Regularity results in free boundary
problems
GOHAR ALEKSANYAN
Doctoral Thesis
Stockholm, Sweden 2016
TRITA-MAT-A 2016:10
ISRN KTH/MAT/A-16/10-SE
ISBN 978-91-7729-161-9
KTH
Institutionen för Matematik
100 44 Stockholm
SWEDEN
Akademisk avhandling som med tillstånd av Kungl Tekniska högskolan
fram-lägges till offentlig granskning för avläggande av Filosofie doktorsexamen i
matematik fredagen den 2 december 2016 kl 13:00 i sal D3, Kungl Tekniska
högskolan, Lindstedtsvägen 5, Stockholm.
c
Gohar Aleksanyan, 2016
Tryck: Universitetsservice US AB
iii
In memory of my father and teacher,
Hakob Aleksanyan
v
Abstract
This thesis consists of three scientific papers, devoted to the
regu-larity theory of free boundary problems. We use iteration arguments to
derive the optimal regularity in the optimal switching problem, and to
analyse the regularity of the free boundary in the biharmonic obstacle
problem and in the double obstacle problem.
In Paper A, we study the interior regularity of the solution to the
optimal switching problem. We derive the optimal C
1,1-regularity of
the minimal solution under the assumption that the zero loop set is the
closure of its interior.
In Paper B, assuming that the solution to the biharmonic obstacle
problem with a zero obstacle is sufficiently close-to the one-dimensional
solution
16(x
n)
3+, we derive the C
1,α-regularity of the free boundary,
under an additional assumption that the noncoincidence set is an
NTA-domain.
In Paper C we study the two-dimensional double obstacle problem
with polynomial obstacles p
1≤ p
2, and observe that there is a new
type of blow-ups that we call double-cone solutions. We investigate
the existence of double-cone solutions depending on the coefficients of
p
1, p
2, and show that if the solution to the double obstacle problem with
obstacles p
1= −|x|
2and p
2= |x|
2is close to a double-cone solution,
then the free boundary is a union of four C
1,α-graphs, pairwise crossing
at the origin.
vi
Sammanfattning
Denna avhandling består av tre vetenskapliga artiklar som ägnas
åt regularitetsteori för frirandsproblem. Vi använder iterationsargument
för att härleda den optimala regulariteten hos optimalväxlingsproblemet
och för att analysera regulariteten av den fria randen i det biharmoniska
hinderproblemet och i dubbelhinderproblemet.
I Artikel A studerar vi den lokala regulariteten hos lösningen till
op-timalväxlingsproblemet. Under antagandet att nolloopmängden är
till-slutningen av sitt inre visar vi C
1,1-regularitet för den minimala
lös-ningen.
I Artikel B, antaget att lösningen till det biharmoniska
hinderpro-blemet med ett nollhinder ligger tillräckligt nära den endimensionell
lösning, härleder vi C
1,α-regularitet för den fria randen, under ett extra
antagande att positivitetsmängden är ett NTA-område.
I Artikel A och Artikel B utför vi analys i
R
nför n ≥ 2, medan i
Arti-kel C studerar vi ett tvådimensionellt hinderproblem. Det huvudsakliga
objektet för undersökningen i Artikel C är så kallade
dubbelkonlösning-ar. Vi visar att om lösningen till dubbelhinderproblemet med hinder
p
1= −|x|
2och p
2= |x|
2är nära en dubbelkonlösning, då består den
fria randen en union av fyra C
1,α-grafer, parvis korsande i origo.
Contents
Contents
vii
I Part: Introduction and summary
1
The obstacle problem
7
2
The double obstacle problem
11
2.1
Summary of Paper C . . . .
12
3
The optimal switching problem
17
3.1
Summary of Paper A . . . .
18
4
The biharmonic obstacle problem
23
4.1
Summary of Paper B . . . .
24
References
27
II Part: Scientific papers
Paper A
Optimal regularity in the optimal switching problem
Ann. I. H. Poincaré-AN (2015), http://dx.doi.org/10.1016/j.anihpc.2015.06.001
Paper B
Regularity of the free boundary in the biharmonic obstacle problem
Preprint: http://arxiv.org/abs/1603.06819
Paper C
Analysis of blow-ups for the double obstacle problem in dimension two
Acknowledgements
First and foremost I would like to thank my advisor, John Andersson, for
the mathematics I have learned from him. This thesis exists due to him. John
has been a great source of knowledge and inspiration to me. I am grateful for
everything.
