Nonlinear Elliptic Equations of the Second Order

Graduate Studies in Mathematics 1st Edition

Qing Han install download

https://ebookmeta.com/product/nonlinear-elliptic-equations-of-
the-second-order-graduate-studies-in-mathematics-1st-edition-
qing-han/

Download more ebook from https://ebookmeta.com

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

Nonlinear Second Order Parabolic Equations 1st Edition

Mingxin Wang

https://ebookmeta.com/product/nonlinear-second-order-parabolic-
equations-1st-edition-mingxin-wang/

Partial Differential Equations III Nonlinear Equations

Third Edition Michael E. Taylor

https://ebookmeta.com/product/partial-differential-equations-iii-
nonlinear-equations-third-edition-michael-e-taylor/

Semigroup Approach to Nonlinear Diffusion Equations 1st

Edition Viorel Barbu

https://ebookmeta.com/product/semigroup-approach-to-nonlinear-
diffusion-equations-1st-edition-viorel-barbu/

Linguistic Landscapes in Language and Teacher

Education: Multilingual Teaching and Learning Inside
and Beyond the Classroom 1st Edition Sílvia Melo-
Pfeifer
https://ebookmeta.com/product/linguistic-landscapes-in-language-
and-teacher-education-multilingual-teaching-and-learning-inside-
and-beyond-the-classroom-1st-edition-silvia-melo-pfeifer/
Hidden Human Computers The Black Women of Nasa Hidden
Heroes Sue Bradford Edwards

https://ebookmeta.com/product/hidden-human-computers-the-black-
women-of-nasa-hidden-heroes-sue-bradford-edwards/

Life After Dictatorship: Authoritarian Successor

Parties Worldwide 1st Edition James Loxton

https://ebookmeta.com/product/life-after-dictatorship-
authoritarian-successor-parties-worldwide-1st-edition-james-
loxton/

Child Psychopathology 8th Edition Eric J Mash

https://ebookmeta.com/product/child-psychopathology-8th-edition-
eric-j-mash/

Spirit of the Earth The Triad Prophecy 1 1st Edition

Adriadne Lefox

https://ebookmeta.com/product/spirit-of-the-earth-the-triad-
prophecy-1-1st-edition-adriadne-lefox/

Badlands C J Box

https://ebookmeta.com/product/badlands-c-j-box/
Pediatrics Morning Report Beyond the Pearls 1st Edition
Adler Salazar Randall Y Chan Michelle Pietzak

https://ebookmeta.com/product/pediatrics-morning-report-beyond-
the-pearls-1st-edition-adler-salazar-randall-y-chan-michelle-
pietzak/
GRADUATE STUDIES
I N M AT H E M AT I C S 171

Nonlinear
Elliptic Equations
of the Second
Order
Qing Han

American Mathematical Society

 https://doi.org/10.1090//gsm/171

GRADUATE STUDIES
I N M AT H E M AT I C S 171

Nonlinear
Elliptic Equations
of the Second
Order

Qing Han

American Mathematical Society

Providence, Rhode Island
 EDITORIAL COMMITTEE
Dan Abramovich
Daniel S. Freed
Rafe Mazzeo (Chair)
Gigliola Staffilani

2010 Mathematics Subject Classification. Primary 35J60, 35J25, 35J93, 35J96.

For additional information and updates on this book, visit

www.ams.org/bookpages/gsm-171

Library of Congress Cataloging-in-Publication Data

Names: Han, Qing, 1964–
Title: Nonlinear elliptic equations of the second order / Qing Han.
Description: Providence, Rhode Island : American Mathematical Society, [2016] | Series: Gradu-
ate studies in mathematics; volume 171 | Includes bibliographical references and index.
Identifiers: LCCN 2015043419 | ISBN 9781470426071 (alk. paper)
Subjects: LCSH: Differential equations, Elliptic. | Differential equations, Nonlinear. | AMS:
Partial differential equations – Elliptic equations and systems – Nonlinear elliptic equations.
msc | Partial differential equations – Elliptic equations and systems – Boundary value problems
for second-order elliptic equations. msc | Partial differential equations – Elliptic equations and
systems – Quasilinear elliptic equations with mean curvature operator. msc | Partial differential
equations – Elliptic equations and systems – Elliptic Monge-Ampère equations. msc
Classification: LCC QA377 .H31825 2016 | DDC 515/.3533–dc23 LC record available at http://
lccn.loc.gov/2015043419

Copying and reprinting. Individual readers of this publication, and nonprofit libraries acting
for them, are permitted to make fair use of the material, such as to copy select pages for use
in teaching or research. Permission is granted to quote brief passages from this publication in
reviews, provided the customary acknowledgment of the source is given.
Republication, systematic copying, or multiple reproduction of any material in this publication
is permitted only under license from the American Mathematical Society. Permissions to reuse
portions of AMS publication content are handled by Copyright Clearance Center’s RightsLink
service. For more information, please visit: http://www.ams.org/rightslink.
Send requests for translation rights and licensed reprints to [email protected].
Excluded from these provisions is material for which the author holds copyright. In such cases,
requests for permission to reuse or reprint material should be addressed directly to the author(s).
Copyright ownership is indicated on the copyright page, or on the lower right-hand corner of the
first page of each article within proceedings volumes.

c 2016 by the author. All rights reserved.
Printed in the United States of America.

∞ The paper used in this book is acid-free and falls within the guidelines
established to ensure permanence and durability.
Visit the AMS home page at http://www.ams.org/
10 9 8 7 6 5 4 3 2 1 21 20 19 18 17 16
To Yansu, Raymond, and Tommy
Contents

Preface vii
Introduction 1
Chapter 1. Linear Elliptic Equations 7
§1.1. The Maximum Principle 8
§1.2. Krylov-Safonov’s Harnack Inequality 23
§1.3. The Schauder Theory 42

Part 1. Quasilinear Elliptic Equations

Chapter 2. Quasilinear Uniformly Elliptic Equations 51
§2.1. Basic Properties 52
§2.2. Interior C 1 -Estimates 55
§2.3. Global C 1 -Estimates 58
§2.4. Interior C 1,α -Estimates 61
§2.5. Global C 1,α -Estimates 68
§2.6. Dirichlet Problems 73
Chapter 3. Mean Curvature Equations 79
§3.1. Principal Curvatures 80
§3.2. Global Estimates 87
§3.3. Interior Gradient Estimates 100
§3.4. Dirichlet Problems 105

v
vi Contents

Chapter 4. Minimal Surface Equations 115

§4.1. Integral Formulas 116
§4.2. Differential Identities 127
§4.3. Interior Gradient Estimates 136
§4.4. Interior Curvature Estimates 141
§4.5. Differential Identities: An Alternative Approach 151

Part 2. Fully Nonlinear Elliptic Equations

Chapter 5. Fully Nonlinear Uniformly Elliptic Equations 163
§5.1. Basic Properties 164
§5.2. Interior C 2 -Estimates 172
§5.3. Global C 2 -Estimates 194
§5.4. Interior C 2,α -Estimates 200
§5.5. Global C 2,α -Estimates 208
§5.6. Dirichlet Problems 213
Chapter 6. Monge-Ampère Equations 219
§6.1. Basic Properties 219
§6.2. Global C 2 -Estimates 223
§6.3. Interior C 2 -Estimates 236
§6.4. The Bernstein Problem 241
Chapter 7. Complex Monge-Ampère Equations 253
§7.1. Basic Properties 253
§7.2. Global C 2 -Estimates 258
Chapter 8. Generalized Solutions of Monge-Ampère Equations 277
§8.1. Monge-Ampère Measures 278
§8.2. Dirichlet Problems 300
§8.3. Global Hölder Estimates 313
§8.4. Interior C 1,α -Regularity 325
§8.5. Interior C 2,α -Regularity 340
Bibliography 355
Index 365
Preface

The theory of nonlinear elliptic partial differential equations of the second or-
der has flourished in the past half-century. The pioneering work of de Giorgi
in 1957 opened the door to the study of general quasilinear elliptic differ-
ential equations. Since then, the nonlinear elliptic differential equation has
become a diverse subject and has found applications in science and engi-
neering. In mathematics, the development of elliptic differential equations
has influenced the development of the Riemannian geometry and complex
geometry. Meanwhile, the study of elliptic differential equations in a geo-
metric setting has provided interesting new questions with fresh insights to
old problems.
This book is written for those who have completed their study of the
linear elliptic differential equations and intend to explore the fascinating field
of nonlinear elliptic differential equations. It covers two classes of nonlinear
elliptic differential equations, quasilinear and fully nonlinear, and focuses
on two important nonlinear elliptic differential equations closely related to
geometry, the mean curvature equation and the Monge-Ampère equation.
This book presents a detailed discussion of the Dirichlet problems for
quasilinear and fully nonlinear elliptic differential equations of the second
order: quasilinear uniformly elliptic equations in arbitrary domains, mean
curvature equations in domains with nonnegative boundary mean curvature,
fully nonlinear uniformly elliptic equations in arbitrary domains, and Monge-
Ampère equations in uniformly convex domains. Global solutions of these
equations are also characterized. The choice of topics is influenced by my
personal taste. Some topics may be viewed by others as too advanced for
a graduate textbook. Among those topics are the curvature estimates for
minimal surface equations, the complex Monge-Ampère equation, and the

vii
viii Preface

generalized solutions of the (real) Monge-Ampère equations. Inclusion of

these topics reflects their importance and their connections to many of the
most active current research areas.
There is an inevitable overlap with the successful monograph by Gilbarg
and Trudinger. This book, designed as a textbook, is more focused on basic
materials and techniques. Many results in this book are presented in special
forms. For example, the quasilinear and fully nonlinear uniformly elliptic
differential equations studied in this book are not in their most general
form. The study of these equations serves as a prerequisite to the study of
the mean curvature equation and the Monge-Ampère equation, respectively.
More notably, our discussion of the Monge-Ampère equations is confined
to the pure Monge-Ampère equations, instead of the Monge-Ampère type
equations.
This book is based on one-semester courses I taught at Peking Univer-
sity in the spring of 2011 and at the University of Notre Dame in the fall
of 2011. Part of it was presented in the Special Lecture Series at Peking
University in the summer of 2007, in the Summer School in Mathematics at
the University of Science and Technology of China in the summer of 2008,
and in a graduate course at Beijing International Center of Mathematical
Research in the spring of 2010.
During the writing of the book, I benefitted greatly from comments
and suggestions of many friends, colleagues, and students in my classes.
Chuanqiang Chen, Xumin Jiang, Weiming Shen, and Yue Wang read the
manuscript at various stages. Chuanqiang Chen and Jingang Xiong helped
write Chapter 8. Bo Guan, Marcus Khuri, Xinan Ma, and Yu Yuan provided
valuable suggestions on the arrangement of the book.
It is with pleasure that I record here my gratitude to my thesis advisor,
Fanghua Lin, who guided me into the fascinating world of elliptic differential
equations more than twenty years ago.
I am grateful to Arlene O’Sean, my editor at the American Mathematical
Society, for reading the manuscript and guiding the effort to turn it into a
book. Last but not least, I thank Sergei Gelfand at the AMS for his help in
bringing the book to press.
The research related to this book was partially supported by grants from
the National Science Foundation.

Qing Han
 https://doi.org/10.1090//gsm/171/01

Introduction

The primary goal of this book is to study nonlinear elliptic differential equa-
tions of the second order, with a focus on quasilinear and fully nonlinear
elliptic differential equations. Chapter 1 is a brief review of linear elliptic
differential equations. Then in Part 1 and Part 2, we study quasilinear el-
liptic differential equations and fully nonlinear elliptic differential equations,
respectively.
In Chapter 1, we review briefly three basic topics in the theory of lin-
ear elliptic equations: the maximum principle, Krylov-Safonov’s Harnack
inequality, and the Schauder theory. These topics form the foundation for
further studies of nonlinear elliptic differential equations.
Part 1 is devoted to quasilinear elliptic differential equations and consists
of three chapters.
In Chapter 2, we discuss quasilinear uniformly elliptic equations. We
derive various a priori estimates for their solutions, the estimates of the L∞ -
norms of solutions and their first derivatives by the maximum principle, and
the estimates of the Hölder semi-norms of the first derivatives by Krylov-
Safonov’s Harnack inequality. As a consequence of these estimates, we solve
the Dirichlet boundary-value problem by the method of continuity.
In Chapter 3, we discuss equations of the prescribed mean curvatures,
or the mean curvature equations. We derive various a priori estimates for
their solutions, in particular, the boundary gradient estimates, the global
gradient estimates, and the interior gradient estimates. As a consequence,
we solve the Dirichlet boundary-value problem by the method of continuity.
Difficulties in studying the mean curvature equations are due to a lack of

1
2 Introduction

the uniform ellipticity. The structure of the equation plays an important

role.
In Chapter 4, we discuss minimal surface equations. Needless to say, the
minimal surface equation is a special class of the mean curvature equations;
namely, the mean curvature vanishes identically. It might appear that this
chapter should be included in the previous one. However, there is a reason
for an independent chapter. Results in this chapter are proved by analysis
“upon surfaces”. In other words, we treat minimal surfaces as submanifolds
in the ambient Euclidean spaces and write equations on these submanifolds.
In this chapter, we will derive an improved interior gradient estimate and
an interior curvature estimate for solutions of the minimal surface equation.
Part 2 is devoted to fully nonlinear elliptic differential equations and
consists of four chapters.
In Chapter 5, we discuss fully nonlinear uniformly elliptic equations. We
derive various a priori estimates for their solutions, the estimates of the L∞ -
norms of solutions and their first and second derivatives by the maximum
principle, and the estimates of the Hölder semi-norms of the second deriva-
tives by Krylov-Safonov’s Harnack inequality. As a consequence of these
estimates, we solve the Dirichlet boundary-value problem by the method of
continuity.
In Chapter 6, we discuss Monge-Ampère equations. We derive various
a priori estimates for their solutions, in particular, the boundary Hessian
estimates, the global Hessian estimates, and the interior Hessian estimates.
As a consequence, we solve the Dirichlet boundary-value problem by the
method of continuity. Difficulties in studying the Monge-Ampère equations
are due to a lack of the uniform ellipticity. The structure of the equation
plays an important role.
In Chapter 7, we extend results in the previous chapter to the complex
case and discuss complex Monge-Ampère equations.
In Chapter 8, we discuss generalized solutions of (real) Monge-Ampère
equations. Such solutions are defined only for convex functions, which are
not assumed to be C 2 to begin with. We prove various regularity results un-
der appropriate assumptions on the corresponding Monge-Ampère measures.
In particular, we prove the strict convexity and the interior C 1,α -regularity
for solutions if the Monge-Ampère measures satisfy a doubling condition.
We also derive the optimal interior C 2,α -regularity for solutions under the
condition that the Monge-Ampère measures are induced by positive Hölder
continuous functions. The discussion is based on the level set approach.
Introduction 3

Concerning the arrangement of this book, Part 1 is not a prerequisite for

Part 2. Those who are interested only in fully nonlinear elliptic equations
can skip Part 1 entirely.

