
with A, of the function u is defined as
   |u|  
u LA (G) := inf λ : A dx ≤ 1 .
G λ
The collection of all measurable functions u for which this norm is finite
is called the Orlicz space L A (G). It turns out to be Banach
function space.
The subspace of L A (G) of those functions u such that G u(x) dx = 0 will
be denoted by L⊥A (G). A Hölder-type inequality in Orlicz spaces takes the
form 
u(x)v(x) dx
v LA (G) ≤ sup G ≤ 2 v LA (G) (1.2.10)
u∈L A (G) u LA (G)
for every v ∈ L A (G). If |G| < ∞, then
L A (G) → L B (G) if and only if A dominates B near infinity . (1.2.11)
The decreasing rearrangement u∗ : [0, ∞) → [0, ∞] of a measurable func-
tion u : G → R is the (unique) non-increasing, right-continuous function
which is equimeasurable with u. Thus,
u∗ (s) = sup{t ≥ 0 : |{x ∈ G : |u(x)| > t}| > s} for s ≥ 0.
The equimeasurability of u and u∗ implies that
u L A (G) = u∗ L A (0,|G|) (1.2.12)
for every u ∈ L A (G).
10 Existence Theory for Generalized Newtonian Fluids

The Lebesgue spaces L p (G), corresponding to the choice Ap (t) = tp , if

p ∈ [1, ∞), and A∞ (t) = ∞ · χ(1,∞) (t), if p = ∞, are a basic example of Or-
licz spaces. Other instances of Orlicz spaces are provided by the Zygmund
spaces L p logα L (G), and by the exponential spaces exp L β (G). If either p > 1
and α ∈ R, or p = 1 and α ≥ 0, then L p logα L (G) is the Orlicz space asso-
ciated with a Young function equivalent to tp (log t)α near infinity. Given
β > 0, exp L β (G) denotes the Orlicz space built upon a Young function
β
equivalent to et near infinity.
An important tool will be the following characterization of Hardy type
inequalities in Orlicz spaces [46, Lemma 1].

Lemma 1.2.1. Let A and B be Young functions, and let L ∈ (0, ∞].
(i) There exists a constant C such that
  s 
1 
 f (r ) dr  ≤C f (1.2.13)
s L A (0,L )
0 L B (0,L )

for every f ∈ L A (0, L ) if and only if either L < ∞ and there exist constants c > 0
and t0 ≥ 0 such that
 t
B(s)
t ds ≤ A(ct) for t ≥ t0 , (1.2.14)
t0 s2
or L = ∞ and (1.2.14) holds with t0 = 0. In particular, in the latter case, the
constant C in (1.2.13) depends only on the constant c appearing in (7.0.1).
(ii) There exists a constant C such that
 
 L
dr 
 f (r ) ≤C f (1.2.15)
 r LB (0,L) L A (0,L )
s

for every f ∈ L A (0, L ) if and only if either L < ∞ and there exist constants c > 0
and t0 ≥ 0 such that
 t
A(s)
t ds ≤ B(ct) for t ≥ t0 , (1.2.16)
t0 s2
or L = ∞ and (1.2.16) holds with t0 = 0. In particular, in the latter case, the
constant C in (1.2.15) depends only on the constant c appearing in (1.2.16).

Assume now that G is an open set. The Orlicz–Sobolev space W 1,A (G)
is the set of all functions in L A (G) whose distributional gradient also be-
longs to L A (G). It is a Banach space endowed with the norm
u W 1,A (G) := u L A (G) + ∇u L A (G) .
 Preliminaries 11

We also define the subspace of W 1,A (G) of those functions which vanish
on ∂ G as
W01,A (G)
= {u ∈ W 1,A (G) : the continuation of u by 0 is weakly differentiable}.
In the case where A(t) = tp for some p ≥ 1, and ∂ G is regular enough, such a
definition of W01,A (G) can be shown to reproduce the usual space W01,p (G)
defined as the closure in W 1,p (G) of the space C0∞ (G) of smooth compactly
supported functions in G. In general, the set of smooth bounded functions
is dense in L A (G) only if A satisfies the 2 -condition (just near infinity
when |G| < ∞), and hence, for arbitrary A, our definition of W01,A (G)
yields a space which can be larger than the closure of C0∞ (G) in W01,A (G)
even for smooth domains. On the other hand, if G is a Lipschitz domain,
namely a bounded open set in Rd which is locally the graph of a Lipschitz
function of d − 1 variables, then
W01,A (G) = W 1,A (G) ∩ W01,1 (G),
where W01,1 (G) is defined as usual.
Lemma 1.2.2. Let A be a Young-function satisfying the 2 -condition and
G ⊂ Rd open.
• The following holds
W 1,A (G)
W01,A (G) = C0∞ (G) = W 1,A (G) ∩ W01,1 (G).
• For every u ∈ W 1,A there is a sequence (uk ) ∈ C ∞ (G) ∩ W 1,A (G) such that
uk → u in W 1,A (G).
Lemma 1.2.3. Let A be a Young function satisfying the 2- and the
∇2 -condition. Then L A (G) is reflexive with
L A (G) ∼
= L Ã (G).