Many thanks go to my co-advisor, Erik Lindgren, for his careful
proofread-ing of the articles included in the thesis and for all the discussions. I want
to thank Diogo Aguiar Gomes and Henrik Shahgholian for introducing me to
the optimal switching problem, and for fruitful discussions on the topic.
During the years of my doctoral studies I found many friends among the
fellow graduate students and young scientists. They have been very kind and
supportive; it is my pleasure and my duty to thank Mariusz Hynek, Jevgenija
Pavlova, my cheerful officemate Ludvig af Klinteberg, Gustav Sædén Ståhl
and many others.
I am grateful to my family for their love. Thank you for letting me travel
the world and impatiently awaiting each of my arrivals at Zvartnots
interna-tional airport. My special thanks go to my brother, Sargis Aleksanyan, for all
our conversations about life and mathematics.
CONTENTS
1
Common notations
N
the set of natural numbers
R
the set of real numbers
R
nthe n-dimensional Euclidiean space
x · y
Pni=1x
iy
i, the scalar product of vectors x, y ∈
R
nB
r(x)
{y ∈
R
n: |y − x| < r}, the open ball in
R
n∂A
the boundary of the set A ⊂
R
nA
the closure of the set A ⊂
R
nA
0the interior of the set A ⊂
R
nA
cthe complement of the set A ⊂
R
na
+(a
+
)
max(a, 0), for a ∈
R
a
−(a
−)
max(−a, 0), for a ∈
R
∇
the gradient vector, ∇u :=
∂x∂u1
, ...,
∂u ∂xn
for a scalar function u
∆
the Laplace operator, ∆u :=
Pni ∂∂x2u2i
for a scalar function u
∆
2the biharmonic operator, ∆
2u := ∆(∆u)
D
2u
the Hessian matrix for u
Part I
Introduction and summary
CONTENTS
5
Introduction
This thesis consists of three articles, and all of them are devoted to the
regu-larity theory of free boundary problems.
The aim of Part I is to provide a brief overview of the obstacle problem
and of the free boundary problems that are investigated in Part II. In Part I
we also summarize the results obtained in the thesis, outlining the background
and techniques used. The core of the thesis is Part II, which is a combination
of three papers, named Papers A, B and C.
In Paper A we investigate the optimal regularity of solutions to the optimal
switching problem (see also Chapter 3). The problem is related both to the
classical obstacle problem (Chapter 1) and to the double obstacle problem
(Chapter 2), while the regularity of solutions to the optimal switching problem
is not covered by the known regularity theory for any of the above mentioned
problems. In fact, we are dealing with a system of obstacle-type equations,
where the obstacles depend on the solution.
Paper A also includes an example of a homogeneous global solution to the
two dimensional double obstacle problem, where the free boundary is a cross.
This example shows a new type of behavior of the free boundary for the double
obstacle problem, which does not happen in the case of the obstacle problem.
Paper C analyses the blow-ups of the solution to the double obstacle problem
in dimension two, with an emphasis on the so-called double-cone solutions.
In Paper B we use a linearization argument and show that if the solution
to the biharmonic obstacle problem is close to a halfspace solution, then the
free boundary is locally a C
1,α-graph. The method does not require the use
of comparison principles and regularity of the solution. For the biharmonic
obstacle problem there are no optimal growth or nondegeneracy properties
known. The comparison principle for harmonic functions plays an important
role in the analysis of solutions to the obstacle problem. While there are
no comparison principles for biharmonic functions, as one can easily see by
looking at the following one-dimensional example; f (x) = x
2in
R, f is a
nonnegative biharmonic function, obtaining a minimum at an interior point
x = 0. Hence we make an additional assumption that the noncoincidence
set is a non-tangentially accessible (NTA) domain. Chapter 4 contains a brief
introduction to the biharmonic obstacle problem and an overview of Paper B.