We now list some basic notations to be used in this book.

We denote by x points in Rn and write x = (x1 , . . . , xn ) in terms of
its coordinates. For any x ∈ Rn , we denote by |x| the standard Euclidean
norm, unless otherwise stated. Namely, for any x = (x1 , . . . , xn ), we have
 n 1
 2

|x| = 2
xi .
i=1

Sometimes, we need to distinguish one particular direction and write points

in Rn as (x , xn ) for x = (x1 , . . . , xn−1 ) ∈ Rn−1 . We also denote by Rn+ the
upper half-space; i.e., Rn = {x ∈ Rn : xn > 0}.
Let Ω be a domain in Rn , that is, an open and connected subset in Rn .
We denote by L∞ (Ω) the collection of all bounded functions in Ω and define
the L∞ -norm by
|u|L∞ (Ω) = sup |u|.
Ω
We denote by C(Ω) the collection of all continuous functions in Ω, by C m (Ω)
the collection of all functions with continuous derivatives up to order m,
for any integer m ≥ 1, and by C ∞ (Ω) the collection of all functions with
continuous derivatives of arbitrary order. For any u ∈ C m (Ω), we denote by
∇m u the collection of all partial derivatives of u of order m. For m = 1 and
m = 2, we usually write ∇m u in special forms. For first-order derivatives,
we write ∇u as a vector of the form
∇u = (∂1 u, . . . , ∂n u).
This is the gradient vector of u. For second-order derivatives, we write ∇2 u
in the matrix form
⎛ ⎞
∂11 u ∂12 u · · · ∂1n u
⎜ ∂21 u ∂2n u · · · ∂2n u ⎟
⎜ ⎟
∇2 u = ⎜ . .. .. .. ⎟ .
⎝ .. . . . ⎠
∂n1 u ∂n2 u · · · ∂nn u
This is a symmetric matrix, called the Hessian matrix of u. For derivatives of
order higher than two, we need to use multi-indices. A multi-index β ∈ Zn+
is given by β = (β1 , . . . , βn ) with nonnegative integers β1 , . . . , βn . We write

n
|β| = βi .
i=1
4 Introduction

The partial derivative ∂ β u is defined by

∂ β u = ∂1β1 · · · ∂nβn u,
and its order is |β|. For any positive integer m, we define
⎛ ⎞1

2

|∇ u| =
m ⎝ β 2⎠
|∂ u| ,
|β|=m

and the C m -norm by


m
|u|C m (Ω) = |∇k u|L∞ (Ω) .
k=0

For a constant α ∈ (0, 1), we denote by C α (Ω) the collection of all Hölder
continuous functions in Ω with the Hölder exponent α and by C m,α (Ω) the
collection of all functions in C m (Ω) whose derivatives of order m are Hölder
continuous in Ω with the Hölder exponent α. We define the Hölder semi-
norm by
|u(x) − u(y)|
[u]C α (Ω) = sup ,
x,y∈Ω |x − y|α
x=y

and the C α -norm by

|u|C α (Ω) = |u|L∞ (Ω) + [u]C α (Ω) .
For any positive integer m and a contant α ∈ (0, 1), we also define the
C m,α -norm by

|u|C m,α (Ω) = |u|C m (Ω) + [∇β u]C α (Ω) .
|β|=m

Accordingly, we can define C(Ω̄), C α (Ω̄), C m (Ω̄), C m,α (Ω̄), and C ∞ (Ω̄) if
∂Ω is appropriately regular and define [ · ]C α (Ω̄) , | · |C α (Ω̄) , | · |C m (Ω̄) , and
| · |C m,α (Ω̄) similarly.
We adopt the summation convention on repeated indices throughout the
book. The general form of the linear equations of the second order is given
by
aij (x)∂ij u + bi (x)∂i u + c(x)u = f (x) in Ω,
where aij , bi , c, and f are given functions in Ω. Very often, we write deriva-
tives as ui = ∂i u and uij = ∂ij u for brevity. In this way, we can express
linear equations in the following form:
aij uij + bi ui + cu = f in Ω.
Introduction 5

Subscripts here have different meanings for coefficients and solutions. Sim-
ilarly, the general forms of the quasilinear equations and the fully nonlinear
equations of the second order are given, respectively, by
aij (x, u, ∇u)uij = f (x, u, ∇u) in Ω
and
F (x, u, ∇u, ∇2 u) = 0 in Ω.
A significant portion of the book is devoted to the derivation of a priori
estimates, where certain norms of solutions are bounded by a positive con-
stant C depending only on a set of known quantities. In a given context,
the same letter C will be used to denote different constants depending on
the same set of quantities.
 https://doi.org/10.1090//gsm/171/02

Chapter 1

Linear Elliptic
Equations

In this chapter, we review briefly three basic topics in the theory of lin-
ear elliptic equations: the maximum principle, Krylov-Safonov’s Harnack
inequality, and the Schauder theory.
In Section 1.1, we review Hopf’s maximum principle. The maximum
principle is an important method to study elliptic differential equations of
the second order. In this section, we review the weak maximum principle and
the strong maximum principle and derive several forms of a priori estimates
of solutions.
In Section 1.2, we review Krylov-Safonov’s Harnack inequality. The
Harnack inequality is an important result in the theory of elliptic differential
equations of the second order and plays a fundamental role in the study of
nonlinear elliptic differential equations.
In Section 1.3, we review the Schauder theory for uniformly elliptic linear
equations. Three main topics are a priori estimates in Hölder norms, the
regularity of arbitrary solutions, and the solvability of the Dirichlet problem.
Among these topics, a priori estimates are the most fundamental and form
the basis for the existence and the regularity of solutions. We will review
both the interior Schauder theory and the global Schauder theory.
These three sections play different roles in the rest of the book. In the
study of quasilinear elliptic equations in Part 1, the maximum principle
will be used to derive estimates of derivatives up to the first order, the
Harnack inequality will be used to derive estimates of the Hölder semi-
norms of derivatives of the first order, and the Schauder theory will be used

7
8 1. Linear Elliptic Equations

to solve the linearized equations. In the study of fully nonlinear elliptic

equations in Part 2, the maximum principle will be used to derive estimates
of derivatives up to the second order, the Harnack inequality will be used to
derive estimates of the Hölder semi-norms of derivatives of the second order,
and the Schauder theory will be used to solve the linearized equations.
It is not our intention to present a complete review of the linear theory.
Notably missing from this short review are the W 2,p -theory for linear equa-
tions of the nondivergence form and the H k -theory and the de Giorgi-Moser
theory for linear equations of the divergence form. Refer to Chapters 2–9 of
[59] for a complete account of the linear theory.

1.1. The Maximum Principle

The maximum principle is an important method to study elliptic differential
equations of the second order. In this section, we review the weak maximum
principle and the strong maximum principle and derive several forms of a
priori estimates of solutions. Refer to Chapter 3 of [59] for details.
Throughout this section, we let Ω be a bounded domain in Rn and let
aij , bi , and c be bounded and continuous functions in Ω, with aij = aji . We
consider the operator L given by
(1.1.1) Lu = aij ∂ij u + bi ∂i u + cu in Ω,
for any u ∈ C 2 (Ω). The operator L is always assumed to be strictly elliptic
in Ω; namely, for any x ∈ Ω and ξ ∈ Rn ,
(1.1.2) aij (x)ξi ξj ≥ λ|ξ|2 ,
for some positive constant λ. For later reference, L is called uniformly elliptic
if, for any x ∈ Ω and ξ ∈ Rn ,
(1.1.3) λ|ξ|2 ≤ aij (x)ξi ξj ≤ Λ|ξ|2 ,
for some positive constants λ and Λ, which are usually called the ellipticity
constants.

1.1.1. The Weak Maximum Principle. In this subsection, we review

the weak maximum principle and its corollaries. We first introduce subso-
lutions and supersolutions.
Definition 1.1.1. For some f ∈ C(Ω), a C 2 (Ω)-function u is called a sub-
solution (or supersolution) of Lw = f if Lu ≥ f (or Lu ≤ f ) in Ω.

If aij = δij , bi = c = 0, and f = 0, subsolutions (or supersolutions) are

subharmonic (or superharmonic).
Now we prove the weak maximum principle for subsolutions. Recall that
u+ is the nonnegative part of u, defined by u+ = max{0, u}.
1.1. The Maximum Principle 9

Theorem 1.1.2. Let Ω be a bounded domain in Rn and L be given by

(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and
(1.1.2). Suppose that u ∈ C(Ω̄) ∩ C 2 (Ω) satisfies Lu ≥ 0 in Ω. Then, u
attains on ∂Ω its nonnegative maximum in Ω̄; i.e.,
max u ≤ max u+ .
Ω̄ ∂Ω

Proof. We first consider the special case Lu > 0 in Ω. If u has a local

nonnegative maximum at a point x0 in Ω, then u(x0 ) ≥ 0, ∇u(x0 ) = 0,
and the Hessian matrix ∇2 u(x0 ) is negative semi-definite. By (1.1.2), the
matrix aij (x0 ) is positive definite. Then,
Lu(x0 ) = (aij ∂ij u + bi ∂i u + cu) (x0 ) ≤ 0.
This leads to a contradiction. Hence, the nonnegative maximum of u in Ω̄
is attained only on ∂Ω.
Now we consider the general case Lu ≥ 0 in Ω. For any ε > 0, consider
w(x) = u(x) + εeμx1 ,
where μ is a positive constant to be determined. Then,
Lw = Lu + εeμx1 (a11 μ2 + b1 μ + c).
Since b1 and c are bounded and a11 ≥ λ > 0 in Ω, by choosing μ > 0 large
enough, we get
a11 μ2 + b1 μ + c > 0 in Ω.
This implies Lw > 0 in Ω. By the special case we just discussed, w attains
its nonnegative maximum only on ∂Ω and hence,
max w ≤ max w+ .
Ω̄ ∂Ω

Then,
max u ≤ max w ≤ max w+ ≤ max u+ + ε max eμx1 .
Ω̄ Ω̄ ∂Ω ∂Ω x∈∂Ω

We have the desired result by letting ε → 0 and using the fact that ∂Ω ⊂ Ω̄.


If c ≡ 0 in Ω, we can draw conclusions about the maximum of u rather

than its nonnegative maximum. A similar remark holds for the strong max-
imum principle.
A continuous function in Ω̄ always attains its maximum in Ω̄. Theorem
1.1.2 asserts that any subsolution continuous up to the boundary attains its
maximum on the boundary ∂Ω, but possibly also in Ω. Theorem 1.1.2 is
10 1. Linear Elliptic Equations

referred to as the weak maximum principle. A stronger version asserts that

subsolutions attain their maximum only on the boundary, unless they are
constant.
As a simple consequence of Theorem 1.1.2, we have the following result.
Corollary 1.1.3. Let Ω be a bounded domain in Rn and L be given by
(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and
(1.1.2). Suppose that u ∈ C(Ω̄) ∩ C 2 (Ω) satisfies Lu ≥ 0 in Ω and u ≤ 0 on
∂Ω. Then, u ≤ 0 in Ω.

More generally, we have the following comparison principle.

Corollary 1.1.4. Let Ω be a bounded domain in Rn and L be given by
(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and
(1.1.2). Suppose that u, v ∈ C(Ω̄) ∩ C 2 (Ω) satisfy Lu ≥ Lv in Ω and u ≤ v
on ∂Ω. Then, u ≤ v in Ω.

