A Beginners Guide to

the Theory of Viscosity Solutions

Shigeaki Koike

August 27, 2010

2nd edition
26 August 2010

Department of Mathematics, Saitama University, 255 Shimo-Okubo, Sakura, Saitama,

338-8570 JAPAN
i
Preface for the 2nd edition
The rst version of this text has been out of print since 2009. It seems that
there is no hope to publish the second version from the publisher. Therefore,
I have decided to put the les of the second version in my Home Page.
Although I corrected many errors in the rst version, there must be some
mistakes in this version. I would be glad if the reader would kindly inform
me errors and typos etc.
Acknowledgment for the 2nd edition
I would like to thank Professors H. Ishii, K. Ishii, T. Nagasawa, M. Ohta,
T. Ohtsuka, and graduate students, T. Imai, K. Kohsaka, H. Mitake, S.
Nakagawa, T. Nozokido for pointing out numerous errors in the First edition.
26 August 2010 Shigeaki Koike
ii
Preface
This book was originally written in Japanese for undergraduate students
in the Department of Mathematics of Saitama University. In fact, the rst
hand-written draft was prepared for a series of lectures on the viscosity so-
lution theory for undergraduate students in Ehime University and Hokkaido
University.
The aim here is to present a brief introduction to the theory of viscosity
solutions for students who have knowledge on Advanced Calculus (i.e. dier-
entiation and integration on functions of several-variables) and hopefully, a
little on Lebesgue Integration and Functional Analysis. Since this is written
for undergraduate students who are not necessarily excellent, I try to give
easy proofs throughout this book. Thus, if you do not feel any diculty
to read Users guide [6], you should try to read that one.
I also try not only to show the viscosity solution theory but also to men-
tion some related classical results.
Our plan of this book is as follows: We begin with our motivation in
section 1. Section 2 introduces the denition of viscosity solutions and their
properties. In section 3, we rst show classical comparison principles and
then, extend them to viscosity solutions of rst- and second-order PDEs,
separately. We establish two kinds of existence results via Perrons method
and representation formulas for Bellman and Isaacs equations in section 4.
We discuss boundary value problems for viscosity solutions in sections 5.
Section 6 is a short introduction to the L
p
-viscosity solution theory, on which
we have an excellent book [4].
In Appendix, which is the hardest part, we give proofs of fundamental
propositions.
In order to learn more on viscosity solutions, I give a list of books:
A popular survey paper [6] by Crandall-Ishii-Lions on the theory of viscos-
ity solutions of second-order, degenerate elliptic PDEs is still a good choice
for undergraduate students to learn rst. However, to my experience, it
seems a bit hard for average undergraduate students to understand.
Bardi-Capuzzo Dolcettas book [1] contains lots of information on viscos-
ity solutions for rst-order PDEs (Hamilton-Jacobi equations) while Fleming-
Soners [10] complements topics on second-order (degenerate) elliptic PDEs
with applications in stochastic control problems.
Barles book [2] is also nice to learn his original techniques and French
language simultaneously !
iii
It has been informed that Ishii would write a book [15] in Japanese on
viscosity solutions in the near future, which must be more advanced than
this.
For an important application via the viscosity solution theory, we refer
to Gigas [12] on curvature ow equations. Also, I recommend the reader to
consult Lecture Notes [3] (Bardi-Crandall-Evans-Soner-Souganidis) not only
for various applications but also for a friendly introduction by Crandall,
who rst introduced the notion of viscosity solutions with P.-L. Lions in early
80s.
If the reader is interested in section 6, I recommend him/her to attack
Caarelli-Cabres [4].
As a general PDE theory, although there are so many books on PDEs, I
only refer to my favorite ones; Gilbarg-Trudingers [13] and Evans [8]. Also
as a textbook for undergraduate students, Han-Lins short lecture notes [14]
is a good choice.
Since this is a text-book, we do not refer the reader to original papers
unless those are not mentioned in the books in our references.
Acknowledgment
First of all, I would like to express my thanks to Professors H. Morimoto
and Y. Tonegawa for giving me the opportunity to have a series of lectures in
their universities. I would also like to thank Professors K. Ishii, T. Nagasawa,
and a graduate student, K. Nakagawa, for their suggestions on the rst draft.
I wish to express my gratitude to Professor H. Ishii for having given me
enormous supply of ideas since 1980.
I also wish to thank the reviewer for several important suggestions.
My nal thanks go to Professor T. Ozawa for recommending me to publish
this manuscript. He kindly suggested me to change the original Japanese title
(A secret club on viscosity solutions).
iv
Contents
1 Introduction 1
1.1 From classical solutions to weak solutions 2
1.2 Typical examples of weak solutions 4
1.2.1 Burgers equation 4
1.2.2 Hamilton-Jacobi equation 6
2 Denition 9
2.1 Vanishing viscosity method 9
2.2 Equivalent denitions 17
3 Comparison principle 23
3.1 Classical comparison principle 24
3.1.1 Degenerate elliptic PDEs 24
3.1.2 Uniformly elliptic PDEs 24
3.2 Comparison principle for rst-order PDEs 26
3.3 Extension to second-order PDEs 31
3.3.1 Degenerate elliptic PDEs 33
3.3.2 Remarks on the structure condition 35
3.3.3 Uniformly elliptic PDEs 38
4 Existence results 40
4.1 Perrons method 40
4.2 Representation formulas 45
4.2.1 Bellman equation 46
4.2.2 Isaacs equation 51
4.3 Stability 58
5 Generalized boundary value problems 62
5.1 Dirichlet problem 65
5.2 State constraint problem 68
5.3 Neumann problem 72
5.4 Growth condition at [x[ 75
v
6 L
p
-viscosity solutions 78
6.1 A brief history 78
6.2 Denition and basic facts 81
6.3 Harnack inequality 84
6.3.1 Linear growth 85
6.3.2 Quadratic growth 87
6.4 Holder continuity estimates 89
6.5 Existence result 92
7 Appendix 95
7.1 Proof of Ishiis lemma 95
7.2 Proof of the ABP maximum principle 100
7.3 Proof of existence results for Pucci equations 105
7.4 Proof of the weak Harnack inequlity 108
7.5 Proof of the local maximum principle 113
References 118
Notation Index 120
Index 121
vi
1 Introduction
Throughout this book, we will work in (except in sections 4.2 and 5.4),
where
R
n
is open and bounded.
We denote by , ) the standard inner product in R
n
, and set [x[ =
_
x, x) for x R
n
. We use the standard notion of open balls: For r > 0
and x R
n
,
B
r
(x) := y R
n
[ [x y[ < r, and B
r
:= B
r
(0).
For a function u : R, we denote its gradient and Hessian matrix at
x , respectively, by
Du(x) :=
_
_
_
_
_
u(x)
x
1
.
.
.
u(x)
xn
_
_
_
_
_
,
D
2
u(x) :=
_
_
_
_
_
_
_
_
_
_
_
_

2
u(x)
x
2
1
j-th

2
u(x)
x
1
xn
.
.
.
.
.
.
.
.
.
i-th

2
u(x)
x
i
x
j

.
.
.
.
.
.
.
.
.
.
.
.

2
u(x)
xnx
1


2
u(x)
x
2
n
_
_
_
_
_
_
_
_
_
_
_
_
.
Also, S
n
denotes the set of all real-valued n n symmetric matrices. Note
that if u C
2
(), then D
2
u(x) S
n
for x .
We recall the standard ordering in S
n
:
X Y X, ) Y , ) for R
n
.
We will also use the following notion in sections 6 and 7: For =
t
(
1
, . . . ,
n
), =
t
(
1
, . . . ,
n
) R
n
, we denote by the n n matrix
whose (i, j)-entry is
i

j
for 1 i, j n;
=
_
_
_
_
_
_
_
_
_
_

1
j-th
1

n
.
.
.
.
.
.
.
.
.
i-th
i

j

.
.
.
.
.
.
.
.
.
.
.
.

1

n

n
_
_
_
_
_
_
_
_
_
_
.
1
We are concerned with general second-order partial dierential equations
(PDEs for short):
F(x, u(x), Du(x), D
2
u(x)) = 0 in . (1.1)
We suppose (except in several sections) that
F : RR
n
S
n
R is continuous
with respect to all variables.
1.1 From classical solutions to weak solutions
As the rst example of PDEs, we present the Laplace equation:
u = 0 in . (1.2)
Here, we dene u := trace(D
2
u). In the literature of the viscosity solution
theory, we prefer to have the minus sign in front of .
Of course, since we do not require any boundary condition yet, all poly-
nomials of degree one are solutions of (1.2). In many textbooks (particularly
those for engineers), under certain boundary condition, we learn how to solve
(1.2) when has some special shapes such as cubes, balls, the half-space or
the whole space R
n
. Here, solve means that we nd an explicit formula of
u using elementary functions such as polynomials, trigonometric ones, etc.
However, the study on (1.2) in such special domains is not applicable
because, for instance, solutions of equation (1.2) represent the density of a
gas in a bottle, which is neither a ball nor a cube.
Unfortunately, in general domains, it seems impossible to nd formulas for
solutions u with elementary functions. Moreover, in order to cover problems
arising in physics, engineering and nance, we will have to study more general
and complicated PDEs than (1.2). Thus, we have to deal with general PDEs
(1.1) in general domains.
If we give up having formulas for solutions of (1.1), how do we investigate
PDEs (1.1) ? In other words, what is the right question in the study of PDEs
?
In the literature of the PDE theory, the most basic questions are as fol-
lows:
2
(1) Existence: Does there exist a solution ?
(2) Uniqueness: Is it the only solution ?
(3) Stability: If the PDE changes a little,
does the solution change a little ?
The importance of the existence of solutions is trivial since, otherwise,
the study on the PDE could be useless.
To explain the signicance of the uniqueness of solutions, let us remem-
ber the reason why we study the PDE. Usually, we discuss PDEs or their
solutions to understand some specic phenomena in nature, engineerings or
economics etc. Particularly, people working in applications want to know
how the solution looks like, moves, behaves etc. For this purpose, it might
be powerful to use numerical computations. However, numerical analysis
only shows us an approximate shapes, movements, etc. Thus, if there
are more than one solution, we do not know which is approximated by the
numerical solution.
Also, if the stability of solutions fails, we could not predict what will hap-
pen from the numerical experiments even though the uniqueness of solutions
holds true.
Now, let us come back to the most essential question:
What is the solution of a PDE ?
For example, it is natural to call a function u : R a solution of
(1.1) if there exist the rst and second derivatives, Du(x) and D
2
u(x), for
all x , and (1.1) is satised at each x when we plug them in the left
hand side of (1.1). Such a function u will be called a classical solution of
(1.1).
However, unfortunately, it is dicult to seek for a classical solution be-
cause we have to verify that it is suciently dierentiable and that it satises
the equality (1.1) simultaneously.
Instead of nding a classical solution directly, we have decided to choose
the following strategy:
(A) Find a candidate of the classical solution,
(B) Check the dierentiability of the candidate.
In the standard books, the candidate of a classical solution is called a
weak solution; if the weak solution has the rst and second derivatives, then
3
it becomes a classical solution. In the literature, showing the dierentiability
of solutions is called the study on the regularity of those.
Thus, with these terminologies, we may rewrite the above with mathe-
matical terms:
(A) Existence of weak solutions,
(B) Regularity of weak solutions.
However, when we cannot expect classical solutions of a PDE to exist,
what is the right candidate of solutions ?
We will call a function the candidate of solutions of a PDE if it is a
unique and stable weak solution under a suitable setting. In section 2,
we will dene such a candidate named viscosity solutions for a large class
of PDEs, and in the proceeding sections, we will extend the denition to
more general (possibly discontinuous) functions and PDEs.
In the next subsection, we show a brief history on weak solutions to
remind what was known before the birth of viscosity solutions.
1.2 Typical examples of weak solutions
In this subsection, we give two typical examples of PDEs to derive two kinds
of weak solutions which are unique and stable.
1.2.1 Burgers equation
We consider Burgers equation, which is a model PDE in Fluid Mechanics:
u
t
+
1
2
(u
2
)
x
= 0 in R(0, ) (1.3)
under the initial condition:
u(x, 0) = g(x) for x R, (1.4)
where g is a given function.
In general, we cannot nd classical solutions of (1.3)-(1.4) even if g is
smooth enough. See [8] for instance.
In order to look for the appropriate notion of weak solutions, we rst
introduce a function space C
1
0
(R[0, )) as a test function space:
C
1
0
(R[0, )) :=
_
C
1
(R[0, ))

there is K > 0 such that

supp [K, K] [0, K]
_
.
4
Here and later, we denote by supp the following set:
supp := (x, t) R[0, ) [ (x, t) ,= 0.
Suppose that u satises (1.3). Multiplying (1.3) by C
1
0
(R [0, ))
and then, using integration by parts, we have
_
R
_

0
_
u

t
+
u
2
2

x
_
(x, t)dtdx +
_
R
u(x, 0)(x, 0)dx = 0.
Since there are no derivatives of u in the above, this equality makes sense if
u
K>0
L
1
((K, K)(0, K)). Hence, we may adapt the following property
as the denition of weak solutions of (1.3)-(1.4).
_

_
_
R
_

0
_
u

t
+
u
2
2

x
_
(x, t)dtdx +
_
R
g(x)(x, 0)dx = 0
for all C
1
0
(R[0, )).
We often call this a weak solution in the distribution sense. As you no-
ticed, we derive this notion by an essential use of integration by parts. We
say that a PDE is in divergence form when we can adapt the notion of
weak solutions in the distribution sense. When the PDE is not in divergence
form, we say that it is in nondivergence form.
We note that the solution of (1.3) may have singularities even though the
initial value g belongs to C

by an observation via characteristic method.

From the denition of weak solutions, we can derive the so-called Rankine-
Hugoniot condition on the set of singularities.
On the other hand, unfortunately, we cannot show the uniqueness of weak
solutions of (1.3)-(1.4) in general while we know the famous Lax-Oleinik
formula (see [8] for instance), which is the expected solution.
In order to obtain the uniqueness of weak solutions, for the denition, we
add the following property (called entropy condition) which holds for the
expected solution given by the Lax-Oleinik formula: There is C > 0 such
that
u(x + z, t) u(x, t)
Cz
t
for all (x, t, z) R(0, ) (0, ). We call u an entropy solution of (1.3)
if it is a weak solution satisfying this inequality. It is also known that such
a weak solution has a certain stability property.
5
We note that this entropy solution satises the above mentioned impor-
tant properties; existence, uniqueness and stability. Thus, it must be a
right denition for weak solutions of (1.3)-(1.4).
1.2.2 Hamilton-Jacobi equations
Next, we shall consider general Hamilton-Jacobi equations, which arise in
Optimal Control and Classical Mechanics:
u
t
+ H(Du) = 0 in (x, t) R
n
(0, ) (1.5)
under the same initial condition (1.4).
In this example, we suppose that H : R
n
R is convex, i.e.
H(p + (1 )q) H(p) + (1 )H(q) (1.6)
for all p, q R
n
, [0, 1].
Remark. Since a convex function is locally Lipschitz continuous in general,
we do not need to assume the continuity of H.
Example. In Classical Mechanics, we often call this H a Hamiltonian.
As a simple example of H, we have H(p) = [p[
2
.
Notice that we cannot adapt the weak solution in the distribution sense
for (1.5) since we cannot use the integration by parts.
We next introduce the Lagrangian L : R
n
R dened by
L(q) = sup
pR
n
p, q) H(p).
When H(p) = [p[
2
, it is easy to verify that the maximum is attained in the
right hand side of the above.
It is surprising that we have a neat formula for the expected solution
(called Hopf-Lax formula) presented by
u(x, t) = min
yR
n
_
tL
_
x y
t
_
+ g(y)
_
. (1.7)
More precisely, it is shown that the right hand side of (1.7) is dierentiable
and satises (1.5) almost everywhere.
6
Thus, we could call u a weak solution of (1.5)-(1.4) when u satises (1.5)
almost everywhere. However, if we decide to use this notion as a weak solu-
tion, the uniqueness of those fails in general. We will see an example in the
next section.
As was shown for Burgers equation, in order to say that the unique
weak solution is given by (1.7), we have to add one more property for the
denition of weak solutions: There is C > 0 such that
u(x + z, t) 2u(x, t) + u(x z, t) C[z[
2
(1.8)
for all x, z R, t > 0. This is called the semi-concavity of u.
We note that (1.8) is a hypothesis on the one-sided bound of second
derivatives of functions u.
In 60s, Kruzkov showed that the limit function of approximate solutions
by the vanishing viscosity method (see the next section) has this property
(1.8) when H is convex. He named u a generalized solution of (1.5) when
it satises (1.5) almost everywhere and (1.8).
To my knowledge, between Kruzkovs works and the birth of viscosity
solutions, there had been no big progress in the study of rst-order PDEs in
nondivergence form.
Remark. The convexity (1.6) is a natural hypothesis when we consider
only optimal control problems where one person intends to minimize some
costs (energy in terms of Physics). However, when we treat game prob-
lems (one person wants to minimize costs while the other tries to maximize
them), we meet non-convex and non-concave (i.e. fully nonlinear)
Hamiltonians. See section 4.2.
In this book, since we are concerned with viscosity solutions of PDEs in
nondivergence form, for which the integration by parts argument cannot be
used to dene the notion of weak solutions in the distribution sense, we shall
give typical examples of such PDEs.
Example. (Bellman and Isaacs equations)
We rst give Bellman equations and Isaacs equations, which arise in
(stochastic) optimal control problems and dierential games, respectively.
As will be seen, those are extensions of linear PDEs.
Let A and B be sets of parameters. For instance, we suppose A and B
are (compact) subsets in R
m
(for some m 1). For a A, b B, x ,
7
r R, p =
t
(p
1
, . . . , p
n
) R
n
, and X = (X
ij
) S
n
, we set
L
a
(x, r, p, X) := trace(A(x, a)X) +g(x, a), p) + c(x, a)r,
L
a,b
(x, r, p, X) := trace(A(x, a, b)X) +g(x, a, b), p) + c(x, a, b)r.
Here A(, a), A(, a, b), g(, a), g(, a, b), c(, a) and c(, a, b) are given functions
for (a, b) AB.
For inhomogeneous terms, we consider functions f(, a) and f(, a, b) in
for a A and b B.
We call the following PDEs Bellman equations:
sup
aA
L
a
(x, u(x), Du(x), D
2
u(x)) f(x, a) = 0 for x . (1.9)
Notice that the supremum over A is taken at each point x .
Taking account of one more parameter set B, we call the following PDEs
Isaacs equations:
sup
aA
inf
bB
L
a,b
(x, u(x), Du(x), D
2
u(x)) f(x, a, b) = 0 for x (1.10)
and
inf
bB
sup
aA
L
a,b
(x, u(x), Du(x), D
2
u(x)) f(x, a, b) = 0 for x . (1.10

)
Example. (Quasi-linear equations)
We say that a PDE is quasi-linear if the coecients of D
2
u contains u
or Du. Although we will not study quasilinear PDEs in this book, we give
some of those which are in nondivergence form.
We rst give the PDE of mean curvature type:
F(x, p, X) :=
_
[p[
2
trace(X) Xp, p)
_
.
Notice that this F is independent of x-variables. We refer to [12] for appli-
cations where this kind of operators appears.
Next, we show a relatively new one called L

-Laplacian:
F(x, p, X) := Xp, p).
Again, this F does not contain x-variables. We refer to Jensens work [16],
where he rst studied the PDE D
2
uDu, Du) = 0 in via the viscosity
solution approach.
8
2 Denition
In this section, we derive the denition of viscosity solutions of (1.1) via the
vanishing viscosity method.
We also give some basic properties of viscosity solutions and equivalent
denitions using semi-jets.
2.1 Vanishing viscosity method
When the notion of viscosity solutions was born, in order to explain the
reason why we need it, many speakers started in their talks by giving the
following typical example called the eikonal equation:
[Du[
2
= 1 in . (2.1)
We seek C
1
functions satisfying (2.1) under the Dirichlet condition:
u(x) = 0 for x . (2.2)
However, since there is no classical solution of (2.1)-(2.2) (showing the non-
existence of classical solutions is a good exercise), we intend to derive a
reasonable denition of weak solutions of (2.1).
In fact, we expect that the following function (the distance from )
would be the unique solution of this problem (see Fig 2.1):
u(x) = dist(x, ) := inf
y
[x y[.
Fig 2.1
1
1
1 0
y
y = u(x)
x
9
If we consider the case when n = 1 and = (1, 1), then the expected
solution is given by
u(x) = 1 [x[ for x [1, 1]. (2.3)
Since this function is C

except at x = 0, we could decide to call u a weak

solution of (2.1) if it satises (2.1) in except at nite points.
Fig 2.2
1
1
1 0
x
y
However, even in the above simple case of (2.1), we know that there are
innitely many such weak solutions of (2.1) (see Fig 2.2); for example, u is
the weak solution and
u(x) =
_

_
x + 1 for x [1,
1
2
),
x for x [
1
2
,
1
2
),
x 1 for x [
1
2
, 1],
. . . etc.
Now, in order to look for an appropriate notion of weak solutions, we
introduce the so-called vanishing viscosity method; for > 0, we consider
the following PDE as an approximate equation of (2.1) when n = 1 and
= (1, 1):
_
u

+ (u

)
2
= 1 in (1, 1),
u

(1) = 0.
(2.4)
The rst term, u

, in the left hand side of (2.4) is called the vanishing

viscosity term (when n = 1) as 0.
By an elementary calculation, we can nd a unique smooth function u

in the following manner: We rst note that if a classical solution of (2.4)

10
exists, then it is unique. Thus, we may suppose that u

(0) = 0 by symmetry.
Setting v

= u

, we rst solve the ODE:

_
v

+ v
2

= 1 in (1, 1),
v

(0) = 0.
(2.5)
It is easy to see that the solution of (2.5) is given by
v

(x) = tanh
_
x

_
.
Hence, we can nd u

by
u

(x) = log
_
_
cosh
_
x

_
cosh
_
1

_
_
_
= log
_
e
x

+ e

e
1

+ e

_
.
It is a good exercise to show that u

converges to the function in (2.3)

uniformly in [1, 1].
Remark. Since u

(x) := u

(x) is the solution of

_
u

+ (u

)
2
= 1 in (1, 1),
u(1) = 0,
we have u(x) := lim
0
u

(x) = u(x). Thus, if we replace u

by +u

,
then the limit function would be dierent in general.
To dene weak solutions, we adapt the properties which hold for the
(uniform) limit of approximate solutions of PDEs with the minus vanishing
viscosity term.
Let us come back to general second-order PDEs:
F(x, u, Du, D
2
u) = 0 in . (2.6)
We shall use the following denition of classical solutions:
Denition. We call u : R a classical subsolution (resp.,
supersolution, solution) of (2.6) if u C
2
() and
F(x, u(x), Du(x), D
2
u(x)) 0 (resp., 0, = 0) in .
11
Remark. If F does not depend on X-variables (i.e. F(x, u, Du) = 0; rst-
order PDEs), we only suppose u C
1
() in the above in place of u C
2
().
Throughout this text, we also suppose the following monotonicity condi-
tion with respect to X-variables:
Denition. We say that F is (degenerate) elliptic if
_
F(x, r, p, X) F(x, r, p, Y )
for all x , r R, p R
n
, X, Y S
n
provided X Y.
(2.7)
We notice that if F does not depend on X-variables (i.e. F = 0 is the
rst-order PDE), then F is automatically elliptic.
We also note that the left hand side F(x, r, p, X) = trace(X) of the
Laplace equation (1.2) is elliptic.
We will derive properties which hold true for the (uniform) limit (as
+0) of solutions of
u + F(x, u, Du, D
2
u) = 0 in ( > 0). (2.8)
Note that since trace(X) + F(x, r, p, X) is uniformly elliptic (see in
section 3 for the denition) provided that F is elliptic, it is easier to solve
(2.8) than (2.6) in practice. See [13] for instance.
Proposition 2.1. Assume that F is elliptic. Let u

C
2
() C()
be a classical subsolution (resp., supersolution) of (2.8). If u

converges to
u C() (as 0) uniformly in any compact sets K , then, for any
C
2
(), we have
F(x, u(x), D(x), D
2
(x)) 0 (resp., 0)
provided that u attains its maximum (resp., minimum) at x .
Remark. When F does not depend on X-variables, we only need to sup-
pose and u

to be in C
1
() as before.
Proof. We only give a proof of the assertion for subsolutions since the
other one can be shown in a symmetric way.
12
Suppose that u attains its maximum at x for C
2
(). Setting

(y) := (y) + [y x[
4
for small > 0, we see that
(u

)( x) > (u

)(y) for y x.
(This tiny technique to replace a maximum point by a strict one will appear
in Proposition 2.2.)
Let x

be a point such that (u

)(x

) = max

(u

). Note
that x

also depends on > 0.