Chapter 1
The obstacle problem
The obstacle problem is a motivating example in the theory of variational
in-equalities and free boundary problems. In this chapter we give the variational
formulation of the problem and a short survey on the known regularity
the-ory of the obstacle problem. This presentation is far from being exhaustive.
We have selected the properties of the obstacle problem which have a strong
relation to the problems studied in the main text.
The chapter is based on the article [5]. The text is also supported by the
book [11].
Let Ω ⊂
R
nbe a given domain with a smooth boundary, and let ϕ ∈ C
2(Ω)
be a given function, called an obstacle. We are looking for the minimizer
u ∈ W
1,2(Ω) to the Dirichlet functional
J [u] =
Z
Ω
|∇u(x)|
2dx,
(1.1)
over admissible functions u ≥ ϕ in Ω, with boundary condition u = g on
∂Ω, where g is a continuous function, satisfying g ≥ ϕ on ∂Ω. Since J is a
strictly convex functional over a convex admissible set, there exists a unique
minimizer. The minimizer u is called the solution to the obstacle problem.
A variational argument easily verifies that u is a superharmonic function in
Ω, and therefore it has a lower semi-continuous equivalent, which we call u.
Denote by
Ω
u:= {x ∈ Ω; u(x) > ϕ(x)},
(1.2)
then Ω
uis an open set, called the noncoincidence set. Denote the free
bound-ary for the obstacle problem by
Γ
u:= ∂Ω
u∩ Ω.
(1.3)
We see that Γ
udepends on the solution u which is not known a priori,
ex-plaining the usage of the word ‘free’.
8
CHAPTER 1. THE OBSTACLE PROBLEM
In the set Ω
u, the function u does not touch the obstacle ϕ, and it locally
minimizes the Dirichlet energy, (1.1). Therefore u is a harmonic function in
Ω
u. Hence we obtain the following Euler-Lagrange equation for the obstacle
problem,
min(−∆u, u − ϕ) = 0.
(1.4)
It has been shown (see for instance [5]) that the solution to the obstacle
problem u is as regular as the obstacle ϕ up to C
1,1, which is the best regularity
that the solution may achieve, since ∆u jumps from zero to ∆ϕ across the
free boundary.
Next we turn to to the regularity of the free boundary in the obstacle
problem. First let us rewrite equation (1.4) in a more convenient form. Denote
by v := u − ϕ, then according to (1.4), v is a nonnegative function solving
∆v = −∆ϕχ
{v>0},
(1.5)
where χ
Ais the characteristic function of a set A ⊂
R
n. It follows from (1.4)
that ∆ϕ ≤ 0 in the coincidence set {u = ϕ}. Taking ∆ϕ ≡ −1, we study the
normalized obstacle problem,
∆u = χ
{u>0}.
(1.6)
It is easy to verify that u minimizes the following energy functional
˜
J [u] =
Z Ω1
2
|∇u(x)|
2+ u(x)dx,
(1.7)
over nonnegative functions with given boundary condition.
Let u be the solution to the normalized obstacle problem in Ω ⊂
R
n,
B
1(0) ⊂⊂ Ω. For a fixed x
0∈ Γ
u∩B
1and r < 1 small, satisfying B
r(x
0) ⊂ B
1,
consider the following rescalings,
u
x0,r(x) :=
u(rx + x
0)
r
2.
(1.8)
The abbreviation u
r:= u
0,ris used. Let us note that the rescalings u
x0,rsolve
the obstacle problem (1.6) in the ball B
1r
(0).
By using comparison principle, we can show the following quadratic growth
estimate
sup
Br(x0)u ≤ C
nr
2,
(1.9)
where x
0∈ Γ
u, B
2r(x
0) ⊂ Ω, and C
nis a dimensional constant. Hence the
rescalings u
x0,rare bounded,
sup
B19
for small r > 0.
Furthermore, at free boundary points the solution cannot decay faster than
quadratically. Let, x
0∈ Ω
u∩ Ω, then
sup
Br(x0)u ≥
1
2n
r
2
,
(1.11)
which can be derived by applying the maximum principle for the harmonic
function v = u −
2n1|x − x
0|
2in Ω
u∩ B
r(x
0). Inequality (1.11) implies that
rescalings u
x0,rpossess the following nondegenaracy property,
sup
B1u
x0,r≥
1
2n
.