The comparison principle provides a reason that functions u satisfying

Lu ≥ f are called subsolutions. They are less than a solution v of Lv = f
with the same boundary value.
In the following, we simply say by the maximum principle when we apply
Theorem 1.1.2, Corollary 1.1.3, or Corollary 1.1.4.
A consequence of the maximum principle is the uniqueness of solutions
of Dirichlet problems.
Corollary 1.1.5. Let Ω be a bounded domain in Rn and L be given by
(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and
(1.1.2). Then for any f ∈ C(Ω) and ϕ ∈ C(∂Ω), there exists at most one
solution u ∈ C(Ω̄) ∩ C 2 (Ω) of
Lu = f in Ω,
u=ϕ on ∂Ω.

1.1.2. The Strong Maximum Principle. The weak maximum principle

asserts that subsolutions of linear elliptic equations attain their nonnegative
maximum on the boundary under suitable conditions. In fact, these subso-
lutions can attain their nonnegative maximum only on the boundary, unless
they are constant. This is the strong maximum principle.
For any C 1 -function u in Ω̄ that attains its maximum on ∂Ω, say at
x0 ∈ ∂Ω, we have ∂u ∂ν (x0 ) ≥ 0, where ν is the exterior unit normal to Ω at
x0 . The Hopf lemma asserts that this normal derivative is in fact positive if
u is a subsolution in Ω.
Theorem 1.1.6. Let B be an open ball in Rn with x0 ∈ ∂B and L be given
by (1.1.1), for some aij , bi , c ∈ L∞ (B) ∩ C(B) satisfying c ≤ 0 in B and
1.1. The Maximum Principle 11

(1.1.2). Suppose that u ∈ C 1 (B̄)∩C 2 (B) satisfies Lu ≥ 0 in B, u(x) < u(x0 )

for any x ∈ B, and u(x0 ) ≥ 0. Then,
∂u
(x0 ) > 0,
∂ν
where ν is the exterior unit normal to B at x0 .

Proof. Without loss of generality, we assume B = BR for some R > 0. By

the continuity of u up to ∂BR , we have, for any x ∈ B̄R ,

u(x) ≤ u(x0 ).

For positive constants μ and ε to be determined, we set

w(x) = e−μ|x| − e−μR

2 2

and
v(x) = u(x) − u(x0 ) + εw(x).
We consider w and v in D = BR \ B̄R/2 .
A direct calculation yields

Lw = e−μ|x| 4μ2 aij xi xj − 2μaij δij − 2μbi xi + c − ce−μR
2 2


≥ e−μ|x|
2
4μ2 aij xi xj − 2μ aij δij + bi xi + c ,

where we used c ≤ 0 in BR . By the strict ellipticity (1.1.2), we have

1
aij (x)xi xj ≥ λ|x|2 ≥ λR2 in D.
4
Hence,

Lw ≥ e−μ|x|
2
μ2 λR2 − 2μ aij δij + bi xi + c ≥ 0 in D

if we choose μ sufficiently large. By c ≤ 0 and u(x0 ) ≥ 0, we obtain, for any

ε > 0,
Lv = Lu + εLw − cu(x0 ) ≥ 0 in D.
Next, we discuss v on ∂D in two cases. First, on ∂BR/2 , we have u−u(x0 ) <
0, and hence u−u(x0 ) < −ε for some ε > 0, by the continuity of u on ∂BR/2 .
Note that w < 1 on ∂BR/2 . Then for such an ε, we obtain v < 0 on ∂BR/2 .
Second, on ∂BR , we have w = 0 and u ≤ u(x0 ). Hence, v ≤ 0 on ∂BR and
v(x0 ) = 0. Therefore, v ≤ 0 on ∂D.
In conclusion, Lv ≥ 0 in D and v ≤ 0 on ∂D. By the maximum principle,
we have
v≤0 in D.
12 1. Linear Elliptic Equations

In view of v(x0 ) = 0, v attains at x0 its maximum in D̄. Hence, we obtain

∂v
(x0 ) ≥ 0,
∂ν
and then
∂u ∂w
(x0 ) = 2εμRe−μR > 0.
2
(x0 ) ≥ −ε
∂ν ∂ν
This is the desired result. 

Theorem 1.1.6 still holds if we substitute for B any bounded C 1 -domain

which satisfies an interior sphere condition at x0 ∈ ∂Ω, namely, if there
exists a ball B ⊂ Ω with x0 ∈ ∂B. This is because such a ball B is tangent
to ∂Ω at x0 . We note that the interior sphere condition always holds for
C 2 -domains.
Now, we are ready to prove the strong maximum principle due to Hopf
[84].
Theorem 1.1.7. Let Ω be a bounded domain in Rn and L be given by
(1.1.1), for some aij , bi , c ∈ C(Ω) satisfying c ≤ 0 in Ω and (1.1.2). Suppose
that u ∈ C(Ω̄) ∩ C 2 (Ω) satisfies Lu ≥ 0 in Ω. Then, u attains only on ∂Ω
its nonnegative maximum in Ω̄ unless u is constant.

Proof. Let M be the nonnegative maximum of u in Ω̄ and set

D = {x ∈ Ω : u(x) = M }.
We prove either D = ∅ or D = Ω by contradiction. Suppose D is a nonempty
proper subset of Ω. It follows from the continuity of u that D is relatively
closed in Ω. Then, Ω \ D is open and we can find an open ball B ⊂ Ω \ D
such that ∂B ∩ D = ∅. In fact, we may choose a point x∗ ∈ Ω \ D with
dist(x∗ , D) < dist(x∗ , ∂Ω) and then take the ball centered at x∗ with the
radius dist(x∗ , D). Suppose x0 ∈ ∂B ∩ D. Obviously, we have
Lu ≥ 0 in B
and
u(x) < u(x0 ) for any x ∈ B and u(x0 ) = M ≥ 0.
By Theorem 1.1.6, we have
∂u
(x0 ) > 0,
∂ν
where ν is the exterior unit normal to B at x0 . On the other hand, x0 is
an interior maximum point of u in Ω. This implies ∇u(x0 ) = 0, which leads
to a contradiction. Therefore, either D = ∅ or D = Ω. In the first case, u
attains only on ∂Ω its nonnegative maximum in Ω̄, while in the second case,
u is constant in Ω. 
1.1. The Maximum Principle 13

The following result improves Corollary 1.1.3.

Corollary 1.1.8. Let Ω be a bounded domain in Rn and L be given by
(1.1.1), for some aij , bi , c ∈ C(Ω) satisfying c ≤ 0 in Ω and (1.1.2). Suppose
that u ∈ C(Ω̄) ∩ C 2 (Ω) satisfies Lu ≥ 0 in Ω and u ≤ 0 on ∂Ω. Then, either
u < 0 in Ω or u ≡ 0 in Ω.

1.1.3. A Priori Estimates. As Corollary 1.1.5 shows, an important ap-

plication of the maximum principle is to prove the uniqueness of solutions
of boundary-value problems. Equally or more important is to derive a priori
estimates. In derivations of a priori estimates, it is essential to construct
auxiliary functions. We will provide proofs of all results in this subsection,
as constructing auxiliary functions is an important technique we will develop
in this book. We point out that we need only the weak maximum principle
in this subsection.
We first derive an estimate for subsolutions. In the next result, we write
u+ = max{u, 0} as before and f − = max{−f, 0}.
Theorem 1.1.9. Let Ω be a bounded domain in Rn and L be given by
(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and
(1.1.2). Suppose that u ∈ C(Ω̄) ∩ C 2 (Ω) satisfies
Lu ≥ f in Ω,
for some f ∈ L∞ (Ω) ∩ C(Ω). Then,
sup u ≤ max u+ + Cd2 sup f − ,
Ω ∂Ω Ω

where d = diam(Ω) and C is a positive constant depending only on n, λ,

and d|bi |L∞ (Ω) , for i = 1, . . . , n.

Proof. Set
F = sup f − , Φ = max u+ .
Ω ∂Ω

Then, Lu ≥ −F in Ω and u ≤ Φ on ∂Ω. Without loss of generality, we

assume Ω ⊂ {0 < x1 < d} for some constant d > 0. For some constant
μ > 0 to be chosen later, set
 μx1

v = Φ + d2 eμ − e d F.

We note that v ≥ Φ in Ω̄. Next, by a straightforward calculation and c ≤ 0

in Ω, we have
μx1
 μx1

Lv = −(a11 μ2 + db1 μ)F e d + cΦ + cd2 eμ − e d F
≤ −(a11 μ2 + db1 μ)F.
14 1. Linear Elliptic Equations

Note that a11 ≥ λ in Ω by the strict ellipticity (1.1.2). We choose μ large

so that
a11 μ2 + db1 μ ≥ 1 in Ω.
Then, Lv ≤ −F in Ω. Therefore,
Lu ≥ Lv in Ω,
u≤v on ∂Ω.
By the maximum principle, we obtain
u≤v in Ω,
and hence, for any x ∈ Ω,
 μx1

u(x) ≤ Φ + d2 eμ − e d F.

This yields the desired result. 

The function v in the proof above is what we called an auxiliary function.

In fact, auxiliary functions were already used in the proof of Theorem 1.1.6.
If L in Theorem 1.1.9 does not involve the lower-order terms, i.e., L =
aij ∂ij , we may take
1
v =Φ+ F (d2 − x21 ).
2λ
In this case, a simple calculation yields
1
Lv = − a11 F ≤ −F,
λ
where we used the strict ellipticity (1.1.2).
By replacing u with −u, Theorem 1.1.9 extends to supersolutions and
solutions of the equation Lu = f .

Theorem 1.1.10. Let Ω be a bounded domain in Rn and L be given by

(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and
(1.1.2). Suppose that u is a C(Ω̄) ∩ C 2 (Ω)-solution of
Lu = f in Ω,
u=ϕ on ∂Ω,
for some f ∈ L∞ (Ω) ∩ C(Ω) and ϕ ∈ C(∂Ω). Then,
sup |u| ≤ max |ϕ| + C sup |f |,
Ω ∂Ω Ω

where C is a positive constant depending only on n, λ, diam(Ω), and the

sup-norms of bi .
1.1. The Maximum Principle 15

Next, we construct barrier functions for a large class of domains Ω and

discuss global properties of solutions. The geometry of ∂Ω plays an impor-
tant role. We consider the case where Ω satisfies an exterior sphere condition
at x0 ∈ ∂Ω in the sense that there exists a ball BR (y0 ) such that
Ω ∩ BR (y0 ) = ∅, Ω̄ ∩ B̄R (y0 ) = {x0 }.
Lemma 1.1.11. Let Ω be a bounded domain in Rn satisfying an exterior
sphere condition at x0 ∈ ∂Ω and L be given by (1.1.1), for some aij , bi ,
c ∈ L∞ (Ω) ∩ C(Ω) satisfying c ≤ 0 in Ω and (1.1.3). Then, there exists a
function wx0 ∈ C(Ω̄) ∩ C 2 (Ω) such that
Lwx0 ≤ −1 in Ω
and, for any x ∈ Ω̄ \ {x0 },
wx0 (x0 ) = 0, wx0 (x) > 0,
where wx0 depends only on n, λ, Λ, the L∞ -norms of bi , diam(Ω), and R
in the exterior sphere condition.

Proof. Set D = diam(Ω). For the given x0 ∈ ∂Ω, consider an exterior ball
BR (y) with B̄R (y) ∩ Ω̄ = {x0 }. Let d(x) be the distance from x to ∂BR (y);
i.e.,
d(x) = |x − y| − R.
Then, for any x ∈ Ω,
0 < d(x) < D.
Consider a C 2 -function ψ defined in [0, D), with ψ(0) = 0 and ψ > 0 in
(0, D). Set
w = ψ(d) in Ω.
We now calculate Lw. A direct calculation yields
xi − yi
∂i d(x) = ,
|x − y|
δij (xi − yi )(xj − yj )
∂ij d(x) = − .
|x − y| |x − y|3
Hence, |∇d| = 1, aij ∂i d∂j d ≥ λ, and
1 1
aij ∂ij d = aij δij − aij ∂i d∂j d
|x − y| |x − y|
nΛ λ nΛ − λ nΛ − λ
≤ − = ≤ .
|x − y| |x − y| |x − y| R
Next,
∂i w = ψ  ∂i d, ∂ij w = ψ  ∂i d∂j d + ψ  ∂ij d.
Then,
Lw = ψ  aij ∂i d∂j d + ψ  (aij ∂ij d + bi ∂i d) + cψ.
16 1. Linear Elliptic Equations

We now require ψ  > 0 and ψ  < 0. Hence,

 
 nΛ − λ
Lw ≤ λψ + + b0 ψ  ,
R
where
 1

n 2

b0 = sup b2i .
Ω i=1
We write this as
Lw ≤ λ ψ  + aψ  + b − 1,
where
nΛ − λ b0 1
a= + , b= .
λR λ λ
We need to find a function ψ in [0, D) such that
ψ  + aψ  + b = 0 in (0, D),
 
ψ < 0, ψ > 0 in (0, D), and ψ(0) = 0.
First, general solutions of the ordinary differential equation above are given
by
b C2 −ad
ψ(d) = − d + C1 − e ,
a a
for some constants C1 and C2 . For ψ(0) = 0, we need C1 = C2 /a. Hence
we have, for some constant C,
b C
ψ(d) = − d + (1 − e−ad ),
a a
which implies
 
b b
ψ  (d) = Ce−ad − = e−ad C − ead ,
a a
ψ  (d) = −Cae−ad .
b aD
In order to have ψ  > 0 in (0, D), we need C ≥ e . Then, ψ  > 0 in
a
(0, D), and hence ψ > ψ(0) = 0 in (0, D). Therefore, we take
b b
ψ(d) = − d + 2 eaD (1 − e−ad )
a
 a 
b 1 aD −ad
= e (1 − e ) − d .
a a
Such a ψ satisfies all the requirements we imposed. 