Since u

converges to u uniformly in B
r
( x) and x is the unique maximum
point of u

, we note that lim

0
x

= x. Thus, we see that x

for
small > 0. Notice that if we argue by instead of

, the limit of x

might
dier from x.
Thus, at x

, we have
u

(x

) + F(x

, u

(x

), Du

(x

), D
2
u

(x

)) 0.
Since D(u

)(x

) = 0 and D
2
(u

)(x

) 0, in view of ellipticity, we
have

(x

) + F(x

, u

(x

), D

(x

), D
2

(x

)) 0.
Sending 0 in the above, we have
F( x, u( x), D

( x), D
2

( x)) 0.
Since D

( x) = D( x) and D
2

( x) = D
2
( x), we conclude the proof. 2
Denition. We call u : R a viscosity subsolution (resp.,
supersolution) of (2.6) if, for any C
2
(),
F(x, u(x), D(x), D
2
(x)) 0 (resp., 0)
provided that u attains its maximum (resp., minimum) at x .
We call u : R a viscosity solution of (2.6) if it is both a viscosity
sub- and supersolution of (2.6).
Remark. Here, we have given the denition to general functions but
we will often suppose that they are (semi-)continuous in Theorems etc.
In fact, in our propositions in sections 2.1, we will suppose that viscosity
sub- and supersolutions are continuous.
13
However, all the proposition in section 2.1 can be proved by replacing up-
per and lower semi-continuity for viscosity subsolutions and supersolutions,
respectively.
We will introduce general viscosity solutions in section 3.3.
Notation. In order to memorize the correct inequality, we will often
say that u is a viscosity subsolution (resp., supersolution) of
F(x, u, Du, D
2
u) 0 (resp., 0) in
if it is a viscosity subsolution (resp., supersolution) of (2.6).
Proposition 2.2. For u : R, the following (1) and (2) are equiva-
lent:
_

_
(1) u is a viscosity subsolution (resp., supersolution) of (2.6),
(2) if 0 = (u )( x) > (u )(x) (resp., < (u )(x))
for C
2
(), x and x x,
then F( x, ( x), D( x), D
2
( x)) 0 (resp., 0).
Fig 2.3
graph of
graph of

graph of () + (u )( x)
x
x
graph of u
Proof. The implication (1) (2) is trivial.
For the opposite implication in the subsolution case, suppose that u
attains a maximum at x . Set

(x) = (x) + [x x[
4
+ (u )( x).
14
See Fig 2.3. Since 0 = (u

)( x) > (u

)(x) for x x, (2) gives

F( x,

( x), D

( x), D
2

( x)) 0,
which implies the assertion. 2
By the next proposition, we recognize that viscosity solutions are right
candidates of weak solutions when F is elliptic.
Proposition 2.3. Assume that F is elliptic. A function u : R
is a classical subsolution (resp., supersolution) of (2.6) if and only if it is a
viscosity subsolution (resp., supersolution) of (2.6) and u C
2
().
Proof. Suppose that u is a viscosity subsolution of (2.6) and u C
2
().
Taking u, we see that u attains its maximum at any points x .
Thus, the denition of viscosity subsolutions yields
F(x, u(x), Du(x), D
2
u(x)) 0 for x .
On the contrary, suppose that u C
2
() is a classical subsolution of
(2.6).
Fix any C
2
(). Assuming that u takes its maximum at x ,
we have
D(u )(x) = 0 and D
2
(u )(x) 0.
Hence, in view of ellipticity, we have
0 F(x, u(x), Du(x), D
2
u(x)) F(x, u(x), D(x), D
2
(x)). 2
We introduce the sets of upper and lower semi-continuous functions: For
K R
n
,
USC(K) := u : K R [ u is upper semi-continuous in K,
and
LSC(K) := u : K R [ u is lower semi-continuous in K.
Remark. Throughout this book, we use the following maximum principle
for semi-continuous functions:
15
An upper semi-continuous function in a compact set attains its maximum.
We give the following lemma which will be used without mentioning it.
Since the proof is a bit technical, the reader may skip it over rst.
Proposition 2.4. Assume that u USC() (resp., u LSC()) is a viscos-
ity subsolution (resp., supersolution) of (2.6) in . Assume also that inf
K
u >
(resp., sup
K
u < ) for any compact sets K .
Then, for any open set

, u is a viscosity subsolution (resp., supersolution)

of (2.6) in

.
Proof. We only show the assertion for subsolutions since the other can be shown
similarly.
For C
2
(

), by Proposition 2.2, we suppose that for some x

,
0 = (u )( x) > (u )(y) for all y

x.
Choose r > 0 such that B
3r
( x)

, and set s := inf

B
2r
( x)
(u ) > 0.
We recall the standard mollier

(R
n
) and supp

: First, we set
(x) :=
_
C
0
e
1
|x|
2
1
for [x[ < 1,
0 for [x[ 1,
where C
0
:=
__
B
1
e
1
|x|
2
1
dx
_
1
.
Next, we dene

by

(x) :=
1

_
x

_
.
Notice that
_
R
n

dx = 1 for > 0.
Now, setting

(x) :=

(x) =
_
R
n (y)

(x y)dy C

(R
n
), where
(x) :=
_
_
_
(x) for x B
r
( x),
2s + sup

u for x B
r
( x),
we easily verify that for 0 < < r,
max
\B
2r
( x)
(u

) 2s.
On the other hand, since we have
(u

)( x) =
_
B( x)

( x y)(( x) (y))dy,
16
there is
0
> 0 such that
(u

)( x) s for (0,
0
).
Therefore, we can choose x

B
2r
( x) such that max

(u

) = (u

)(x

).
Thus, taking a subsequence of x

>0
denoted by x

again (if necessary),

we nd z B
2r
( x) such that lim
0
x

= z. Hence, sending 0, we have

max
B
2r
( x)
(u ) = (u )(z), which gives z = x
By the denition, we have
F(x

, u(x

), D

(x

), D
2

(x

)) 0.
Note that lim
0
u(x

) = u( x). Therefore, sending 0 in the above, we conclude

the proof. 2
2.2 Equivalent denitions
We present equivalent denitions of viscosity solutions. However, since we
will need those in the proof of uniqueness for second-order PDEs,
the reader may postpone this subsection until section 3.3.
First, we introduce semi-jets of functions u : R at x by
J
2,+
u(x) :=
_

_
(p, X) R
n
S
n

u(y) u(x) +p, y x)

+
1
2
X(y x), y x)
+o([y x[
2
) as y x
_

_
and
J
2,
u(x) :=
_

_
(p, X) R
n
S
n

u(y) u(x) +p, y x)

+
1
2
X(y x), y x)
+o([y x[
2
) as y x
_

_
.
Note that J
2,
u(x) = J
2,+
(u)(x).
Remark. We do not impose any continuity for u in these denitions.
We recall the notion of small order o in the above: For k 1,
f(x) o([x[
k
) (resp., o([x[
k
)) as x 0

_
there is C([0, ), [0, )) such that (0) = 0, and
sup
xBr\{0}
f(x)
[x[
k
(r)
_
resp., inf
xBr\{0}
f(x)
[x[
k
([x[)
_
17
In the next proposition, we give some basic properties of semi-jets: (1)
is a relation between semi-jets and classical derivatives, and (2) means that
semi-jets are dened in dense sets of .
Proposition 2.5. For u : R, we have the following:
(1) If J
2,+
u(x) J
2,
u(x) ,= , then Du(x) and D
2
u(x) exist and,
J
2,+
u(x) J
2,
u(x) = (Du(x), D
2
u(x)).
(2) If u USC() (resp., u LSC()), then
=
_
x

x
k
such that J
2,+
u(x
k
) ,= , lim
k
x
k
= x
_
_
resp., =
_
x

x
k
such that J
2,
u(x
k
) ,= , lim
k
x
k
= x
__
.
Proof. The proof of (1) is a direct consequence from the denition.
We give a proof of the assertion (2) only for J
2,+
.
Fix x and choose r > 0 so that B
r
(x) . For > 0, we can choose
x

B
r
(x) such that u(x

)
1
[x

x[
2
= max
yBr(x)
(u(y)
1
[y x[
2
).
Since [x

x[
2
(max
B
r
(x)
u(x)), we see that x

converges to x B
r
(x)
as 0. Thus, we may suppose that x

B
r
(x) for small > 0. Hence, we
have
u(y) u(x

) +
1

([y x[
2
[x

x[
2
) for all y B
r
(x).
It is easy to check that (2(x

x)/, 2
1
I) J
2,+
u(x

). 2
We next introduce a sort of closure of semi-jets:
J
2,
u(x) :=
_

_
(p, X) R
n
S
n

x
k
and (p
k
, X
k
) J
2,
u(x
k
)
such that (x
k
, u(x
k
), p
k
, X
k
)
(x, u(x), p, X) as k
_

_
.
Proposition 2.6. For u : R, the following (1), (2), (3) are equiva-
lent.
_

_
(1) u is a viscosity subsolution (resp., supersolution) of (2.6).
(2) For x and (p, X) J
2,+
u(x) (resp., J
2,
u(x)),
we have F(x, u(x), p, X) 0 (resp., 0).
(3) For x and (p, X) J
2,+
u(x) (resp., J
2,
u(x)),
we have F(x, u(x), p, X) 0 (resp., 0).
18
Proof. Again, we give a proof of the assertion only for subsolutions.
Step 1: (2) = (3). For x and (p, X) J
2,+
u(x), we can nd (p
k
, X
k
)
J
2,+
u(x
k
) with x
k
such that lim
k
(x
k
, u(x
k
), p
k
, X
k
) = (x, u(x), p, X)
and
F(x
k
, u(x
k
), p
k
, X
k
) 0,
which implies (3) by sending k .
Step 2: (3) = (1). For C
2
(), suppose also (u)(x) = max(u).
Thus, the Taylor expansion of at x gives
u(y) u(x)+D(x), yx)+
1
2
D
2
(x)(yx), yx)+o([xy[
2
) as y x.
Thus, we have (D(x), D
2
(x)) J
2,+
u(x) J
2,+
u(x).
Step 3: (1) = (2). For (p, X) J
2,+
u(x) (x ), we can nd nonde-
creasing, continuous : [0, ) [0, ) such that (0) = 0 and
u(y) u(x) +p, y x) +
1
2
X(y x), y x) +[y x[
2
([y x[) (2.9)
as y x. In fact, by the denition of o, we nd
0
C([0, ), [0, )) such
that
0
(0) = 0, and

0
(r) sup
yBr(x)\{x}
1
[x y[
2
_
u(y) u(x) p, y x)
1
2
X(y x), y x)
_
,
we verify that (r) := sup
0tr

0
(t) satises (2.9).
Now, we dene by
(y) := p, y x) +
1
2
X(y x), y x) + ([x y[),
where
(t) :=
_

3t
t
__
2s
s
(r)dr
_
ds t
2
(t).
It is easy to check that
(D(x), D
2
(x)) = (p, X) and (u )(x) (u )(y) for y .
Therefore, we conclude the proof. 2
19
Remark. In view of the proof of Step 3, we verify that for x ,
J
2,+
u(x) =
_
(D(x), D
2
(x)) R
n
S
n

C
2
() such that u
attains its maximum at x
_
,
J
2,
u(x) =
_
(D(x), D
2
(x)) R
n
S
n

C
2
() such that u
attains its minimum at x
_
.
Thus, we intuitively know J
2,
u(x) from their graph.
Example. Consider the function u C([1, 1]) in (2.3). From the graph
below, we may conclude that J
2,
u(0) = , and J
2,+
u(0) = (1 [0, ))
(1 [0, )) ((1, 1) R). See Fig 2.4.1 and 2.4.2.
We omit how to obtain J
2,
u(0) of this and the next examples.
We shall examine J
2,
for discontinuous functions. For instance, consider
the Heaviside function:
u(x) :=
_
1 for x 0,
0 for x < 0.
We see that J
2,
u(0) = and J
2,+
u(0) = (0 [0, )) ((0, ) R). See
Fig 2.5.
Fig 2.4.1
y = (x)
y = x + 1
y
y = u(x)
x
In order to deal with boundary value problems in section 5, we prepare
some notations: For a set K R
n
, which is not necessarily open, we dene
20
Fig 2.4.2
y = (x)
y
y = u(x)
x
y = ax + 1 ([a[ < 1)
Fig 2.5
y = (x)
y
y = u(x)
x
1
0
y = ax + 1
(a > 0)
21
semi-jets of u : K R at x K by
J
2,+
K
u(x) :=
_

_
(p, X) R
n
S
n

u(y) u(x) +p, y x)

+
1
2
X(y x), y x)
+o([y x[
2
) as y K x
_

_
,
J
2,
K
u(x) :=
_

_
(p, X) R
n
S
n

u(y) u(x) +p, y x)

+
1
2
X(y x), y x)
+o([y x[
2
) as y K x
_

_
,
and
J
2,
K
u(x) :=
_

_
(p, X) R
n
S
n

x
k
K and (p
k
, X
k
) J
2,
K
u(x
k
)
such that (x
k
, u(x
k
), p
k
, X
k
)
(x, u(x), p, X) as k
_

_
.
Remark. It is obvious to verify that
x = J
2,

u(x) = J
2,

u(x) and J
2,

u(x) = J
2,

u(x).
For x , we shall simply write J
2,
u(x) (resp., J
2,
u(x)) for J
2,

u(x) =
J
2,

u(x) (resp, J
2,

u(x) = J
2,

u(x)).
Example. Consider u(x) 0 in K := [0, 1]. It is easy to observe that
J
2,+
u(x) = J
2,+
K
u(x) = 0 [0, ) provided x (0, 1). It is also easy to
verify that
J
2,+
K
u(0) = (0 [0, )) ((0, ) R),
and
J
2,
K
u(0) = (0 (, 0]) ((, 0) R).
We nally give some properties of J
2,

and J
2,

. Since the proof is easy,

we omit it.
Proposition 2.7. For u : R, C
2
() and x , we have
J
2,

(u + )(x) = (D(x), D
2
(x)) +J
2,

u(x)
and
J
2,

(u + )(x) = (D(x), D
2
(x)) +J
2,

u(x).
22
3 Comparison principle
In this section, we discuss the comparison principle, which implies the unique-
ness of viscosity solutions when their values on coincide (i.e. under the
Dirichlet boundary condition). In the study of the viscosity solution theory,
the comparison principle has been the main issue because the uniqueness of
viscosity solutions is harder to prove than existence and stability of them.
First, we recall some classical comparison principles and then, show
how to modify the proof to a modern viscosity version.
In this section, the comparison principle roughly means that
Comparison principle
viscosity subsolution u
viscosity supersolution v
u v on
_

_
= u v in
Modifying our proofs of comparison theorems below, we obtain a slightly
stronger assertion than the above one:
viscosity subsolution u
viscosity supersolution v
_
= max

(u v) = max

(u v)
We remark that the comparison principle implies the uniqueness of (con-
tinuous) viscosity solutions under the Dirichlet boundary condition:
Uniqueness for the Dirichlet problem
viscosity solutions u and v
u = v on
_
= u = v in
Proof of the comparison principle implies the uniqueness.
Since u (resp., v) and v (resp., u), respectively, are a viscosity subsolution
and supersolution, by u = v on , the comparison principle yields u v
(resp., v u) in . 2
In this section, we mainly deal with the following PDE instead of (2.6).
u + F(x, Du, D
2
u) = 0 in , (3.1)
where we suppose that
0, (3.2)
and
F : R
n
S
n
R is continuous. (3.3)
23
3.1 Classical comparison principle
In this subsection, we show that if one of viscosity sub- and supersolutions
is a classical one, then the comparison principle holds true. We call this the
classical comparison principle.
3.1.1 Degenerate elliptic PDEs
We rst consider the case when F is (degenerate) elliptic and > 0.
Proposition 3.1. Assume that > 0 and (3.3) hold. Assume also
that F is elliptic. Let u USC() (resp., v LSC()) be a viscosity
subsolution (resp., supersolution) of (3.1) and v LSC() C
2
() (resp.,
u USC() C
2
()) a classical supersolution (resp., subsolution) of (3.1).
If u v on , then u v in .
Proof. We only prove the assertion when u is a viscosity subsolution of
(3.1) since the other one can be shown similarly.
Set max

(u v) =: and choose x such that (u v)( x) = .

Suppose that > 0 and then, we will get a contradiction. We note that
x because u v on .
Thus, the denition of u and v respectively yields
u( x) + F( x, Dv( x), D
2
v( x)) 0 v( x) + F( x, Dv( x), D
2
v( x)).
Hence, by these inequalities, we have
= (u v)( x) 0,
which contradicts > 0. 2
3.1.2 Uniformly elliptic PDEs
Next, we present the comparison principle when = 0 but F is uniformly
elliptic in the following sense. Notice that if > 0 and F is uniformly ellip-
tic, then Proposition 3.1 yields Proposition 3.3 below because our uniform
ellipticity implies (degenerate) ellipticity.
Throughout this book, we freeze the uniform ellipticity constants:
0 < .
24
With these constants, we introduce the Puccis operators: For X S
n
,
T
+
(X) := maxtrace(AX) [ I A I for A S
n
,
T

(X) := mintrace(AX) [ I A I for A S

n
.
We give some properties of T

. We omit the proof since it is elementary.

Proposition 3.2. For X, Y S
n
, we have the following:
(1) T
+
(X) = T

(X),
(2) T

(X) = T

(X) for 0,
(3) T
+
is convex, T

is concave,
(4)
_
T

(X) +T

(Y ) T

(X + Y ) T

(X) +T
+
(Y )
T
+
(X + Y ) T
+
(X) +T
+
(Y ).
Denition. We say that F : R
n
S
n
R is uniformly elliptic
(with the uniform ellipticity constants 0 < ) if
T

(X Y ) F(x, p, X) F(x, p, Y ) T
+
(X Y )
for x , p R
n
, and X, Y S
n
.
We also suppose the following continuity on F with respect to p R
n
:
There is > 0 such that
[F(x, p, X) F(x, p

, X)[ [p p

[ (3.4)
for x , p, p

R
n
, and X S
n
.
Proposition 3.3. Assume that (3.2), (3.3) and (3.4) hold. Assume also
that F is uniformly elliptic. Let u USC() (resp., v LSC()) be a
viscosity subsolution (resp., supersolution) of (3.1) and v LSC() C
2
()
(resp., u USC() C
2
()) a classical supersolution (resp., subsolution) of
(3.1).
If u v on , then u v in .
Proof. We give a proof only when u is a viscosity subsolution and v a
classical supersolution of (3.1).
25
Suppose that max

(u v) =: > 0. Then, we will get a contradiction

again.
For > 0, we set

(x) = e
x
1
, where := max( + 1)/, + 1 > 0.
We next choose > 0 so small that
max
x
e
x
1

2
Let x be the point such that (uv +

)( x) = max

(uv +

) .
By the choice of > 0, since u v on , we see that x .
From the denition of viscosity subsolutions, we have
u( x) + F( x, D(v

)( x), D
2
(v

)( x)) 0.
By the uniform ellipticity and (3.4), we have
u( x) + F( x, Dv( x), D
2
v( x)) +T

(D
2

( x)) [D

( x)[ 0.
Noting that [D

( x)[ e
x
1
and T

(D
2

( x))
2
e
x
1
, we have
u( x) + F( x, Dv( x), D
2
v( x)) +( )e
x
1
0. (3.5)
Since v is a classical supersolution of (3.1), by (3.5) and ( + 1)/, we
have
(u v)( x) + e
x
1
0.
Hence, we have
(

( x)) e
x
1
,
which gives a contradiction because + 1. 2
3.2 Comparison principle for rst-order PDEs
In this subsection, without assuming that one of viscosity sub- and supersolu-
tions is a classical one, we establish the comparison principle when F in (3.1)
does not depend on D
2
u; rst-order PDEs. We will study the comparison
principle for second-order ones in the next subsection.
In the viscosity solution theory, Theorem 3.4 below was the rst surprising
result.
Here, instead of (3.1), we shall consider the following PDE:
u + H(x, Du) = 0 in . (3.6)
26
We shall suppose that
> 0, (3.7)
and that there is a continuous function
H
: [0, ) [0, ) such that

H
(0) = 0 and
[H(x, p) H(y, p)[
H
([x y[(1 +[p[)) for x, y and p R
n
. (3.8)
In what follows, we will call
H
in (3.8) a modulus of continuity. For
notational simplicity, we use the following notation:
/ := : [0, ) [0, ) [ () is continuous, (0) = 0.
Theorem 3.4. Assume that (3.7) and (3.8) hold. Let u USC() and
v LSC() be a viscosity sub- and supersolution of (3.6), respectively.
If u v on , then u v in .
Proof. Suppose max

(u v) =: > 0 as usual. Then, we will get a

contradiction.
Notice that since both u and v may not be dierentiable, we cannot use
the same argument as in Proposition 3.1.
Now, we present the most important idea in the theory of viscosity solu-
tions to overcome this diculty.
Setting

(x, y) := u(x) v(y) (2)

1
[x y[
2
for > 0, we choose
(x

, y

) such that

(x

, y

) = max
x,y

(x, y).
Noting that

(x

, y

) max
x

(x, x) = , we have
[x

[
2
2
u(x

) v(y

) . (3.9)
Since is compact, we can nd x, y , and
k
> 0 such that lim
k

k
= 0
and lim
k
(x

k
, y

k
) = ( x, y).
We shall simply write for
k
(i.e. in what follows, 0 means that

k
0 when k ).
Setting M := max

u min

v, by (3.9), we have
[x

[
2
2M 0 (as 0).
27
Thus, we have x = y.
Since (3.9) again implies
0 liminf
0
[x

[
2
2
limsup
0
[x

[
2
2
limsup
0
(u(x

) v(y

))
(u v)( x) 0,
we have
lim
0
[x

[
2

= 0. (3.10)
Moreover, since (u v)( x) = > 0, we have x from the assumption
u v on . Thus, for small > 0, we may suppose that (x

, y

) .
Furthermore, ignoring the left hand side in (3.9), we have
liminf
0
(u(x

) v(y

)). (3.11)
Taking (x) := v(y

) + (2)
1
[x y

[
2
, we see that u attains its
maximum at x

. Hence, from the denition of viscosity subsolutions, we

have
u(x

) + H
_
x

,
x

_
0.
On the other hand, taking (y) := u(x

) (2)
1
[y x

[
2
, we see that
v attains its minimum at y

. Thus, from the denition of viscosity

supersolutions, we have
v(y

) + H
_
y

,
x

_
0.
The above two inequalities yield
(u(x

) v(y

))
H
_
[x

[ +
[x

[
2

_
.
Sending 0 in the above together with (3.10) and (3.11), we have 0,
which is a contradiction. 2
Remark. In the above proof, we could show that lim
0
u(x

) = u( x) and
lim
0
v(y

) = v( x) although we do not need this fact. In fact, by (3.9), we

have
v(y

) u(x

) ,
28
which implies that
v( x) liminf
0
v(y

) liminf
0
u(x

) limsup
0
u(x

) u( x) ,
and
v( x) liminf
0
v(y

) limsup
0
v(x

) limsup
0
u(x

) u( x) .
Hence, since all the inequalities become the equalities, we have
u( x) = liminf
0
u(x

) = limsup
0
u(x

) and v( x) = liminf
0
v(y

) = limsup
0
v(y

).
We remark here that we cannot apply Theorem 3.4 to the eikonal equation
(2.1) because we have to suppose > 0 in the above proof.
We shall modify the above proof so that the comparison principle for
viscosity solutions of (2.1) holds.
To simplify our hypotheses, we shall consider the following PDE:
H(x, Du) f(x) = 0 in . (3.12)
Here, we suppose that H has homogeneous degree > 0 with respect to the
second variable; there is > 0 such that
H(x, p) =

H(x, p) for x , p R
n
and > 0. (3.13)
To recover the lack of assumption > 0, we suppose the positivity of f
C(); there is > 0 such that
min
x
f(x) =: > 0. (3.14)
Example. When H(x, p) = [p[
2
(i.e. = 2) and f(x) 1 ( i.e. = 1),
equation (3.12) becomes (2.1).
The second comparison principle for rst-order PDEs is as follows:
Theorem 3.5. Assume that (3.8), (3.13) and (3.14) hold. Let u
USC() and v LSC() be a viscosity sub- and supersolution of (3.12),
respectively.
If u v on , then u v in .
29
Proof. Suppose that max

(u v) =: > 0 as usual. Then, we will get a

contradiction.
If we choose (0, 1) so that
(1 ) max

u

2
,
then we easily verify that
max

(u v) =:

2
.
We note that for any z such that (u v)(z) = , we may suppose
z . In fact, otherwise (i.e. z ), if we further suppose that < 1 is
close to 1 so that (1 ) min

v /4, then the assumption (u v on

) implies

2
= u(z) v(z) ( 1)v(z)

4
,
which is a contradiction. For simplicity, we shall omit writing the dependence
on for and (x

, y

) below.
At this stage, we shall use the idea in the proof of Theorem 3.4: Consider
the mapping

: R dened by

(x, y) := u(x) v(y)

[x y[
2
2
.
Choose (x

, y

) such that max

x,y

(x, y) =

(x

, y

). Note
that

(x

, y

) /2.
As in the proof of Theorem 3.4, we may suppose that lim
0
(x

, y

) =
( x, y) for some ( x, y) (by taking a subsequence if necessary). Also,
we easily see that
[x

[
2
2
u(x

) v(y

) M

:= max

u min

v. (3.15)
Thus, sending 0, we have x = y. Hence, (3.15) implies that u( x)
v( x) = , which yields x because of the choice of . Thus, we see that
(x

, y

) for small > 0.