(1.12)
From (1.10) we can obtain that ku
x0,rk
C1,1(B1/2)is uniformly bounded.
Hence through a subsequence u
x0,rconverges to a function u
0in C
1,α
(B
1/2),
where 0 < α < 1. Let us emphasize that, at this point we do not have any
tools which would guarantee the existence of the lim
r→0+u
x0,rin any function
space. Any limit of the rescaling u
x0,rover a sequence r
j→ 0+ as j → ∞ is
called a blow-up of the solution u. The blow-ups are defined in
R
n.
Further-more, the blow-ups are global solutions, i.e. solutions in
R
nto the normalized
obstacle problem (1.6). It follows from the nondegenaracy property (1.12) and
the C
1,α-convergence that the blow-ups are not identically zero.
The analysis of the possible blow-ups and the proof of the uniqueness of
the blow-up are far from being easy, and we cannot include it in this short
overview of the obstacle problem. Let us only state the following important
theorem.
Theorem 1.1. Let u be the solution to the normalized obstacle problem, (1.6),
and let x
0∈ Γ
ube a free boundary point. Then the blow-up of u at x
0is unique,
that is the limit u
0(x) = lim
j→∞u(rjx+x0)
r2
j
does not depend on the sequence
r
j→ 0+. Furthermore, there are only two types of possible blow-ups; either
u
0is a half-space solution,
u
0(x) =
1
2
(x · e)
2 +, where e ∈
R
nis a unit vector,
(1.13)
or u
0is a polynomial solution,
u
0(x) =
1
2
(x · Ax), where A is an n × n nonnegative
symmetric matrix with trA = 1.
10
CHAPTER 1. THE OBSTACLE PROBLEM
If u
x0,r→
1 2
(x · e)
2
+
, then the vector e is the approximate unit normal to Γ
uat x
0, and x
0is called a regular point. It has been shown that almost every
free boundary point is regular, and that the set of regular points is a relatively
open subset of Γ
u. Furthermore, in a neighborhood of a regular point, the free
boundary is a C
1,α-graph, moreover, it is real analytic.
If the blow-up of u at x
0is a polynomial, then x
0is called a singular point.
It has been shown that the set of singular points is a closed set contained in
a lower dimensional C
1manifold.
Chapter 2
The double obstacle problem
Let Ω be a bounded open set in
R
nwith smooth boundary. The solution to
the double obstacle problem in Ω is the minimizer of the functional
J (u) =
Z
Ω
|∇u(x)|
2dx
over functions u ∈ W
1,2(Ω), ψ
1≤ u ≤ ψ
2, satisfying the boundary condition
u = g on ∂Ω. For the problem to be well defined we assume that ψ
1≤ ψ
2in
Ω, and ψ
1≤ g ≤ ψ
2on ∂Ω. The functions ψ
1and ψ
2are called respectively
the lower and the upper obstacles.
If u < ψ
2, then u solves the obstacle problem with ψ
1, and therefore
−∆u ≥ 0 by (1.4). Also observe that if u > ψ
1, then −u solves the obstacle
problem with −ψ
2, and ∆u ≥ 0 by (1.4). Therefore we may conclude that u
satisfies the following inequalities,
ψ
1≤ u ≤ ψ
2, ∆u ≥ 0 if u > ψ
1and ∆u ≤ 0 if u < ψ
2.
(2.1)
It has been shown that the solution to the double obstacle problem is locally
C
1,1under the assumption ψ
i∈ C
2(Ω), see for instance [3, 8]. Hence we can
rewrite (2.1) as follows,
ψ
1≤ u ≤ ψ
2and ∆u = ∆ψ
1χ
{u=ψ1}
+ ∆ψ
2χ
{u=ψ2}− ∆ψ
1χ
{ψ1=ψ2}a.e.,
where χ
Ais the characteristic function of a set A ⊂
R
n. Let us observe that
∆u = ∆ψ
1in the set {u = ψ
1} ∩ {u < ψ
2}, hence (2.1) implies that ∆ψ
1≤ 0,
and ψ
1is a superharmonic function. By a similar argument, we obtain that
ψ
2is a subharmonic function in the set {u = ψ
2} ∩ {u > ψ
1}.