Now we estimate the modulus of continuity of solutions with the help of

barrier functions constructed in Lemma 1.1.11.
1.1. The Maximum Principle 17

Theorem 1.1.12. Let Ω be a bounded domain in Rn satisfying an exterior

sphere condition at x0 ∈ ∂Ω and L be given by (1.1.1), for some aij , bi ,
c ∈ L∞ (Ω) ∩ C(Ω) satisfying (1.1.3). Suppose that u is a C(Ω̄) ∩ C 2 (Ω)-
solution of
Lu = f in Ω,
u=ϕ on ∂Ω,
for some f ∈ L∞ (Ω) ∩ C(Ω) and ϕ ∈ C(∂Ω). Then, for any x ∈ Ω,
|u(x) − u(x0 )| ≤ ω(|x − x0 |),
where ω is a nondecreasing continuous function in (0, D), with D = diam(Ω)
and limr→0 ω(r) = 0, depending only on n, λ, Λ, the L∞ -norms of bi and c,
diam(Ω), R in the exterior sphere condition, supΩ |u|, max∂Ω |ϕ|, supΩ |f |,
and the modulus of continuity of ϕ on ∂Ω.

Proof. Set
L0 = aij ∂ij + bi ∂i .
Then, L0 u = f − cu in Ω. Let w = wx0 be the function in Lemma 1.1.11 for
L0 , i.e.,
L0 w ≤ −1 in Ω,
and, for any x ∈ ∂Ω \ {x0 },
w(x0 ) = 0, w(x) > 0.
We set
F = sup |f − cu|, Φ = max |ϕ|.
Ω ∂Ω

Then,
L0 (±u) ≥ −F in Ω.
Let ε be an arbitrary positive constant. By the continuity of ϕ at x0 , there
exists a positive constant δ such that, for any x ∈ ∂Ω ∩ Bδ (x0 ),
|ϕ(x) − ϕ(x0 )| ≤ ε.
We then choose K sufficiently large so that K ≥ F and
Kw ≥ 2Φ on ∂Ω \ Bδ (x0 ).
We point out that K depends on ε through the positive lower bound of w
on ∂Ω \ Bδ (x0 ). Then,
L0 (Kw) ≤ −F in Ω,
and
|ϕ − ϕ(x0 )| ≤ ε + Kw on ∂Ω.
18 1. Linear Elliptic Equations

Therefore,

L0 ± (u − ϕ(x0 )) ≥ L0 (ε + Kw) in Ω,
± u − ϕ(x0 ) ≤ ε + Kw on ∂Ω.

By the maximum principle, we have ± u−ϕ(x0 ) ≤ ε +Kw in Ω, and hence

|u − ϕ(x0 )| ≤ ε + Kw in Ω.

Note that the second term in the right-hand side converges to zero as x → x0 .
Then, there exists a positive constant δ  < δ such that

|u − ϕ(x0 )| ≤ 2ε in Ω ∩ Bδ (x0 ).

This yields the desired result. 

Remark 1.1.13. It is clear from the proof that Theorem 1.1.12 is a local
result. If we assume that ϕ is continuous at x0 ∈ ∂Ω and, in addition, u
is bounded in a neighborhood of x0 , then we can estimate the modulus of
continuity of u at x0 .

By a similar method as in the proof of Theorem 1.1.12, we can derive a

boundary gradient estimate.

Theorem 1.1.14. Let Ω be a bounded domain in Rn satisfying an exterior

sphere condition at x0 ∈ ∂Ω and L be given by (1.1.1), for some aij , bi ,
c ∈ L∞ (Ω) ∩ C(Ω) satisfying (1.1.3). Suppose that u is a C(Ω̄) ∩ C 2 (Ω)-
solution of

Lu = f in Ω,
u=ϕ on ∂Ω,

for some f ∈ L∞ (Ω) ∩ C(Ω) and ϕ ∈ C 2 (Ω̄). Then, for any x ∈ Ω,

 
|u(x) − u(x0 )| ≤ C sup |u| + |ϕ|C 2 (Ω̄) + sup |f | |x − x0 |,
Ω Ω

where C is a positive constant depending only on n, λ, Λ, the L∞ -norms of

bi and c, diam(Ω), and R in the exterior sphere condition.

If ∂Ω is C 1 at x0 and u is C 1 at x0 , taking the normal derivative at x0 ,

we obtain
   
 ∂u 
 (x0 ) ≤ C sup |u| + |ϕ| 2 + sup |f | .
 ∂ν  C (Ω̄)
Ω Ω
1.1. The Maximum Principle 19

Proof. Set
L0 = aij ∂ij + bi ∂i .
Then, L0 u = f − cu in Ω. By setting v = u − ϕ, we have
L0 v = f − cu − L0 ϕ in Ω,
v = 0 on ∂Ω.
Next, we set
F = sup |f − cu − L0 ϕ|.
Ω
Then,
L0 (±v) ≥ −F in Ω,
±v = 0 on ∂Ω.
Let w = wx0 be the function in Lemma 1.1.11 for L0 , i.e.,
L0 w ≤ −1 in Ω,
and, for any x ∈ ∂Ω \ {x0 },
w(x0 ) = 0, w(x) > 0.
Then,
L0 (±v) ≥ L0 (F w) in Ω,
±v ≤ F w on ∂Ω.
By the maximum principle, we have ±v ≤ F w in Ω, and hence
|v| ≤ F w in Ω.
With v = u − ϕ and u(x0 ) = ϕ(x0 ), we obtain
|u − u(x0 )| ≤ |u − ϕ| + |ϕ − ϕ(x0 )| ≤ F w + |ϕ − ϕ(x0 )|.
This implies the desired result. 

A remark similar to Remark 1.1.13 holds for Theorem 1.1.14.

It does not seem optimal to require ϕ to be C 2 in Theorem 1.1.14. It
is natural to ask whether the result holds if ϕ ∈ C 1 (∂Ω). However, C 1 -
boundary values in general do not yield boundary gradient estimates, even
for harmonic functions in balls. It is a good exercise to derive the best
modulus of continuity for harmonic functions in balls with C 1 -boundary
values.
To end this section, we derive an estimate of the Hölder semi-norms near
the boundary.
20 1. Linear Elliptic Equations

Theorem 1.1.15. Let α ∈ (0, 1) be a constant, Ω be a bounded domain in

Rn satisfying an exterior sphere condition at x0 ∈ ∂Ω, and L be given by
(1.1.1), for some aij , bi , c ∈ L∞ (Ω) ∩ C(Ω) satisfying (1.1.3). Suppose that
u is a C(Ω̄) ∩ C 2 (Ω)-solution of
Lu = f in Ω,
u=ϕ on ∂Ω,
for some f ∈ L∞ (Ω) ∩ C(Ω) and ϕ ∈ C(∂Ω) satisfying, for any x ∈ ∂Ω,
|ϕ(x) − ϕ(x0 )| ≤ Φα |x − x0 |α ,
for some positive constant Φα . Then, for any x ∈ Ω,
 
α
|u(x) − u(x0 )| ≤ C sup |u| + Φα + sup |f | |x − x0 | 1+α ,
Ω Ω

where C is a positive constant depending only on n, α, λ, Λ, the L∞ -norms

of bi and c, diam(Ω), and R in the exterior sphere condition.

Proof. Set
L0 = aij ∂ij + bi ∂i .
Then, L0 u = f − cu in Ω. Next, we set
M = sup |u|, F = sup |f − cu|.
Ω Ω

Hence,
L0 ± (u − u(x0 )) ≥ −F in Ω.
Now, we take a ball BR (y0 ) ⊂ Rn
such that Ω̄ ∩ B̄R (y0 ) = {x0 }. Then, for
any r ∈ (0, R], we take y ∈ R such that Ω̄ ∩ B̄r (y) = {x0 }. Note that
n

r < |x − y| < 2r for any x ∈ Ω ∩ B2r (y).

For any x ∈ ∂Ω ∩ B2r (y), we have
|u(x) − u(x0 )| ≤ Φα |x − x0 |α ≤ (3r)α Φα ,
where we used the triangle inequality |x − x0 | ≤ |x − y| + |y − x0 | ≤ 3r. On
Ω ∩ ∂B2r (y), we simply have
|u − u(x0 )| ≤ 2M.
Therefore,
L0 ± (u − u(x0 )) ≥ −F in Ω ∩ B2r (y),
± u − u(x0 ) ≤ (3r) Φα α
on ∂Ω ∩ B2r (y),
± u − u(x0 ) ≤ 2M on Ω ∩ ∂B2r (y).
It is worth pointing out that r will vary later on.
1.1. The Maximum Principle 21

Set, for some positive constant μ to be determined,

rμ
v(x) = 1 − .
|x − y|μ
Then, 0 < v < 1 in Ω ∩ B2r (y). A straightforward calculation yields, for any
i, j = 1, . . . , n,
∂i v = μrμ |x − y|−μ−2 (xi − yi ),
 
(xi − yi )(xj − yj )
∂ij v = μrμ |x − y|−μ−2 −(μ + 2) + δ ij ,
|x − y|2
and hence,

−μ−2 aij (xi − yi )(xj − yj )
L0 v = μr |x − y|
μ
− (μ + 2)
|x − y|2

+ δij aij + bi (xi − yi ) .

By the uniform ellipticity (1.1.3), we get

μrμ
L0 v ≤ − (μ + 2)λ + nΛ + 2rb0 in Ω ∩ B2r (y),
|x − y|μ+2
where

n 1
2
b0 = sup b2i .
Ω i=1
By choosing μ ≥ 1 sufficiently large, depending only on n, λ, Λ, b0 , and R,
we have
rμ 1 1
L0 v ≤ − ≤ − μ+2 2 ≤ − μ+2 2 in Ω ∩ B2r (y).
|x − y| μ+2 2 r 2 R
For some positive constant A to be determined, we set
w = Av + (3r)α Φα .
First, we have
A
L0 w = AL0 v ≤ − in Ω ∩ B2r (y).
2μ+2 R2
To have L0 ± (u − u(x0 )) ≥ L0 w in Ω ∩ B2r (y), we require A ≥ 2μ+2 R2 F .
Next, it is obvious that ±(u − u(x0 )) ≤ w on ∂Ω ∩ B2r (y) for any A > 0
since v ≥ 0 there. Last, for any x ∈ Ω ∩ ∂B2r (y), we have
rμ 1 1
v(x) = 1 − =1− μ ≥ ,
|x − y|μ 2 2
as long as μ ≥ 1. Hence, to have ±(u − u(x0 )) ≤ w on Ω ∩ ∂B2r (y), we
require A ≥ 4M . Therefore, we take
A = 2μ+2 R2 F + 4M.
22 1. Linear Elliptic Equations

Then,
L0 ± (u − u(x0 )) ≥ L0 w in Ω ∩ B2r (y),
± u − u(x0 ) ≤ w on ∂(Ω ∩ B2r (y)).
By the maximum principle, we obtain ± u − u(x0 ) ≤ w in Ω ∩ B2r (y), and
hence
|u − u(x0 )| ≤ w in Ω ∩ B2r (y).
This implies, for any x ∈ Ω ∩ B2r (y),
 
rμ
|u(x) − u(x0 )| ≤ A 1 − + 3rα Φα .
|x − y|μ
By the inequality 1 − tμ ≤ μ(1 − t) for any t ∈ (0, 1), we obtain, for any
x ∈ Ω ∩ B2r (y),
 
rμ r μ(|x − y| − r)
1− ≤ μ 1 − ≤
|x − y|μ |x − y| |x − y|
μ|x − x0 | μ|x − x0 |
≤ ≤ ,
|x − y| r
where we used the triangle inequality and |x0 − y| = r. Therefore, for any
x ∈ Ω ∩ B2r (y),
|x − x0 |
|u(x) − u(x0 )| ≤ μA + 3rα Φα .
r

Now, we fix an x ∈ Ω and set R0 = min{1, R}. If |x − x0 | < R01+α , take

r such that |x − x0 | = r1+α . Obviously, r < R0 . With the ball Br (y) chosen
as above with Ω̄ ∩ B̄r (y) = {x0 }, we have

|x − y| ≤ |x − x0 | + |x0 − y| ≤ r1+α + r < 2r.

Therefore, x ∈ Ω ∩ B2r (y) and hence

α
|u(x) − u(x0 )| ≤ (μA + 3Φα )rα = (μA + 3Φα )|x − x0 | 1+α .