Moreover, (3.15) again implies
lim
0
[x

[
2

= 0. (3.16)
30
Now, taking (x) := (v(y

) +(2)
1
[xy

[
2
)/, we see that u attains
its maximum at x

. Thus, we have
H
_
x

,
x

_
f(x

).
Hence, by (3.13), we have
H
_
x

,
x

f(x

). (3.17)
On the other hand, taking (y) = u(x

) (2)
1
[y x

[
2
, we see that
v attains its minimum at y

. Thus, we have
H
_
y

,
x

_
f(y

). (3.18)
Combining (3.18) with (3.17), we have
f(y

f(x

) H
_
y

,
x

_
H
_
x

,
x

_

H
_
[x

[
_
1 +
[x

__
.
Sending 0 in the above with (3.16), we have
(1

)f( x) 0,
which contradicts (3.14). 2
3.3 Extension to second-order PDEs
In this subsection, assuming a key lemma, we will present the comparison
principle for fully nonlinear, second-order, (degenerate) elliptic PDEs (3.1).
We rst remark that the argument of the proof of the comparison principle
for rst-order PDEs cannot be applied at least immediately.
Let us have a look at the diculty. Consider the following simple PDE:
u u = 0, (3.19)
where > 0. As one can guess, if the argument does not work for this
easiest PDE, then it must be hopeless for general PDEs.
31
However, we emphasize that the same argument as in the proof of The-
orem 3.4 does not work. In fact, let u USC() and v LSC() be a
viscosity sub- and supersolution of (3.19), respectively, such that u v on
. Setting

(x, y) := u(x) v(y) (2)

1
[x y[
2
as usual, we choose
(x

, y

) so that max
x,y

(x, y) =

(x

, y

) > 0 as before.
We may suppose that (x

, y

) converges to ( x, x) (as 0) for

some x such that (u v)( x) > 0. From the denitions of u and v, we
have
u(x

)
n

0 v(y

) +
n

.
Hence, we only have
(u(x

) v(y

))
2n

,
which does not give any contradiction as 0.
How can we go beyond this diculty ?
In 1983, P.-L. Lions rst obtained the uniqueness of viscosity solutions
for elliptic PDEs arising in stochastic optimal control problems (i.e. Bell-
man equations; F is convex in (Du, D
2
u)). However, his argument heavily
depends on stochastic representation of viscosity solutions as value func-
tions. Moreover, it seems hard to extend the result to Isaacs equations; F
is fully nonlinear.
The breakthrough was done by Jensen in 1988 in case when the coe-
cients on the second derivatives of the PDE are constant. His argument relies
purely on real-analysis and can work even for fully nonlinear PDEs.
Then, Ishii in 1989 extended Jensens result to enable us to apply to
elliptic PDEs with variable coecients. We present here the so-called Ishiis
lemma, which will be proved in Appendix.
Lemma 3.6. (Ishiis lemma) Let u and w be in USC(). For
C
2
( ), let ( x, y) be a point such that
max
x,y
(u(x) + w(y) (x, y)) = u( x) + w( y) ( x, y).
Then, for each > 1, there are X = X(), Y = Y () S
n
such that
(D
x
( x, y), X) J
2,+

u( x), (D
y
( x, y), Y ) J
2,+

w( y),
32
and
( +|A|)
_
I 0
0 I
_

_
X 0
0 Y
_
A +
1

A
2
,
where A = D
2
( x, y) S
2n
.
Remark. We note that if we suppose that u, w C
2
() and ( x, y)
in the hypothesis, then we easily have
X = D
2
u( x), Y = D
2
w( y), and
_
X 0
0 Y
_
A.
Thus, the last matrix inequality means that when u and w are only contin-
uous, we get some error term
1
A
2
, where > 1 will be large.
We also note that for (x, y) := [x y[
2
/(2), we have
A := D
2
( x, y) =
1

_
I I
I I
_
and |A| =
2

. (3.20)
For the last identity, since
|A|
2
:= sup
__
A
_
x
y
_
, A
_
x
y
__

[x[
2
+[y[
2
= 1
_
,
the triangle inequality yields |A|
2
= 2
2
sup[x y[
2
[ [x[
2
+ [y[
2
= 1
4/
2
. On the other hand, taking x = y (i.e. [x[
2
= 1/2) in the supremum
of the denition of |A|
2
in the above, we have |A|
2
4/
2
.
Remark. The other way to show the above identity, we may use the fact
that for B S
n
, in general,
|B| = max[
k
[ [
k
is the eigen-value of B.
3.3.1 Degenerate elliptic PDEs
Now, we give our hypotheses on F, which is called the structure condition.
Structure condition
There is an
F
/ such that if X, Y S
n
and > 1 satisfy
3
_
I 0
0 I
_

_
X 0
0 Y
_
3
_
I I
I I
_
,
then F(y, (x y), Y ) F(x, (x y), X)

F
([x y[(1 +[x y[)) for x, y .
(3.21)
33
In section 3.3.2, we will see that if F satises (3.21), then it is elliptic.
We rst prove the comparison principle when (3.21) holds for F using
this lemma. Afterward, we will explain why assumption (3.21) is reasonable.
Theorem 3.7. Assume that > 0 and (3.21) hold. Let u USC()
and v LSC() be a viscosity sub- and supersolution of (3.1), respectively.
If u v on , then u v in .
Proof. Suppose that max

(u v) =: > 0 as usual. Then, we will get a

contradiction.
Again, for > 0, consider the mapping

: R dened by

(x, y) = u(x) v(y)

1
2
[x y[
2
.
Let (x

, y

) be a point such that max

x,y

(x, y) =

(x

, y

)
. As in the proof of Theorem 3.4, we may suppose that
lim
0
(x

, y

) = ( x, x) for some x (i.e. x

, y

for small > 0).

Moreover, since we have (u v)( x) = ,
lim
0
[x

[
2

= 0, (3.22)
and
liminf
0
(u(x

) v(y

)). (3.23)
In view of Lemma 3.6 (taking w := v, := 1/, (x, y) = [xy[
2
/(2))
and its Remark, we nd X, Y S
n
such that
_
x

, X
_


J
2,+
u(x

),
_
x

, Y
_


J
2,
v(y

),
and

_
I 0
0 I
_

_
X 0
0 Y
_

3

_
I I
I I
_
.
Thus, the equivalent denition in Proposition 2.6 implies that
u(x

) + F
_
x

,
x

, X
_
0 v(y

) + F
_
y

,
x

, Y
_
.
34
Hence, by virtue of our assumption (3.21), we have
(u(x

) v(y

))
F
_
[x

[ +
[x

[
2

_
. (3.24)
Taking the limit inmum, as 0, together with (3.22) and (3.23) in the
above, we have
0,
which is a contradiction. 2
3.3.2 Remarks on the structure condition
In order to ensure that assumption (3.21) is reasonable, we rst present some
examples. For this purpose, we consider the Isaacs equation as in section
1.2.2.
F(x, p, X) := sup
aA
inf
bB
L
a,b
(x, p, X) f(x, a, b),
where
L
a,b
(x, p, X) := trace(A(x, a, b)X) +g(x, a, b), p) for (a, b) AB.
If we suppose that A and B are compact sets in R
m
(for some m 1),
and that the coecients in the above and f(, a, b) satisfy the hypotheses
below, then F satises (3.21).
_

_
(1) M
1
> 0 and
ij
(, a, b) : R such that A
ij
(x, a, b) =
m

k=1

ik
(x, a, b)
jk
(x, a, b), and [
jk
(x, a, b)
jk
(y, a, b)[ M
1
[x y[
for x, y , i, j = 1, . . . , n, k = 1, . . . , m, a A, b B,
(2) M
2
> 0 such that [g
i
(x, a, b) g
i
(y, a, b)[ M
2
[x y[ for x, y ,
i = 1, . . . , n, a A, b B,
(3)
f
/ such that
[f(x, a, b) f(y, a, b)[
f
([x y[) for x, y , a A, b B.
We shall show (3.21) only when
F(x, p, X) :=
n

i,j=1
m

k=1

ik
(x, a, b)
jk
(x, a, b)X
ij
35
for a xed (a, b) AB because we can modify the proof below to general
F.
Thus, we shall omit writing indices a and b.
To verify assumption (3.21), we choose X, Y S
n
such that
_
X 0
0 Y
_
3
_
I I
I I
_
.
Setting
k
=
t
(
1k
(x), . . . ,
nk
(x)) and
k
=
t
(
1k
(y), . . . ,
nk
(y)) for any
xed k 1, 2, . . . , m, we have
__
X 0
0 Y
__

k

k
_
,
_

k

k
__
3
__
I I
I I
__

k

k
_
,
_

k

k
__
= 3[
k

k
[
2
3nM
2
1
[x y[
2
.
Therefore, taking the summation over k 1, . . . , m, we have
F(y, (x y), Y ) F(x, (x y), X)
n

i,j=1
(A
ij
(y)Y
ij
+ A
ij
(x)X
ij
)
=
m

k=1
(Y
k
,
k
) +X
k
,
k
))
3mnM
2
1
[x y[
2
. 2
We next give other reasons why (3.21) is a suitable assumption. The
reader can skip the proof of the following proposition if he/she feels that the
above reason is enough to adapt (3.21).
Proposition 3.8. (1) (3.21) implies ellipticity.
(2) Assume that F is uniformly elliptic. If / satises that sup
r0
(r)/(r +
1) < , and
[F(x, p, X) F(y, p, X)[ ([x y[(|X| +[p[ + 1)) (3.25)
for x, y , p R
n
, X S
n
, then (3.21) holds for F.
Proof. For a proof of (1), we refer to Remark 3.4 in [6].
For the readers convenience, we give a proof of (2) which is essentially used
in a paper by Ishii-Lions (1990). Let X, Y S
n
satisfy the matrix inequality in
(3.21). Note that X Y .
36
Multiplying
_
I I
I I
_
to the last matrix inequality from both sides, we
have
_
X Y X + Y
X + Y X Y
_
12
_
0 0
0 I
_
.
Thus, multiplying
_

s
_
for s R and , R
n
with [[ = [[ = 1, we see that
0 (12 (X Y ), ))s
2
2(X + Y ), )s (X Y ), ).
Hence, we have
[(X + Y ), )[
2
[(X Y ), )[(12 +[(X Y ), )[),
which implies
|X + Y | |X Y |
1/2
(12 +|X Y |)
1/2
.
Thus, we have
|X|
1
2
(|X Y | +|X + Y |) |X Y |
1/2
(6 +|X Y |)
1/2
.
Since X Y (i.e. the eigen-values of X Y are non-positive), we see that
F(y, p, X) F(y, p, Y ) T

(X Y ) |X Y |. (3.26)
For the last inequality, we recall Remark after Lemma 3.6.
Since we may suppose is concave, for any xed > 0, there is M

> 0 such
that (r) + M

r and (r) = inf

>0
( + M

r) for r 0. By (3.25) and (3.26),

since |X| 3 and |Y | 3, we have
F(y, p, Y ) F(x, p, X)
+ M

[x y[([p[ + 1) + sup
0t6
_
M

[x y[t
1/2
(6 + t)
1/2
t
_
.
Noting that
M

[x y[t
1/2
(6 + t)
1/2
t
3

M
2

[x y[
2
,
we have
F(y, (x y), Y ) F(x, (x y), X)
+ M

[x y[([x y[ + 1) + 3
1
M
2

[x y[
2
,
which implies the assertion by taking the inmum over > 0. 2
37
3.3.3 Uniformly elliptic PDEs
We shall give a comparison result corresponding to Proposition 3.3; F is
uniformly elliptic and 0.
Theorem 3.9. Assume that (3.2), (3.3), (3.4) and (3.21) hold. Assume
also that F is uniformly elliptic. Let u USC() and v LSC() be a
viscosity sub- and supersolution of (3.1), respectively.
If u v on , then u v in .
Remark. As in Proposition 3.3, we may suppose = 0.
Proof. Suppose that max

(u v) =: > 0.
Setting := ( + 1)/, we choose > 0 so that
max
x
e
x
1

2
.
We then set := max
x
(u(x) v(x) + e
x
1
) > 0.
Putting (x, y) := (2)
1
[x y[
2
e
x
1
, we let (x

, y

) be the
maximum point of u(x) v(y) (x, y) over .
By the compactness of , we may suppose that (x

, y

) ( x, y)
as 0 (taking a subsequence if necessary). Since u(x

) v(y

) (x

, y

),
we have [x

[
2
2(max

umin

v+2
1
) and moreover, x = y. Hence,
we have
u( x) v( x) + e
x
1
,
which implies x because of our choice of . Thus, we may suppose that
(x

, y

) for small > 0. Moreover, as before, we see that

lim
0
[x

[
2

= 0. (3.27)
Applying Lemma 3.6 to u(x) := u(x)+e
x
1
and v(y), we nd X, Y S
n
such that ((x

)/, X) J
2,+
u(x

), ((x

)/, Y ) J
2,
v(y

), and

_
I O
O I
_

_
X O
O Y
_

3

_
I I
I I
_
.
We shall simply write x and y for x

and y

, respectively.
38
Note that Proposition 2.7 implies
_
x y

e
x
1
e
1
, X
2
e
x
1
I
1
_
J
2,+
u(x),
where e
1
R
n
and I
1
S
n
are given by
e
1
:=
_
_
_
_
_
_
1
0
.
.
.
0
_
_
_
_
_
_
and I
1
:=
_
_
_
_
_
_
1 0 0
0 0 0
.
.
.
.
.
.
.
.
.
0 0 0
_
_
_
_
_
_
.
Setting r := e
x
1
, from the denition of u and v, we have
0 F
_
y,
x y

, Y
_
F
_
x,
x y

re
1
, X rI
1
_
.
In view of the uniform ellipticity and (3.4), we have
0 r + rT
+
(I
1
) + F
_
y,
x y

, Y
_
F
_
x,
x y

, X
_
.
Hence, by (3.21) and the denition of T
+
, we have
0 r( ) +
F
_
[x y[ +
[x y[
2

_
,
which together with (3.27) yields 0 e
x
1
(). This is a contradiction
because of our choice of > 0. 2
39
4 Existence results
In this section, we present some existence results for viscosity solutions of
second-order (degenerate) elliptic PDEs.
We rst present a convenient existence result via Perrons method, which
was established by Ishii in 1987.
Next, for Bellman and Isaacs equations, we give representation formulas
for viscosity solutions. From the dynamic programming principle below, we
will realize how natural the denition of viscosity solutions is.
4.1 Perrons method
In order to introduce Perrons method, we need the notion of viscosity solu-
tions for semi-continuous functions.
Denition. For any function u : R, we denote the upper and
lower semi-continuous envelope of u by u

and u

, respectively, which are

dened by
u

(x) = lim
0
sup
yB(x)
u(y) and u

(x) = lim
0
inf
yB(x)
u(y).
We give some elementary properties for u

and u

without proofs.
Proposition 4.1. For u : R, we have
(1) u

(x) u(x) u

(x) for x ,
(2) u

(x) = (u)

(x) for x ,
(3) u

(resp., u

) is upper (resp., lower) semi-continuous in , i.e.

limsup
yx
u

(y) u

(x), (resp., liminf

yx
u

(y) u

(x)) for x ,
(4) if u is upper (resp., lower) semi-continuous in ,
then u(x) = u

(x) (resp., u(x) = u

(x)) for x .
With these notations, we give our denition of viscosity solutions of
F(x, u, Du, D
2
u) = 0 in . (4.1)
Denition. We call u : R a viscosity subsolution (resp., superso-
lution) of (4.1) if u

(resp., u

) is a viscosity subsolution (resp., supersolution)

of (4.1).
40
We call u : R a viscosity solution of (4.1) if it is both a viscosity
sub- and supersolution of (4.1).
Remark. We note that we supposed that viscosity sub- and supersolu-
tions are, respectively, upper and lower semi-continuous in our comparison
principle in section 3. Adapting the above new denition, we omit the semi-
continuity for viscosity sub- and supersolutions in Propositions 3.1, 3.3 and
Theorems 3.4, 3.5, 3.7, 3.9.
In what follows,
we use the above denition.
Remark. We remark that the comparison principle Theorem 3.7 implies
the continuity of viscosity solutions.
Continuity of viscosity solutions
viscosity solution u
satises u

= u

on
_
= u C()
Proof of the continuity of u. Since u

and u

are, respectively, a viscosity

subsolution and a viscosity supersolution and u

on , Theorem 3.7
yields u

in . Because u

u u

in , we have u = u

= u

in ;
u C(). 2
We rst show that the point-wise supremum (resp., inmum) of viscos-
ity subsolutions (resp., supersolution) becomes a viscosity subsolution (resp.,
supersolution).
Theorem 4.2. Let o be a non-empty set of upper (resp., lower) semi-
continuous viscosity subsolutions (resp., supersolutions) of (4.1).
Set u(x) := sup
vS
v(x) (resp., u(x) := inf
vS
v(x)). If sup
xK
[u(x)[ <
for any compact sets K , then u is a viscosity subsolution (resp.,
supersolution) of (4.1).
Proof. We only give a proof for subsolutions since the other can be proved
in a symmetric way.
For x , we suppose that 0 = (u

)( x) > (u

)(x) for x x
and C
2
(). We shall show that
F( x, ( x), D( x), D
2
( x)) 0. (4.2)
41
Let r > 0 be such that B
2r
( x) . We can nd s > 0 such that
max
Br( x)
(u

) s. (4.3)
We choose x
k
B
r
( x) such that lim
k
x
k
= x, u

( x) k
1
u(x
k
)
and [(x
k
) ( x)[ < 1/k. Moreover, we select upper semi-continuous u
k
o
such that u
k
(x
k
) + k
1
u(x
k
).
By (4.3), for 3/k < s, we have
max
Br( x)
(u
k
) < (u
k
)(x
k
).
Thus, for large k > 3/s, there is y
k
B
r
( x) such that u
k
attains its
maximum over B
r
( x) at y
k
. Hence, we have
F(y
k
, u
k
(y
k
), D(y
k
), D
2
(y
k
)) 0. (4.4)
Taking a subsequence if necessary, we may suppose z := lim
k
y
k
. Since
(u

)( x) (u
k
)(x
k
) +
3
k
(u
k
)(y
k
) +
3
k
(u

)(y
k
) +
3
k
by the upper semi-continuity of u

, we have
(u

)( x) (u

)(z),
which yields z = x, and moreover, lim
k
u
k
(y
k
) = u

( x) = ( x). Therefore,
sending k in (4.4), by the continuity of F, we obtain (4.2). 2
Our rst existence result is as follows.
Theorem 4.3. Assume that F is elliptic. Assume also that there are
a viscosity subsolution USC() L

loc
() and a viscosity supersolution
LSC() L

loc
() of (4.1) such that
in .
Then, u(x) := sup
vS
v(x) (resp., u(x) = inf
w

S
w(x)) is a viscosity solu-
tion of (4.1), where
o :=
_
v USC()

v is a viscosity subsolution
of (4.1) such that v in
_
42
_
resp.,

o :=
_
w LSC()

w is a viscosity supersolution
of (4.1) such that w in
__
.
Sketch of proof. We only give a proof for u since the other can be shown
in a symmetric way.
First of all, we notice that o , = since o.
Due to Theorem 4.2, we know that u is a viscosity subsolution of (4.1).
Thus, we only need to show that it is a viscosity supersolution of (4.1).
Assume that u LSC(). Assuming that 0 = (u )( x) < (u )(x)
for x x and C
2
(), we shall show that
F( x, ( x), D( x), D
2
( x)) 0.
Suppose that this conclusion fails; there is > 0 such that
F( x, ( x), D( x), D
2
( x)) 2.
Hence, there is r > 0 such that
F(x, (x) + t, D(x), D
2
(x)) for x B
r
( x) and [t[ r. (4.5)
First, we claim that ( x) < ( x). Indeed, otherwise, since u in
, attains its minimum at x . See Fig 4.1.
Fig 4.1
y = (x)
y = u(x)
y = (x)
x
x
y = (x)

Hence, from the denition of supersolution , we get a contradiction to

(4.5) for x = x and t = 0.
43
We may suppose that ( x) < ( x) since, otherwise, = = at x.
Setting 3 := ( x) u( x) > 0, from the lower and upper semi-continuity of
and , respectively, we may choose s (0, r] such that
(x) + (x) + 2 (x) for x B
2s
( x).
Moreover, we can choose (0, s) and
0
(0, min , r) such that
(x) + 2
0
u(x) for x B
s+
( x) B
s
( x).
If we can dene a function w o such that w( x) > u( x), then we nish
our proof because of the maximality of u at each point.
Now, we set
w(x) :=
_
maxu(x), (x) +
0
in B
s
( x),
u(x) in B
s
( x).
See Fig 4.2.
Fig 4.2
y = (x)
y = u(x)
y = (x)
y = w(x)
x
x
y = (x) +
0

3
It suces to show that w o. Because of our choice of
0
, s > 0, it is
easy to see w in . Thus, we only need to show that w is a viscosity
subsolution of (4.1).
To this end, we suppose that (w

)(x) (w

)(z) = 0 for x ,
and then we will get
F(z, w

(z), D(z), D
2
(z)) 0. (4.6)
If z B
s
( x) =:

, by Proposition 2.4, then u

attains its maximum

at z

, we get (4.6).
44
If z B
s
( x), then (4.6) holds again since w = u in B
s+
( x) B
s
( x).
It remains to show (4.6) when z B
s
( x). Since +
0
is a viscosity
subsolution of (4.1) in B
s
( x), Theorem 4.2 with := B
s
( x) yields (4.6). 2
Correct proof, which the reader may skip rst. Since we do not suppose that
u LSC() here, we have to work with u

.
Suppose that 0 = (u

)( x) < (u

)(x) for x x for some C

2
(),
x , > 0 and
F( x, ( x), D( x), D
2
( x)) 2.
Hence, we get (4.5) even in this case.
We also show that the w dened in the above is a viscosity subsolution of (4.1).
It only remains to check that sup

(w u) > 0.
In fact, choosing x
k
B
1/k
( x) such that
u

( x) +
1
k
u(x
k
),
we easily verify that if 1/k min
0
/2, s and [( x)(x
k
)[ <
0
/2, then we have
w(x
k
) (x
k
) +
0
> ( x) +

0
2
= u

( x) +

0
2
u(x
k
). 2
4.2 Representation formula
In this subsection, for given Bellman and Isaacs equations, we present the
expected solutions, which are called value functions. In fact, via the dy-
namic programming principle for the value functions, we verify that they are
viscosity solutions of the corresponding PDEs.
Although this subsection is very important to learn how the notion of
viscosity solutions is the right one from a view point of applications in optimal
control and games,
if the reader is more interested in the PDE theory than these applications,
he/she may skip this subsection.
We shall restrict ourselves to
investigate the formulas only for rst-order PDEs
45
because in order to extend the results below to second-order ones, we need
to introduce some terminologies from stochastic analysis. However, this is
too much for this thin book.
As will be seen, we study the minimization of functionals associated with
ordinary dierential equations (ODEs for short), which is called a deter-
ministic optimal control problem. When we adapt stochastic dierential
equations instead of ODEs, those are called stochastic optimal control
problems. We refer to [10] for the later.
Moreover, to avoid mentioning the boundary condition, we will work on
the whole domain R
n
.
Throughout this subsection, we also suppose (3.7); > 0.
4.2.1 Bellman equation
We x a control set A R
m
for some m N. We dene / by
/ := : [0, ) A [ () is measurable.
For x R
n
and /, we denote by X(; x, ) the solution of
_
X

(t) = g(X(t), (t)) for t > 0,

X(0) = x,
(4.7)
where we will impose a sucient condition on continuous functions g : R
n

A R
n
so that (4.7) is uniquely solvable.
For given f : R
n
A R, under suitable assumptions (see (4.8) below),
we dene the cost functional for X(; x, ):
J(x, ) :=
_

0
e
t
f(X(t; x, ), (t))dt.
Here, > 0 is called a discount factor, which indicates that the right hand
side of the above is nite.
Now, we shall consider the optimal cost functional, which is called the
value function in the optimal control problem;
u(x) := inf
A
J(x, ) for x R
n
.
Theorem 4.4. (Dynamic Programming Principle) Assume that
_