1.3 BASICS ON LIPSCHITZ TRUNCATION

The purpose of the Lipschitz truncation technique is to approximate a
Sobolev function u ∈ W 1,p by λ-Lipschitz functions uλ that coincide with u
up to a set of small measure. The functions uλ are constructed nonlinearly
by modifying u on the level set of the Hardy–Littlewood maximal function
of the gradient ∇ u. This idea goes back to Acerbi and Fusco [1–3]. Lips-
chitz truncations are used in various areas of analysis: calculus of variations,
12 Existence Theory for Generalized Newtonian Fluids

in the existence theory of partial differential equations, and in regularity

theory. We refer to [62] for a longer list of references.
The basic idea is to take a function u ∈ W 1,p (Rd ), where p ≥ 1, and
cut values where the maximal function of its gradient is large. The Hardy–
Littlewood maximal operator is defined by

M (v)(x) = sup − |v| dy
B: x∈B B

for v ∈ 1 (Rd )
Lloc which can be extended to vector- (or matrix-)valued
functions by setting M (v) = M (|v|). Basic properties of the maximal op-
erator are summarized in the following lemma (see e.g. [134] and [141,
Lemma 3.2] for d)).
Lemma 1.3.1. a) Let v ∈ Lloc1 (Rd ) and λ > 0. The level-set {x ∈ Rd :

|M (v)(x)| > λ} is open.

b) The strong-type estimate
M (v) L p (Rd ) ≤ cp v L p (Rd ) ∀v ∈ L p (Rd )
holds for all p ∈ (1, ∞].
c) The weak-type estimate
  v pp
Ld {x ∈ Rd : |M (v)(x)| > λ} ≤ cp p ∀v ∈ L p (Rd )
λ
holds for all p ∈ [1, ∞) and all λ > 0.
d) We have the estimate
 
|v|p dx ≤ cp |v|p dx ∀p ∈ [1, ∞)
{M (v)>λ} {|v|>λ/2}

for all λ > 0.

The Lipschitz truncation will be defined via the maximal function of
the gradient. For u ∈ W 1,p (Rd ) the “bad set” is defined by
Oλ := {x ∈ Rd : M (∇ u)(x) > λ}, (1.3.17)
where λ ≥ 0. If we have u ∈ for G =
W 1,p (G) Rd
it has to be extended to
u ∈ W (R ). This can be done in an obvious way if u ∈ W01,p (G) where
1,p d

the extension is zero outside G. In the general case we may apply [5,
Thm. 4.26]. Now, for x, y ∈ Rd \ Oλ we have a.e.
 
|u(x) − u(y)| ≤ c |x − y| M (∇ u)(x) + M (∇ u)(y) ≤ 2c λ|x − y|,
see, e.g., [112]. Hence u is Lipschitz-continuous in Rd \ Oλ with Lipschitz
constant proportional to λ. By a standard extension theorem (see e.g. [71,
 Preliminaries 13

p. 201]) we can extend u (defined in Rd \ Oλ ) to uλ (defined in Rd ) such

that the Lipschitz constant is preserved. (When dealing with this simple
extension it is necessary to cut large values of M (u) as well. We neglect this
for brevity.) This means we have
|∇ uλ | ≤ c λ in Rd . (1.3.18)
Moreover, by construction we have
{x ∈ Rd : u = uλ } ⊂ Oλ .
This and Lemma 1.3.1 c) imply
  c ∇u p
p
L {x ∈ R : u = uλ } ≤ L (Oλ ) ≤
d d d
. (1.3.19)
λp
Combining (1.3.18) and (1.3.19) shows
  
|∇ uλ | dx =
p
|∇ uλ | dx +
p
|∇ uλ |p dx
Rd Rd \Oλ Oλ

 
≤ |∇ u|p dx + c λp Ld Oλ
Rd \Oλ

≤c |∇ u|p dx.
Rd

We obtain the following stability result

∇ uλ p ≤ c ∇ u p. (1.3.20)
The basic properties (1.3.18)–(1.3.20) are already enough to make the Lips-
chitz truncation a powerful tool for numerous applications. We present two
rather classical ones.

Lower semi-continuity in W 1,p .

Let G ⊂ Rd be an open and bounded with Lipschitz boundary. Suppose
further that F : G × Rd×D → [0, ∞) is a continuous function with p-growth
(p > 1), i.e.,
F (x, Q) ≤ c |Q|p + g(x) ∀Q ∈ Rd×D (1.3.21)
with a constant c ≥ 0 and a non-negative function g ∈ L 1 (G). We are inter-
ested in minimizing the functional

GF [w] = F (x, ∇ w) dx
G

defined for functions w : G → RD . An important concept in showing the

existence of minimizers is the lower semi-continuity of GF with respect to
14 Existence Theory for Generalized Newtonian Fluids

an appropriate topology. The functional GF is called W 1,p -weakly lower

semi-continuous if
GF [v] ≤ lim inf GF [vn ] (1.3.22)
n→∞

provided vn v in W 1,p (G) for m → ∞. The functional GF is called

1,∞ ∗
W -weakly lower semi-continuous if
GF [v] ≤ lim inf GF [vn ]
n→∞
∗
provided vn v in for n → ∞.
W 1,∞ (G)
Due to (1.3.21) the right concept for the functional GF is W 1,p -weak
lower semi-continuity. It can be deduced from W 1,∞ -weak∗ lower semi-
continuity by the Lipschitz truncation, see Lemma 1.3.3 below. Using this
idea Acerbi and Fusco [1] showed the W 1,p lower semi-continuity of GF in
the case where F is only quasi-convex. Note that W 1,∞ -weak∗ lower semi-
continuity is a consequence of the definition of quasi-convexity, see [1,
Thm. 2.1]. For brevity we do not discuss the concept of quasi-convexity
and refer instead to the fundamental papers [16] and [115].