If |x − x0 | ≥ R01+α , then
2M α
|u(x) − u(x0 )| ≤ 2M ≤ α |x − x0 | 1+α .
R0
Hence, we have the desired estimate. 

The power α/(1 + α) does not seem optimal. However, the present form
is sufficient for applications later on.
1.2. Krylov-Safonov’s Harnack Inequality 23

1.2. Krylov-Safonov’s Harnack Inequality

The Harnack inequality is an important result in the theory of elliptic dif-
ferential equations of the second order and plays a fundamental role in the
study of nonlinear elliptic differential equations. In this section, we review
the Harnack inequality due to Krylov and Safonov. Refer to Chapter 9 of
[59] for details.

1.2.1. Aleksandrov’s Maximum Principle. We first prove a maximum

principle due to Aleksandrov, which yields an estimate of solutions in terms
of integral norms of nonhomogeneous terms. To do this, we need to introduce
the concept of contact sets.
For any u ∈ C(Ω), we define

Γ+ = y ∈ Ω : u(x) ≤ u(y) + p · (x − y)

for any x ∈ Ω and some p = p(y) ∈ Rn .

The set Γ+ is called the upper contact set of u. Clearly, u is concave if and
only if Γ+ = Ω. If u ∈ C 1 (Ω), then p(y) = ∇u(y) for y ∈ Γ+ . In this case,
any support hyperplane must be a tangent plane to the graph, and

Γ+ = {y ∈ Ω : u(x) ≤ u(y) + ∇u(y) · (x − y) for any x ∈ Ω} .

If u ∈ C 2 (Ω), the Hessian matrix ∇2 u = (∂ij u) is negative semi-definite in

Γ+ . Lower contact sets can be defined similarly.
We first prove a result on the maximum of positive parts of C 2 -functions.

Lemma 1.2.1. Let Ω be a bounded domain in Rn and g ∈ L1loc (Rn ) be

nonnegative. Then, for any u ∈ C(Ω̄) ∩ C 2 (Ω),
 
g(p) dp ≤ g(∇u)| det ∇2 u| dx,
BM/d Γ+

where Γ+ is the upper contact set of u, M = supΩ u − max∂Ω u+ , and

d = diam(Ω).

Proof. Without loss of generality, we assume u ≤ 0 on ∂Ω. Set Ω+ =

{u > 0}. By the area-formula for ∇u in Γ+ ∩ Ω+ ⊂ Ω, we have
 
(1) g(p) dp ≤ g(∇u)| det(∇2 u)| dx,
∇u(Γ+ ∩Ω+ ) Γ+ ∩Ω+

where | det(∇2 u)| is the Jacobian of the map ∇u : Ω → Rn . In fact, we may

consider χε = ∇u−εId : Ω → Rn , where Id is the identity map in Rn . Then,
∇χε = ∇2 u − εI, which is negative definite in Γ+ and hence injective. By a
24 1. Linear Elliptic Equations

change of variables, we have

 
g(p) dp = g(χε )| det(∇2 u − εI)| dx.
χε (Γ+ ∩Ω+ ) Γ+ ∩Ω+

This implies (1) if we let ε → 0.

We assume M = supΩ u > 0. Now, we claim
(2) BM/d ⊂ ∇u(Γ+ ∩ Ω+ ).
In other words, for any a ∈ Rn with |a| < M/d, there exists an x ∈ Γ+ ∩ Ω+
such that a = ∇u(x).
We assume u attains its maximum M at 0 ∈ Ω; i.e.,
u(0) = M.
For any a ∈ Rn with |a| < M/d, consider
ua (x) = u(x) − a · x.
First, we have ua (0) = M and, for any x ∈ ∂Ω,
ua (x) ≤ −a · x < M.
Hence, ua attains its maximum at some xa ∈ Ω, with ua (xa ) ≥ M . Then,
u(xa ) = ua (xa ) + a · xa ≥ M + a · xa > 0,
and, for any x ∈ Ω,
u(x) = ua (x) + a · x ≤ ua (xa ) + a · x = u(xa ) + a · (x − xa ).
Hence, xa ∈ Γ+ ∩ Ω+ and a = ∇u(xa ).
By combining (1) and (2), we have
 
g(p) dp ≤ g(∇u)| det(∇2 u)| dx.
BM/d Γ+ ∩Ω+

This is the desired result. 

In fact, the proof above can be interpreted geometrically. For any a ∈

Rn \ {0} with |a| < M/d, we consider the affine function l(x) = M − a · x;
the graph of l has the slope a and the value M at the origin. Note that l
assumes positive values on the boundary by the choice of a. By moving the
graph of l upward vertically, we assume it touches the graph of u at xa for
the last time. In other words, if we move the graph of l further upward,
then it will never touch the graph of u. Then, xa ∈ Γ+ .
We note that the integral domain Γ+ in Lemma 1.2.1 can be replaced
by  
Γ ∩ x ∈ Ω : u(x) > max u .
+ +
∂Ω
This is clear from the proof.
1.2. Krylov-Safonov’s Harnack Inequality 25

A special case of Lemma 1.2.1 is given by the following result.

Corollary 1.2.2. Let Ω be a bounded domain in Rn . Then, for any u ∈
C(Ω̄) ∩ C 2 (Ω),
 1
d n
sup u ≤ max u + √
+
| det ∇ u| dx
2
,
Ω ∂Ω n ωn Γ+

where Γ+ is the upper contact set of u, d = diam(Ω), and ωn is the volume

of the unit ball B1 in Rn .

Proof. The desired result follows from Lemma 1.2.1 by taking g = 1. 

In Corollary 1.2.2, we estimated the sup-norm of a function. We now

improve such an estimate for values at any point in terms of its distance to
the boundary. For simplicity, we consider only concave functions in convex
domains.
Lemma 1.2.3. Let Ω be a bounded convex domain in Rn and u be a concave
C(Ω̄) ∩ C 2 (Ω)-function with u = 0 on ∂Ω. Then, for any x0 ∈ Ω,

n
u(x0 ) ≤ C(diam(Ω))n−1 dist(x0 , ∂Ω) | det ∇2 u| dx,
Ω
where C is a positive constant depending only on n.

Proof. By the concavity, we have u ≥ 0 in Ω. We assume u > 0 in Ω. By

taking g = 1 in (1) in the proof of Lemma 1.2.1, we have

(1) |∇u(Ω)| ≤ | det(∇2 u)| dx.
Ω
For any given x0 ∈ Ω, let v be the concave function whose graph is the cone
with vertex (x0 , u(x0 )) and base ∂Ω, with v = 0 on ∂Ω. In fact, v can be
defined in the following way. For any x  ∈ ∂Ω and t ∈ [0, 1], define
v tx0 + (1 − t)
x = tu(x0 ).
In particular, v(x0 ) = u(x0 ). By the concavity of u and u = v = 0 on ∂Ω,
we have v ≤ u in Ω. Set
S = {p ∈ Rn : v(x) ≤ v(x0 ) + p · (x − x0 ) in Ω}.
In other words, S is the collection of slopes of the supporting planes of the
graph of v at (x0 , v(x0 )). As in the proof of Lemma 1.2.1, we can prove
(2) S ⊂ ∇u(Ω).
Set D = diam(Ω) and d = dist(x0 , ∂Ω). We claim
 
v(x0 ) n−1 v(x0 )
(3) · ≤ C|S|,
D d
26 1. Linear Elliptic Equations

where C is a positive constant depending only on n. Then, the desired result

follows from (1), (2), and (3).
We now prove (3). First, by the definition of v, we have, for any p ∈
Bv(x0 )/D and any x ∈ Ω,
v(x) ≤ v(x0 ) + p · (x − x0 ).
Hence,
(4) Bv(x0 )/D ⊂ S.
Next, we note that
(5) there exists a p0 ∈ S such that |p0 | = v(x0 )/d.
To see this, we take a x  ∈ ∂Ω with | x − x0 | = d and a supporting plane
H of Ω at x  in Rn . The existence of such an H follows from the convexity
of Ω. The hyperplane in Rn+1 generated by H and the point (x0 , v(x0 )) is
a supporting plane of the graph of v with the slope p0 . This verifies (5).
Last, by the definition of S, we note that S is a convex set. By combining
(4), (5), and |p0 | ≥ v(x0 )/D, we conclude that S contains the convex hull
of Bv(x0 )/D and p0 . A geometric argument yields
 
v(x0 ) n−1
C · |p0 | ≤ |S|.
D
This proves (3). 

Next, we relate det ∇2 u in Lemma 1.2.1 to subsolutions of linear differ-

ential equations.
Lemma 1.2.4. Let Ω be a bounded domain in Rn and g ∈ L1loc (Rn ) be
nonnegative. Assume aij ∈ C(Ω), with aij = aji , such that aij (x) is

positive definite for any x ∈ Ω, with D ∗ = n det(aij ). Suppose that u ∈
C(Ω̄) ∩ C 2 (Ω) satisfies
aij ∂ij u ≥ f in Ω,
for some f ∈ C(Ω) with f /D ∗ ∈ Ln (Ω). Then,
−
   − n
f
g(p) dp ≤ g(∇u) ,
BM/d Γ+ nD ∗
where Γ+ is the upper contact set of u, M = supΩ u − max∂Ω u+ , and
d = diam(Ω).

Proof. Since the matrix A = (aij ) is positive definite, we have

 
−aij ∂ij u n
det(A) · det(−∇ u) ≤
2
on Γ+ .
n
Then, the desired result follows from Lemma 1.2.1. 
1.2. Krylov-Safonov’s Harnack Inequality 27

Now, we are ready to prove the maximum principle due to Alexandrov

[1], [2], which yields an estimate of the L∞ -norms of solutions in terms of
Ln -norms of nonhomogeneous terms. This is different from the maximum
principle we discussed in Section 1.1, where we estimated the L∞ -norms of
solutions in terms of L∞ -norms of nonhomogeneous terms.
Theorem 1.2.5. Let Ω be a bounded domain in Rn . Assume aij ∈ C(Ω),
with aij = aji , such that aij (x) is positive definite for any x ∈ Ω, with

D ∗ = n det(aij ). Suppose that u ∈ C(Ω̄) ∩ C 2 (Ω) satisfies
aij ∂ij u ≥ f in Ω,
for some f ∈ C(Ω) with f − /D ∗ ∈ Ln (Ω). Then,
 −
f 
sup u ≤ max u + Cd 
+ 
 D∗  n + ,
Ω ∂Ω L (Γ )

where Γ+ is the upper contact set of u, d = diam(Ω), and C is a positive

constant depending only on n.

Proof. We simply take g = 1 in Lemma 1.2.4. 

Remark 1.2.6. (i) The integral domain Γ+ in Theorem 1.2.5 can be re-
placed by  
Γ ∩ x ∈ Ω : u(x) > max u .
+ +
∂Ω

(ii) The uniform ellipticity is not needed in Theorem 1.2.5. This is

important in applications to certain nonlinear elliptic equations, including
the mean curvature equation and the Monge-Ampère equation.

1.2.2. The Harnack Inequality. In this subsection, we derive the Har-

nack inequality for positive solutions of uniformly elliptic differential equa-
tions and discuss its corollaries. We will focus on a priori estimates instead
of the regularity. Our assumptions are more than what we need.
We let Ω be a bounded domain in Rn and let aij be continuous functions
in Ω, with aij = aji . We consider the linear differential operator L given by
(1.2.1) Lu = aij ∂ij u in Ω,
for any u ∈ C 2 (Ω). The operator L is always assumed to be uniformly
elliptic in Ω; namely, for any x ∈ Ω and ξ ∈ Rn ,
(1.2.2) λ|ξ|2 ≤ aij (x)ξi ξj ≤ Λ|ξ|2 ,
for some positive constants λ and Λ, which are usually called the ellipticity
constants. This means that all eigenvalues of the matrix aij (x) are between
√
λ and Λ. We also set D = det(aij ) and D ∗ = n D. Hence, D ∗ is the
28 1. Linear Elliptic Equations

geometric mean of the eigenvalues of (aij ). The uniform ellipticity (1.2.2)

implies
(1.2.3) λ ≤ D ∗ ≤ Λ.

First, we derive an upper bound for subsolutions.

Theorem 1.2.7. Let BR be a ball in Rn and L be given by (1.2.1) in BR , for
some aij ∈ C(BR ) satisfying (1.2.2). Suppose that u ∈ L∞ (BR ) ∩ C 2 (BR )
satisfies Lu ≥ f in BR , for some f ∈ Ln (BR )∩C(BR ). Then, for any p > 0,
 n 
−
sup u ≤ C R p u+ Lp (BR ) + R f − Ln (BR ) ,
BR/2

where C is a positive constant depending only on n, p, λ, and Λ.