_
(1) sup
aA
_
|f(, a)|
L

(R
n
)
+|g(, a)|
W
1,
(R
n
)
_
< ,
(2) sup
aA
[f(x, a) f(y, a)[
f
([x y[) for x, y R
n
,
(4.8)
46
where
f
/.
For any T > 0, we have
u(x) = inf
A
_
_
T
0
e
t
f(X(t; x, ), (t))dt + e
T
u(X(T; x, ))
_
.
Proof. For xed T > 0, we denote by v(x) the right hand side of the
above.
Step 1: u(x) v(x). Fix any > 0, and choose

/ such that
u(x) +
_

0
e
t
f(X(t; x,

),

(t))dt.
Setting x = X(T; x,

) and

/ by

(t) =

(T + t) for t 0, we have
_

0
e
t
f(X(t; x,

),

(t))dt =
_
T
0
e
t
f(X(t; x,

),

(t))dt
+e
T
_

0
e
t
f(X(t; x,

),

(t))dt.
Here and later, without mentioning, we use the fact that
X(t + T; x, ) = X(t; x, ) for T > 0, t 0 and /,
where
(t) := (t + T) (t 0) and x := X(T; x, ).
Indeed, the above relation holds true because of the uniqueness of solutions
of (4.7) under assumptions (4.8). See Fig 4.3.
Thus, taking the inmum in the second term of the right hand side of the
above among /, we have
u(x) +
_
T
0
e
t
f(X(t; x, ), (t))dt + e
T
u( x),
which implies one-sided inequality by taking the inmum over / since > 0
is arbitrary.
Step 2: u(x) v(x). Fix > 0 again, and choose

/ such that
v(x) +
_
T
0
e
t
f(X(t; x,

),

(t))dt + e
T
u( x),
47
Fig 4.3
X(t; x, )
X(t + T; x, ) = X(t; x, )
t = 0
x
x
t = T
x = X(T; x, )
where x := X(T; x,

). We next choose
1
/ such that
u( x) +
_

0
e
t
f(X(t; x,
1
),
1
(t))dt.
Now, setting

0
(t) :=
_

(t) for t [0, T),

1
(t T) for t T,
we see that
v(x) + 2
_

0
e
t
f(X(t; x,
0
),
0
(t))dt,
which gives the opposite inequality by taking the inmum over
0
/ since
> 0 is arbitrary again. 2
Now, we give an existence result for Bellman equations.
Theorem 4.5. Assume that (4.8) holds. Then, u is a viscosity solution
of
sup
aA
u g(x, a), Du) f(x, a) = 0 in R
n
. (4.9)
Sketch of proof. In Steps 1 and 2, we give a proof when u USC(R
n
)
and u LSC(R
n
), respectively.
Step 1: Subsolution property. Fix C
1
(R
n
), and suppose that 0 =
(u )( x) (u )(x) for some x R
n
and any x R
n
.
Fix any a
0
A, and set
0
(t) := a
0
for t 0 so that
0
/.
48
For small s > 0, in view of Theorem 4.4, we have
( x) e
s
(X(s; x,
0
)) u( x) e
s
u(X(s; x,
0
))

_
s
0
e
t
f(X(t; x,
0
), a
0
)dt.
Setting X(t) := X(t; x,
0
) for simplicity, by (4.7), we see that
e
t
(X(t)) g(X(t),
0
), D(X(t))) =
d
dt
_
e
t
(X(t))
_
. (4.10)
Hence, we have
0
_
s
0
e
t
(X(t)) g(X(t), a
0
), D(X(t))) f(X(t), a
0
)dt.
Therefore, dividing the above by s > 0, and then sending s 0, we have
0 ( x) g( x, a
0
), D( x)) f( x, a
0
),
which implies the desired inequality of the denition by taking the supremum
over A.
Step 2: Supersolution property. To show that u is a viscosity supersolu-
tion, we argue by contradiction.
Suppose that there are x R
n
, > 0 and C
1
(R
n
) such that 0 =
(u )( x) (u )(x) for x R
n
, and that
sup
aA
( x) g( x, a), D( x)) f( x, a) 2.
Thus, we can nd > 0 such that
sup
aA
(x) g(x, a), D(x)) f(x, a) for x B

( x). (4.11)
By assumption (4.8) for g, setting t
0
:= /(sup
aA
|g(, a)|
L

(R
n
)
+1) > 0,
we easily see that
[X(t; x, ) x[
_
t
0
[X

(s; x, )[ds for t [0, t

0
] and /.
Hence, by setting X(t) := X(t; x, ) for any xed /, (4.11) yields
(X(t)) g(X(t), (t)), D(X(t))) f(X(t), (t)) (4.12)
49
for t [0, t
0
]. Since (4.10) holds for in place of
0
, multiplying e
t
in
(4.12), and then integrating it over [0, t
0
], we obtain
( x) e
t
0
(X(t
0
))
_
t
0
0
e
t
f(X(t), (t))dt

(1 e
t
0
).
Thus, setting
0
= (1 e
t
0
)/ > 0, which is independent of /, we
have
u( x)
_
t
0
0
e
t
f(X(t), (t))dt + e
t
0
u(X(t
0
))
0
.
Therefore, taking the inmum over /, we get a contradiction to Theorem
4.4. 2
Correct proof, which the reader may skip rst.
Step 1: Subsolution property. Assume that there are x R
n
, > 0 and
C
1
(R
n
) such that 0 = (u

)( x) (u

)(x) for x R
n
and that
sup
aA
( x) g( x, a), D( x)) f( x, a) 2.
In view of (4.8), there are a
0
A and r > 0 such that
(x) g(x, a
0
), D(x)) f(x, a
0
) for x B
2r
( x). (4.13)
For large k 1, we can choose x
k
B
1/k
( x) such that u

( x) u(x
k
) + k
1
and [( x) (x
k
)[ < 1/k. We will only use k such that 1/k r.
Setting
0
(t) := a
0
, we note that X
k
(t) := X(t; x
k
,
0
) B
2r
( x) for t [0, t
0
]
with some t
0
> 0 and for large k.
On the other hand, by Theorem 4.4, we have
u(x
k
)
_
t
0
0
e
t
f(X
k
(t), a
0
)dt + e
t
0
u(X
k
(t
0
)).
Thus, we have
(x
k
)
2
k
( x)
1
k
u(x
k
)
_
t
0
0
e
t
f(X
k
(t), a
0
)dt + e
t
0
(X
k
(t
0
)).
Hence, by (4.13) as in Step 1 of Sketch of proof, we see that

2
k

_
t
0
0
e
t
f(X
k
(t), a
0
) +g(X
k
(t), a
0
), D(X
k
(t))) (X
k
(t))dt

(1 e
t
0
),
50
which is a contradiction for large k.
Step 2: Supersolution property. Assume that there are x R
n
, > 0 and
C
1
(R
n
) such that 0 = (u

)( x) (u

)(x) for x R
n
and that
sup
aA
( x) g( x, a), D( x)) f( x, a) 2.
In view of (4.8), there is r > 0 such that
(x) g(x, a), D(x)) f(x, a) for x B
2r
( x) and a A. (4.14)
For large k 1, we can choose x
k
B
1/k
( x) such that u

( x) u(x
k
) k
1
and [( x) (x
k
)[ < 1/k. In view of (4.8), there is t
0
> 0 such that
X
k
(t; x
k
, ) B
2r
( x) for all k
1
r
, / and t [0, t
0
].
Now, we select
k
/ such that
u(x
k
) +
1
k

_
t
0
0
e
t
f(X(t; x
k
,
k
),
k
(t))dt + e
t
0
u(X(t
0
; x
k
,
k
)).
Setting X
k
(t) := X(t; x
k
,
k
), we have
(x
k
) +
3
k
( x) +
2
k
u(x
k
) +
1
k

_
t
0
0
e
t
f(X
k
(t),
k
(t))dt + e
t
0
(X
k
(t)).
Hence, we have
3
k

_
t
0
0
e
t
g(X
k
(t),
k
(t)), D(X
k
(t))) + f(X
k
(t),
k
(t)) (X
k
(t))dt.
Putting (4.14) with
k
in the above, we have
3
k

_
t
0
0
e
t
dt,
which is a contradiction for large k 1. 2
4.2.2 Isaacs equation
In this subsection, we study fully nonlinear PDEs (i.e. p R
n
F(x, p) is
neither convex nor concave) arising in dierential games.
51
We are given continuous functions f : R
n
A B R and g : R
n

AB R
n
such that
_

_
(1) sup
(a,b)AB
_
|f(, a, b)|
L

(R
n
)
+|g(, a, b)|
W
1,
(R
n
)
_
< ,
(2) sup
(a,b)AB
[f(x, a, b) f(y, a, b)[
f
([x y[) for x, y R
n
,
(4.15)
where
f
/.
Under (4.15), we shall consider Isaacs equations:
sup
aA
inf
bB
u g(x, a, b), Du) f(x, a, b) = 0 in R
n
, (4.16)
and
inf
bB
sup
aA
u g(x, a, b), Du) f(x, a, b) = 0 in R
n
. (4.16

)
As in the previous subsection, we shall derive the expected solution.
We rst introduce some notations: While we will use the same notion /
as before, we set
B := : [0, ) B [ () is measurable.
Next, we introduce the so-called sets of non-anticipating strategies:
:=
_

_
: / B

for any T > 0, if

1
and
2
/ satisfy
that
1
(t) =
2
(t) for a.a. t (0, T),
then [
1
](t) = [
2
](t) for a.a. t (0, T)
_

_
and
:=
_

_
: B /

for any T > 0, if

1
and
2
B satisfy
that
1
(t) =
2
(t) for a.a. t (0, T),
then [
1
](t) = [
2
](t) for a.a. t (0, T)
_

_
.
Using these notations, we will consider maximizing-minimizing problems
of the following cost functional: For /, B, and x R
n
,
J(x, , ) :=
_

0
e
t
f(X(t; x, , ), (t), (t))dt,
52
where X(; x, , ) is the (unique) solutions of
_
X

(t) = g(X(t), (t), (t)) for t > 0,

X(0) = x.
(4.17)
The expected solutions for (4.16) and (4.16

), respectively, are given by

u(x) = sup

inf
A
_

0
e
t
f(X(t; x, , []), (t), [](t))dt,
and
v(x) = inf

sup
B
_

0
e
t
f(X(t; x, [], ), [](t), (t))dt.
We call u and v upper and lower value functions of this dierential game,
respectively. In fact, under appropriate hypotheses, we expect that v u,
which cannot be proved easily. To show v u, we rst observe that u and v
are, respectively, viscosity solutions of (4.16) and (4.16

). Noting that
sup
aA
inf
bB
rg(x, a, b), p)f(x, a, b) inf
bB
sup
aA
rg(x, a, b), p)f(x, a, b)
for (x, r, p) R
n
RR
n
, we see that u (resp., v) is a viscosity supersolution
(resp., subsolution) of (4.16

) (resp., (4.16)). Thus, the standard comparison

principle implies v u in R
n
(under suitable growth condition at [x[
for u and v).
We shall only deal with u since the corresponding results for v can be
obtained in a symmetric way.
To show that u is a viscosity solution of the Isaacs equation (4.16), we rst
establish the dynamic programming principle as in the previous subsection:
Theorem 4.6. (Dynamic Programming Principle) Assume that (4.15)
hold. Then, for T > 0, we have
u(x) = sup

inf
A
_
_
_
T
0
e
t
f(X(t; x, , []), (t), [](t))dt
+e
T
u(X(T; x, , []))
_
_
.
Proof. For a xed T > 0, we denote by w(x) the right hand side of the
above.
Step 1: u(x) w(x). For any > 0, we choose

such that
u(x) inf
A
_

0
e
t
f(X(t; x, ,

[]), (t),

[](t))dt =: I

.
53
For any xed
0
/, we dene the mapping T
0
: / / by
T
0
[] :=
_

0
(t) for t [0, T),
(t T) for t [T, )
for /.
Thus, for any /, we have
I

_
T
0
e
t
f(X(t; x,
0
,

[
0
]),
0
(t),

[
0
](t))dt
+
_

T
e
t
f(X(t; x, T
0
[],

[T
0
[]]), T
0
[](t),

[T
0
[]](t))dt
=: I
1

+ I
2

.
We next dene by
[](t) :=

[T
0
[]](t + T) for t 0 and /.
Note that belongs to .
Setting x := X(T; x,
0
,

[
0
]), we have
I
2

= e
T
_

0
e
t
f(X(t; x, , []), (t), [](t))dt.
Taking the inmum over /, we have
u(x) I
1

+ e
T
inf
A
_

0
e
t
f(X(t; x, , []), (t), [](t))dt
=: I
1

+

I
2

.
Since

I
2

e
T
u( x), we have
u(x) I
1

+ e
T
u( x),
which implies u(x) w(x) by taking the inmum over
0
/ and then,
the supremum over . Therefore, we get the one-sided inequality since > 0
is arbitrary.
Step 2: u(x) w(x). For > 0, we choose
1

such that
w(x) inf
A
_
_
_
T
0
e
t
f(X(t; x, ,
1

[]), (t),
1

[](t))dt
+e
T
u(X(T; x, ,
1

[]))
_
_
.
54
For any xed
0
/, setting x = X(T; x,
0
,
1

[
0
]), we have
w(x)
_
T
0
e
t
f(X(t; x,
0
,
1

[
0
]),
0
(t),
1

[
0
](t))dt + e
T
u( x).
Next, we choose
2

such that
u( x) inf
A
_

0
e
t
f(X(t; x, ,
2

[]), (t),
2

[](t))dt. =: I.
For /, we dene the mapping T
1
: / / by
T
1
[](t) := (t + T) for t 0.
Thus, we have
I
_

0
e
t
f(X(t; x, T
1
[
0
],
2

[T
1
[
0
]]), T
1
[
0
](t),
2

[T
1
[
0
]](t))dt =:

I.
Now, for /, setting
[](t) :=
_

1

[](t) for t [0, T),

[T
1
[]](t T) for t [T, ),
and

X(t) := X(t; x, T
1
[
0
],
2

[T
1
[
0
]]), we have

I =
_

T
e
(tT)
f(

X(t T), T
1
[
0
](t T),
2

[T
1
[
0
]](t T))dt
= e
T
_

T
e
t
f(

X(t T),
0
(t), [
0
](t))dt.
Since
X(t; x,
0
, [
0
]) =
_
X(t; x,
0
,
1

[
0
]) for t [0, T),

X(t T) for t [T, ),

we have
w(x) 2
_

0
e
t
f(X(t; x,
0
, [
0
]),
0
(t), [
0
](t))dt.
Since
0
is arbitrary, we have
w(x) 2 inf
A
_

0
e
t
f(X(t; x, , []), (t), [](t))dt,
55
which yields the assertion by taking the supremum over and then, by
sending 0. 2
Now, we shall verify that the value function u is a viscosity solution of
(4.16).
Since we only give a sketch of proofs, one can skip the following theorem.
For a correct proof, we refer to [1], originally by Evans-Souganidis (1984).
Theorem 4.7. Assume that (4.15) holds.
(1) Then, u is a viscosity subsolution of (4.16).
(2) Assume also the following properties:
_

_
(i) A R
m
is compact for some integer m 1.
(ii) there is an
A
/ such that
[f(x, a, b) f(x, a

, b)[ +[g(x, a, b) g(x, a

, b)[
A
([a a

[)
for x R
n
, a, a

A and b B.
(4.18)
Then, u is a viscosity supersolution of (4.16).
Remark. To show that v is a viscosity subsolution of (4.16

), instead of (4.18),
we need to suppose the following hypotheses:
_

_
(i) B R
m
is compact for some integer m 1.
(ii) there is an
B
/ such that
[f(x, a, b) f(x, a, b

)[ +[g(x, a, b) g(x, a, b

)[
B
([b b

[)
for x R
n
, b, b

B and a A,
(4.18

)
while to verify that v is a viscosity supersolution of (4.16

), we only need (4.15).

Sketch of proof. We shall only prove the assertion assuming that u USC(R
n
)
and u LSC(R
n
) in Step 1 and 2, respectively.
To give a correct proof without the semi-continuity assumption, we need a bit
careful analysis similar to the proof for Bellman equations. We omit the correct
proof here.
Step 1: Subsolution property. Suppose that the subsolution property fails; there
are x R
n
, > 0 and C
1
(R
n
) such that 0 = (u )(x) (u )(y) (for all
y R
n
) and
sup
aA
inf
bB
u(x) g(x, a, b), D(x)) f(x, a, b) 3.
We note that X(; x, , []) are uniformly continuous for any (, ) /
in view of (4.15).
56
Thus, we can choose that a
0
A such that
inf
bB
(x) g(x, a
0
, b), D(x)) f(x, a
0
, b) 2.
For any , setting
0
(t) = a
0
for t 0, we simply write X() for
X(; x,
0
, [
0
]). Thus, we nd small t
0
> 0 such that
(X(t)) g(X(t), a
0
, [
0
](t)), D(X(t))) f(X(t), a
0
, [
0
](t))
for t [0, t
0
]. Multiplying e
t
in the above and then, integrating it over [0, t
0
],
we have

(1 e
t
0
)
_
t
0
0
_
d
dt
_
e
t
(X(t))
_
+ e
t
f(X(t), a
0
, [
0
](t))
_
dt
= (x) e
t
0
(X(t
0
))
_
t
0
0
e
t
f(X(t), a
0
, [
0
](t))dt.
Hence, we have
u(x)

(1 e
t
0
)
_
t
0
0
e
t
f(X(t), a
0
, [
0
](t))dt + e
t
0
u(X(t
0
)) =:

I.
Taking the inmum over /, we have

I inf
A
_
_
_
t
0
0
e
t
f(X(t; x, , []), (t), [](t))dt
+e
t
0
u(X(t
0
; x, , []))
_
_
.
Therefore, since is arbitrary, we have
u(x)

(1 e
t
0
) sup

inf
A
_
_
_
t
0
0
e
t
f(X(t; x, , []), (t), [](t))dt
+e
t
0
u(X(t
0
; x, , []))
_
_
,
which contradicts Theorem 4.6.
Step 2: Supersolution property. Suppose that the supersolution property fails;
there are x R
n
, > 0 and C
1
(R
n
) such that 0 = (u )(x) (u )(y)
for y R
n
, and
sup
aA
inf
bB
u(x) g(x, a, b), D(x)) f(x, a, b) 3.
For any a A, there is b(a) B such that
u(x) g(x, a, b(a)), D(x)) f(x, a, b(a)) 2.
57
In view of (4.18), there is (a) > 0 such that if [a a

[ < (a) and [x y[ < (a),

then we have
(y) g(y, a

, b(a)), D(y)) f(y, a

, b(a)) .
From the compactness of A, we may select a
k

M
k=1
such that
A =
M
_
k=1
A
k
,
where
A
k
:= a A [ [a a
k
[ < (a
k
).
Furthermore, we set

A
1
= A
1
, and inductively,

A
k
:= A
k

k1
j=1
A
j
;

A
k


A
j
=
for k ,= j. We may also suppose that

A
k
,= for k = 1, . . . , M.
For /, we dene

0
[](t) := b(a
k
) provided (t)

A
k
.
Now, setting X(t) := X(t; x, ,
0
[]), we nd t
0
> 0 such that
(X(t)) g(X(t), (t),
0
[](t)), D(X(t))) f(X(t), (t),
0
[](t))
for t [0, t
0
]. Multiplying e
t
in the above and then, integrating it, we obtain
(x) e
t
0
(X(t
0
))
_
t
0
0
e
t
f(X(t), (t),
0
[](t))dt

(1 e
t
0
).
Since / is arbitrary, we have
u(x) +

(1 e
t
0
) inf
A
_
_
_
t
0
0
e
t
f(X(t; x, ,
0
[]), (t),
0
[](t))dt
+e
t
0
u(X(t
0
; x, ,
0
[]))
_
_
,
which contradicts Theorem 4.6 by taking the supremum over . 2
4.3 Stability
In this subsection, we present a stability result for viscosity solutions, which
is one of the most important properties for solutions as noted in section 1.
Thus, this result justies our notion of viscosity solutions.
However, since we will only use Proposition 4.8 below in section 7.3, the
reader may skip the proof.
58
First of all, for possibly discontinuous F : RR
n
S
n
R, we are
concerned with
F(x, u, Du, D
2
u) = 0 in . (4.19)
We introduce the following notation:
F

(x, r, p, X) := lim
0
inf
_
F(y, s, q, Y )

y B

(x), [s r[ < ,
[q p[ < , |Y X| <
_
,
F

(x, r, p, X) := lim
0
sup
_
F(y, s, q, Y )

y B

(x), [s r[ < ,
[q p[ < , |Y X| <
_
.
Denition. We call u : R a viscosity subsolution (resp., super-
solution) of (4.19) if u

(resp., u

) is a viscosity subsolution (resp., super-

soluion) of
F

(x, u, Du, D
2
u) 0
_
resp., F

(x, u, Du, D
2
u) 0
_
in .
We call u : R a viscosity solution of (4.19) if it is both a viscosity
sub- and supersolution of (4.19).
Now, for given continuous functions F
k
: RR
n
S
n
R, we set
liminf
k

F
k
(x, r, p, X)
:= lim
k
inf
_

_
F
j
(y, s, q, Y )

[y x[ < 1/k, [s r[ < 1/k,

[q p[ < 1/k, |Y X| < 1/k
and j k
_

_
,
limsup
k

F
k
(x, r, p, X)
:= lim
k
sup
_

_
F
j
(y, s, q, Y )

[y x[ < 1/k, [s r[ < 1/k,

[q p[ < 1/k, |Y X| < 1/k
and j k
_

_
.
Our stability result is as follows.
Proposition 4.8. Let F
k
: R R
n
S
n
R be continuous
functions. For (x, r, p, X) RR
n
S
n
, we set
F(x, r, p, X) := liminf
k

F
k
(x, r, p, X)
59
_
resp., F(x, r, p, X) := limsup
k

F
k
(x, r, p, X)
_
.
Let u
k
: R be a viscosity subsolution (resp., supersolution) of
F
k
(x, u
k
, Du
k
, D
2
u
k
) = 0 in .
Setting u (resp., u) by
u(x) := lim
k
sup(u
j
)

(y) [ y B
1/k
(x) , j k
_
resp., u(x) := lim
k
inf(u
j
)

(y) [ y B
1/k
(x) , j k
_
for x , then u (resp., u) is a viscosity subsolution (resp., supersolution)
of
F(x, u, Du, D
2
u) 0 (resp., F(x, u, Du, D
2
u) 0) in .
Remark. We note that u USC(), u LSC(), F LSC( R
R
n
S
n
) and F USC( RR
n
S
n
).
Proof. We only give a proof for subsolutions since the other can be shown
similarly.
Given C
2
(), we let x
0
be such that 0 = (u)(x
0
) > (u)(x)
for x x
0
. We shall show that F(x
0
, u(x
0
), D(x
0
), D
2
(x
0
)) 0.
We may choose x
k
B
r
(x
0
) (for a subsequence if necessary), where r
(0,dist(x
0
, )), such that
lim
k
x
k
= x
0
and lim
k
(u
k
)

(x
k
) = u(x
0
). (4.20)
We select y
k
B
r
(x
0
) such that ((u
k
)

)(y
k
) = sup
Br(x
0
)
((u
k
)

).
We may also suppose that lim
k
y
k
= z for some z B
r
(x
0
) (taking
a subsequence if necessary). Since ((u
k
)

)(y
k
) ((u
k
)

)(x
k
), (4.20)
implies
0 = liminf
k
((u
k
)

)(x
k
) liminf
k
((u
k
)

)(y
k
)
liminf
k
(u
k
)

(y
k
) (z)
limsup
k
(u
k
)

(y
k
) (z) (u )(z).
60
Thus, this yields z = x
0
and lim
k
(u
k
)

(y
k
) = u(x
0
). Hence, we see that
y
k
B
r
(x
0
) for large k 1. Since (u
k
)

attains a maximum over B

r
(x
0
)
at y
k
B
r
(x
0
), by the denition of u
k
(with Proposition 2.4 for

= B
r
(x
0
)),
we have
F
k
(y
k
, (u
k
)

(y
k
), D(y
k
), D
2
(y
k
)) 0,
which concludes the proof by taking the liminf
k

. 2
61
5 Generalized boundary value problems
In order to obtain the uniqueness of solutions of an ODE, we have to suppose
certain initial or boundary condition. In the study of PDEs, we need to
impose appropriate conditions on for the uniqueness of solutions.
Following the standard PDE theory, we shall treat a few typical boundary
conditions in this section.
Since we are mainly interested in degenerate elliptic PDEs, we cannot
expect solutions to satisfy the given boundary condition on the whole
of . The simplest example is as follows: For := (0, 1), consider the
degenerate elliptic PDE

du
dx
+ u = 0 in (0, 1).
Note that it is impossible to nd a solution u of the above such that u(0) =
u(1) = 1.
Our plan is to propose a denition of generalized solutions for boundary
value problems. For this purpose, we extend the notion of viscosity solutions
to possibly discontinuous PDEs on while we normally consider those in .
For general G : RR
n
S
n
R, we are concerned with
G(x, u, Du, D
2
u) = 0 in . (5.1)
As in section 4.3, we dene
G