Lower integrability for the p-Laplace system.

Consider the system
div(|∇ v|p−2 ∇ v) = div F in G, (1.3.23)
v=0 on ∂ G. (1.3.24)
Here, G is an open set in Rd ,
with d ≥ 2, the exponent p ∈ (1, ∞) and the
×D
function F :  → R d is given. A weak solution to (1.3.23) is a function
v ∈ W01,p (G) such that
 
p−2
|∇ v| ∇ v : ϕ dx = F : ∇ϕ dx
G G

for all ϕ ∈ W01,p (G). Its existence can be shown via standard methods pro-
vided F ∈ L p (G). We are concerned here with the question of how the
regularity of F transfers to v (particularly to |∇ v|p−2 ∇ v). In the linear
case p = 2 this is answered by the classical theory of Calderón and Zyg-
mund [43]. It says that F ∈ L q (G) implies ∇ v ∈ L q (G) for all q ∈ (1, ∞).
Note that the case q < 2, where q is below the duality exponent p , is in-
cluded. In that situation existence of weak solutions is not clear a priori.
There has been a great deal of effort in obtaining a corresponding result for
the nonlinear case p = 2 such that
F ∈ L q (G) ⇒ |∇ v|p−2 ∇ v ∈ L q (G) ∀q ∈ (1, ∞) (1.3.25)
 Preliminaries 15

together with a corresponding estimate. This has been positively answered

in the fundamental paper by Iwaniec [97] provided q ≥ p . An improvement
to q > p − δ for some small δ > 0 has been carried out in [98] by different
methods (for an overview and further references see [113]). We remark that
the case q ∈ (1, p − δ) is still open. Based on the Lipschitz truncation we can
give a relatively easy proof for the estimate in the case q ∈ (p − δ, p ) using
the approach in [141] (see also [40] for a more general setting and [101] for
the parabolic problem), see Lemma 1.3.4 below.
Before we give proofs of these applications we present an important
improvement of the Lipschitz truncation which firstly appeared in [62]. It
concerns the smallness of the level-sets. Similar ideas have been used earlier
for the L ∞ -truncation in [78].

Lemma 1.3.2. Let v ∈ L p (Rd ) with p ∈ (1, ∞). Then there exist j0 ∈ N and a
j +1
sequence λj ∈ R with 22 ≤ λj ≤ 22 −1 such that
j

 
λj Ld {x ∈ Rd : M (v) > λj } ≤ c p 2−j v
p p
p

for all j ≥ j0 where c = cp is the constant in Lemma 1.3.1 b).

Proof. We have
  ∞
M (v) pp = ϑ p−1 χ{|v|>ϑ} dϑ dx
Rd 0
  2m+1
= ϑ p−1 χ{|M (v)|>ϑ} dϑ dx
Rd m∈Z 2m

 p (1.3.26)
≥ 2m χ{|M (v)|>2m+1 } dx
Rd m∈Z

2j+1 −1 
 k p
≥ 2 χ{|M (v)|>2·2k } dx.
j∈N Rd
k=2j

The continuity of M on L p (Rd ), see Lemma 1.3.1 b), implies

2j+1 −1 
 k p
2 χ{|M (v)|>2·2k } dx ≤ c p v pp .
j∈N Rd
k=2j

In particular, for all j ∈ N

2j+1 −1 
 k p
2 χ{|M (v)|>2·2k } dx ≤ c p v pp .
Rd
k=2j
16 Existence Theory for Generalized Newtonian Fluids

Since the sum contains 2j summands, there is at least one index kj such that

 p
2kj χ{|M (v)|>2·2kj } dx ≤ c p v pp 2−j . (1.3.27)
Rd

Define λj := 2kj and we conclude from (1.3.27) that


 p p −j
λj χ{|M (v)|>2 λj } dx ≤ c p v p2 .
Rd
This proves the claim.
Lemma 1.3.2 shows that there is a particular sequence of levels (λj ) such
that
 
p
λj Ld Oλj ≤ κj ∇ v p
p (1.3.28)
with κj → 0 for j → ∞. This improves the estimate (1.3.19) and does not
follow from the original results by Acerbi and Fusco. It allows us to simplify
the original proof of W 1,p -lower semi-continuity from [1].
Lemma 1.3.3. Assume that the functionals GF defined in (1.3.22) are
W 1,∞ -weakly∗ lower semi-continuous for any choice of F satisfying (1.3.21). Then
they are W 1,p -weakly lower semi-continuous.
Proof. Let (vn ) ⊂ W 1,p (G) be a sequence with weak limit v such that
un := vn − v 0 in W 1,p (G).
We take the sequence (λj ) in accordance with Lemma 1.3.2 for the level-
sets of ∇ u. We apply the Lipschitz truncation to the sequence (un ) with
level λ = λj , see the construction after (1.3.17), and obtain for the double
sequence (un,j := un,λj )
∇ un,j ∞ ≤ c λj (1.3.29)
∗ 1,∞
un,j 0 in W (G) (1.3.30)
 
p
λj L d {x ∈ Rd : un = un,j } ≤ κj (1.3.31)
due to (1.3.18)–(1.3.20), where κj → 0 for j → ∞. We obtain

lim inf G [vn ] = lim inf F (·, ∇ v + ∇ un ) dx
m→∞ n→∞ G
 
≥ lim inf F (·, ∇ v + ∇ un ) dx = lim inf F (∇ v + ∇ un,j ) dx
n→∞ G\Oλj n→∞ G\Oλj
 