Proof. Without loss of generality, we consider R = 1. We first introduce a

cutoff function. For any constant β ≥ 1, we define
η(x) = (1 − |x|2 )β in B1 .
By simple differentiations, we obtain
∂i η(x) = −2βxi (1 − |x|2 )β−1
and
∂ij η(x) = −2βδij (1 − |x|2 )β−1 + 4β(β − 1)xi xj (1 − |x|2 )β−2 .
Hence,
1− β1
|∇η(x)| = 2β|x|(1 − |x|2 )β−1 ≤ 2βη ,
and
Lη = 4β(β − 1)aij xi xj (1 − |x|2 )β−2 − 2βaij δij (1 − |x|2 )β−1 .
Setting w = ηu, we have
Lw = ηLu + uLη + 2aij ∂i η∂j u.
Consider the upper contact set Γ+ of w in B1 . We clearly have w > 0 in Γ+
since w = 0 on ∂B1 . Hence, for any x ∈ Γ+ ,
w(x) 2w(x)
|∇w(x)| ≤ ≤ ,
1 − |x| 1 − |x|2
which implies
 
1 1 2w −1− β1
|∇u(x)| = |∇w − u∇η| ≤ + u|∇η| ≤ 2(1 + β)η w.
η η 1 − |x|2
Then in Γ+ , we have
− β2
|aij ∂i η∂j u| ≤ Λ|∇u||∇η| ≤ 4β(1 + β)Λη w.
1.2. Krylov-Safonov’s Harnack Inequality 29

Note that we also have u > 0 in Γ+ . Hence,

uLη = −2βuaij δij (1 − |x|2 )β−1 + 4β(β − 1)uaij xi xj (1 − |x|2 )β−2
≥ −2βuaij δij (1 − |x|2 )β−1
1− β1 − β1
≥ −2βnΛuη = −2nΛβwη
− β2
≥ −2nΛβwη .
Therefore, we obtain
− β2 − β2
Lw ≥ −Cη w − ηf − ≥ − Cη w + f− on Γ+ ,
where C is a positive constant depending only on n, β, λ, and Λ. For β ≥ 2,
by applying Theorem 1.2.5, we obtain
 
−2
sup w+ ≤ C η β w+ Ln (B1 ) + f − Ln (B1 ) .
B1

1− 2 2
Now writing w+ = (w+ ) β (w+ ) β , we have
 1− 2 
β 2
sup w ≤ C
+
sup w +
(u+ ) β Ln (B1 ) + f− Ln (B1 ) .
B1 B1

By choosing β > 2 and applying Young’s inequality, we get

 
2 β
−
sup w ≤ C (u ) Ln (B1 ) + f Ln (B1 )
+ + β 2
B1
 
+ −
=C u 2n + f Ln (B1 ) .
L β (B )
1

For β > 2, we have 2n/β ∈ (0, n). Hence, for any p ∈ (0, n),

sup w+ ≤ C u+ Lp (B1 ) + f − Ln (B1 ) .
B1

We apply the Hölder inequality to get the desired result for p ≥ n. 

By replacing u with −u, Theorem 1.2.7 extends to supersolutions and

solutions of the equation Lu = f.
Corollary 1.2.8. Let BR be a ball in Rn and L be given by (1.2.1) in BR ,
for some aij ∈ C(BR ) satisfying (1.2.2). Suppose that u is an L∞ (BR ) ∩
C 2 (BR )-solution of
Lu = f in BR ,
for some f ∈ Ln (BR ) ∩ C(BR ). Then, for any p > 0,
 n 
−
|u|L∞ (BR/2 ) ≤ C R p u Lp (BR ) + R f Ln (BR ) ,

where C is a positive constant depending only on n, p, λ, and Λ.

30 1. Linear Elliptic Equations

Now, we turn our attention to supersolutions. We first introduce the

Calderon-Zygmund decomposition. For any x ∈ Rn and r > 0, we define
 r 
Qr (x) = y ∈ Rn : |yi − xi | ≤ for any i = 1, . . . , n .
2
For x = 0, we simply write Qr instead.
Take the unit cube Q1 . Cut it equally into 2n cubes, which we take as the
first generation. Do the same cutting for these small cubes to get the second
generation. Continue this process. These cubes (from all generations) are
called dyadic cubes. Any cube Q in the (k + 1)th-generation arises from
some cube Q  in the kth-generation, which is called the predecessor of Q.

Lemma 1.2.9. Suppose that f is a nonnegative L1 (Q1 )-function and that

δ is a constant such that

1
f dx < δ.
|Q1 | Q1
Then, there exists a sequence of (nonoverlapping) dyadic cubes {Qj } in Q1
such that 
f ≤ δ a.e. in Q1 \ Qj
j
and 
1
δ≤ f dx < 2n δ.
|Qj | Qj

Proof. Cut Q1 into 2n dyadic cubes and keep the cube Q if


1
(1) f dx ≥ δ.
|Q| Q
Continue cutting for others, and always keep the cube Q if (1) is satisfied
and cut the rest. Let {Qj } be the cubes we have kept during this infinite
process. We need only verify

(2) f ≤ δ a.e. in Q1 \ Qj .
j

For any x ∈ Q1 \ j Qj , from the way we collect {Qj }, there exists a sequence

of cubes Qj containing x such that

1
f dx < δ
|Qj  | Qj 
and

diam(Qj ) → 0 as j  → ∞.
Then, (2) follows from the Lebesgue density theorem. 

Next, we prove a consequence of the Calderon-Zygmund decomposition,

which will be needed in the proof of the weak Harnack inequality.
1.2. Krylov-Safonov’s Harnack Inequality 31

Lemma 1.2.10. Suppose that A and B are measurable sets such that A ⊂
B ⊂ Q1 and
(1) |A| < δ for some δ ∈ (0, 1);
 ⊂ B for the
(2) for any dyadic cube Q, |A ∩ Q| ≥ δ|Q| implies Q
 of Q.
predecessor Q
Then,
|A| ≤ δ|B|.

Proof. We apply Lemma 1.2.9 to f = χA . By the assumption (1), we

obtain a sequence of dyadic cubes {Qj } such that

A⊂ Qj except for a set of measure zero,
j

|A ∩ Qj |
δ≤ < 2n δ,
|Qj |
 j of Qj ,
and, for any predecessor Q
j |
|A ∩ Q
< δ.
j |
|Q
 j ⊂ B for each j. Hence,
By the assumption (2), we have Q

A⊂  j ⊂ B except for a set of measure zero.
Q
j

 j } so that they are nonoverlapping. Then, we get

We relabel {Q
 
|A| ≤ |A ∩ Qj | ≤ δ  j | ≤ δ|B|.
|Q
j j

This yields the desired estimate. 

Now we are ready to prove the weak Harnack inequality for nonnegative
supersolutions.
Theorem 1.2.11. Let BR be a ball in Rn and L be given by (1.2.1) in BR ,
for some aij ∈ C(BR ) satisfying (1.2.2). Suppose that u ∈ L∞ (BR )∩C 2 (BR )
satisfies u ≥ 0 and Lu ≤ f in BR , for some f ∈ Ln (BR ) ∩ C(BR ). Then,
 
−n
R p u Lp (B2τ R ) ≤ C inf u + R f Ln (BR ) ,
Bτ R
√
where τ = (8 n)−1 and p and C are positive constants depending only on
n, λ, and Λ.
32 1. Linear Elliptic Equations

The proof consists of several steps. In the first step, we prove that if
a positive supersolution is small somewhere in Q3 , then it has an upper
bound in a good portion of Q1 . This step is the key ingredient in the weak
Harnack inequality. In the second step, we iterate to get a power decay of
distribution functions for positive supersolutions.
√
Proof. We consider the case R = 2 n and prove
 
u Lp (B1/2 ) ≤ C inf u + f Ln (B2√n ) .
B1/4

The proof consists of several steps.

Step 1. We prove that there exist constants ε0 > 0, μ ∈ (0, 1), and
M > 1, depending only on n, λ, and Λ, such that if
(1) inf u ≤ 1, f Ln (B2√n ) ≤ ε0 ,
Q3

then
|{u ≤ M } ∩ Q1 | > μ.
The basic idea of the proof is to construct a function ϕ, which is very concave
outside Q1 , such that if we correct u by ϕ, the lower contact set of u + ϕ
occurs in Q1 and occupies a large portion of Q1 . In other words, we localize
where the contact occurs by choosing suitable functions.
Note that B1/4 ⊂ B1/2 ⊂ Q1 ⊂ Q3 ⊂ B2√n . For some large constant
β > 0 to be determined and some M > 0, define
 β
|x|2
ϕ(x) = −M 1 − in B2√n .
4n
Then, ϕ = 0 on ∂B2√n . We choose M , according to β, such that
(2) ϕ ≤ −2 in Q3 .

Set
w = u + ϕ in B2√n .
We will prove, by choosing β large,
(3) Lw ≤ f + η in B2√n ,
for some η ∈ C0∞ (Q1 ) and 0 ≤ η ≤ C, for a positive constant C depending
only on n, λ, and Λ. To do this, we calculate the Hessian matrix of ϕ. A
straightforward calculation yields
 β−1
M |x|2
∂ij ϕ(x) = βδij 1 −
2n 4n
 β−2
M |x|2
− β(β − 1)xi xj 1 − ,
(2n)2 4n
1.2. Krylov-Safonov’s Harnack Inequality 33

and hence
 β−2   
M |x|2 |x|2 1
Lϕ = β 1− 1− aij δij − (β − 1)aij xi xj
2n 4n 4n 2n
 β−2   
M |x|2 |x|2 1
≤ β 1− 1− nΛ − (β − 1)λ|x| .
2
2n 4n 4n 2n
Therefore for |x| ≥ 1/4, we have Lϕ ≤ 0 if we choose β large, depending
only on n, λ, and Λ. Hence,
Lw ≤ f in B2√n \ Q1 .
This finishes the proof of (3).
Now we apply Theorem 1.2.5 to −w in B2√n . Note that inf Q3 w ≤ −1
by (1) and (2) and w ≥ 0 on ∂B2√n . In view of Remark 1.2.6(i), we obtain
1 ≤ C |f | + η Ln (B2√n ∩Γ− ∩Ω− )

+ C|Γ− ∩ Ω− ∩ Q1 | n ,
1
≤C f Ln (B2√n )

where Γ− is the lower contact set of w and Ω− = {x ∈ B2√n : w(x) < 0}.
Choosing ε0 small enough, we get
1  1
≤ C|Γ− ∩ Ω− ∩ Q1 | n ≤ C {u ≤ M } ∩ Q1  n
1

2
since w(x) ≤ 0 implies u(x) ≤ −ϕ(x) ≤ M in Ω− .
Step 2. We prove that there exist positive constants ε0 , γ, and C,
depending only on n, λ, and Λ, such that if (1) holds, then, for any t > 0,
|{u > t} ∩ Q1 | ≤ Ct−γ .
In fact, under the assumption (1), we will prove, for any k = 1, 2, . . .,
 
 
(4) {u > M k } ∩ Q1  ≤ (1 − μ)k ,
where M and μ are as in Step 1.
For k = 1, (4) is implied by Step 1. Now suppose (4) holds for k − 1, for
some k ≥ 2. Set
A = {u > M k } ∩ Q1 , B = {u > M k−1 } ∩ Q1 .
Clearly, A ⊂ B ⊂ Q1 and |A| ≤ |{u > M } ∩ Q1 | ≤ 1 − μ by Step 1. We
claim that if Qr (x0 ) is a cube in Q1 , with r ∈ (0, 1/2), such that
(5) |A ∩ Qr (x0 )| > (1 − μ)|Qr (x0 )|,
then Q3r (x0 ) ∩ Q1 ⊂ B. Assuming the claim, we are in a position to apply
Lemma 1.2.10 to get
|A| ≤ (1 − μ)|B|.
Then, (4) follows.
34 1. Linear Elliptic Equations

To prove the claim, we consider the transform

x = x0 + ry for y ∈ Q1 and x ∈ Qr (x0 ).
Set
r2

aij (y) = aij (x), f(y) = f (x),
M k−1
and
1
(y) =
u u(x).
M k−1
 ≥ 0 and
Then, u
  ≤ f in B2√n .
aij ∂yi yj u
Moreover,
r
f Ln (B2√n ) ≤ f Ln (B2√n ) ≤ f Ln (B2√n ) ≤ ε0 .
M k−1
, we get
Writing (5) in terms of u
 
 
u(y) > M } ∩ Q1 | = r−n {u(x) > M k } ∩ Qr (x0 ) > 1 − μ,
|{

and hence,
u(y) ≤ M } ∩ Q1 | ≤ μ.
|{
, we have
By applying what we proved in Step 1 to u
 > 1.
inf u
Q3

Hence, u > M k−1 in Q3r (x0 ), and in particular Q3r (x0 ) ∩ Q1 ⊂ B. This
finishes the proof of the claim.
Step 3. We prove that there exist positive constants γ and C, depending
only on n, λ, and Λ, such that, for any t > 0,
 γ
−γ
|{x ∈ Q1 : u(x) > t}| ≤ Ct inf u + f Ln (B2√n ) .
Q3

To prove this, we consider, for any δ > 0,

 −1
1
uδ = inf u + δ + f Ln (B2√n ) u.
Q3 ε0
By applying Step 2 to uδ and then letting δ → 0, we obtain the desired
result. We point out that δ is introduced only for the case f ≡ 0 and
inf Q3 u = 0.
Step 4. Now we prove
 
u Lp (Q 1)
≤ C inf u + f Ln (B2√n ) ,
Q3

where p and C are positive constants depending only on n, λ, and Λ.