(x, r, p, X) := lim
0
inf
_
G(y, s, q, Y )

y B

(x), [s r[ < ,
[q p[ < , |Y X| <
_
,
G

(x, r, p, X) := lim
0
sup
_
G(y, s, q, Y )

y B

(x), [s r[ < ,
[q p[ < , |Y X| <
_
.
Denition. We call u : R a viscosity subsolution (resp., superso-
lution) of (5.1) if, for any C
2
(),
G

(x, u

(x), D(x), D
2
(x)) 0
_
resp., G

(x, u

(x), D(x), D
2
(x)) 0
_
provided that u

(resp., u

) attains its maximum (resp., minimum)

at x .
62
We call u : R a viscosity solution of (5.1) if it is both a viscosity
sub- and supersolution of (5.1).
Our comparison principle in this setting is as follows:
Comparison principle in this setting
viscosity subsolution u of (5.1)
viscosity supersolution v of (5.1)
_
= u v in
Note that
the boundary condition is contained in the denition.
Using the above new denition, we shall formulate the boundary value
problems in the viscosity sense. Given F : R R
n
S
n
R and
B : RR
n
S
n
R, we investigate general boundary value problems
_
F(x, u, Du, D
2
u) = 0 in ,
B(x, u, Du, D
2
u) = 0 on .
(5.2)
Setting G by
G(x, r, p, X) :=
_
F(x, r, p, X) for x ,
B(x, r, p, X) for x ,
we give the denition of boundary value problems (5.2) in the viscosity sense.
Denition. We call u : R a viscosity subsolution (resp., superso-
lution) of (5.2) if it is a viscosity subsolution (resp., supersolution) of (5.1),
where G is dened in the above.
We call u : R a viscosity solution of (5.2) if it is both a viscosity
sub- and supersolution of (5.2).
Remark. When F and B are continuous and G is given as above, G

and
G

can be expressed in the following manner:

G

(x, r, p, X) =
_
F(x, r, p, X) for x ,
minF(x, r, p, X), B(x, r, p, X) for x ,
G

(x, r, p, X) =
_
F(x, r, p, X) for x ,
maxF(x, r, p, X), B(x, r, p, X) for x .
63
It is not hard to extend the existence and stability results corresponding
to Theorem 4.3 and Proposition 4.8, respectively, to viscosity solutions in
the above sense. However, it is not straightforward to show the comparison
principle in this new setting. Thus, we shall concentrate our attention to
the comparison principle, which implies the uniqueness (and continuity) of
viscosity solutions.
The main diculty to prove the comparison principle is that we have to
avoid the boundary conditions for both of viscosity sub- and supersolu-
tions.
To explain this, let us consider the case when G is given by (5.2). Let u
and v be, respectively, a viscosity sub- and supersolution of (5.1). We shall
observe that the standard argument in Theorem 3.7 does not work.
For > 0, suppose that (x, y) u(x) v(y) (2)
1
[x y[
2
attains its
maximum at (x

, y

) . Notice that there is NO reason to verify that

(x

, y

) .
The worst case is that (x

, y

) . In fact, in view of Lemma 3.6,

we nd X, Y S
n
such that ((x

)/, X) J
2,+

u(x

), ((x

)/, Y )
J
2,

v(y

), the matrix inequalities in Lemma 3.6 hold for X, Y . Hence, we

have
min
_
F
_
x

, u(x

),
x

, X
_
, B
_
x

, u(x

),
x

, X
__
0
and
max
_
F
_
y

, v(y

),
x

, Y
_
, B
_
y

, v(y

),
x

, Y
__
0.
However, even if we suppose that (3.21) holds for F and B in , we cannot
get any contradiction when
F
_
x

, u(x

),
x

, X
_
0 B
_
y

, v(y

),
x

, Y
_
or
B
_
x

, u(x

),
x

, X
_
0 F
_
y

, v(y

),
x

, Y
_
.
It seems impossible to avoid this diculty as long as we use [x y[
2
/(2) as
test functions.
64
Our plan to go beyond this diculty is to nd new test functions

(x, y)
(instead of [xy[
2
/(2)) so that the function (x, y) u(x) v(y)

(x, y)
attains its maximum over at an interior point (x

, y

) . To
this end, since we will use several perturbation techniques, we suppose two
hypotheses on F: First, we shall suppose the following continuity of F with
respect to (p, X)-variables.
_

_
There is an
0
/ such that
[F(x, p, X) F(x, q, Y )[
0
([p q[ +|X Y |)
for x , p, q R
n
, X, Y S
n
.
(5.3)
The next assumption is a bit stronger than the structure condition (3.21):
_

_
There is
F
/ such that
if X, Y S
n
and > 1 satisfy
3
_
I 0
0 I
_

_
X 0
0 Y
_
3
_
I I
I I
_
, then
F(y, p, Y ) F(x, p, X)
F
([x y[(1 +[p[ + [x y[))
for x, y , p R
n
, X, Y S
n
.
(5.4)
5.1 Dirichlet problem
First, we consider Dirichlet boundary value problems (Dirichlet problems for
short) in the above sense.
Assuming that viscosity sub- and supersolutions are continuous on ,
we will obtain the comparison principle for them.
We now recall the classical Dirichlet problem
_
u + F(x, Du, D
2
u) = 0 in ,
u g = 0 on .
(5.5)
Note that the Dirichlet problem of (5.5) in the viscosity sense is as follows:
subsolution
_
u + F(x, Du, D
2
u) 0 in ,
minu + F(x, Du, D
2
u), u g 0 on ,
and
supersolution
_
u + F(x, Du, D
2
u) 0 in ,
maxu + F(x, Du, D
2
u), u g 0 on .
65
We shall suppose the following property on the shape of , which may
be called an interior cone condition (see Fig 5.1):
_
For each z , there are r, s (0, 1) such that
x rn(z) + r for x B
r
(z), r (0, r) and B
s
(0).
(5.6)
Here and later, we denote by n(z) the unit outward normal vector at z .
Fig 5.1
x
rn(z)
n(z)
r
sr
z

Theorem 5.1. Assume that > 0, (5.3), (5.4) and (5.6) hold. For
g C(), we let u and v : R be, respectively, a viscosity sub- and
supersolution of (5.5) such that
liminf
xz
u

(x) u

(z) and limsup

xz
v

(x) v

(z) for z . (5.7)

Then, u

in .
Remark. Notice that (5.7) implies the continuity of u

and v

on .
Proof. Suppose that max

(u

) =: > 0. We simply write u and v

for u

and v

, respectively.
Case 1: max

(u v) = . We choose z such that (u v)(z) = .

We shall divide three cases:
Case 1-1: u(z) > g(z). For , (0, 1), where > 0 will be xed later,
setting (x, y) := (2
2
)
1
[xy n(z)[
2
[xz[
2
, we let (x

, y

)
be the maximum point of (x, y) := u(x) v(y) (x, y) over .
66
Since z n(z) for small > 0 by (5.6), (x

, y

) (z, z n(z))
implies that
[x

n(z)[
2
2
2
u(x

) v(y

) u(z) +v(z n(z)) [x

z[
2
. (5.8)
Since [x

[ M, where M :=

2(max

umin

v u(z) +v(z) +1)

1/2
,
for small > 0, we may suppose that (x

, y

) ( x, x) and (x

)/ z
for some x and z R
n
as 0 along a subsequence (denoted by
again). Thus, from the continuity (5.7) of v at z , (5.8) implies that
u( x) v( x) [ x z[
2
,
which yields x = z. Moreover, we have
lim
0
[x

n(z)[
2

2
= 0,
which implies that
lim
0
[x

= . (5.9)
Furthermore, we note that y

= x

n(z) + o() because of (5.6).

Applying Lemma 3.6 with Proposition 2.7 to u(x) +
1
n(z), x) [x
z[
2
2
1

2
and v(y) +
1
n(z), y), we nd X, Y S
n
such that
_
x

n(z) + 2(x

z), X + 2I
_
J
2,+

u(x

), (5.10)
_
x

n(z), Y
_
J
2,

v(y

), (5.11)
and

2
_
I O
O I
_

_
X O
O Y
_

3

2
_
I I
I I
_
.
Putting p

:=
2
(x

)
1
n(z), by (5.3), we have
F(x

, p

, X) F(x

, p

+2(x

z), X +2I)
0
(2[x

z[ +2). (5.12)
Since y

and u(x

) > g(x

) for small > 0 provided x

, in view
of (5.10) and (5.11), we have
(u(x

) v(y

)) F(y

, p

, Y ) F(x

, p

+ 2(x

z), X + 2I).
67
Combining this with (5.12) , by (5.4), we have
(u(x

)v(y

))
0
(2[x

z[+2)+
F
_
[x

[
_
1 +[p

[ +
[x

2
__
.
Sending 0 together with (5.9) in the above, we have

0
(2) +
F
(2
2
),
which is a contradiction for small > 0, which only depends on and .
Case 1-2: v(z) < g(z). To get a contradiction, we argue as above replacing
(x, y) by (x, y) := (2
2
)
1
[x y + n(z)[
2
[x z[
2
so that x

= y

n(z) + o() for small > 0. Note that we need here the continuity of
u on in (5.7) while the other one in (5.7) is needed in Case 1-1. (See also
the proof of Theorem 5.3 below.)
Case 1-3: u(z) g(z) and v(z) g(z). This does not occur because 0 <
= (u v)(z) 0.
Case 2: sup

(u v) < . In this case, using the standard test function

[x y[
2
/(2) (without [x z[
2
term), we can follow the same argument as
in the proof of Theorem 3.7. 2
Remark. Unfortunately, without assuming the continuity of viscosity so-
lutions on , the comparison principle fails in general.
In fact, setting F(x, r, p, X) r and g(x) 1, consider the function
u(x) :=
_
0 for x ,
1 for x .
Note that u

0 and u

u in , which are respectively a viscosity sub- and

supersolution of G(x, u, Du, D
2
u) = 0 in . Therefore, this example shows
that the comparison principle fails in general without assumption (5.7).
5.2 State constraint problem
The state constraint boundary condition arises in a typical optimal control
problem. Thus, if the reader is more interested in the PDE theory, he/she
may skip Proposition 5.2 below, which explains why we will adapt the state
constraint boundary condition in Theorem 5.3.
68
To explain our motivation, we shall consider Bellman equations of rst-order.
sup
aA
u g(x, a), Du) f(x, a) = 0 in .
Here, we use the notations in section 4.2.1.
We introduce the following set of controls: For x ,
/(x) := () / [ X(t; x, ) for t 0.
We shall suppose that
/(x) ,= for all x . (5.13)
Also, we suppose that
_

_
(1) sup
aA
_
|f(, a)|
L

()
+|g(, a)|
W
1,
()
_
< ,
(2) sup
aA
[f(x, a) f(y, a)[
f
([x y[) for x, y ,
(5.14)
where
f
/.
We are now interested in the following the optimal cost functional:
u(x) := inf
A(x)
_

0
e
t
f(X(t; x, ), (t))dt.
Proposition 5.2. Assume that > 0, (5.13) and (5.14) hold. Then, we have
(1) u is a viscosity subsolution of
sup
aA
u g(x, a), Du) f(x, a) 0 in ,
(2) u is a viscosity supersolution of
sup
aA
u g(x, a), Du) f(x, a) 0 in .
Remark. We often say that u satises the state constraint boundary condition
when it is a viscosity supersolution of
F(x, u, Du, D
2
u) 0 in .
Proof. In fact, at x , it is easy to verify that the dynamic programming
principle (Theorem 4.4) holds for small T > 0. Thus, we may show Theorem 4.5
replacing R
n
by .
69
Hence, it only remains to show (2) on . Thus, suppose that there are x ,
> 0 and C
1
() such that (u

)( x) = 0 (u

)(x) for x , and

sup
aA
( x) g( x, a), D( x)) f( x, a) 2.
Then, we will get a contradiction.
Choose x
k
B
1/k
( x) such that u

( x) + k
1
u(x
k
) and [( x) (x
k
)[ <
1/k. In view of (5.14), there is t
0
> 0 such that for any /(x
k
) and large k 1,
we have
(X
k
(t)) g(X
k
(t), (t)), D(X
k
(t))) f(X
k
(t), (t)) for t (0, t
0
),
where X
k
(t) := X(t; x
k
, ). Thus, multiplying e
t
and then, integrating it over
(0, t
0
), we have
(x
k
) e
t
0
(X
k
(t
0
)) +
_
t
0
0
e
t
f(X
k
(t), (t))dt

(1 e
t
0
).
Since we have
u(x
k
)
2
k
+ e
t
0
u(X
k
(t
0
)) +
_
t
0
0
e
t
f(X
k
(t), (t))dt

(1 e
t
0
),
taking the inmum over /(x
k
), we apply Theorem 4.4 to get
0
2
k

(1 e
t
0
),
which is a contradiction for large k. 2
Motivated by this proposition, we shall consider more general second-order
elliptic PDEs.
Theorem 5.3. Assume that > 0, (5.3), (5.4), (5.6) and (5.12) hold. Let
u : R be, respectively, a viscosity sub- and supersolution of
u + F(x, Du, D
2
u) 0 in ,
and
v + F(x, Dv, D
2
v) 0 in .
Assume also that
liminf
xz
u

(x) u

(z) for z . (5.15)

70
Then, u

in .
Remark. In 1986, Soner rst treated the state constraint problems for deter-
ministic optimal control (i.e. rst-order PDEs) by the viscosity solution approach.
We note that we do not need continuity of v on while we need it for
Dirichlet problems. For further discussion on the state constraint problems, we
refer to Ishii-Koike (1996).
We also note that the proof below is easier than that for Dirichlet problems
in section 5.1 because we only need to avoid the boundary condition for viscosity
subsolutions.
Proof. Suppose that max

(u

) =: > 0. We shall write u and v for u

and v

, respectively, again.
We may suppose that max

(u v) = since otherwise, we can use the

standard procedure to get a contradiction.
Now, we proceed the same argument in Case 1-2 in the proof of Theorem 5.1
(although it is not precisely written).
For , > 0, setting (x, y) := (2
2
)
1
[x y + n(z)[
2
+[x z[
2
, where n is
the unit outward normal vector at z , we let (x

, y

) the maximum
point of u(x) v(y) (x, y) over . As in the proof of Theorem 3.4, we have
lim
0
(x

, y

) = (z, z) and lim

0
[x

= . (5.16)
Since x

= y

n(z) + o() for small > 0, in view of Lemma 3.6 with

Proposition 2.7, we can nd X, Y S
n
such that
_
x

2
+

n(z) + 2(x

z), X + 2I
_
J
2,+

u(x

),
_
x

2
+

n(z), Y
_
J
2,

v(y

),
and

2
_
I O
O I
_

_
X O
O Y
_

2
_
I I
I I
_
.
Setting p

:=
2
(x

) +
1
n(z), we have
(u(x

) v(y

)) F(y

, p

, Y ) F(x

, p

+ 2(x

z), X + 2I)

0
(2[x

z[ + 2) +
F
_
[x

[
_
1 +[p

[ +
[x

__
.
71
Hence, sending 0 with (5.16), we have

0
(2) +
F
(2
2
),
which is a contradiction for small > 0. 2
5.3 Neumann problem
In the classical theory and modern theory for weak solutions in the distribu-
tion sense, the (inhomogeneous) Neumann condition is given by
n(x), Du(x)) g(x) = 0 on ,
where n(x) denotes the unit outward normal vector at x .
In Dirichlet and state constraint problems, we have used a test function
which forces one of x

and y

to be in . However, in the Neumann boundary

value problem (Neumann problem for short), we have to avoid the boundary
condition for viscosity sub- and supersolutions simultaneously. Thus, we need
a new test function dierent from those in sections 5.1 and 5.2.
We rst dene the signed distance function from by
(x) :=
_
inf[x y[ [ y for x
c
,
inf[x y[ [ y for x .
In order to obtain the comparison principle for the Neumann problem,
we shall impose a hypothesis on (see Fig 5.2):
_

_
(1) There is r > 0 such that
(B
r
(z + rn(z)))
c
for z .
(2) There is a neighborhood N of such that
C
2
(N), and D(x) = n(x) for x .
(5.17)
Remark. This assumption (1) is called the uniform exterior sphere con-
dition. Since [x z rn(z)[ r for z and x , we have
n(z), x z)
[x z[
2
2 r
for z and x . (5.18)
It is known that when is smooth enough, (2) of (5.17) holds true.
72
Fig 5.2
rn(z)

r
z

We shall consider the inhomogeneous Neumann problem:

_
u + F(x, Du, D
2
u) = 0 in ,
n(x), Du) g(x) = 0 on .
(5.19)
Remember that we adapt the denition of viscosity solutions of (5.19) for
the corresponding G in (5.2).
Theorem 5.4. Assume that > 0, (5.3), (5.4) and (5.17) hold. For
g C(), we let u and v : R be a viscosity sub- and supersolution of
(5.19), respectively.
Then, u

in .
Remark. We note that we do not need any continuity of u and v on .
Proof. As before, we write u and v for u

and v

, respectively.
As in the proof of Theorem 3.7, we suppose that max

(u v) =: > 0.
Also, we may suppose that max

(u v) = .
Let z be a point such that (u v)(z) = . For small > 0, we see
that the mapping x u(x) v(y) [x z[
2
takes its strict maximum
at z.
For small , > 0, where > 0 will be xed later, setting (x, y) :=
(2)
1
[x y[
2
g(z)n(z), x y) + ((x) + (y) + 2) + [x z[
2
, we let
(x

, y

) be the maximum point of (x, y) := u(x) v(y) (x, y)

over N N, where N is in (5.17).
Since (x

, y

) (z, z), as before, we may extract a subsequence, which

is denoted by (x

, y

) again, such that (x

, y

) ( x, x). We may suppose

x . Since ( x, x) limsup
0
(x

, y

), we have
u( x) v( x) [ x z[
2
,
73
which yields x = z. Moreover, we have
lim
0
[x

[
2

= 0. (5.20)
Applying Lemma 3.6 to u(x) ((x) +1) g(z)n(z), x) [xz[
2
and
v(y) ((y) + 1) +g(z)n(z), y), we nd X, Y S
n
such that
_
p

+ n(x

) + 2(x

z), X + D
2
(x

) + 2I
_
J
2,+

u(x

), (5.21)
_
p

n(y

), Y D
2
(y

)
_
J
2,

v(y

), (5.22)
where p

:=
1
(x

) + g(z)n(z), and

_
I 0
0 I
_

_
X 0
0 Y
_

3

_
I I
I I
_
.
When x

, by (5.18), we calculate in the following manner:

n(x

), D
x
(x

, y

)) = n(x

), p

+ n(x

) + 2(x

z))

[x

[
2
2 r
+ g(z)n(x

), n(z)) + 2[x

z[.
Hence, given > 0, we see that
n(x

), D
x
(x

, y

)) g(x

)

2
for small > 0.
Thus, by (5.21), this yields
u(x

) + F(x

, p

+ n(x

) + 2(x

z), X + D
2
(x

) + 2I) 0. (5.23)
Of course, if x

, then the above inequality holds from the denition.

On the other hand, similarly, if y

, then
n(y

), D
y
(x

, y

)) g(y

2
for small > 0.
Hence, by (5.22), we have
v(y

) + F(y

, p

n(y

), Y D
2
(y

)) 0. (5.24)
74
Using (5.3) and (5.4), by (5.23) and (5.24), we have
(u(x

) v(y

)) F(y

, p

, Y ) F(x

, p

, X) + 2
0
(M)

F
_
[x

[
_
1 +[p

[ +
[x

__
+ 2
0
(M),
where M := 3 + sup
xN
(2[x z[ + [D
2
(x)[). Sending 0 with (5.20)
in the above, we have
2
0
(M),
which is a contradiction for small > 0. 2
5.4 Growth condition at [x[
In the standard PDE theory, we often consider PDEs in unbounded domains,
typically, in R
n
. In this subsection, we present a technique to establish the
comparison principle for viscosity solutions of
u + F(x, Du, D
2
u) = 0 in R
n
. (5.25)
We remind the readers that in the proofs of comparison results we always
suppose max

(u v) > 0, where u and v are, respectively, a viscosity sub-

and supersolution. However, considering := R
n
, the maximum of u v
might attain its maximum at [x[ . Thus, we have to choose a test
function (x, y), which forces u(x) v(y) (x, y) to takes its maximum at
a point in a compact set.
For this purpose, we will suppose the linear growth condition (for sim-
plicity) for viscosity solutions.
We rewrite the structure condition (3.21) for R
n
:
_

_
There is an
F
/ such that if X, Y S
n
and > 1 satisfy
3
_
I 0
0 I
_

_
X 0
0 Y
_
3
_
I I
I I
_
,
then F(y, (x y), Y ) F(x, (x y), X)

F
([x y[(1 +[x y[)) for x, y R
n
.
(5.26)
We will also need the Lipschitz continuity of (p, X) F(x, p, X), which
is stronger than (5.3).
_
There is
0
> 0 such that [F(x, p, X) F(x, q, Y )[

0
([p q[ +|X Y |) for x R
n
, p, q R
n
, X, Y S
n
.
(5.27)
75
Proposition 5.5. Assume that > 0, (5.26) and (5.27) hold. Let u and
v : R
n
R be, respectively, a viscosity sub- and supersolution of (5.25).
Assume also that there is C
0
> 0 such that
u

(x) C
0
(1 +[x[) and v

(x) C
0
(1 +[x[) for x R
n
. (5.28)
Then, u

in R
n
.
Proof. We shall simply write u and v for u

and v

, respectively.
For > 0, we set

:= sup
xR
n(u(x) v(x) 2(1 +[x[
2
)). We note that
(5.28) implies that there is z

R
n
such that

= u(z

)v(z

)2(1+[z

[
2
).
Set := limsup
0

R .
When 0, since
(u v)(x) 2(1 +[x[
2
) +

for > 0 and x R

n
,
we have u v in R
n
.
Thus, we may suppose (0, ]. Setting

(x, y) := u(x) v(y)

(2)
1
[x y[
2
(1 + [x[
2
) (1 + [y[
2
) for , > 0, where > 0 will be
xed later, in view of (5.28), we can choose (x

, y

) R
n
R
n
such that

(x

, y

) = max
(x,y)R
n
R
n

(x, y)

.
As before, extracting a subsequence if necessary, we may suppose that
lim
0
[x

[
2

= 0. (5.29)
By Lemma 3.6 with Proposition 2.7, putting p

:= (x

)/, we nd
X, Y S
n
such that
(p

+ 2x

, X + 2I) J
2,+
u(x

),
(p

2y

, Y 2I) J
2,
v(y

),
and

_
I O
O I
_

_
X O
O Y
_

3

_
I I
I I
_
.
Hence, we have
(u(x

) v(y

))
F(y

, p

2y

, Y 2I) F(x

, p

+ 2x

, X + 2I)
F(y

, p

, Y ) F(x

, p

, X) + 2
0
(2 +[x

[ +[y

[)

F
_
[x

[
_
1 +
[x

__
+ (2 +[x

[
2
+[y

[
2
) + C,
76
where C = C(
0
, ) > 0 is independent of , > 0. For the last inequality,
we used 2ab a
2
+
1
b
2
for > 0.
Therefore, we have

F
_
[x

[
_
1 +
[x

__
+ C.
Sending 0 in the above together with (5.29), we get C, which is
a contradiction for small > 0. 2
77
6 L
p
-viscosity solutions
In this section, we discuss the L
p
-viscosity solution theory for uniformly
elliptic PDEs:
F(x, Du, D
2
u) = f(x) in , (6.1)
where F : R
n
S
n
R and f : R are given. Since we will use the
fact that u + C (for a constant C R) satises the same (6.1), we suppose
that F does not depend on u itself. Furthermore, to compare with classical
results, we prefer to have the inhomogeneous term (the right hand side of
(6.1)).
The aim in this section is to obtain the a priori estimates for L
p
-viscosity
solutions without assuming any continuity of the mapping x F(x, q, X),
and then to establish an existence result of L
p
-viscosity solutions for Dirichlet
problems.
Remark. In general, without the continuity assumption of x F(x, p, X),
even if X F(x, p, X) is uniformly elliptic, we cannot expect the unique-
ness of L
p
-viscosity solutions. Because Nadirashvili (1997) gave a counter-
example of the uniqueness.
6.1 A brief history
Let us simply consider the Poisson equation in a smooth domain with
zero-Dirichlet boundary condition:
_
u = f in ,
u = 0 on .
(6.2)
In the literature of the regularity theory for uniformly elliptic PDEs of
second-order, it is well-known that
if f C

() for some (0, 1), then u C

2,
(). (6.3)
Here, C

(U) (for a set U R

n
) denotes the set of functions f : U R such
that
sup
xU
[f(x)[ + sup
x,yU,x=y
[f(x) f(y)[
[x y[

< .
78
Also, C
k,
(U), for an integer k 1, denotes the set of functions f : U R
so that for any multi-index = (
1
, . . . ,
n
) 0, 1, 2, . . .
n
with [[ :=

n
i=1

i
k, D

f C

(U), where
D

f :=

||
f
x

1
1
x
n
n
.
These function spaces are called Holder continuous spaces and the implication
in (6.3) is called the Schauder regularity (estimates). Since the PDE in
(6.2) is linear, the regularity result (6.3) may be extended to
if f C
k,
() for some (0, 1), then u C
k+2,
(). (6.4)
Moreover, we obtain that (6.4) holds for the following PDE:
trace(A(x)D
2
u(x)) = f(x) in , (6.5)
where the coecient A() C

(, S
n
) satises that
[[
2
A(x), ) [[
2
for R
n
and x .
Furthermore, we can obtain (6.4) even for linear second-order uniformly
elliptic PDEs if the coecients are smooth enough.
Besides the Schauder estimates, we know a dierent kind of regularity
results: For a solution u of (6.5), and an integer k 0, 1, 2, . . .,
if f W
k,p
() for some p > 1, then u W
k+2,p
(). (6.6)
Here, for an open set O R
n
, we say f L
p
(O) if [f[
p
is integrable in O,
and f W
k,p
(O) if for any multi-index with [[ k, D

f L
p
(O). Notice
that L
p
() = W
0,p
().
This (6.6) is called the L
p
regularity (estimates). For a later con-
venience, for p 1, we recall the standard norms of L
p
(O) and W
k,p
(O),
respectively:
|u|
L
p
(O)
:=
__
O
[u(x)[
p
dx
_
1/p
, and |u|
W
k,p
(O)
:=

||k
|D

u|
L
p
(O)
.
In Appendix, we will use the quantity |u|
L
p
()
even for p (0, 1) although
this is not the norm (i.e. the triangle inequality does not hold).
79
We refer to [13] for the details on the Schauder and L
p
regularity theory
for second-order uniformly elliptic PDEs.
As is known, a diculty occurs when we drop the smoothness of A
ij
.
An extreme case is that we only suppose that A
ij
are bounded (possibly
discontinuous, but still satisfy the uniform ellipticity). In this case, what can
we say about the regularity of solutions of (6.5) ?
The extreme case for PDEs in divergence form is the following:

i,j=1

x
i
_
A
ij
(x)
u
x
j
(x)
_
= f(x) in . (6.7)
De Giorgi (1957) rst obtained Holder continuity estimates on weak so-
lutions of (6.7) in the distribution sense; for any C

0
(),
_

(A(x)Du(x), D(x)) f(x)(x)) dx = 0.