≥ lim inf F (·, ∇ v + ∇ un,j ) dx − lim sup F (·, ∇ v + ∇ un,j ) dx.
n→∞ G n→∞ O λj

(1.3.32)
 Preliminaries 17

We can use the functional GF̃ with

F̃ (x, Q) = F (x, ∇ v + Q), Q ∈ R d ×D .
The function F̃ has p-growth as required in (1.3.21) (by ∇ v ∈ L p ()) such
that

lim inf F (·, ∇ v + ∇ un,j ) dx = lim inf GF̃ [un,λj ] ≥ GF̃ [0] = GF [v]. (1.3.33)
n→∞ G n→∞

This is a consequence of the W 1,∞ -weak∗ lower semi-continuity of GF̃ .

Finally we have
  
 
F (·, ∇ v + ∇ un,j ) dx ≤ c |∇ v|p + g dx + c |∇ un,j |p dx
O λj O λj O λj

   
≤c |∇ v|p + g dx + c λd Ld Oλj
O λj

such that

lim sup lim sup F (·, ∇ v + ∇ un,j ) dx = 0 (1.3.34)
λ→∞ n→∞ O λj

by (1.3.31) and |∇ v|p + g ∈ L 1 (G). Combining (1.3.32)–(1.3.34) shows that

GF [v] ≤ lim inf GF [vn ],
m→∞
i.e. GF is W 1,p -weakly lower semi-continuous.
We now turn to the proof of the lower integrability for the p-Laplace
system. In addition to the Lipschitz truncation crucial ingredients are the
following integral identities. Let 0 <  < ∞, 0 ≤ δ <  < δ and (X, , μ)
be a measure space. There holds for every μ-measurable function f with
|f | ∈ L 1 (X, , μ)
 ∞  
1
ϑ −1−δ |f |δ dμ dϑ = |f | dμ, (1.3.35)
0 {|f |>ϑ} −δ X
 ∞  
1
ϑ −1−δ |f |δ dμ dϑ = |f | dμ. (1.3.36)
0 {|f |≤ϑ} δ− X

Both equalities are easy consequences of Fubini’s Theorem. As we will

apply (1.3.35) and (1.3.36) several times it is important that all estimates
hold for any λ > 0. So Lemma 1.3.2 is no use.
Lemma 1.3.4. There is a number δ > 0 such that for all q ∈ (p −δ) the following
holds. Let v ∈ W01,q(p−1) (G) be a weak solution to (1.3.23) with F ∈ L q (G). Then
we have
 
 
|∇ v|p−2 ∇ vq dx ≤ c |F|q dx.
G G
18 Existence Theory for Generalized Newtonian Fluids

Proof. Take the solution v to (1.3.23) and use its Lipschitz truncation vλ
as a test-function, see the construction after (1.3.17). Note that the Lips-
chitz truncation can preserve zero boundary values at least in the case of a
Lipschitz boundary, cf. [62, Thm. 3.2]. We obtain
   
p−2 p−2
|∇ v| ∇ v : ∇ vλ dx = − |∇ v| ∇ v : ∇ vλ dx + F : ∇ vλ dx
Rd \Oλ Oλ Rd

which implies by (1.3.18)

  
p−1
|∇ v| dx ≤ c λ
p
|∇ v| dx + |F||∇ vλ | dx. (1.3.37)
Rd \Oλ Oλ Rd

Furthermore, the following holds

  
|∇ v| dx =p
|∇ v| dx + p
|∇ v|p dx
{|∇ v|≤λ} {|∇ v|≤λ}∩Oλ {|∇ v|≤λ}\Oλ
 
≤λ |∇ v|p−1 dx + |∇ v|p dx.
Oλ Rd \O λ

Inserting this into (1.3.37) yields

  
|∇ v|p dx ≤ c λ |∇ v|p−1 dx + |F||∇ vλ | dx.
{|∇ v|≤λ} Oλ Rd

As a consequence of Lemma 1.3.1 d) we deduce

  
|∇ v|p dx ≤ c λ |∇ v|p−1 dx + |F||∇ vλ | dx.
{|∇ v|≤λ} {|∇ v|>λ/2} Rd

After multiplying with λq−1−p and integrating we have

 ∞   ∞ 
q−1−p q−p
λ |∇ v| dx dλ ≤ c
p
λ |∇ v|p−1 dx dλ
0 {|∇ v|≤λ} 0 {|∇ v|>λ}
 ∞ 
q−1−p
+ λ |F||∇ vλ | dx dλ.
0 Rd

Setting χλ := Rd |F||∇ vλ | dx we obtain on account of (1.3.35) and (1.3.36)
   ∞
1 c
|∇ v| dx ≤
q
|∇ v| dx +
q
λq−1−p χ (λ) dλ.
p−q q−p+1 0