1.2. Krylov-Safonov’s Harnack Inequality 35

Set, for any t > 0,

A(t) = {x ∈ Q1 : u(x) > t} .
First, we have, for any p > 0 and τ > 0,
  ∞
p
u dx = p tp−1 |A(t)| dt
Q1 0
 τ  ∞
=p t |A(t)| dt + p
p−1
tp−1 |A(t)| dt = I + II.
0 τ
For I, we have easily
 τ
I ≤ p |Q1 | tp−1 dt = τ p .
0
For II, we get, by Step 3,
 ∞
II = p tp−1 |A(t)| dt
τ
 ∞  γ
≤ Cp t p−γ−1
dt inf u + f Ln (B2√n )
τ Q3
 γ
p−γ
= Cτ inf u + f Ln (B2√n )
Q3

if we choose p < γ. Combining these two estimates, we obtain

  
γ
u dx ≤ C τ + τ
p p p−γ
inf u + f Ln (B2√n ) .
Q1 Q3

Next, we choose
τ = inf u + f Ln (B2√n ) .
Q3
This implies the desired result. 

We point out that p in Theorem 1.2.11 may be less than 1.

Now, we can prove the Harnack inequality due to Krylov and Safonov
[99], [100].
Theorem 1.2.12. Let BR be a ball in Rn and L be given by (1.2.1) in BR ,
for some aij ∈ C(BR ) satisfying (1.2.2). Suppose that u is a nonnegative
L∞ (BR ) ∩ C 2 (BR )-solution of
Lu = f in BR ,
for some f ∈ Ln (BR ) ∩ C(BR ). Then,
 
sup u ≤ C inf u + R f Ln (BR ) ,
BR/2 BR/2

where C is a positive constant depending only on n, λ, and Λ.

Exploring the Variety of Random
Documents with Different Content
polished, and mounted in rings and other jewelry. A number of
specimens of this Mason County topaz are displayed in museums.

Tourmaline
Tourmaline is a complex silicate of boron and aluminum. Other
elements, such as magnesium, sodium, lithium, calcium, iron, 95
or fluorine, also may be present. This mineral has a glassy to
resinous luster. Only the dark-colored varieties of tourmaline have
been found in Texas. One is a black variety called schorl, and
another is a brown variety called dravite. Other kinds of tourmaline,
although not found in Texas, are colorless or some shade of blue,
yellow, red, pink, or green. Some crystals even show more than one
color.

Tourmaline is too hard to scratch with a steel file, it has a specific

gravity of 3 to 3.25, and it has a conchoidal to uneven fracture. Very
little light passes through the dark varieties, and some fragments of
schorl look like shiny, black coal.

Tourmaline occurs as masses without crystal shapes, but crystals are

commonly found. The crystals are prism-shaped and have small
vertical grooves, called striations, on the prism faces. When you look
at some crystals from an end, you will see that the cross section is a
triangle with the sides bowed outward.
 Black tourmaline crystals with milky quartz from north of Llano, Llano
County, Texas.

Both the black and the brown varieties of tourmaline have been
found at several places in the Llano uplift of central Texas. One well-
known locality is at Town Mountain north of Llano in Llano County.
Here, the tourmaline occurs in milky quartz that is associated with
Precambrian granite rocks. In west Texas, in Culberson and
Hudspeth counties, black tourmaline occurs in pegmatite rocks in the
Van Horn Mountains, the Carrizo Mountains, and the Wylie
Mountains. In the Eagle Mountains of Hudspeth County, it is found in
metamorphic rocks as well as in pegmatites.
Some tourmaline formed from hot fluids containing boron that were
given off by magmas far below the earth’s surface. These fluids
traveled up through cracks and other openings in overlying rocks. As
the fluids reacted with other elements and compounds, the
tourmaline formed.

The clear, light-colored varieties of tourmaline are much admired,

and they are more widely used as gemstones than are the dark-
colored varieties. Some collectors, however, find that the dark-
colored Texas tourmalines, when cut and polished, make shiny,
attractive gemstones.

Some tourmaline is used as grinding material, but no Texas

tourmaline is produced for this purpose.

Travertine. See Calcite.

Uranium Minerals (Carnotite, Uranophane,

Pitchblende)
In 1945, the world suddenly became aware of the awesome power
of atomic energy when the element uranium was used to produce
some of the first atomic bombs. Uranium does not occur alone in
nature but is found combined with other elements in a number of
minerals.

All of the uranium minerals are radioactive. The uranium they

contain is gradually breaking down and changing into a series of 13
other elements, called daughter elements. Each daughter element
breaks down and changes into the next daughter element of the
series. While breaking down, these elements give off particles and
rays of energy.
This energy or radioactivity is made up of what are called alpha
particles, beta particles, and gamma rays. You cannot see, hear,
taste, smell, or feel them. The alpha and beta particles are 96
weak and do not travel far. The gamma rays, however, can
travel farther and can pass through seemingly solid material.
Scientists have found that these rays can move through about 1 foot
of rock, 2½ feet of water, and several hundred feet of air.

Prospectors searching for uranium minerals carry instruments that

are able to detect this radioactivity. The uranium itself gives off only
alpha particles, but some of its daughter elements give off gamma
rays. These daughter elements are normally found with the uranium,
and it is their strong gamma rays that the instruments are most apt
to detect.
 A Geiger counter is used to detect radioactivity.

One of the instruments used is the Geiger counter. It indicates

radioactivity by means of a meter, a flashing light, or a clicking
sound, which can be heard through earphones. Another instrument
for detecting radioactivity is the scintillation counter. It is more
sensitive than the Geiger counter and it can detect radioactivity from
a greater distance. The scintillation counter can be used from an
automobile or an airplane, but the Geiger counter must be quite
close to the source of radioactivity to be of use. 97

Various uranium minerals have been found, mostly in small amounts,

in a number of places in Texas. Some of these minerals, such as
uraninite or pitchblende, are heavy and dark colored. Others,
including carnotite, tyuyamunite, autunite, and uranophane, are a
shade of yellow or green. They are quite soft. Deposits of the light-
colored uranium minerals have been mined from two areas of Texas.
One of these areas is in Garza County on the Texas High Plains, and
the other is in Karnes and Live Oak counties in the Gulf Coastal
Plain.

One of the light-colored uranium minerals, carnotite, is a potassium-

uranium vanadate, which has a bright canary-yellow or lemon-yellow
color. This mineral is transparent to translucent and has an earthy or
a pearly luster. Carnotite usually is found as crusts and as powdery
masses. It is quite soft and can be scratched with a fingernail.

Carnotite, along with tyuyamunite, autunite, and several other soft,

yellowish or greenish uranium minerals, is found in the Texas Gulf
Coastal Plain. These minerals occur in the Jackson, Catahoula, and
Oakville strata (which are Tertiary in age) in an area extending from
Gonzales County to the Rio Grande (in parts of the area indicated by
no. 2 and no. 3 on the geologic map, pp. 4-5). The largest deposits
in this district have been found in the Karnes County area.

The Gulf Coastal Plain uranium minerals occur mostly with

sandstones and clays in a sequence of strata that contains volcanic
ash. It is believed that small scattered amounts of uranium
compounds that were present in the volcanic ash sediments were
dissolved by seeping underground water. These waters then moved
into the sandstones and clays where they deposited the uranium as
carnotite and as other uranium minerals.
Another uranium mineral, uranophane (calcium-uranium silicate),
also occurs in Texas. Uranophane has a yellow to yellow-orange
color and a pearly to greasy luster. When rubbed across a streak
plate, it leaves a light yellow to a light yellow-orange streak. It is soft
enough to be scratched by a copper penny. Uranophane has been
found in extrusive igneous rocks in northwestern Presidio County in
west Texas.

A dark-colored uranium mineral, pitchblende, is a variety of the

mineral uraninite, uranium dioxide. Pitchblende does not occur with
a crystal shape but rather as rounded and irregular-shaped masses.
It is brownish black, greenish black, or black. If you rub it across a
streak plate, pitchblende leaves a brownish-black streak. This
mineral is heavy (it has a specific gravity of 6.5 to 8.5) and hard (a
pocket knife will not scratch it, although a steel file will). Pitchblende
has a submetallic luster and looks dull, greasy, or like pitch or tar.

Small amounts of pitchblende have been found at several places in

Texas. One of these localities is a few miles west of Burnet in Burnet
County in central Texas. Here, the pitchblende occurs in Precambrian
igneous rocks that are associated with gneiss. In south Texas, some
fine, scattered particles of pitchblende have been found about 325
feet below the surface in Tertiary (Pliocene) sediments that cover the
Palangana salt dome in Duval County. No pitchblende is mined in
Texas.

Uranophane. See Uranium Minerals.

Vitrophyre. See Obsidian and Vitrophyre.

Volcanic Ash (Pumicite)

Volcanic ash deposits, which also are known as pumicite, are loose
and powdery. They are made up mostly of material that is thrown
into the air when volcanoes erupt. If a volcano erupts with a violent
explosion, the nearby rocks are blown into powder. Molten lava also
is hurled into the air, where some of it immediately cools to become
tiny bubbles and particles of glass. The winds may carry some of this
fine material far away before depositing it.

Deposits of volcanic ash are white, bluish, greenish, yellowish, 98

or grayish, and some of them glisten like snow in the sunlight.
They feel rough and gritty. When examined under a microscope, this
material shows the tiny curved and sharp-cornered particles of the
broken volcanic glass. Deposits of volcanic ash may also contain clay,
silt, sand, or other impurities.

Volcanoes, which may have been located in the Davis Mountains and
in other areas of west Texas and in northern Mexico, erupted during
Tertiary time. The volcanic ash that we find at the surface today in
some of the Tertiary formations in Texas could have come from
these volcanoes. Tertiary volcanic ash deposits occur in the Texas
Gulf Coastal Plain (such as in Brazos, Fayette, Karnes, Polk, Starr,
Trinity, and other counties) and in the Trans-Pecos country of west
Texas.

Volcanic ash deposits of Quaternary (Pleistocene) age, which are

less than a million years old, are found in a number of counties on
the Texas High Plains. Farther to the east, ash deposits occur in
Baylor, Dickens, Kent, and Wilbarger counties. This volcanic ash may
have come from a volcano that erupted in northern New Mexico
during Quaternary time.

Volcanic ash or pumicite has several commercial uses. Some is used

to make pozzolan cement, and some is used in sweeping
compounds, cleansing and scouring powders, and abrasive soaps.
Pumicite has been mined in Dickens, Scurry, Starr, and several other
counties of Texas.
Wad. See Manganese Minerals.

Wood Opal. See Opal.

99
 COMPOSITION, HARDNESS, AND SPECIFIC
GRAVITY OF SOME TEXAS MINERALS

For convenient reference, the Texas minerals described in this book

are listed below, together with their chemical compositions, specific
gravities, and hardness. You will be able to find similar information
about additional minerals in mineralogy textbooks such as those
noted on page 24.

Mineral Composition Specific Hardness

Gravity
Albite NaAlSi₃O₈ 2.62 6
Almandite Fe₃Al₂ (SiO₄)₂ 4.2 7
Amphibole Ca₂Mg₅Si₈O₂₂(OH)₂ 3.0-3.3 1-2½
asbestos
Anhydrite CaSO₄ 2.9 3-3½
Argentite Ag₂S 7.3 2-2½
Azurite Cu₃(CO₃)₂(OH)₂ 3.77 3½-4
Barite BaSO₄ 4.5 3-3½
Biotite K(Mg, Fe)₃AlSi₃O₁₀(OH)₂ 2.8-3.2 2½-3
Braunite 3MnMnO₃MnSiO₃ 4.75-4.82 6-6½
Calcite CaCO₃ 2.72 3
Carnotite K₂O·2UO₃·V₂O₅·nH₂O 5.03 2
Cassiterite SnO₂ 6.8-7.1 6-7
Celestite SrSO₄ 3.95-3.97 3-3½
Cerargyrite AgCl 5.5 1-1½
Chalcocite Cu₂S 5.5-5.8 2½-3
Chalcopyrite CuFeS₂ 4.1-4.3 3½-4
Cinnabar HgS 8.10 2½
Dolomite CaMg(CO₃)₂ 2.85 3½-4
Feldspar (see Albite, Microcline, Orthoclase)
Fluorite CaF₂ 3.18 4
Galena PbS 7.4-7.6 2½
Garnet (see Almandite, Grossularite)
Gold Au 15.0-19.3 2½-3
Graphite C 2.2 1-2
Grossularite Ca₃Al₂(SiO₄)₃ 3.53 6½
Gypsum CaSO₄·2H₂O 2.32 2
Halite NaCl 2.16 2½
Hematite Fe₂O₃ 5.26 1-6½
Hollandite MnBaMn₁₆O₁₄ 4.7-5 4-6
Limonite FeO(OH)·nH₂O 3.6-4.0 1-5½
Magnetite Fe₃O₄ 5.18 6
Malachite Cu₂CO₃(OH)₂ 3.9-4.03 3½-4
Mica (see Muscovite, Biotite)
Microcline KAlSi₃O₈ 2.54-2.57 6
Muscovite KAl₃Si₃O₁₀(OH)₂ 2.76-3.1 2-2½
Opal SiO₂·nH₂O 1.9-2.2 5-6
Orthoclase KAlSi₃O₈ 2.57 6
Pitchblende UO₂ 6.5-8.5 5½
Pyrite FeS₂ 5.02 6-6½
Pyrolusite MnO₂ 4.75 1-2
Quartz SiO₂ 2.65 7
Serpentine Mg₃Si₂O₅(OH)₄ 2.48 3-4
Silver Ag 10.5 2½-3
Sulfur S 2.05-2.09 1½-2½
Talc Mg₃Si₄O₁₀(OH)₂ 2.7-2.8 1
Topaz Al₂SiO₄(F,OH)₂ 3.4-3.6 8
Tourmaline Complex silicate of boron and 3.0-3.25 7-7½
aluminum
Uranophane CaO·2UO₃·2SiO₂·7H₂O 3.8-3.9 2-3
100
 BOOKS ABOUT ROCKS AND MINERALS

Many books have been written about rocks and minerals. Some are
listed below, and it is likely that your librarian will be able to suggest
others.