Here, we set
C

0
() :=
_
: R

() is innitely many times dierentiable,

and supp is compact in
_
.
We refer to [14] for the details of De Giorgis proof and, a dierent proof
by Moser (1960).
Concerning the corresponding PDE in nondivergence form, by a stochas-
tic approach, Krylov-Safonov (1979) rst showed the Holder continuity esti-
mates on strong solutions of
trace(A(x)D
2
u(x)) = f(x) in . (6.8)
Afterward, Trudinger (1980) (see [13]) gave a purely analytic proof for it.
Since these results appeared before the viscosity solution was born, they
could only deal with strong solutions, which satisfy PDEs in the a.e. sense.
In 1989, Caarelli proved the same Holder estimate for viscosity solutions
of fully nonlinear second-order uniformly elliptic PDEs.
To show Holder continuity of solutions, it is essential to prove the follow-
ing Harnack inequality for nonnegative solutions. In fact, to prove the
Harnack inequality, we split the proof into two parts:
80
weak Harnack inequality
for supersolutions
local maximum principle
for subsolutions
_

_
=
Harnack inequality
for solutions
In section 6.4, we will show that L
p
-viscosity solutions satisfy the (inte-
rior) Holder continuous estimates.
6.2 Denition and basic facts
We rst recall the denition of L
p
-strong solutions of general PDEs:
F(x, u, Du, D
2
u) = f(x) in . (6.9)
We will use the following function space:
W
2,p
loc
() := u : R [ u W
2,p
() for all C

0
().
Throughout this section, we suppose at least
p >
n
2
so that u W
2,p
loc
() has the second-order Taylor expansion at almost all
points in , and that u C().
Denition. We call u C() an L
p
-strong subsolution (resp., super-
solution, solution) of (6.9) if u W
2,p
loc
(), and
F(x, u(x), Du(x), D
2
u(x)) f(x) (resp., f(x), = f(x)) a.e. in .
Now, we present the denition of L
p
-viscosity solutions of (6.9).
Denition. We call u C() an L
p
-viscosity subsolution (resp., su-
persolution) of (6.9) if for W
2,p
loc
(), we have
lim
0
ess. inf
B(x)
_
F(y, u(y), D(y), D
2
(y)) f(y)
_
0
81
_
resp. lim
0
ess. sup
B(x)
_
F(y, u(y), D(y), D
2
(y)) f(y)
_
0
_
provided that u takes its local maximum (resp., minimum) at x .
We call u C() an L
p
-viscosity solution of (6.9) if it is both an L
p
-
viscosity sub- and supersolution of (6.9).
Remark. Although we will not explicitly utilize the above denition, we
recall the denition of ess. sup
A
and ess. inf
A
of h : A R, where A R
n
is a measurable set:
ess. sup
A
h(y) := infM R [ h M a.e. in A,
and
ess. inf
A
h(y) := supM R [ h M a.e. in A.
Coming back to (6.1), we give a list of assumptions on F : R
n
S
n

R:
_

_
(1) F(x, 0, O) = 0 for x ,
(2) x F(x, q, X) is measurable for (q, X) R
n
S
n
,
(3) F is uniformly elliptic.
(6.10)
We recall the uniform ellipticity condition of X F(x, q, X) with the con-
stants 0 < from section 3.1.2.
For the right hand side f : R, we suppose that
f L
p
() for p n. (6.11)
We will often suppose the Lipschitz continuity of F with respect to q
R
n
;
_
there is 0 such that [F(x, q, X) F(x, q

, X)[ [q q

[
for (x, q, q

, X) R
n
R
n
S
n
.
(6.12)
Remark. We note that (1) in (6.10) and (6.12) imply that F has the linear
growth in Du;
[F(x, q, O)[ [q[ for x and q R
n
.
Remark. We note that when x F(x, q, X) and x f(x) are continu-
ous, the denition of L
p
-viscosity subsolution (resp., supersolution) of (6.1)
82
coincides with the standard one under assumption (6.10) and (6.12). For a
proof, we refer to a paper by Caarelli-Crandall-Kocan-

Swi ech [5].

In this book, we only study the case of (6.11) but most of results can be
extended to the case when p > p

= p

(, , n) (n/2, n), where p

is the
so-called Escauriazas constant (see the references in [4]).
The following proposition is obvious but it will be very convenient to
study L
p
-viscosity solutions of (6.1) under assumptions (6.10), (6.11) and
(6.12).
Proposition 6.1. Assume that (6.10), (6.11) and (6.12) hold. If u
C() is an L
p
-viscosity subsolution (resp., supersolution) of (6.1), then it is
an L
p
-viscosity subsolution (resp., supersolution) of
T

(D
2
u) [Du[ f in
_
resp., T
+
(D
2
u) + [Du[ f in
_
.
We recall the Aleksandrov-Bakelman-Pucci (ABP for short) maximum
principle, which will play an essential role in this section (and also Appendix).
To this end, we introduce the notion of upper contact sets: For u :
O R, we set
[u, O] :=
_
x O

there is p R
n
such that
u(y) u(x) +p, y x) for all y O
_
.
Proposition 6.2. (ABP maximum principle) For 0, there is C
0
:=
C
0
(, , n, , diam()) > 0 such that if for f L
n
(), u C() is an
L
n
-viscosity subsolution (resp., supersolution) of
T

(D
2
u) [Du[ f in
+
[u]
(resp., T
+
(D
2
u) + [Du[ f in
+
[u]),
then
max

u max

u
+
+ diam()C
0
|f
+
|
L
n
([u,]
+
[u])
_
resp., max

(u) max

(u)
+
+ diam()C
0
|f

|
L
n
([u,]
+
[u])
_
,
83
Fig 6.1
[u, ]
y = u(x)

x
where

+
[u] := x [ u(x) > 0.
The next proposition is a key tool to study L
p
-viscosity solutions, partic-
ularly, when f is not supposed to be continuous. The proof will be given in
Appendix.
Proposition 6.3. Assume that (6.11) holds for p n. For any 0,
there are an L
p
-strong subsolution u and an L
p
-strong supersolution v
C(B
1
) W
2,p
loc
(B
1
), respectively, of
_
T
+
(D
2
u) + [Du[ f in B
1
,
u = 0 on B
1
,
and
_
T

(D
2
v) [Dv[ f in B
1
,
v = 0 on B
1
.
Moreover, we have the following estimates: for w = u or w = v, and small
(0, 1), there is

C =

C(, , n, , ) > 0 such that
|w|
W
2,p
(B

)


C

|f|
L
p
(B
1
)
.
Remark. In view of the proof (Step 2) of Proposition 6.2, we see that
C|f

|
L
n
(B
1
)
w C|f
+
|
L
n
(B
1
)
in B
1
, where w = u, v.
6.3 Harnack inequality
In this subsection, we often use the cube Q
r
(x) for r > 0 and x =
t
(x
1
, . . . , x
n
)
R
n
;
Q
r
(x) := y =
t
(y
1
, . . . , y
n
) [ [x
i
y
i
[ < r/2 for i = 1, . . . , n,
84
and Q
r
:= Q
r
(0). Notice that
B
r/2
(x) Q
r
(x) B
r

n/2
(x) for r > 0.
We will prove the next two propositions in Appendix.
Proposition 6.4. (Weak Harnack inequality) For 0, there are p
0
=
p
0
(, , n, ) > 0 and C
1
:= C
1
(, , n, ) > 0 such that if u C(B
2

n
) is a
nonnegative L
p
-viscosity supersolution of
T
+
(D
2
u) + [Du[ 0 in B
2

n
,
then we have
|u|
L
p
0(Q
1
)
C
1
inf
Q
1/2
u.
Remark. Notice that p
0
might be smaller than 1.
Proposition 6.5. (Local maximum principle) For 0 and q > 0,
there is C
2
= C
2
(, , n, , q) > 0 such that if u C(B
2

n
) is an L
p
-viscosity
subsolution of
T

(D
2
u) [Du[ 0 in B
2

n
,
then we have
sup
Q
1
u C
2
|u
+
|
L
q
(Q
2
)
.
Remark. Notice that we do not suppose that u 0 in Proposition 6.5.
6.3.1 Linear growth
The next corollary is a direct consequence of Propositions 6.4 and 6.5.
Corollary 6.6. For 0, there is C
3
= C
3
(, , n, ) > 0 such that if
u C(B
2

n
) is a nonnegative L
p
-viscosity sub- and supersolution of
T

(D
2
u) [Du[ 0 and T
+
(D
2
u) + [Du[ 0 in B
2

n
,
respectively, then we have
sup
Q
1
u C
3
inf
Q
1
u.
85
In order to treat inhomogeneous PDEs, we will need the following corol-
lary:
Corollary 6.7. For 0 and f L
p
(B
3

n
) with p n, there is
C
4
= C
4
(, , n, ) > 0 such that if u C(B
3

n
) is a nonnegative L
p
-
viscosity sub- and supersolution of
T

(D
2
u) [Du[ f and T
+
(D
2
u) + [Du[ f in B
2

n
,
respectively, then we have
sup
Q
1
u C
4
_
inf
Q
1
u +|f|
L
p
(B
3

n
)
_
.
Proof. According to Proposition 6.3, we nd v, w C(B
3

n
)W
2,p
loc
(B
3

n
)
such that
_
T
+
(D
2
v) + [Dv[ f
+
a.e. in B
3

n
,
v = 0 on B
3

n
,
and
_
T

(D
2
w) [Dw[ f

a.e. in B
3

n
,
w = 0 on B
3

n
.
In view of Proposition 6.3 and its Remark, we can choose

C =

C(, , n, ) >
0 such that
0 v

C|f
+
|
L
p
(B
3

n
)
in B
3

n
, |v|
W
2,p
(B
2

n
)


C|f
+
|
L
p
(B
3

n
)
,
and
0 w

C|f

|
L
p
(B
3

n
)
in B
3

n
, |w|
W
2,p
(B
2

n
)


C|f

|
L
p
(B
3

n
)
.
Since v, w W
2,p
(B
2

n
), it is easy to verify that u
1
:= u + v and
u
2
:= u + w are, respectively, an L
p
-viscosity sub- and supersolution of
T

(D
2
u
1
) [Du
1
[ 0 and T
+
(D
2
u
2
) + [Du
2
[ 0 in B
2

n
.
Since v 0 in B
3

n
, applying Proposition 6.5 to u
1
, for any q > 0, we nd
C
2
(q) > 0 such that
sup
Q
1
u sup
Q
1
u
1
+

C|f
+
|
L
p
(B
3

n
)
C
2
(q)|(u
1
)
+
|
L
q
(Q
2
)
+

C|f
+
|
L
p
(B
3

n
)
C
2
(q)|u|
L
q
(Q
2
)
+

C|f
+
|
L
p
(B
3

n
)
.
(6.13)
86
On the other hand, applying Proposition 6.4 to u
2
, there is p
0
> 0 such
that
|u|
L
p
0(Q
2
)
|u
2
|
L
p
0(Q
2
)
C
1
inf
Q
1
u
2
C
1
_
inf
Q
1
u +

C|f

|
L
p
(B
3

n
)
_
. (6.14)
Therefore, combining (6.14) with (6.13) for q = p
0
, we can nd C
4
> 0 such
that the assertion holds. 2
Corollary 6.8. (Harnack inequality, nal version) Assume that (6.10),
(6.11) and (6.12) hold. If u C() is an L
p
-viscosity solution of (6.1), and
if B
3

nr
(x) for r (0, 1], then
sup
Qr(x)
u C
4
_
inf
Qr(x)
u + r
2
n
p
|f|
L
p
()
_
,
where C
4
> 0 is the constant in Corollary 6.7.
Proof. By translation, we may suppose that x = 0.
Setting v(x) := u(rx) for x B
3

n
, we easily see that v is an L
p
-viscosity
subsolution and supersolution of
T

(D
2
v) [Dv[ r
2

f and T
+
(D
2
v) + [Dv[ r
2

f, in B
3

n
,
respectively, where

f(x) := f(rx). Note that |

f|
L
p
(B
3

n
)
= r

n
p
|f|
L
p
(B
3

nr
)
.
Applying Corollary 6.7 to v and then, rescaling v to u, we conclude the
assertion. 2
6.3.2 Quadratic growth
Here, we consider the case when q F(x, q, X) has quadratic growth. We
refer to [10] for applications where such quadratic nonlinearity appears.
We present a version of the Harnack inequality when F has a quadratic
growth in Du in place of (6.12);
_
there is 0 such that [F(x, q, X) F(x, q

, X)[
([q[ +[q

[)[q q

[ for (x, q, q

, X) R
n
R
n
S
n
,
(6.15)
which together with (1) of (6.10) implies that
[F(x, q, O)[ [q[
2
for (x, q) R
n
.
87
The associated Harnack inequality is as follows:
Theorem 6.9. For 0 and f L
p
(B
3

n
) with p n, there is
C
5
= C
5
(, , n, ) > 0 such that if u C(B
3

n
) is a nonnegative L
p
-
viscosity sub- and supersolution of
T

(D
2
u) [Du[
2
f and T
+
(D
2
u) + [Du[
2
f in B
3

n
,
respectively, then we have
sup
Q
1
u C
5
e
2

M
_
inf
Q
1
u +|f|
L
p
(B
3

n
)
_
,
where M := sup
B
3

n
u.
Proof. Set := /. Fix any (0, 1).
We claim that v := e
u
1 and w := 1 e
u
are, respectively, a non-
negative L
p
-viscosity sub- and supersolution of
T

(D
2
v) (e
M
+ )f
+
and T
+
(D
2
w) (1 +)f

in B
3

n
.
We shall only prove this claim for v since the other for w can be obtained
similarly.
Choose W
2,p
loc
(B
3

n
) and suppose that u attains its local maximum
at x B
3

n
. Thus, we may suppose that v(x) = (x) and v in B
r
(x),
where B
2r
(x) B
3

n
. Note that 0 v e
M
1 in B
3

n
.
For any (0, 1), in view of W
2,p
(B
r
(x)) C

0
(B
r
(x)) with some

0
(0, 1), we can choose
0
(0, r) such that
v + in B

0
(x).
Setting (y) :=
1
log((y) + 1) for y B

0
(x) (extending W
2,p
in
B
3

n
B

0
(x) if necessary), we have
lim
0
ess inf
B(x)
_
T

(D
2
) [D[
2
f
+
_
0.
Since
D =
D
( + 1)
and D
2
=
D
2

( + 1)

D D
( + 1)
2
,
88
the above inequality yields
lim
0
ess inf
B(x)
_
T

(D
2
)
( + 1)
f
+
_
0.
Since 0 < 1 + 1 e
M
+ in B

0
(x), we have
lim
0
ess inf
B(x)
_
T

(D
2
) (e
M
+ )f
+
_
0.
Since u v ue
M
and ue
M
w u, using the same argument
to get (6.13) and (6.14), we have
sup
Q
1
u
1

sup
Q
1
v C
6
_
|v|
L
p
0(Q
2
)
+ (e
M
+ )|f
+
|
L
p
(B
3

n
)
_
C
7
_
e
2M
|w|
L
p
0(Q
2
)
+ (e
M
+ )|f
+
|
L
p
(B
3

n
)
_
C
8
_
e
2M
inf
Q
1
w + (e
2M
+ )|f|
L
p
(B
3

n
)
_
C
9
e
2M
_
inf
Q
1
u + (1 +)|f|
L
p
(B
3

n
)
_
.
Since C
k
(k = 6, . . . , 9) are independent of > 0, sending 0, we conclude
the proof. 2
Remark. We note that the same argument by using two dierent trans-
formations for sub- and supersolutions as above can be found in [14] for
uniformly elliptic PDEs in divergence form with the quadratic nonlinearity.
6.4 Holder continuity estimates
In this subsection, we show how the Harnack inequality implies the Holder
continuity.
Theorem 6.10. Assume that (6.10), (6.11) and (6.12) hold. For each
compact set K , there is = (, , n, , p, dist(K, ), |f|
L
p
()
)
(0, 1) such that if u C() is an L
p
-viscosity solution of (6.1), then there is

C =

C(, , n, , p, dist(K, ), max

[u[, |f|
L
p
()
) > 0
[u(x) u(y)[

C[x y[

for x, y K.
Remark. We notice that is independent of sup

[u[.
89
In our proof below, we may relax the dependence max

[u[ in

C by
sup[u(x)[ [ dist(x, K) < for small > 0.
Proof. Setting r
0
:= min1,dist(K, )/(3

n) > 0, we may suppose

that there is C
4
> 1 such that if w C() is a nonnegative L
p
-viscosity
sub- and supersolution of
T

(D
2
w) [Dw[ f and T
+
(D
2
w) + [Dw[ f in ,
respectively, then we see that for any r (0, r
0
] and x K (i.e. B
3

nr
(x)
),
sup
Qr(x)
w C
4
_
inf
Qr(x)
w + r
2
n
p
|f|
L
p
()
_
.
For simplicity, we may suppose x = 0 K.
Now, we set
M(r) := sup
Qr
u, m(r) := inf
Qr
u and osc(r) := M(r) m(r).
It is sucient to nd C > 0 and (0, 1) such that
M(r) m(r) Cr

for small r > 0.

We denote by S(r) the set of all nonnegative w C(B
3

nr
), which is,
respectively, an L
p
-viscosity sub- and supersolution of
T

(D
2
w) [Dw[ [f[ and T
+
(D
2
w) + [Dw[ [f[ in B
3

nr
.
Setting v
1
:= um(r) and w
1
:= M(r) u, we see that v
1
and w
1
belong
to S(r). Hence, setting C
10
:= maxC
4
|f|
L
p
()
, C
4
, 4 > 3, we have
sup
Q
r/2
v
1
C
10
_
inf
Q
r/2
v
1
+ r
2
n
p
_
and sup
Q
r/2
w
1
C
10
_
inf
Q
r/2
w
1
+ r
2
n
p
_
.
Thus, setting := 2
n
p
> 0, we have
M(r/2) m(r) C
10
_
m(r/2) m(r) + (r/2)

_
,
M(r) m(r/2) C
10
_
M(r) M(r/2) + (r/2)

_
.
90
Hence, adding these inequalities, we have
(C
10
+ 1)(M(r/2) m(r/2)) (C
10
1)(M(r) m(r)) + 2C
10
(r/2)

.
Therefore, setting := (C
10
1)/(C
10
+1) (1/2, 1) and C
11
:= 2C
10
/(C
10
+
1), we see that
osc(r/2) osc(r) + C
11
(r/2)

.
Moreover, changing r/2
k1
for integers k 2, we have
osc(r/2
k
)
k
osc(r) + C
11
r

j=1
2
j

k
osc(r
0
) +
C
11
2

1
r

C
12
(
k
+ r

),
where C
12
:= maxosc(r
0
), C
11
/(2

1).
For r (0, r
0
), by setting s = r

, where = log /(log log 2) (0, 1),

there is a unique integer k 1 such that
s
2
k
r <
s
2
k1
,
which yields
log(s/r)
log 2
k <
log(s/r)
log 2
+ 1.
Hence, recalling (1/2, 1), we have
osc(r) osc(s/2
k1
) C
12
(
k
+ (2s)

) 2

C
12
_

(1) log r/ log 2

+ r

_
.
Setting := ( 1) log / log 2 (0, 1) (because (1/2, 1)), we have

(1) log r/ log 2

= r

and r

= r

.
Thus, setting C
13
:= 2

C
12
, we have
osc(r) C
13
r

. 2 (6.16)
Remark. We note that we may derive (6.16) when p > n/2 by taking
= 2
n
p
> 0.
We shall give the corresponding Holder continuity for PDEs with quadratic
nonlinearity (6.15). Since we can use the same argument as in the proof of
91
Theorem 6.1 using Theorem 6.9 instead of Corollaries 6.7 and 6.8, we omit
the proof of the following:
Corollary 6.11. Assume that (6.10), (6.11) and (6.15) hold. For each
compact set K , there are

C =

C(, , n, , p, dist(K, ), sup

[u[) > 0
and = (, , n, , p, dist(K, ), sup

[u[) (0, 1) such that if an L

p
-
viscosity solution u C() of (6.1), then we have
[u(x) u(y)[

C[x y[

for x, y K.
Remark. Note that both of and

C depend on sup

[u[ in this quadratic

case.
6.5 Existence result
For the existence of L
p
-viscosity solutions of (6.1) under the Dirichlet condition,
we only give an outline of proof, which was rst shown in a paper by Crandall-
Kocan-Lions-

Swi ech in [7] (1999).