If q is close enough to p, say p − q < δ̃ , we have

c 1
<
q−p+1 p−q
and hence
  ∞
|∇ v|q dx ≤ c λq−1−p χ (λ) dλ.
0
 Preliminaries 19

We split χ (λ) = χ1 (λ) + χ2 (λ) where

 
χ1 (λ) = |F||∇ vλ | dx, χ1 (λ) = |F||∇ vλ | dx.
{M (∇ v)≤λ} {M (∇ v)>λ}

Note that we have |∇ vλ | = |∇ v| ≤ M (∇ v) on {M (∇ v) ≤ λ}. Setting μ =

|F|Ld and using (1.3.36) (with f = M (∇ v), δ = 1 and  = q − p + 1) as well
as Hölder’s inequality we obtain
 ∞  ∞ 
q−1−p q−1−p
λ χ1 (λ) dλ ≤ λ M (∇ v) dμ dλ
{M (∇ v)≤λ}
0 0

1
= M (∇ v)q−p+1 dμ
p − q Rd

1
= M (∇ v)q−p+1 |F| dx
p − q Rd
1
≤ F q/(p−1) M (∇ v) qq−p+1
p−q
≤ c F q/(p−1) ∇ v qq−p+1
and similarly by (1.3.35) (with f = 1, δ = 0 and  = q − p + 1)
 ∞ 
 ∞
λq−1−p χ2 (λ) dλ ≤ c λq−p dμ dλ
0 0
 {M (∇ v)>λ}
c
= M (∇ v)q−p+1 dμ
q − p + 1 Rd
≤ c F q/(p−1) ∇ v qq−p+1 .
Combining the estimates above implies
 
q
|∇ v|q dx ≤ c |F| p−1 dx ∀q ∈ (p − δ̃, p)
G G

or equivalently on setting δ = p−1

δ̃

 
 
|∇ v|p−2 ∇ vq dx ≤ c |F|q dx ∀q ∈ (p − δ, p ).
G G

We now turn to an alternative approach for the extension of u|Rd \Oλ

into the “bad set” which has been used in [33] and [60]. Instead of using
classical extension theorems as in the definition after (1.3.17) one can work
with a Whitney covering of the “bad set” and local approximations. This is
much more flexible and allows for instants to cut only parts of the gradient
(in particular the symmetric gradient) or to work with higher derivatives,
cf. Chapter 3. In fact, this is so far the only successful method for parabolic
problems, cf. Section 5.2. The following lemma shows how to decompose
Another random document with
no related content on Scribd:
[Contents]

120. The Boy and his Master. [Story]

See Grimm 68, The Thief and his Master, Bolte u. Polívka 2: 60–69;
and compare Tremearne, 223–224; Mason and Espinosa, New
Mexico, JAFL 24: 423–424.

[Contents]

121. The Language of Beasts. [Story]

See Grimm 17, The White Snake, Bolte u. Polívka 1: 131–134; and
Aarne’s study, Der Tiersprachen verstehende Mann, in FF
Communications No. 15. Compare Koelle, 143–145; Basset 2: 119–
124; Junod, 314–317; Chatelain, 219–223; Smith, 565.

[Contents]

122. The Three Pieces of Advice. [Story]

Compare Steere, 413; Mason & Espinosa, JAFL 24: 408–411:

discussed by Espinosa, JAFL 27: 213–214.
[Contents]

123. The Brothers and the Life-tree. [Story]

See Grimm 107, The Two Travellers; Bolte u. Polívka 2: 468–482

and compare Dayrell, 58–60; Espinosa, JAFL 27: 191–195. [285]

[Contents]

124. The Skillful Brothers. [Story]

See Grimm 129, Four Skillful Brothers, Bolte u. Polívka 2:165–169

and compare Cronise and Ward, 200–205; Renel 1:215–223;
Dennett, 33–34; Parsons, Sea Islands, 75.

[Contents]

125. The Three Sillies. [Story]

See Grimm 34, Clever Elsie, Bolte u. Polívka 1:335–342, and

Clouston, Book of Noodles, 7. Compare Parsons, Andros Island,
128–129; Sea Islands, 94.

[Contents]
 126. A Misunderstanding. [Story]

See Grimm 84, Hans Married, Bolte u. Polívka 2:203–204.

[Contents]

127. Big-head, Big-belly and Little-foot. [Story]

The story is very common in Jamaica. See Grimm 18, The Straw, the
Coal and the Bean, Bolte u. Polívka 1:135–137, and compare
Parsons, Andros Island, 147.

[Contents]

128. The Goat in the Lion’s Den. [Story]

[Contents]

129. Donkey, Cat and the Lion’s Head. [Story]

The familiar episode of the Wolf’s head which occurs early in the
Reynard cycle (see Percy Society Publications 12, Introduction,
pages xxxiii–xxxiv) is, in African stories, often combined with that of
the Goat in the Lion’s den (or the Hyena’s). Compare Rattray,
Chinyanje, 149–152; Tremearne, 227–229; FL 22:63–65.

[Contents]

130. Clever Molly May. [Story]

See Grimm 77, Clever Gretel, Bolte u. Polívka 2:129–131; and

Parsons, Sea Islands, 140. From this point in the group of stories
Anansi is introduced in the role of hero.

[Contents]

131. Dancing to Anansi’s Fiddle. [Story]

See Grimm 110, The Jew among Thorns, Bolte u. Polívka 2:490–
503; and compare Bundy, JAFL 32:412–413.