Nontechnical Books for Beginners

Getting Acquainted With Minerals, by George L. English and David E.
Jensen. McGraw-Hill Book Company, Inc., New York, N. Y.
(second edition, 1958).

The Rock Book, by Carroll L. Fenton and Mildred A. Fenton.

Doubleday & Company, Inc., Garden City, N. Y. (1940).

Mineral Collector’s Guide, by David E. Jensen. Ward’s Natural Science

Establishment, Inc., Rochester, N. Y. (1953).

My Hobby is Collecting Rocks and Minerals, by David E. Jensen. Hart

Book Company, New York, N. Y. (1955).

Rocks and Minerals, by Richard M. Pearl. Barnes & Noble, New York,
N. Y. (1956).

1001 Questions Answered About the Mineral Kingdom, by Richard M.

Pearl. Dodd, Mead & Company, New York, N. Y. (1959).
Rocks and Minerals, by Herbert S. Zim and Paul R. Schaffer. Simon
and Schuster, Inc., New York, N.Y. (1957).

Textbooks and Other Reference Books

Economic Mineral Deposits, by Alan M. Bateman. John Wiley & Sons,
Inc., New York, N. Y. (second edition, 1950).

A Textbook of Mineralogy, by Edward S. Dana, revised by William E.

Ford. John Wiley & Sons, Inc., New York, N. Y. (fourth edition,
1932).

Industrial Minerals and Rocks (Nonmetallics Other Than Fuels),

Joseph L. Gillson, Editor-in-Chief. The American Institute of
Mining, Metallurgical, and Petroleum Engineers, New York, N. Y.
(third edition, 1960).

Dana’s Manual of Mineralogy, revised by Cornelius S. Hurlbut, Jr.

John Wiley & Sons, Inc., New York, N. Y. (seventeenth edition,
1959).

Mineralogy, by Edward H. Kraus, Walter F. Hunt, and Lewis S.

Ramsdell. McGraw-Hill Book Company, Inc., New York, N. Y.
(fifth edition, 1959).

Nonmetallic Minerals, by Raymond B. Ladoo and W. M. Meyers.

McGraw-Hill Book Company, Inc., New York, N. Y. (second
edition, 1951).

Rocks and Rock Minerals, by Louis V. Pirsson, revised by Adolph

Knopf. John Wiley and Sons, Inc., New York, N. Y. (third edition,
1947).

A Field Guide to Rocks and Minerals, by Frederick H. Pough.

Houghton Mifflin Company, Boston, Mass. (third edition, 1960).
Mineral Facts and Problems, by the Staff of the Bureau of Mines. U.
S. Bureau of Mines Bulletin 585. U. S. Government Printing
Office, Washington, D. C. (1960).

Selected References on Texas Rocks and

Minerals
Entries marked with an asterisk are published by the Bureau
of Economic Geology, The University of Texas, Austin. Those
not out of print are distributed at nominal sale price, and a list
of publications will be sent on request. These publications can
be consulted at many public libraries and Chamber of
Commerce offices.

*Report on the Pavitte Silver-Copper Prospect in Burnet County,

Texas, by V. E. Barnes. Univ. Texas, Bureau Econ. Geol. Mineral
Resource Survey Circ. 5 (1936).

*Report on the Sheridan Copper Prospect in Burnet County, Texas,

by V. E. Barnes. Univ. Texas, Bur. Econ. Geol. Mineral Resource
Survey Circ. 9 (1936).

*Building Stones of Central Texas, by V. E. Barnes, R. F. Dawson,

and G. A. Parkinson. Univ. Texas Pub. 4246 (1947).

*Iron Ore in the Llano Region, Central Texas, by V. E. Barnes. Univ.

Texas, Bur. Econ. Geol. Rept. Inves. No. 5 (1949).

*Utilization of Texas Serpentine, by V. E. Barnes, D. A. Shock, and W.

A. Cunningham. Univ. Texas Pub. 5020 (1950).

*Lead Deposits in the Upper Cambrian of Central Texas, by V. E.

Barnes. Univ. Texas, Bureau Econ. Geol. Rept. Inves. No. 26
(1956).
*Mineral Resources of the Colorado River Industrial Development
Association Area, by J. W. Dietrich and J. T. Lonsdale. Univ.
Texas, Bureau Econ. Geol. Rept. Inves. No. 37 (1958).

*Some Uranium Occurrences in West Texas, by D. H. Eargle. Univ.

Texas, Bureau Econ. Geol. Rept. Inves. No. 27 (1956).

*A Preliminary Report on the Stratigraphy of the Uranium-Bearing

Rocks of the Karnes County Area, South-Central Texas, by D. H.
Eargle and J. L. Snider. Univ. Texas, Bureau Econ. Geol. Rept.
Inves. No. 30 (1957).

The Brown Iron Ores of Eastern Texas, by E. B. Eckel. U. S. Geol.

Survey Bull. 902 (1938).

*The Rustler Springs Sulphur Deposits as a Source of Fertilizer, by G.

L. Evans. Univ. Texas, Bureau Econ. Geol. Rept. Inves. No. 1
(1946).

Origin of the Gulf Coast Salt-Dome Sulphur Deposits, by Herbert W.

Feely and J. Lawrence Kulp. Bull. Amer. Assoc. Petrol. Geol., vol.
41, pp. 1802-1853 (1957).

*Pegmatites of the Van Horn Mountains, Texas, by P. T. Flawn. Univ.

Texas, Bureau Econ. Geol. Rept. Inves. No. 9 (1951).

*The Hazel Copper-Silver Mine, Culberson County, Texas, by P. T.

Flawn. Univ. Texas, Bureau Econ. Geol. Rept. Inves. No. 16
(1952).

*Basement Rocks of Texas and Southeast New Mexico, by P. T.

Flawn. Univ. Texas Pub. 5605 (1956).

*Texas Miners Boost Talc Output, by P. T. Flawn. Univ. Texas, 101

Bureau Econ. Geol. Rept. Inves. No. 35 (1958).
*Geology and Mineral Deposits of Pre-Cambrian Rocks of the Van
Horn Area, Texas, by P. B. King and P. T. Flawn. Univ. Texas Pub.
5301 (1953).

*Igneous Rocks of the Balcones Fault Region of Texas, by J. T.

Lonsdale. Univ. Texas Bull. 2744 (1927).

Mineral Resources of the Llano-Burnet Region, Texas, with an

Account of the Pre-Cambrian Geology, by Sidney Paige. U. S.
Geol. Survey Bull. 450 (1911).

*Mineral Resources of the Texas Coastal Plain (Preliminary Report),

by J. M. Perkins and J. T. Lonsdale. Univ. Texas, Bureau Econ.
Geol. Mineral Resource Circ. 38 (1955).

Geology and Ore Deposits of the Shafter Mining District, Presidio

County, Texas, by C. P. Ross. U. S. Geol. Survey Bull. 928-B
(1943).

*The Geology of Texas, Vol. II, Structural and Economic Geology, by

E. H. Sellards, C. L. Baker, and others. Univ. Texas Bull. 3401
(1935).

*Texas Mineral Resources, by E. H. Sellards and others. Univ. Texas

Pub. 4301 (1946).

*Geological Resources of the Trinity River Tributary Area in Texas

and Oklahoma, by H. B. Stenzel, A. E. Weissenborn, and others.
Univ. Texas Pub. 4824 (1948).

Uranium at Palangana Salt Dome, Duval County, Texas, by A. D.

Weeks and D. H. Eargle. In U. S. Geol. Survey Prof. Paper 400-B
(1960).

Geology of the Quicksilver Deposits of the Terlingua District, Texas,

by R. G. Yates and G. A. Thompson. U. S. Geol. Survey Prof.
Paper 312 (1959).
 102

GLOSSARY

Amorphous—without crystalline structure and therefore without

regular form.

Balcones fault zone—a system of faults extending from north of

Waco in McLennan County, through Travis and Bexar counties,
to near Del Rio in Val Verde County (see p. 42).

Boulder—a large rock or mineral fragment that has a diameter

greater than 256 millimeters (about 10 inches).

Breccia—a rock made up of sharp-cornered, cemented fragments

with diameters greater than 2 millimeters (about ⁸/₁₀₀ of an
inch).

Cambrian—the earliest period of the Paleozoic Era (see p. 3).

Cenozoic—the present era, one of the great divisions of geologic

time (see p. 3). This era began about 63 million years ago.

Clastic—made up of broken fragments of rocks or minerals.

Cleavage—occurs when minerals split along smooth flat surfaces

that are parallel to possible crystal faces. These planes as well
as crystal faces are controlled by the crystal lattice or atomic
structures of the minerals.
Cleavage fragment—a mineral specimen that has been broken
along its planes of cleavage.

Cobble—a rock or mineral fragment that has a diameter between

64 and 256 millimeters (about 2½ and 10 inches).

Conchoidal—a curved fracture surface shaped like the inside of a

shell or spoon.

Conglomerate—a rock composed of cemented, rounded rock or

mineral fragments, most of which are of gravel size.

Cretaceous—the third and latest period of the Mesozoic Era (see p.

3).

Cryptocrystalline—made up of tiny crystalline particles that are

too small to be distinguished even under high magnification.

Crystalline—having a definite, orderly internal structure.

Cube—a solid that has six equal, square sides.

Dodecahedron—a solid that has twelve plane, four-sided faces.

Element—a basic building block of all matter, which cannot be

separated into different substances by ordinary chemical means.

Eocene—the second epoch of the Tertiary Period (see p. 3).

Epoch—a unit of geologic time that is a subdivision of a period.

Era—a major division of geologic time, which consists of several

periods.

Extrusive rocks—igneous rocks formed from magma that was

extruded on the earth’s surface.
Fault—a break in the rocks or strata of the earth’s crust along which
movement or slippage has taken place.

Fluid—a substance made up of particles that can move freely about;

it can be a liquid or a gas.

Formation—rocks or strata that are recognized and mapped as a

unit.

Fracture—the kind of surface obtained if a mineral is broken in a

different direction from that of the cleavage or parting.
Commonly, fracture surfaces are rough, uneven, or curved,
whereas cleavage surfaces are smooth.

Geologic map (areal)—shows the extent and distribution of

formations exposed at the earth’s surface.

Granular—the texture of a rock or mineral that is made up of

visible grains. If all the grains are about the same size, the term
equigranular is used.

Granule—a rock or mineral fragment that has a diameter of from 2

to 4 millimeters (about ⁸/₁₀₀ to ¹⁵/₁₀₀ of an inch).

Gravel—uncemented rock or mineral fragments that have diameters

greater than 2 millimeters (about ⁸/₁₀₀ of an inch).

Gulf Coastal Plain—an area that extends, in Texas, from the Gulf
of Mexico to the Balcones fault zone and in which Quaternary,
Tertiary, and Upper Cretaceous strata crop out at the surface
(see p. 42).

High Plains—an area in northwest Texas extending from the Pecos

River valley north to the Oklahoma-Texas boundary (see p. 42).

Igneous rocks—rocks formed by the cooling and hardening of hot,

molten rock material.
Intrusive rocks—igneous rocks that have formed below the
surface of the earth.

Lava—molten rock material that has poured out onto the earth’s
surface from volcanoes; also the rock that is formed after the
molten material has cooled and hardened.

Llano uplift—an area in central Texas where Precambrian and early

Paleozoic rocks occur at the earth’s surface (see p. 42).

Magma—hot, molten rock material from which igneous rocks form.

Massive—in a mass, without a regular or complete form.

Mesozoic—an era, one of the great divisions of geologic time (see

p. 3). This era began about 230 million years ago and lasted
until about 63 million years ago.

Metamorphic rock—rock formed from igneous or sedimentary

rocks that are altered by heat, pressure, and fluids below the
earth’s surface.

Miocene—the fourth epoch of the Tertiary Period (see p. 3).

Mississippian—the fifth period of the Paleozoic Era (see p. 3).

Nodule—a small, rounded mass or lump.

Octahedron—a solid that has eight triangular faces.

Opaque—no light can pass through.

Ordovician—the second period of the Paleozoic Era (see p. 103

3).

Paleozoic—an era, one of the great divisions of geologic time (see

p. 3). This era began at the end of Precambrian time and lasted
until about 230 million years ago.