Theorem 6.12. Assume that (6.10), (6.11) and (6.12) hold. Assume also that
(1) of (5.17) holds.
For given g C(), there is an L
p
-viscosity solution u C() of (6.1) such
that
u(x) = g(x) for x . (6.17)
Remark. We may relax assumption (1) of (5.17) so that the assertion holds for
which may have some concave corners. Such a condition is called uniform
exterior cone condition.
Sketch of proof.
Step1: We rst solve approximate PDEs, which have to satisfy a sucient
condition in Step 3; instead of (6.1), under (6.17), we consider
F
k
(x, Du, D
2
u) = f
k
in , (6.18)
where smooth F
k
and f
k
approximate F and f, respectively. In fact, F
k
and f
k
are given by F
1/k
and f
1/k
, where
1/k
is the standard mollier with respect
to x-variables. We remark that F
1/k
means the convolution of F(, p, X) and

1/k
.
92
We nd a viscosity solution u
k
C() of (6.18) under (6.17) via Perrons
method for instance. At this stage, we need to suppose the smoothness of to
construct viscosity sub- and supersolutions of (6.18) with (6.17). Remember that
if F and f are continuous, then the notion of L
p
-viscosity solutions equals to that
of the standard ones (see Proposition 2.9 in [5]).
In view of (1) of (5.17) (i.e. the uniform exterior sphere condition), we can
construct viscosity sub- and supersolutions of (6.18) denoted by USC() and
LSC() such that = = g on . To show this fact, we only note that we
can modify the argument in Step 1 in section 7.3.
Step 2: We next obtain the a priori estimates for u
k
so that they converge to
a continuous function u C(), which is the candidate of the original PDE.
For this purpose, after having established the L

estimates via Proposition

6.2, we apply Theorem 6.10 (interior Holder continuity) to u
k
in Step 1 because
(6.10)-(6.12) hold for approximate PDEs with the same constants , , .
We need a careful analysis to get the equi-continuity up to the boundary .
See Step 1 in section 7.3 again.
Step 3: Finally, we verify that the limit function u is the L
p
-viscosity solution
via the following stability result, which is an L
p
-viscosity version of Proposition
4.8.
To state the result, we introduce some notations: For B
2r
(x) with r > 0
and x , and W
2,p
(B
r
(x)), we set
G
k
[](y) := F
k
(y, D(y), D
2
(y)) f
k
(y),
and
G[](y) := F(y, D(y), D
2
(y)) f(y)
for y B
r
(x).
Proposition 6.13. Assume that F
k
and F satisfy (6.10) and (6.12) with
, > 0 and 0. For f, f
k
L
p
() with p n, let u
k
C() be an L
p
-viscosity
subsolution (resp., supersolution) of (6.18). Assume also that u
k
converges to u
uniformly on any compact subsets of as k , and that for any B
2r
(x)
with r > 0 and x , and W
2,p
(B
r
(x)),
lim
k
|(G[] G
k
[])
+
|
L
p
(Br(x))
= 0
_
resp., lim
k
|(G[] G
k
[])

|
L
p
(Br(x))
= 0
_
.
Then, u C() is an L
p
-viscosity subsolution (resp., supersolution) of (6.1).
93
Proof of Proposition 6.13. We only give a proof of the assertion for subsolu-
tions.
Suppose the contrary: There are r > 0, > 0, x and W
2,p
(B
2r
(x))
such that B
3r
(x) , 0 = (u )(x) (u )(y) for y B
2r
(x), and
u in B
2r
(x) B
r
(x), (6.19)
and
G[](y) a.e. in B
2r
(x). (6.20)
For simplicity, we shall suppose that r = 1 and x = 0.
It is sucient to nd
k
W
2,p
(B
2
) such that lim
k
sup
B
2
[
k
[ = 0, and
G
k
[ +
k
](y) a.e. in B
2
.
Indeed, since u
k
( +
k
) attains its maximum over B
2
at an interior point
z B
2
by (6.19), the above inequality contradicts the fact that u
k
is an L
p
-
viscosity subsolution of (6.18).
Setting h(x) := G[](x) and h
k
(x) := G
k
[](x), in view of Proposition 6.3, we
can nd
k
C(B
2
) W
2,p
loc
(B
2
) such that
_

_
T

(D
2

k
) [D
k
[ (h h
k
)
+
a.e. in B
2
,

k
= 0 on B
2
,
0
k
C|(h h
k
)
+
|
L
p
(B
2
)
in B
2
,
|
k
|
W
2,p
(B
1
)
C|(h h
k
)
+
|
L
p
(B
2
)
.
We note that our assumption together with the third inequality in the above yields
lim
k
sup
B
2
[
k
[ = 0.
Using (6.10), (6.12) and (6.20), we have
G
k
[ +
k
] T

(D
2

k
) [D
k
[ + h
k
(h h
k
)
+
+ (h h
k
)
a.e. in B
2
. 2
94
7 Appendix
In this appendix, we present proofs of the propositions, which appeared in the
previous sections. However, to prove them, we often need more fundamental
results, for which we only give references. One of such results is the following
Area formula, which will be employed in sections 7.1 and 7.2. We refer to
[9] for a proof of a more general Area formula.
Area formula
C
1
(R
n
, R
n
),
g L
1
(R
n
),
A R
n
measurable
_

_
=
_
(A)
[g(y)[dy
_
A
[g((x))[[det(D(x))[dx
We note that the Area formula is a change of variable formula when
[det(D)[ may vanish. In fact, the equality holds if [det(D)[ > 0 and is
injective.
7.1 Proof of Ishiis lemma
First of all, we recall an important result by Aleksandrov. We refer to the
Appendix of [6] and [10] for a functional analytic proof, and to [9] for a
measure theoretic proof.
Lemma 7.1. (Theorem A.2 in [6]) If f : R
n
R is convex, then for
a.a. x R
n
, there is (p, X) R
n
S
n
such that
f(x + h) = f(x) +p, h) +
1
2
Xh, h) + o([h[
2
) as [h[ 0.
(i.e., f is twice dierentiable at a.a. x R
n
.)
We next recall Jensens lemma, which is a version of the ABP maximum
principle in 7.2 below.
Lemma 7.2. (Lemma A.3 in [6]) Let f : R
n
R be semi-convex
(i.e. x f(x) + C
0
[x[
2
is convex for some C
0
R). Let x R
n
be a strict
maximum point of f. Set f
p
(x) := f(x) p, x) for x R
n
and p R
n
.
Then, for r > 0, there are C
1
,
0
> 0 such that
[
r,
[ C
1

n
for (0,
0
],
95
where

r,
:=
_
x B
r
( x)

p B

such that f
p
(y) f
p
(x) for y B
r
( x)
_
.
Proof. By translation, we may suppose x = 0.
For integers m, we set f
m
(x) = f
1/m
(x), where
1/m
is the mollier.
Note that x f
m
(x) + C
0
[x[
2
is convex.
Setting

m
r,
=
_
x B
r

p B

such that f
m
p
(y) f
m
p
(x) for y B
r
_
,
where f
m
p
(x) = f
m
(x)p, x), we claim that there are C
1
,
0
> 0, independent
of large integers m, such that
[
m
r,
[ C
1

n
for (0,
0
].
We remark that this concludes the assertion. In fact, setting A
m
:=

k=m

k
r,
,
we have

m=1
A
m

r,
. Because, for x

m=1
A
m
, we can select p
k
B

and m
k
such that lim
k
m
k
= , and
max
Br
f
m
k
p
k
= f
m
k
p
k
(x).
Hence, sending k (along a subsequence if necessary), we nd p B

such that max

B
r
f
p
= f
p
(x), which yields x
r,
.
Therefore, we have
C
1

n
lim
m
[A
m
[ = [

m=1
A[ [
r,
[.
Now we shall prove our claim. First of all, we notice that x f
m
(x) +
C
0
[x[
2
is convex.
Since 0 is the strict maximum of f, we nd
0
> 0 such that

0
= f(0) max
B
4r/3
\B
r/3
f.
Fix p B

0
, where
0
=
0
/(3r). For m 3/r, we note that
f
m
(x) p, x) f(0)
0
+
0
r f(0)
2
0
3
in B
r
B
2r/3
.
96
On the other hand, for large m, we verify that
f
m
(0) f(0)
f
(m
1
) > f(0)

0
3
,
where
f
denotes the modulus of continuity of f. Hence, in view of these
observations, for any p B

0
, if max
Br
f
m
p
= f
m
p
(x) for x B
r
, then x B
r
.
In other words, we see that
B

= Df
m
(
m
r,
) for (0,
0
].
Thanks to the Area formula, we have
[B

[ =
_
Df
m
(
m
r,
)
dy
_

m
r,
[detD
2
f
m
[dx (2C
0
)
n
[
m
r,
[.
Here, we have employed that 2C
0
I D
2
f
m
O in
m
r,
. 2
Although we can nd a proof of the next proposition in [6], we recall the
proof with a minor change for the readers convenience.
Proposition 7.3. (Lemma A.4 in [6]) If f C(R
m
), B S
m
,
f() + (/2)[[
2
is convex and max
R
m
f() 2
1
B, ) = f(0), then
there is an X S
m
such that
(0, X) J
2,+
f(0) J
2,
f(0) and I X B.
Proof. For any > 0, setting f

() := f() 2
1
B, ) [[
2
, we notice
that the semi-convex f

attains its strict maximum at = 0.

In view of Lemmas 7.1 and 7.2, there are

, q

such that
f

() +q

, ) has a maximum at

, at which f is twice dierentiable.

It is easy to see that Df(

) 0 (as 0) and, moreover, from the

convexity of f() + (/2)[[
2
,
I D
2
f(

) B + 2I.
Noting (Df(

), D
2
f(

)) J
2,+
f(

) J
2,
f(

), we conclude the assertion

by taking the limit as 0. 2
We next give a magic property of sup-convolutions. For the readers
convenience, we put the proof of [6].
97
Lemma 7.4. (Lemma A.5 in [6]) For v USC(R
n
) with sup
R
n v <
and > 0, we set
v() := sup
xR
n
_
v(x)

2
[x [
2
_
.
For , q R
n
, Y S
n
, and (q, Y ) J
2,+
v(), we have
(q, Y ) J
2,+
v( +
1
q) and v() +
[q[
2
2
= v( +
1
q).
In particular, if (0, Y ) J
2,+
v(0), then (0, Y ) J
2,+
v(0).
Proof. For (q, Y ) J
2,+
v(), we choose y R
n
such that
v() = v(y)

2
[y [
2
.
Thus, from the denition, we see that for any x, R
n
,
v(x)

2
[ x[
2
v() v() +q, )
+
1
2
Y ( ), ) + o([ [
2
)
= v(y)

2
[y [
2
+q, )
+
1
2
Y ( ), ) + o([ [
2
).
Taking = x y + in the above, we have (q, Y ) J
2,+
v(y).
To verify that y = +
1
q, putting x = y and = (( y) + q)
for > 0 in the above again, we have
[( y) + q[
2
o(),
which concludes the proof. 2
Proof of Lemma 3.6. First of all, extending upper semi-continuous func-
tions u, w in into R
n
by in R
n
, we shall work in R
n
R
n
instead
of .
By translation, we may suppose that x = y = 0, at which u(x) + w(y)
(x, y) attains its maximum.
98
Furthermore, replacing u(x), w(y) and (x, y), respectively, by
u(x) u(0) D
x
(0, 0), x), w(y) w(0) D
y
(0, 0), y)
and
(x, y) (0, 0) D
x
(0, 0), x) D
y
(0, 0), y),
we may also suppose that (0, 0) = u(0) = w(0) = 0 and D(0, 0) = (0, 0)
R
n
R
n
.
Since (x, y) =
_
A
2
_
x
y
_
,
_
x
y
__
+o([x[
2
+[y[
2
), where A := D
2
(0, 0)
S
2n
, for each > 0, we see that the mapping (x, y) u(x) + w(y)
1
2
_
(A+ I)
_
x
y
_
,
_
x
y
__
attains its (strict) maximum at 0 R
2n
.
We will show the assertion for A+I in place of A. Then, sending 0,
we can conclude the proof. Therefore, we need to prove the following:
Simplied version of Ishiis lemma.
For upper semi-continuous functions u and w in R
n
, we suppose that
u(x) + w(y)
_
A
2
_
x
y
_
,
_
x
y
__
u(0) +w(0) = 0 in R
n
R
n
.
Then, for each > 1, there are X, Y S
n
such that (0, X) J
2,+
u(0),
(0, Y ) J
2,+
w(0) and ( +|A|)
_
I O
O I
_

_
X O
O Y
_
A+
1

A
2
.
Proof of the simplied version of Lemma 3.6. Since Holders inequality im-
plies
_
A
_
x
y
_
,
_
x
y
__

__
A+
1

A
2
__

_
,
_

__
+( +|A|)([x [
2
+[y [
2
)
for x, y, , R
n
and > 0, setting = +|A|, we have
u(x)

2
[x [
2
+ w(y)

2
[y [
2

1
2
__
A +
1

A
2
__

_
,
_

__
.
Using the notation in Lemma 7.4, we denote by u and w the sup-convolution
of u and w, respectively, with the above > 0. Thus, we have
u() + w()
1
2
__
A+
1

A
2
__

_
,
_

__
for all , R
n
.
99
Since u(0) u(0) = 0 and w(0) w(0) = 0, the above inequality implies
u(0) = w(0) = 0.
In view of Proposition 7.3 with m = 2n, f(, ) = u() + w() and
B = A+
1
A
2
, there is Z S
2n
such that (0, Z) J
2,+
f(0, 0) J
2,
f(0, 0)
and I Z B.
Hence, from the denition of J
2,
, it is easy to verify that there are X, Y
S
n
such that (0, X) J
2,+
u(0) J
2,
u(0), (0, Y ) J
2,+
w(0) J
2,
w(0), and
Z =
_
X O
O Y
_
.
Applying the last property in Lemma 7.4 to u and w, we see that
(0, X) J
2,+
u(0) and (0, Y ) J
2,+
w(0). 2
7.2 Proof of the ABP maximum principle
First of all, we remind the readers of our strategy in this and the next sub-
sections.
We rst show that the ABP maximum principle holds under f L
n
()
C() in Steps 1 and 2 of this subsection. Next, using this fact, we estab-
lish the existence of L
p
-strong solutions of Pucci equations in the next
subsection when f L
p
().
Employing this existence result, in Step 3, we nally prove Proposition
6.2; the ABP maximum principle when f L
n
().
ABP maximum principle for f L
n
() C() (Section 7.2)

Existence of L
p
-strong solutions of Pucci equations (Section 7.3)

ABP maximum principle for f L

n
() (Section 7.2)
Proof of Proposition 6.2. We give a proof in [5] for the subsolution asser-
tion of Proposition 6.2.
By scaling, we may suppose that diam() 1.
Setting
r
0
:= max

u max

u
+
,
we may also suppose that r
0
> 0 since otherwise, the conclusion is obvious.
100
We rst introduce the following notation: For u : R and r 0,

r
:=
_
x

p B
r
such that u(y) u(x) +p, y x) for y
_
.
Recalling the upper contact set in section 6.2, we note that
[u, ] =
_
r>0

r
.
Step 1: u C
2
() C(). We rst claim that for r (0, r
0
),
_
(i) B
r
= Du(
r
),
(ii) D
2
u O in
r
.
(7.1)
To show (i), for p B
r
, we take x such that u( x) p, x) =
max
x
(u(x) p, x)). Since u(x) u( x) r < r
0
for x , taking the
maximum over , we have x . Hence, we see p = Du( x), which concludes
(i).
For x
r
, Taylors formula yields
u(y) = u(x) +Du(x), y x) +
1
2
D
2
u(x)(y x), y x) + o([y x[
2
).
Hence, we have 0 D
2
u(x)(y x), y x) + o([y x[
2
), which shows (ii).
Now, we introduce functions g

(p) :=
_
[p[
n/(n1)
+
n/(n1)
_
1n
for > 0.
We shall simply write g for g

.
Thus, for r (0, r
0
), we see that
_
Du(r)
g(p)dp
_
r
g(Du(x))[det(D
2
u(x))[dx
=
_
r
_
[Du[
n/(n1)
+
n/(n1)
_
1n
[detD
2
u(x)[dx.
Recalling (7.1), we utilize [detD
2
u[ (trace(D
2
u)/n)
n
in
r
to nd
C > 0 such that
_
Br
g(p)dp C
_
r
_
[Du[
n/(n1)
+
n/(n1)
_
1n
(trace(D
2
u))
n
dx. (7.2)
Thus, since ([Du[ +f
+
)
n
g(Du)
1
(
n
+
n
(f
+
)
n
) by Holders inequality,
we have
_
Br
g(p)dp C
_
r
_

n
+
_
f
+

_
n
_
dx. (7.3)
101
On the other hand, since ([p[
n
+
n
)
1
g(p), we have
log
__
r

_
n
+ 1
_
C
_
Br
1
[p[
n
+
n
dp C
_
Br
g(p)dp.
Hence, noting
r

+
[u] for r (0, r
0
), by (7.3), we have
r
_
exp
_
C
_
[u,]
+
[u]
_

n
+
_
f
+

_
n
_
dx
_
1
_
1/n
. (7.4)
When |f
+
|
L
n
([,]
+
[u])
= 0, then sending 0, we get a contradiction.
Thus, we may suppose that |f
+
|
L
n
([,]
+
[u])
> 0.
Setting := |f
+
|
L
n
([u,]
+
[u])
and r := r
0
/2, we can nd C > 0,
independent of u, such that r
0
C|f
+
|
L
n
([u,]
+
[u])
.
Remark. We note that we do not need to suppose f to be continuous in
Step 1 while we need it in the next step.
Step 2: u C() and f L
n
() C(). First of all, because of f
C(), we remark that u is a standard viscosity subsolution of
T

(D
2
u) [Du[ f in
+
[u].
(See Proposition 2.9 in [5].)
Let u

be the sup-convolution of u for > 0;

u

(x) := sup
y
_
u(y)
[x y[
2
2
_
.
Note that u

is semi-convex and thus, twice dierentiable a.e. in R

n
.
We claim that for small > 0, u

is a viscosity subsolution of
T

(D
2
u

) [Du

[ f

in

, (7.5)
where f

(x) := supf
+
(y) [ [x y[ 2(|u|
L

()
)
1/2
and

:= x

+
[u] [ dist(x,
+
[u]) > 2(|u|
L

()
)
1/2
. Indeed, for x

and (q, X)
J
2,+
u

(x), choosing x such that u

(x) = u( x) (2)
1
[x x[
2
, we easily
verify that [q[ =
1
[ x x[ 2
_
|u|
L

()
/. Thus, by Lemma 7.4, we see
that (q, X) J
2,+
u(x + q). Hence, we have
T

(X) [q[ f
+
(x + q) f

(x).
102
We note that for small > 0, we may suppose that
r

:= max

max

(u

)
+
> 0. (7.6)
Here, we list some properties on upper contact sets: For small > 0, we
set

:= x [ dist(x, ) > .
Lemma 7.5. Let v

C(

) and v C() satisfy that v

v uniformly
on any compact sets in as 0. Assume that r := max

vmax

v
+
> 0.
Then, for r (0, r), we have the following properties:
_

_
(1)
r
[v, ] is a compact set in
+
[v],
(2) limsup
0

r
[v

]
r
[v, ],
(3) for small > 0, there is

such that
0<

r
[v

r
,
where

r
:= x [ dist(x,
r
[v, ]) < .
Proof of Lemma 7.5. To show (1), we rst need to observe that for r
(0, r), dist(
r
[v, ], ) > 0. Suppose the contrary; if there is x
k

r
[v, ]
such that x
k
x , then there is p
k
B
r
such that v(y)
v(x
k
) +p
k
, y x
k
) for y . Hence, sending k , we have
max

v max

v
+
r < r,
which is a contradiction. Thus, we can nd a compact set K such that

r
[v, ] K.
Moreover, if v(x) 0 for x
r
[v, ], then we get a contradiction:
r max

v r < r.
Next, choose x limsup
0

r
[v

]. Then, for any k 1, there are

k
(0, 1/k) and p
k
B
r
such that
v

k
(y) v

k
(x) +p
k
, y x) for y

k
.
We may suppose p
k
p for some p B
r
taking a subsequence if necessary.
Sending k in the above, we see that x
r
[v, ].
103
If (3) does not hold, then there are
0
> 0,
k
(0, 1/k) and x
k

r
[v

k
,

k
]

0
r
. We may suppose again that lim
k
x
k
= x for some x .
When x , since there is p
k
B
r
such that v

k
(y) v

k
(x
k
) +p
k
, y x
k
)
for y , we have r < r, which is a contradiction. Thus, we may suppose
that x and, then x
r
[v, ]. Thus, there is k
0
1 such that x
k

0
r
for k k
0
, which is a contradiction. 2
For > 0, we set u

:= u

, where

is the standard mollier. We set

,
r
:=
r
[u

] for r (0, r

), where r

:= max

max

(u

)
+
. Notice
that for small > 0, r

> 0.
In view of the argument to derive (7.2) in Step 1, we have
_
Br
g(p)dp C
_

,
r
_
[Du

[
n/(n1)
+
n/(n1)
_
1n
(trace(D
2
u

))
n
dx
for small r > 0.
Also, by the same argument for (ii) in (7.1), we can show that D
2
u

(x)
O in

,
r
. Furthermore, from the denition of u

, we verify that
1
I
D
2
u

(x) in

.
Hence, sending 0 with Lemma 7.5 (3), we have
_
Br
g(p)dp C
_
r[u

,]
_
[Du

[
n/(n1)
+
n/(n1)
_
1n
(trace(D
2
u

))
n
dx
C
_
r[u

,]
_

n
+
_
f

_
n
_
dx.
Therefore, sending 0 (again with Lemma 7.5 (3)), we obtain (7.4), which
implies the conclusion.
Remark. Using the ABP maximum principle in Step 2 (i.e. f C()), we
can give a proof of Proposition 6.3, which will be seen in section 7.3. Thus,
in Step 3 below, we will use Proposition 6.3.
Step 3: u C() and f L
n
(). Let f
k
C() be nonnegative func-
tions such that |f
k
f
+
|
L
n
()
0 as k .
In view of Proposition 6.3, we choose
k
C() W
2,n
loc
() such that
_

_
T
+
(D
2

k
) + [D
k
[ = f
k
f
+
a.e. in ,

k
= 0 on ,
|
k
|
L

()
C|f
k
f
+
|
L
n
()
.
104
Setting w
k
:= u +
k
|
k
|
L

()
, we easily verify that w
k
is an L
n
-viscosity
subsolution of
T

(D
2
w
k
) [Dw
k
[ f
k
in .
Note that
+
[w
k
]
+
[u].
Thus, by Step 2, we have
max

w
k
max

w
k
+ C|(f
k
)
+
|
L
n
(r[w
k
,]
+
[u])
.
Therefore, sending k with Lemma 7.5 (2), we nish the proof. 2
7.3 Proof of existence results for Pucci equations
We shall solve Pucci equations under the Dirichlet condition in . For sim-
plicity of statemants, we shall treat the case when is a ball though we will
need the existence result in smooth domains later. To extend the result for
general with smooth boundary, we only need to modify the function v
z
in
the argument below.
For 0 and f L
p
() with p n,
_
T

(D
2
u) [Du[ f in B
1
,
u = 0 on B
1
,
(7.7)
and
_
T
+
(D
2
u) + [Du[ f in B
1
,
u = 0 on B
1
.
(7.8)
Note that the rst estimate of (7.10) is valid by Proposition 6.2 when the
inhomogeneous term is continuous.
Sketch of proof. We only show the assertion for (7.8).
Step 1: f C

(B
1
). We shall consider the case when f C

(B
1
).
Set o
,
:= A := (A
ij
) S
n
[ I A I. We can choose a countable
set o
0
:= A
k
:= (A
k
ij
) o
,

k=1
such that o
0
= o
,
.
Noting that [q[ = maxb, q) [ b B

for q R
n
, we choose B
0
:=
b
k
B

k=1
such that B
0
= B

.
According to Evans result in 1983, we can nd classical solutions u
N

C() C
2
() of
_
_
_
max
k=1,...,N
_
trace(A
k
D
2
u) +b
k
, Du)
_
= f in B
1
,
u = 0 on B
1
.
(7.9)
105
Moreover, we nd = () (0, 1), C

> 0 (for each (0, 1)) and C

1
> 0,
which are independent of N 1, such that
|u
N
|
L

(B
1
)
C
1
|f|
L
n
(B
1
)
and |u
N
|
C
2,
(B
1
)
C

. (7.10)
Note that the rst estimate of (7.10) is valid by Proposition 6.2 when the
inhomogeneous term is continuous.
More precisely, by the classical comparison principle, Proposition 3.3, we
have
u
N
u
1
in B
1
. (7.11)
Furthermore, we can construct a subsoluion of (7.9) for any N 1 in
the following manner: Fix z B
1
. Set v
z
(x) := (e
|x2z|
2
e

), where
, > 0 (independent of z B
1
) will be chosen later. We rst note that
v
z
(z) = 0 and v
z
(x) 0 for x B
1
.
Setting L
k
w(x) := trace(A
k
D
2
w(x)) +b
k
, Dw(x)), we verify that
L
k
v
z
(x) 2e
|x2z|
2
(n 2[x 2z[
2
+ [x 2z[)
2e
9
(n 2 + 3).
Thus, xing := (n+3+1)/(2), we have L
k
v
z
(x) 2e
9
. Hence,
taking > 0 large enough so that 2e
9
|f|
L

(B
1
)
, we have
max
k=1,2,...,N
L
k
v
z
(x) f(x) in B
1
.
Now, putting V (x) := sup
zB
1
v
z
(x), in view of Theorem 4.2, we see that
V is a viscosity subsolution of
max
k=1,2,...,N
L
k
u(x) f(x) 0 in B
1
.
Moreover, it is easy to check that V

(x) = 0 for x B
1
. Thus, by Proposi-
tion 3.3 again, we obtain that
V u
N
in B
1
. (7.12)
Therefore, in view of (7.10)-(7.12), we can choose a sequence N
k
and
u C
2
(B
1
) such that lim
k
N
k
= ,
(u
N
k
, Du
N
k
, D
2
u
N
k
) (u, Du, D
2
u) uniformly in B
1
106
for each (0, 1), and
V u u
1
in B
1
. (7.13)
We note that (7.13) implies that u