[Contents]

132. Anansi Claims the Dinner. [Story]

Compare Nassau, 42–44; Tremearne, FL 21:212; Krug, JAFL

25:106–107.
[Contents]

133. Anansi seeks his Fortune. [Story]

See note to number 106, and Grimm 104, Wise Folks, Bolte u.
Polívka, 2:440–451. One version from Parsons, Andros Island,
[286]93–94, connects this episode with those of the frightened
robbers and the tongue-cutting in number 109.

[Contents]

134. The Pannier Jar. [Story]

See note to number 106. This is episode F‴ in Bolte u. Polívka’s

analysis of Grimm’s Little Peasant. Compare Parsons, Sea Islands,
89; JAFL 32: 372, and note for references.

[Contents]

135. Anansi kills his Grandmother. [Story]

See note to number 106. This is episode G″ in Bolte u. Polívka’s

analysis of Grimm’s “Little Peasant.” Compare Parsons, Andros
Island, 87 and note for references.
[Contents]

136. White-belly and Anansi. [Story]

See note to number 106. The trick corresponds to F′ or G′ in Bolte u.

Polívka’s analysis of The Little Peasant.

[Contents]

137. Monkey hunts Anansi. [Story]

See Boas, JAFL 25: 223–226, where the Devil is the rival sorcerer.

[Contents]

138. Anansi and the Pig. [Story]

See Grimm 72a, Bolte u. Polívka, 2: 100–106; and compare

Parsons, Andros Island, 108 and note; discussion of Spanish forms
by Boas, JAFL 25: 252, note; by Espinosa, JAFL 27: 222–227.

[Contents]

139. The Fifer. [Story]

The story is common in Jamaica. See Jekyll, 98–99. It was told me
as a “speak-acting” story, but as I could get no other of exactly the
same character, I do not know how common it used to be to present
a Nansi story in this way. The Nansi story is now given in the form of
a dramatic monologue or rehearsed simply as a tale.

For the story of “The Fifer,” six actors were required, one to
represent the boy, one the father, and four others the “wild beasts.”
“Anansi,” “Dry-head,” “Tacoomah” and “Tiger” were the “beasts.” Roe
said that “the one who takes the son’s part tells the story.” The
dramatization went on much like a school exercise performed by
grown men, with improvised action and (probably) extemporized
dialogue. It ended in a dance in which all six joined.

Compare Tremearne, 301; Harris, Nights, 370–373; Edwards, 87–

88; Parsons, Andros Island, 137–138. [287]

The story seems to be drawn from such prohibitions against

whistling at night or whistling more than twice when walking at night
or through a haunted forest as are quoted by Sebillot, Le Folk-lore
de France 1: 159, 283. He tells a Breton story of a lad who forgot the
prohibition and found himself mocked and followed by the Devil, who
bore him off just as he had reached home. Compare number 66,
note.

[Contents]

141. Tacoomah makes a Dance. [Story]

Medleys of this character seem to have been a popular form of

entertainment and may still be common, though the examples I have
were given me in every case by old men. They are composed of
scraps of song or whole scenes from well-known Nansi stories,
together with game-songs, imitations of animal sounds, and
“rhyming,” strung together much like our own musical medleys—the
last line of one suggesting the first of the next. In this example, story-
songs from numbers 97 and 86 are followed by a game dialogue;
next by some animal imitations; last, by a specimen of Jamaica
“rhyming.” Other examples of this kind of improvised “rhyming” are:

“Mr. Might, jump up a height, after a kite,

And knock his eye, upon his hog-sty, and cry out ‘hi!
oh, my! why should I die’.”

“There is a boat, and in the boat, is a goat, and has

a long coat, catch him under the throat.”

[Contents]

142. Anansi makes a Dance. [Story]

The songs of this medley at first follow the story of Goat’s escape
from the dance, number 40, combined with the parallel story of Rat’s
escape from Puss’s dance. The song is taken from a popular game
in which one player represents the cat, another the rat; all the others
form a line with clasped hands, and Puss tries to catch Rat through
the line, while all sing the song. The bull’s song belongs to number
88 or 89. Anansi’s fifing is possibly taken from 139.
[Contents]

143. Red Yam. [Story]

Old Mary Roden was bed-ridden and lived in a one-roomed hut, the
floor of which was falling in. The little grandchild, when prompted to
“make a figure,” danced quite spontaneously to the rhythm of the
grandmother’s quavering song. The same is true of the next two
numbers. Songs sung to be danced to in this fashion have rather the
monotonous rhythm of a drum-beat than any melody in our sense of
the word. For the story, see number 23. [288]

[Contents]

145. Fowl and Pretty Poll. [Story]

Literally this means, Fowl wants to be married to Parrot in church,

but Parrot has no good clothes. Parrot wants Crow to marry her in
church, but he says he can’t because of his peelhead (or perhaps he
wants to in spite of this peelhead). Compare the witticism vi.

[Contents]

146. The Cumbalo. [Story]

Sarah Findley was an old-time negress who lived in a little hut far out
in the bush. She danced to the song with a queer jumping motion
like boys playing leap-frog and with all the agility of a young girl. The
dance as a wake game is performed upon two parallel bars held by
four men. One informant called it dancing “Calimbe.”

[Contents]

149. Animal Talk. [Story]

Again there are, in this medley which imitates animal sounds,

reminiscences of consecutive Nansi stories—Crab’s words, in
number 54; Ground Dove’s in 50. For the fasting contest, see
number 61.