Since the optimal regularity of the solution is known, our main interest
becomes the analysis of the free boundary. Denote by
Ω
1:= {u > ψ
1}, Ω
2:= {u < ψ
2}, and Ω
12:= Ω
1∩ Ω
2,
(2.2)
12
CHAPTER 2. THE DOUBLE OBSTACLE PROBLEM
then Ω = Ω
1∪ Ω
2∪ Λ, where Λ := {x ∈ Ω : ψ
1(x) = ψ
2(x)}. We call Ω
12the noncoincidence set. It follows from (2.1) that ∆u ≥ 0 and ∆u ≤ 0 in Ω
12.
Hence u is a harmonic function in the noncoincidence set, as in the ”single”
obstacle problem.
Define the free boundary for the double obstacle problem
Γ
u:= ∂Ω
12∩ Ω,
(2.3)
and let
Γ
i:= ∂Ω
i∩ Ω, i = 1, 2, then Γ
u⊂ Γ
1∪ Γ
2.
(2.4)
Assume that
ψ
1≤ ψ
2, and ψ
i∈ C
2(Ω).
(2.5)
If ψ
1< ψ
2, then Γ
1
∩ Γ
2= ∅, and x
0∈ Γ
uimplies that x
0∈ Γ
1or x
0∈ Γ
2. Let
x
0∈ Γ
1, then there exists a small ball B
r(x
0) such that B
r(x
0)∩{u = ψ
2} = ∅.
Hence u solves the obstacle problem with obstacle ψ
1in the ball B
r(x
0), and
the known regularity theory for the obstacle problem can be applied to analyse
the free boundary in a neighborhood of x
0. Therefore we assume that
Λ = {x ∈ Ω : ψ
1(x) = ψ
2(x)} 6= ∅.
(2.6)
Let x
0∈ Γ
ube a free boundary point, if x
0∈ Λ, then in a neighborhood
/
of x
0, we have ψ
1< ψ
2, this case has already been discussed above.
Assume that x
0∈ Γ
u∩ Λ. First let us observe that x
0cannot be an interior
point of Λ. Otherwise, if x
0∈ Λ
0, then there exists B
r(x
0) ⊂ Λ, which is the
same as ψ
1= ψ
2in B
r(x
0). Since ψ
1≤ u ≤ ψ
2, we obtain u = ψ
1= ψ
2in B
r(x
0), hence x
0∈ Γ
/
u. So we are interested in the behavior of the free
boundary at the points x
0∈ ∂Λ ⊂ Γ
u. First we want to understand the
behavior of the free boundary when x
0∈ Λ is an isolated point. Paper C
partially answers this question in dimension two with polynomial obstacles.
2.1
Summary of Paper C
In Paper C we study the following normalized double obstacle problem in
dimension two,
∆u = λ
1χ
{u=p1}+ λ
2χ
{u=p2}(2.7)
with polynomial obstacles p
1≤ p
2, where λ
1
= ∆p
1≤ 0 and λ
2= ∆p
2≥ 0
are constants. We assume that the obstacles p
1and p
2meet at a single point
x
0, i.e. p
1(x) = p
2(x) iff x = x
0.
The work is inspired by the following example in
R
2,
2.1. SUMMARY OF PAPER C
13
It is easy to see that u
0solves the double obstacle problem (2.1), with obstacles
p
1(x) = −x
21
− x
22and p
2(x) = x
21+ x
22. Observe that Λ = {0}, and the free
boundary consists of two lines crossing at the origin. The function u
0is a
motivational example for double-cone solutions, which are the novelty of this
paper.
Let u be the solution to the normalized double obstacle problem (2.7).
Assume that x
0= 0, we can always come to this situation with a change
of variables. By subtracting a first order polynomial from p
1, p
2and u, and
taking into account that u ∈ C
1,1, we get
u(0) = p
1(0) = p
2(0) = 0, and ∇u(0) = ∇p
1(0) = ∇p
2(0) = 0.