= u

on B
1
.
By virtue of the stability result (Proposition 4.8), we see that u is a
viscosity solution of
T
+
(D
2
u) + [Du[ f = 0 in B
1
since sup
k1
trace(A
k
X) + b
k
, p) = T
+
(X) + [p[. Hence, Theorem 3.9
yields u C(B
1
).
Therefore, by Proposition 2.3, we see that u C(B
1
) C
2
(B
1
) is a
classical solution of (7.8).
Step 2: f L
p
(B
1
). (Lemma 3.1 in [5]) Choose f
k
C

(B
1
) such that
|f
k
f|
L
p
()
0 as k .
Let u
k
C(B
1
) C
2
(B
1
) be a classical solution of
T
+
(D
2
u) + [Du[ f
k
= 0 in B
1
such that u
k
= 0 on B
1
.
We rst claim that u
k

k=1
is a Cauchy sequence in L

(B
1
). Indeed,
since (1) and (4) of Proposition 3.2 imply that
T

(D
2
(u
j
u
k
)) [D(u
j
u
k
)[
T
+
(D
2
u
j
) +T

(D
2
u
k
) + [Du
j
[ [Du
k
[
= f
j
f
k
T
+
(D
2
u
j
) T
+
(D
2
u
k
) + [D(u
j
u
k
)[
T
+
(D
2
(u
j
u
k
)) +[D(u
j
u
k
)[,
using Proposition 6.2 when the inhomogeneous term is continuous, we
have
max
B
1
[u
j
u
k
[ C|f
j
f
k
|
L
n
(B
1
)
.
Recalling p n, we thus have
|u
j
u
k
|
L

(B
1
)
C|f
j
f
k
|
L
p
(B
1
)
.
Hence, we nd u C(B
1
) such that u
k
converges to u uniformly in B
1
as
k .
107
Therefore, by the standard covering and limiting arguments with weakly
convergence in W
2,p
locally, it suces to nd C > 0, independent of k 1,
such that
|u
k
|
W
2,p
(B
1/2
)
C.
Moreover, we see that C|f

|
L
p
(B
1
)
u C|f
+
|
L
p
(B
1
)
in B
1
.
For (0, 1/2), we select :=

C
2
(B
1
) such that
_

_
(i) 0 1 in B
1
,
(ii) = 0 in B
1
B
1
,
(iii) = 1 in B
12
,
(iv) [D[ C
0

1
, [D
2
[ C
0

2
in B
1
,
where C
0
> 0 is independent of (0, 1/2).
Now, we recall Caarellis result (1989) (see also [4]): There is a universal
constant

C > 0 such that
|D
2
(u
k
)|
L
p
(B
1
)


C|T
+
(D
2
(u
k
))|
L
p
(B
1
)
.
Hence, we nd C
1
> 0 such that for 0 < < 1/4,
|D
2
u
k
|
L
p
(B
12
)
|D
2
(u
k
)|
L
p
(B
1
)


C|T
+
(D
2
(u
k
))|
L
p
(B
3/4
)
C
1
_
|f
k
|
L
p
(B
1
)
+
1
|Du
k
|
L
p
(B
1
)
+
2
|u
k
|
L
p
(B
1
)
_
Multiplying
2
> 0 in the above, we get

2
|D
2
u
k
|
L
p
(B
12
)
C
1
(|f
k
|
L
p
(B
1
)
+
1
(u
k
) +
0
(u
k
)),
where
j
(u
k
) := sup
0<<1/2

j
|D
j
u
k
|
L
p
(B
1
)
for j = 0, 1, 2.
Therefore, in view of the interpolation inequality (see [13] for example),
i.e. for any > 0, there is C

> 0 such that

1
(u
k
)
2
(u
k
) + C

0
(u
k
),
we nd C
3
> 0 such that

2
(u
k
) C
3
_
|f
k
|
L
p
(B
1
)
+
0
(u
k
)
_
.
On the other hand, since we have L

-estimates for u
k
, we conclude the
proof. 2
Remark. It is possible to show that the uniform limit u in Step 2 is
an L
p
-viscosity solution of (7.8) by Proposition 6.13. Moreover, since it is
known that if L
p
-viscosity supersolution of (7.8) belongs to W
2,p
loc
(B
1
), then
it is an L
p
-strong supersolution (see [5]), u satises T
+
(D
2
u) +[Du[ = f(x)
a.e. in B
1
.
108
7.4 Proof of the weak Harnack inequality
We need a modication of Lemma 4.1 in [4] since our PDE (7.14) below has
the rst derivative term.
Lemma 7.6. (cf. Lemma 4.1 in [4]) There are C
2
(B
2

n
) and
C(B
2

n
) such that
_

_
(1) T

(D
2
) [D[ in B
2

n
,
(2) (x) 2 for x Q
3
,
(3) (x) = 0 for x B
2

n
,
(4) (x) = 0 for x B
2

n
B
1/2
.
Proof. Set
0
(r) := A1 (2

n/r)

for A, > 0 so that

0
(2

n) = 0.
Fig 7.1
y = (x)
y =
0
(x)
2

n
2

n
2
0
Q
3
1/2
r
1/2
Since
_
D
0
([x[) = A(2

n)

[x[
2
x,
D
2

0
([x[) = A(2

n)

[x[
4
[x[
2
I ( + 2)x x,
we caluculate in the following way: At x ,= 0, we have
T

(D
2

0
([x[)) [D
0
([x[)[ A(2

n)

[x[
2
( + 2) n [x[.
109
Setting :=
1
(n + 2

n) 2 so that > 0 for n 2, we see that

the right hand side of the above is nonnegative for x B
2

n
0. Thus,
taking C
2
(B
2

n
) such that (x) =
0
([x[) for x B
2

n
B
1/2
and
(x)
0
(3

n/2) for x B
3

n/2
, we can choose a continuous function
satisfying (1) and (4). See Fig 7.1.
Moreover, taking A := 2/(4/3)

1 so that
0
(3

n/2) = 2, we see
that (2) holds. 2
We now present an important cube decomposition lemma.
We shall explain a terminology for the lemma: For a cube

Q := Q
r
(x) with
r > 0 and x R
n
, we call Q a dyadic cube of

Q if it is one of cubes Q
k

2
n
k=1
so that Q
k
:= Q
r/2
(x
k
) for some x
k


Q, and
2
n
k=1
Q
k


Q
2
n
k=1
Q
k
.
Lemma 7.7. (Lemma 4.2 in [4]) Let A B Q
1
be measurable sets
and 0 < < 1 such that
(a) [A[ ,
(b) Assume that if a dyadic cube Q of

Q Q
1
satises [A Q[ > [Q[,
then

Q B.
Then, [A[ [B[.
Proof of Proposition 6.4. Assuming that u C(B
2

n
) is a nonnegative
viscosity supersolution of
T
+
(D
2
u) + [Du[ 0 in B
2

n
, (7.14)
we shall show that for some constants p
0
> 0 and C
1
> 0,
|u|
L
p
0(Q
1
)
C
1
inf
Q
1/2
u.
To this end, it is sucient to show that if u C(B
2

n
) satises that
inf
Q
1/2
u 1, then we have |u|
L
p
0(Q
1
)
C
1
for some constants p
0
, C
1
> 0.
Indeed, by taking v(x) := u(x)
_
inf
Q
1/2
u +
_
1
for any > 0 in place of u,
we have |v|
L
p
0(Q
1
)
C
1
, which implies the assertion by sending 0.
Lemma 7.8. There are > 0 and M > 1 such that if u C(B
2

n
) is a
nonnegative L
p
-viscosity supersolution of (7.14) such that
inf
Q
3
u 1, (7.15)
110
then we have
[x Q
1
[ u(x) M[ .
Remark. In our setting of proof of Proposition 7.4, assumption (7.15) is
automatically satised.
Proof of Lemma 7.8. Choose C
2
(B
2

n
) and C(B
2

n
) from Lemma
7.6. Using (4) of Proposition 3.2, we easily see that w := u + is an L
n
-
viscosity supersolution of
T
+
(D
2
w) + [Dw[ in B
2

n
.
Since inf
Q
3
w 1 and w 0 on B
2

n
by (2) and (3) in Lemma 7.6,
respectively, by Proposition 6.2, we nd

C > 0 such that
1 sup
Q
3
(w) sup
B
2

n
(w)

C||
L
n
([w,B
2

n
]B
+
2

n
[w])
. (7.16)
In view of (4) of Lemma 7.6, (7.16) implies that
1

C max
B
1/2
[[[x Q
1
[ (u + )(x) < 0[.
Since
u(x) (x) max
B
2

n
() =: M for x B
2

n
.
Therefore, setting = (

C sup
Q
1
[[)
1
> 0 and M = sup
B
2

n
() 2, we
have
[x Q
1
[ u(x) M[. 2
We next show the following:
Lemma 7.9. Under the same assumptions as in Lemma 7.8, we have
[x Q
1
[ u(x) > M
k
[ (1 )
k
for all k = 1, 2, . . .
Proof. Lemma 7.8 yields the assertion for k = 1.
Suppose that it holds for k 1. Setting A := x Q
1
[ u(x) > M
k
and
B := x Q
1
[ u(x) > M
k1
, we shall show [A[ (1 )[B[.
111
Since A B Q
1
and [A[ [x Q
1
[ u(x) > M[ := 1 , in
view of Lemma 7.8, it is enough to check that property (b) in Lemma 7.7
holds.
To this end, let Q := Q
1/2
j (z) be a dyadic cube of

Q := Q
1/2
j1 ( z) (for
some z, z Q
1
and j 1) such that
[A Q[ > [Q[ =
1
2
jn
. (7.17)
It remains to show

Q B.
Assuming that there is x

Q such that x / B; i.e. u( x) M
k1
.
Set v(x) := u(z + 2
j
x)/M
k1
for x B
2

n
. Since [ x
i
z
i
[ 3/2
j+1
, we
see that inf
Q
3
v u( x)/M
k1
1. Furthermore, since z Q
1
, z + 2
j
x
B
2

n
for x B
2

n
.
Fig 7.2
z
x

Q
Q
z
1
2
j
1
2
j1
z +
1
2
j
Q
3
Thus, since v is an L
p
-viscosity supersolution of
T
+
(D
2
v) + [Dv[ 0,
Lemma 7.8 yields [x Q
1
[ v(x) M[ . Therefore, we have
[x Q [ u(x) M
k
[

2
jn
= [Q[.
Thus, we have [Q A[ [Q[. Hence, in view of (7.17), we have
[Q[ = [A Q[ +[Q A[ > [Q[ + [Q[ = [Q[,
112
which is a contradiction. 2
Back to the proof of Proposition 6.4. A direct consequence of Lemma 7.9
is that there are

C, > 0 such that
[x Q
1
[ u(x) t[

Ct

for t > 0. (7.18)

Indeed, for t > M, we choose an integer k 1 so that M
k+1
t > M
k
.
Thus, we have
[x Q
1
[ u(x) t[ [x Q
1
[ u(x) > M
k
[ (1 )
k

C
0
t

,
where

C
0
:= (1 )
1
and := log(1 )/ log M > 0.
Since 1 M

for 0 < t M, taking

C := max

C
0
, M

, we obtain
(7.18).
Now, recalling Fubinis theorem,
_
Q
1
u
p
0
(x)dx
_
{xQ
1
| u(x)1}
u
p
0
(x)dx + 1
= p
0
_

1
t
p
0
1
[x Q
1
[ u(x) t[dt + 1,
(see Lemma 9.7 in [13] for instance), in view of (7.18), for any p
0
(0, ), we
can nd C(p
0
) > 0 such that |u|
L
p
0(Q
1
)
C(p
0
). 2
7.5 Proof of the local maximum principle
Although our proof is a bit technical, we give a modication of Trudingers
proof in [13] (Theorem 9.20), in which he observed a precise estimate for
strong subsolutions on the upper contact set. Recently, Fok in [11] (1996)
gave a similar proof to ours.
We note that we can nd a dierent proof of the local maximum principle
in [4] (Theorem 4.8 (2)).
Proof of Proposition 6.5. We give a proof only when q (0, 1] because it
is immediate to show the assertion for q > 1 by Holders inequality.
Let x
0
Q
1
be such that max
Q
1
u = u(x
0
). It is sucient to show that
max
B
1/4
(x
0
)
u C
2
|u
+
|
L
q
(B
1/2
(x
0
))
113
since B
1/2
(x
0
) Q
2
. Thus, by considering u((x x
0
)/2) instead of u(x), it
is enough to nd C
2
> 0 such that
max
B
1/2
u C
2
|u
+
|
L
q
(B
1
)
.
We may suppose that
max
B
1
u > 0 (7.19)
since otherwise, the conclusion is trivial.
Furthermore, by the continuity of u, we can choose (0, 1/4) such that
1 2 1/2 and
max
B
12
u > 0.
We shall consider the sup-convolution of u again: For (0, ),
u

(x) := sup
yB
1
_
u(y)
[x y[
2
2
_
.
By the uniform convergence of u

to u, (7.19) yields
max
B
1
u

> 0 for small > 0. (7.20)

For small > 0, we can choose := () (0, ) such that lim
0
= 0,
and
T

(D
2
u

) [Du

[ 0 a.e. in B
1
.
Putting

(x) := (1 )
2
[x[
2

for := 2n/q 2, we dene v

(x) :=

(x)u

(x). We note that

r

:= max
B
1
v

> 0.
Fix r (0, r

) and set

r
:=
r
[v

, B
1
]. By (1) in Lemma 7.5, we see

r
B
+
1
[v

].
For later convenience, we observe that
Dv

(x) = 2x(x)
(1)/
u

(x) + (x)Du

(x), (7.21)
D
2
v

(x) = 2(x)
(1)/
u

(x)I + x Du

(x) + Du

(x) x
+4( 1)(x)
(2)/
u

(x)x x + (x)D
2
u

(x).
(7.22)
114
Since u

is twice dierentiable almost everywhere, we can choose a mea-

surable set N

B
1
such that [N

[ = 0 and u

is twice dierentiable at
x B
1
N

. Of course, v

is also twice dierentiable at x B

1
N

.
By (7.22), we have
T

(D
2
v

) T

(D
2
u

) + 2
(1)/
nu

(x Du

+ Du

x)
in B
+
1
[v

]. By using (7.21), the last term in the above can be estimated

from above by
C
2/
(v

)
+
+
1/
[Dv

[.
Moreover, using (7.21) again, we have
T

(D
2
u

) [Du

[
1
[Dv

[ + C
1/
(u

)
+
.
Hence, we nd C > 0 such that
T

(D
2
v

) C
1/
[Dv

[ + C
2/
(v

)
+
=: g

in B
1
N

. (7.23)
We next claim that there is C > 0 such that
[Dv

(x)[ C
1/
(x)v

(x) for x

r
N

. (7.24)
First, we note that at x

r
N

, v

(y) v

(x) + Dv

(x), y x) for
y B
1
.
To show this claim, since we may suppose [Dv

(x)[ > 0 to get the esti-

mate, setting y := x tDv

(x)[Dv

(x)[
1
B
1
for t [1 [x[, 1
+[x[], we see that
0 = v

(y) v

(x) t[Dv

(x)[,
which implies
[Dv

(x)[ Cv

(x)
1/
(x) in

r
N

. (7.25)
Here, we use Lemma 2.8 in [5], which will be proved in the end of this
subsection for the readers convenience:
Lemma 7.10. Let w C() be twice dierentiable a.e. in , and satisfy
T

(D
2
w) g a.e. in ,
115
where g L
p
() with p n. If C
1
I D
2
w(x) O a.e. in for some
C
1
> 0, then w is an L
p
-viscosity subsolution of
T

(D
2
w) g in . (7.26)
Since u

is Lipschitz continuous in B
1
, by (7.22), we see that v

is an
L
n
-viscosity subsolution of
T

(D
2
v

) g

in B
1
.
Noting (7.25), in view of Proposition 6.2, we have
max
B
1
v

C|
2/
(v

)
+
|
L
n
(

r
)
C
_
max
B
1
(v

)
+
_2

|((u

)
+
)
2/
|
L
n
(B
1
)
,
which together with our choice of yields
max
B
1
v

C|(u

)
+
|
L
q
(B
1
)
.
Therefore, by (7.20), we have
max
B
1/2
u

C max
B
1
v

C|(u

)
+
|
L
q
(B
1
)
,
Therefore, sending 0 in the above, we nish the proof. 2
Proof of Lemma 7.10. In order to show that w C() is an L
p
-viscosity
subsolution of (7.26), we suppose the contrary; there are , r > 0, x and
W
2,p
loc
() such that 0 = (w )( x) = max

(w ), B
2r
( x) , and
T

(D
2
) g 2 a.e. in B
r
( x).
We may suppose that x = 0 . Setting (x) := (x) + [x[
4
for small
> 0, we observe that
h := T

(D
2
) g a.e. in B
r
.
Notice that 0 = (w )(0) > (w )(x) for x B
r
0.
116
Moreover, we observe
T

(D
2
(w )) a.e. in B
r
. (7.27)
Consider w

:= w

, where

is the standard mollier for > 0. From

our assumption, we see that, as 0,
_
(1) w

w uniformly in B
r
,
(2) D
2
w

D
2
w a.e. in B
r
.
By Lusins Theorem, for any > 0, we nd E

B
r
such that [B
r
E

[ < ,
_
Br\E
(1 +[T

(D
2
)[)
p
dx < ,
and
D
2
w

D
2
w uniformly in E

(as 0).
Setting h

:= T

(D
2
(w

)), we nd C > 0 such that

h

C +T

(D
2
)
because of our hypothesis. Hence, we have
|(h

)
+
|
p
L
p
(Br)
C
_
Br\E
(1 +[T

(D
2
)[)
p
dx +
_
E
[(h

)
+
[
p
dx.
Sending 0 in the above, by (7.27), we have
limsup
0
|(h

)
+
|
L
p
(Br)
C|(1 +[T

(D
2
)[)|
L
p
(Br\E)
C. (7.28)
On the other hand, in view of Proposition 6.2, we see that
max
Br
(w

) max
Br
(w

) + C|(h

)
+
|
L
p
(Br)
.
Hence, by sending 0, this inequality together with (7.28) implies that
0 = max
Br
(w ) max
Br
(w ) + C for any > 0.
This is a contradiction since max
Br
(w ) < 0. 2
117
References
[1] M. Bardi & I. Capuzzo Dolcetta, Optimal Control and Viscosity
Solutions of Hamilton-Jacobi-Bellman Equations, Systems & Control,
Birkauser, 1997.
[2] G. Barles, Solutions de Viscosite des

Equations de Hamilton-Jacobi,
Mathematiques et Applications 17, Springer, 1994.
[3] M. Bardi, M. G. Crandall, L. C. Evans, H. M. Soner & P. E.
Souganidis, Lecture Notes in Mathematics, 1660, Springer, 1997.
[4] L. A. Caffarelli & X. Cabre, Fully Nonlinear Elliptic Equations,
Colloquium Publications 43, AMS, 1995.
[5] L. A. Caffarelli, M. G. Crandall, M. Kocan & A.

Swiech,
On viscosity solutions of fully nonlinear equations with measurable in-
gredients, Comm. Pure Appl. Math. 49 (1996), 365-397.
[6] M. G. Crandall, H. Ishii & P.-L. Lions, Users guide to viscos-
ity solutions of second order partial dierential equations, Bull. Amer.
Math. Soc., 27 (1992), 1-67.
[7] M. G. Crandall, M. Kocan, P.-L. Lions & A.

Swiech, Existence
results for boundary problems for uniformly elliptic and parabolic fully
nonlinear equations, Electron. J. Dierential Equations, 1999 (1999),
1-22.
[8] L. C. Evans, Partial Dierential Equations, GSM 19, Amer. Math.
Soc., 1998.
[9] L. C. Evans & R. F. Gariepy, Measure Theory and Fine Properties
of Functions, Studies in Advanced Mathematics, CRC Press, 1992.
[10] W. H. Fleming & H. M. Soner, Controlled Markov Processes and
Viscosity Solutions, Applications of Mathematics 25, Springer, 1992.
[11] P.-K. Fok, Some maximum principles and continuity estimates for fully
nonlinear elliptic equations of second order, Dissertation, University of
California at Santa Barbara, 1996.
118
[12] Y. Giga, Surface Evolution Equations a level set method Preprint,
2002.
[13] D. Gilbarg & N. S. Trudinger, Elliptic Partial Dierential Equa-
tions of Second Order, 2nd edition, GMW 224, Springer, 1983.
[14] Q. Han & F. H. Lin, Elliptic Partial Dierential Equations, Courant
Institute Lecture Note 1, Amer. Math. Soc., 1997.
[15] H. Ishii, Viscosity Solutions, (in Japanese).
[16] R. Jensen, Uniqueness of Lipschitz extensions: Minimizing the sup
norm of the gradient, Arch. Rational Mech. Anal., 123 (1993), 51-74.
119
Notation Index
/, 46
B, 52
C
k,
, 79
C

, 78
C

0
, 80
, 52
ess. inf, 80
ess. sup, 80
G

, G

, 62
, 52
[u, O], 83
J
2,
, 17, 22
J
2,
, 18, 22
liminf
k

F
k
, 59
limsup
k

F
k
, 59
, 2
L
p
, 79
LSC, 15
/, 27
n, 66
o, 17
T

, 25
supp, 5
USC, 15
u

, u

, 40
W
2,p
, 79
W
2,p
loc
, 81
120
Index
Aleksandrov A. D., 95
Area formula, 95
Bellman equation, 8
Burgers equation, 4
Caarelli L. A., 80
-Crandall-Kocan-

Swi ech, 83
comparison principle
for boundary value problems,
63
classical , 24, 25
counter-example of , 68
for Dirichlet problems, 23, 66
for rst-order PDEs, 27, 29
for Neumann problems, 73
for second-order PDEs, 34, 38
for state constraint problems,
70
in unbounded domains, 76
continuity
of viscosity solutions, 41
convex, 6
cost functional, 46, 52
optimal , 46, 53
for state constraints, 69
Crandall M. G.
-Kocan-Lions-

Swi ech, 92
cube decomposition lemma, 110
De Giorgi E., 80
divergence form, 5
dyadic cube, 110
dynamic programming principle, 46,
53
eikonal equation, 9
elliptic
degenerate , 12
uniformly , 25
constants, 24
entropy condition, 5
Escauriaza L., 83
Evans L. C., 105
-Souganidis, 56
existence, 3
of L
p
-strong solutions, 84
of L
p
-viscosity solutions, 92
of viscosity solutions, 42
of Isaacs equations, 56
of state constraint problems,
69
of Bellman equations, 48
Fok P.-K., 113
fully nonlinear, 7
Hamilton-Jacobi equation, 6
Harnack inequality, 80, 85, 86, 88
, nal version, 87
weak , 81, 85, 109
Heaviside function, 20
Holder continuity, 79
of L
p
-viscosity solutions, 89,
92
homogeneous degree, 29
Hopf-Lax formula, 6
interior cone condition, 66
Isaacs equation, 8
Ishii H., 32, 40
-Koike, 71
121
s lemma, 32, 95
simplied , 99
-Lions, 36
Jensen R., 8, 32
s lemma, 95
Kruzkov S. N., 7
Krylov N. V.
-Safonov, 80
Laplace equation, 2
Lax-Oleinik formula, 5
L

-Laplacian, 8
Lions P.-L., 32
L
p
regularity, 79
maximum principle
Aleksandrov-Bakelman-Pucci ,
83, 100
local , 81, 85, 113
for semi-continuous functions,
15
mean curvature, 8
modulus of continuity, 27
mollier, 16
Moser J., 80
multi-index, 79
Nadirashvili N., 78
Neumann
condition, 72
problem, 73
non-anticipating strategy, 52
nondivergence form, 5
Poisson equation, 78
Puccis operator, 25
quasi-linear, 8
Rankine-Hugoniot condition, 5
Schauder regularity, 79
semi-concave, 7
semi-continuous envelope, 40
semi-convex, 95
semi-jet, 17
solution
classical , 3, 11
L
p
-strong , 81
L
p
-viscosity , 81
viscosity , 13
of semi-continuous PDEs, 59
semi-continuous , 40
for boundary value problems,
62, 63
weak , 3
distribution sense, 5
Soner H. M., 71
stability, 3
of L
p
-viscosity solutions, 93
of viscosity solutions, 59
structure condition, 33
in R
n
, 75
subsolution see solution, 11, 13, 40,
59, 62, 63, 81
supersolution see solution, 11, 13, 40,
59, 62, 63, 81
Trudinger N. S., 80, 113
uniform exterior sphere condition, 72
uniform exterior cone condition, 92
uniqueness, 3
for Dirichlet problem, 23
unit outward normal, 66
upper contact set, 83
value function, 46, 53
122
vanishing viscosity method, 10
123