Tremearne, 28, says that the imitation of animal cries is a favorite

device in African story-telling. Compare Hollis, Nandi, 109–111,
where a great many examples are given of this kind of
entertainment.

[Contents]

Witticisms.

I & II. These old-fashioned slave stories are from old Vassel Edwards
at Retirement, in the Cock-Pit country. They belong to the “nager-
trick” stories quoted by Lewis.
III. The Congo negro is said to be duller-witted than negroes from the
Gold Coast. To call a man a “Congo” is hence a term of ridicule.

IV. This witticism is common. In one version, the man was said to be
“walking in Kingston.” Mrs. Elizabeth Hilton gave me a version she
learned from Henry Roe, school-master at Retirement, which bears
the marks of having been put together by some literary entertainer.

“Massa Peter was a funny sort of a buckra massa. He was “mustafenia”

(white by law). Massa Peter an’ me, we go to school together. We were
readin’ in a ‘pellin’ (book) an’ we were doin’ jumba fraction sum.

“From the day me leave school me never see Marse Peter any more till
one day we buck up. A glad to see him till a couldn’t glad any more. Marse
Peter went a tell me somet’ing, a laugh till me belly nearly pop.

“Marse Peter was the sort of boy used to go out after hours. Him ma tell
him if him (she) been dead before him, she will show him token (frighten
him). [289]But Marse Peter never will believe her. One night, Marse Peter
go out. When him coming back, he catch right at the cross-road where
dem Taylor boy used to sit down a day-time, an’ smell somet’ing funny, but
he never know wha’. He been ’fraid, but afterward he no ’fraid again. An’
see one man come wid litt’e fire. He say, ‘I beg you a light, sah!’ The man
give him a light. The man has some teeth a his mouth, they long like a
Jack-ass a laugh a sun-hot. Marse Peter pass the man. He meet up
another man. He say, ‘Look here, me frien’, I meet a man jus’ roun’ the
turning, have teeth long like a Jack-ass a laugh a sun-hot.’ The man said,
‘Teeth like these do they long?’ Marse Peter run an’ he run an’ never stop
runnin’ till he meet up a mother bed. From that, Marse Peter never go af’er
no girl again. Marse Peter behave a good buckra massa af’er this.”

V. The witticism is used in a good many connections. In one story, a

man finds a boy by the roadside and takes him home. When he asks
the boy to blow the fire, the duppy says, “Me kyant blow de fire, for
me dead long time an’ dirt eat out all me teet’.” The man beats him
and he runs away crying, “Lor! me dead two time.” In another
version, “Rolling Calf” takes possession of a house. While he is
asleep, the owner makes an iron fork red hot and catches him about
the neck.

VI. See number 145.

VIII. Compare Cundall, FL 15:91, where the “Rolling Calf,” afraid of

the moon, tumbles over into the stream and sprains his foot. He
says, “A don’t mind the wet, a wet, but the ’prain a ’prain me foot’.”

X. In Tremearne, FL 22:222–223, Lizard and Mouse both court a

woman. Mouse tells her that Lizard is blind, can’t see at night; Cock
tells her that Rat is a thief, can’t be seen in the market.

In Koelle, 174–177, Toad and Rat have a wager to see if one can do
what the other cannot. Toad passes a crowd with a whole skin; Rat is
pursued with sticks and stones.

XI. See number 48.

XII. From Alexander Archibald, near Mandeville.

XIII. From Mrs. Matilda Hall, Harmony Hall. See number 4.

XVI. This and the next two witticisms were written out by some
young lads in Bethlehem, Santa Cruz Mountains.

1 See supplementary note, p. 289. ↑

2 See supplementary note, p. 290. ↑
3 See supplementary note, p. 290. ↑
4 See supplementary note, p. 290. ↑
5 See supplementary note, p. 290. ↑
[Contents]
 SUPPLEMENTARY NOTES.

[Contents]

46. Why Tumble-bug Rolls in the Dung. [Story]

In Seidel’s story of the “Miracle of the Sidi” (Geschichten und Lieder

der Afrikaner, 105), the devil dares the Sidi to marry a slave to a
princess. The father of the princess has set to her wooers the
supposedly impossible task of filling a bag with hyacinths out [290]of
hyacinth season. The Sidi fills the bag with stones and bids the slave
empty it out before the king, when the stones are by miracle turned
into hyacinths.

[Contents]

48. Why Dog is always Looking. [Story]

A Jamaica negro proverb runs, “Darg say befo’ him plant yam fe look
like masquita’ foot, him satisfy fe tun beggar.” See Cundall’s
collection (Kingston, 1910), 211.

[Contents]

56. Rat’s Wedding. [Story]

It is not the wooden foot-bridge but any drain beside the road—the
gutter—which Jamaicans call a ‘water table.’

[Contents]

66. Simon Tootoos. [Story]

For the music of these songs see Publications of the Modern

Language Association of America, 39 (1924): 482.

[Contents]

97. Leap, Timber, Leap. [Story]

An old man over eighty who was present at the recital of this story
remembered hearing it when he was a little boy. Hauling lumber was
in old days accompanied by song. The story turns upon a theme
common in American Indian hero cycles, that of a trickster’s claim to
magical powers which he does not possess. [291]

[Contents]
 INDEX TO INFORMANTS.

1. Alexander, Emily, aged 15. She came to my room in the evening

after her work was done at the hotel and recited to me more than
twenty stories which she had from her father, who was a native of
Mandeville, and with which she was in the habit of entertaining the
other young people employed at the hotel.