Hence
p
1(x) = a
1x
21+ 2b
1x
1x
2+ c
1x
22and p
2
(x) = a
2
x
21+ 2b
2x
1x
2+ c
2x
22.
Furthermore, we may assume that b
1= b
2= 0, by rotating the coordinate
system and subtracting a harmonic polynomial from p
1, p
2and from u, thus
obtaining
p
1(x) = a
1x
21+ c
1x
22and p
2
(x) = a
2
x
21+ c
2x
22.
(2.9)
Observe that p
1, p
2are second order homogeneous polynomials; which means
p
1(rx) = r
2p
1(x), and p
2(rx) = r
2p
2(x).
(2.10)
As in the classical obstacle problem, any limit of
u(rx)r2as r → 0+, is called
a blow-up of the solution u to the double obstacle problem at the origin.
We show that the blow-ups of a solution to the normalized double obstacle
problem are homogeneous of degree two functions via Weiss’ monotonicity
formula. This result is not surprising at all, since we already saw in Theorem
1.1 that the blow-ups of the solution to the obstacle problem are homogeneous
degree two functions.
Knowing that the blow-ups are homogeneous global solutions, we make
a complete characterization of possible blow-ups in dimension n = 2.
In
particular we see that there exist blow-ups of a new type. We call these
solutions double-cone solutions, since the noncoincidence set is a union of two
cones with a common vertex at the origin. We show that there exist
double-cone solutions if and only if the following polynomial
P = P (x
1, x
2) ≡ p
1(x
1, x
2) + p
2(x
2, x
1) = (a
1+ c
2)x
21+ (a
2+ c
1)x
22(2.11)
has zeroes other than x = 0. If P ≡ 0, there are infinitely many double-cone
solutions, and if P 6≡ 0, but P = 0 on a line, there are finitely many
double-cone solutions. This result is quite surprising and unexpected. In the next
paragraph we describe how (2.11) affects the stability of the free boundary.
14
CHAPTER 2. THE DOUBLE OBSTACLE PROBLEM
Let ε be an arbitrary number, |ε| << 1. Then for polynomials p
1(x) =
−x
21
− x
22, p
2(x) = x
21+ x
22there exist infinitely many double-cone solutions.
While when we look at the double obstacle problem with p
1= −x
21
− x
22and
˜
p
2= (1 − ε)x
21
+ (1 + ε)x
22there are only four double-cone solutions, and
for p
1= −x
21
− x
22and ¯
p
2= (1 + ε)x
21+ (1 + ε)x
22there are none. This
property reveals the instability of the double obstacle problem in the sense
that changing the obstacles slightly, may change the solution and the free
boundary significantly.
In this paper we also study the regularity of the free boundary in the case
when P ≡ 0. Since in this case we have infinitely many rotational invariant
double-cone solutions, we can use a flatness improvement argument to show
that the free boundary is locally a union of four C
1,α-graphs. We have already
used a similar argument in Paper B, see also Chapter 4.
The case P ≡ 0, can be reduced to the case p
1= −|x|
2, and p
2= |x|
2. A
general double-cone solution in this case can be written in polar coordinates
as follows,
µ = µ
φ1,φ2(r, θ) :=
r
2,
if − φ
2≤ 2θ ≤ φ
1r
2cos(2θ − φ
1),
if φ
1≤ 2θ ≤ π + φ
1r
2cos(2θ + φ
2),
if − π − φ
2≤ 2θ ≤ −φ
2−r
2,
otherwise,
(2.12)
where 0 ≤ φ
1, φ
2≤ π are parameters describing a double-cone solution µ.
The main result in Paper C is the following theorem.
Theorem 2.1 (Theorem 4.7 in Paper C). Let u be the solution to the
two-dimensional double obstacle problem with obstacles p
1= −x
21− x
22
and p
2=
x
21
+ x
22. Assume that ku − µk
L2(B2)
= δ is sufficiently small, where µ is a
double-cone solution, and is not a halfspace solution. Then in a small ball B
r0the free boundary consists of four C
1,γ- graphs meeting at the origin, denoted
by Γ
1+, Γ
1−, Γ
2+, Γ
2−. Neither Γ
1= Γ
1+∪ Γ
−
1
nor Γ
2= Γ
2+∪ Γ
−2
has a normal
at the origin. The curves Γ
1+and Γ
2+cross at a right angle, the same is true
for Γ
1−and Γ
2−.
We show that if the solution is close to a double-cone solution in B
1,
then the blow-up at the origin is unique. Furthermore, employing the known
regularity theory for the free boundary in the obstacle problem, we derive
that the free boundary Γ for the double obstacle problem is a union of four
C
1,γ-graphs meeting at the origin. Neither Γ
1nor Γ
2is flat at the origin, and
2.1. SUMMARY OF PAPER C
15
x
1x
2Γ
2+Γ
1+Γ
2−Γ
1−∆u = 0
u = p
2u = p
1∆u = 0
S
1S
2x
0 ν(0) ν(x0) ν(0)Figure 2.1: The behavior of the free boundary, with obstacles touching at a single point
We mentioned that if P 6≡ 0, but P = 0 on a line, there are finitely many
double-cone solutions. We aim to describe the behavior of the free boundary
in this case in a future publication.
Chapter 3
The optimal switching problem
Let Ω ⊂
R
nbe a bounded domain with smooth boundary. Consider a power
plant with several modes (states) of energy production in Ω. When running
the production mode 1 ≤ i ≤ m, we pay a running cost f
i(x) depending also
on the present position x ∈ Ω. We can switch from mode i to another mode
1 ≤ j ≤ m in order to minimize the cost of energy production, but we have
to pay a switching cost ψ
ij(x) depending also on the present state x. Assume
that ψ
ii= 0, which is reasonable; if we do not switch, there is no need to pay
a switching cost.
The solution to the optimal switching problem is a vector valued function
u = (u
1, u
2, ..., u
m) and it represents the cost of an optimal strategy for energy
production. As it has been discussed in the literature [10], [4], the negative
of the solution to the optimal switching problem, −u, solves the following
system:
min(−L
iu
i+ f
i, min
j
(u
i
− u
j+ ψ
ij)) = 0, in Ω,
(3.1)
where L
iis an elliptic operator, corresponding to the mode i. Equation (3.1)
says that if min
j(u
i− u
j+ ψ
ij) > 0, it is not optimal to switch to any other
mode j 6= i, and we continue running mode i.
For the optimal switching problem to be well defined, we need to impose
the nonnegative loop condition: Let i
0, i
1, . . . , i
l= i
0be any loop of length
l, i.e. including l number of states. Assume that (u
1, u
2, ..., u
m) is a solution
to system (3.1), then u
i− u
j+ ψ
ij≥ 0 for any i, j ∈ {1, 2, ..., m}, and after
summing the equations over the loop, we get
l X
j=1
ψ
ij−1,ij≥ 0.
(3.2)
Condition (3.2) is a necessary assumption for the existence of a solution to
(3.1). Without (3.2) we can make the energy production cost arbitrary small
18
CHAPTER 3. THE OPTIMAL SWITCHING PROBLEM
by looping. For example, if l = 2 and ψ
12(x) + ψ
21(x) < 0 for some x, then
at the point x we can switch from mode 1 to mode 2, and immediately switch
back to mode 1, thus paying a negative cost. This way we can decrease the
cost as much as we want by switching many times. Therefore the assumption
ψ
12+ ψ
21≥ 0 is needed in order to obtain a finite cost function.
The uniqueness and C
1,1-regularity of the solution to system (3.1) have
been studied in the literature under the assumption that the switching costs
ψ
iare nonnegative constants, [7], [10], [4].
In Paper A we consider a system, arising in a model optimal switching
problem with only two states, i. e. m = 2, and L
1= L
2= ∆,
min(−∆u
1+ f
1, u
1− u
2+ ψ
1) = 0
min(−∆u
2+ f
2, u
2− u
1+ ψ
2) = 0,
(3.3)
with given Dirichlet boundary conditions u
i= g
ion ∂Ω. The switching costs
ψ
1, ψ
2are given functions, satisfying the nonnegative loop assumption (3.2),
which is the same as
ψ
1+ ψ
2≥ 0,
(3.4)
since m = 2. Denote by
L := {x ∈ Ω; ψ
1(x) + ψ
2(x) = 0},
and call it a zero loop set.