See 108, 109, 110, 111, 130.

2. Archibald, Alexander, aged 62. He was “tea-meeting chairman” in

the district of New Green, near Mandeville. I visited him at his home
one late afternoon. Seated on his own door-step in the midst of a
circle of his neighbors, he recited six stories with much wit and good
humor.

See 32b, 36.

3. Bailey, Vivian, a lad, also of Mandeville.

See 1b.

4. Baker, Maud, aged 21. She called upon me with her stories, which
she had from her father, a native of Dry River, though she herself
had been educated in Kingston.

See 102, 106.

5. Barrett, Eliza, aged 30. She was one of a group of women who
were friends of the colored housekeeper at Harmony Hall.

See 92b (1).

6. Barrett, George, aged 60 or over. He visited me at Harmony Hall
with a group of men from Maroon Town. They would spend a whole
morning or even all day telling stories in this way, first one and then
another taking his turn and each making way for the other with a fine
sense of fair play.

See 57b (2), 76a, 84a.

7. Brown, Arthur, aged 23. He was a friend of the chauffeur for the
hotel at St. Ann’s Bay and took me down to his mother’s house at
Steeretown, where he gathered a group for story-telling, each one
reciting one or two stories in turn.

See 127a.

8. Brown, Margaret, aged 55, mother of Arthur.

See 47a.

9. Brown, Philip, aged 19, a jolly contingent of Caledonia, near

Mandeville.

See 75, 76b.

10. Brown, T., another contingent, a Claremont lad who had picked
up a quantity of stories but recited them in a slovenly way, without wit
or point.

See 117.

11. Christie, Samuel, over 60. He was one of the group at

Steeretown, near St. Ann’s Bay, and a good story-teller.

See 5b, 12a, 14. [292]

12. Collins, Benjamin, a crippled lad of George’s Valley, near
Mandeville.

See 11a.

13. Daley, Edward, part East Indian. He was in the prison-gang

whose overseer I was interviewing.

See 132.

14. Dodd, Emiline, under 30. She visited the house at Lacovia where
I was staying.

See 86b.

15. Doran, Grace, very old. She was from Whitehall, near Harmony
Hall. She interpolated her stories with songs in the old style, but
talked so rapidly I was unable to follow except in snatches.

See 27a.

16. Edwards, Vassel, over 80. His father and mother had been
slaves in the same district, one at Retirement, where he himself had
lived all his life and was now deacon in the Scotch Presbyterian
church.

See 134 and the first two witticisms.

17. Falconer, Simeon, aged 47. He was an intelligent and

resourceful man, a church member but nevertheless a frequenter of
wakes, where he learned his stories. He dictated his stories to me at
his home, without audience, and on various visits. His little sitting-
room held a mahogany table set against the wall, at each end two
mahogany chairs, in which we sat, and a curiously carved chair
which a friend had brought him from Africa.
 See 10a, 16, 17a, 50, 62, 64.

18. Findley, Sarah, “over 50.” She was mother to one of the house-
girls at Bethlehem, in the Santa Cruz Mountains, an old-time woman
and quite illiterate.

See 146.

19. Forbes, William, over 75. He came from Dry River and was one
of my best story-tellers. He had been song-leader and “Tea-meeting
chairman” for his district and was much respected for his intelligence
and sense of order. He came at several different times and sang or
told stories with equal ease and with a freshness and delight which
was contagious, never failing to bring a small gift from his garden
and never arguing over the pay. He had a very long-shaped head
and beaming eyes.

See 2a, 3, 7, 8, 11b, 24, 25b, 46, 70, 85b, 86a, 101, 140, 141,
142.

20. Ford, May, a young girl. She was of the better class, daughter of
the lodging-house keeper at Newmarket, in Westmoreland.

See 44.

21. Foster, Alexander, aged 40, one of the Maroon story-tellers.

See 88.

22. Gentle, Julia, over 70. She came to me twice at Bethlehem in the
Santa Cruz mountains and recited the stories with great rapidity as if
she knew them by heart in a fixed form, among them some English
ballads of second rate interest.

See 18, 60b, 61, 69b, 77, 112.

23. Hall, Matilda, aged 50, one of the women who came to me at
Harmony Hall, and a good singer and story-teller.

See 57c, 71a, 85b, 148. [293]

24. Harris, William, a young man. He came from a shop at Maggotty

and was scarcely able to speak from bashfulness.

See 73b.

25. Hendricks, Moses, over 60. He was a white man, but lived with
his negro family like one of the race; an excellent story-teller,
dictating some fifteen stories to me at three different visits to his
house.

See 13b, 15, 25a, 26, 29, 35, 48, 69a, 99, 138.

26. Hilton, Elizabeth, aged 41. She was care-taker at Harmony-Hall

and a fine intelligent type. In the evening when her work was done,
she recited to me some thirteen stories learned from her mother who
was brought up at Harmony Hall, and from an old school-master at
Retirement.

See 37a, 40b (1), 84b, 104, 105.

27. Hilton, Norman, aged 13, son to Elizabeth.

See 5.

28. Hilton, James Anderson, aged 33, one of the Maroon men.

See 149.

29. Iron, Adolphus, about 50. A reputed humorist from Golden Grove
near Claremont, but disappointing in frock coat at the lodging house.