3.1
Summary of Paper A
In Paper A our aim is to investigate if solutions to (3.3) are C
1,1. Assume that
f
1, f
2∈ C
α, and ψ
1, ψ
2∈ C
2,α, for some 0 < α < 1.
(3.5)
Then there exist v
1, v
2∈ C
2,αloc
solving the Poisson equation ∆v
i= f
iin Ω.
Denote u
i0
:= u
i− v
i, then u
i0is as regular as u
iup to C
2,α, and (u
10, u
20) solves
the following system
min(−∆u
1, u
1− u
2+ ϕ
1) = 0,
min(−∆u
2, u
2− u
1+ ϕ
2) = 0,
(3.6)
where ϕ
1= v
1− v
2+ ψ
1and ϕ
2= v
2− v
1+ ψ
2are the new switching cost
functions preserving the loop condition, since ϕ
1+ ϕ
2≡ ψ
1+ ψ
2.
We see that u
1solves the obstacle problem with the obstacle u
2− ϕ
1, and
3.1. SUMMARY OF PAPER A
19
the obstacle u
1− ϕ
2. Thus (3.7) is a system of obstacle-type equations, where
the obstacles depend on the solution (u
1, u
2). Hence the regularity of the
solution to system (3.6) cannot be obtained directly by the known regularity
theory for the obstacle problem. We see that (u
1, u
2) cannot have a better
regularity than C
1,1which is the best regularity for the obstacle problem.
If
L = ∅, that is ϕ
1+ϕ
2> 0, then for any x ∈ Ω, u
1(x)−u
2(x)+ϕ
1(x) > 0
or u
2(x) − u
1(x) + ϕ
2(x) > 0. Otherwise, if u
1(x) − u
2(x) + ϕ
1(x) = 0 and
u
2(x) − u
1(x) + ϕ
2(x) = 0, we obtain ϕ
1(x) + ϕ
2(x) = 0, contradicting
L = ∅.
Observe that if u
1− u
2+ ϕ
1> 0, then u
1is a harmonic function, and u
2solves
the obstacle problem with the obstacle u
1− ϕ
2. Hence we may conclude that
u
1is an infinitely differentiable function, and u
2∈ C
loc1,1.
Thus, if
L = ∅, we obtain (u
1, u
2) ∈ C
1,1loc
by the regularity theory for the
obstacle problem. This result is interesting, but not enough, since we want to
know if the C
1,1-regularity holds in case
L 6= ∅.
In
L we have that ϕ
1= −ϕ
2, hence we obtain from (3.6) that u
1−u
2+ϕ
1≥
0 and u
2− u
1− ϕ
1≥ 0, which implies u
1− u
2+ ϕ
1≡ 0 in L . Take any
subharmonic function u
1∈ W
2,p, for n < p < ∞ and let u
2= u
1+ ϕ
1, then
(u
1, u
2) solves (3.6) in
L . Hence system (3.6) may not have a unique solution
if
L 6= ∅, and (3.6) admits solutions that are not C
1,1. Instead we consider
the following system,
min(−∆u
1, u
1− u
2+ ϕ
1) = 0,
min(−∆u
2, u
2− u
1+ ϕ
2) = 0,
min(−∆u
1, −∆u
2) = 0,
(3.7)
which already has a unique solution. By using comparison principles, we show
that the solution to system (3.7) is the smallest among the solutions to (3.6).
The third equation in system (3.7) has a natural implication in the optimal
switching setting; it says that we always run one of the modes i = 1 or i = 2,
even when we can switch for free. Let us also observe that the extra (third)
equation in (3.7) holds automatically if
L = ∅.
Now let us study the relation between the system (3.7) and the double
obstacle problem. It is easy to see that U = u
1− u
2solves the following
double obstacle problem
−ϕ
1≤ U ≤ ϕ
2, −∆U ≤ 0 a.e. if U > −ϕ
1, −∆U ≥ 0 a.e. if U < ϕ
2(3.8)
with obstacles −ϕ
1≤ ϕ
2. Assuming that ϕ
i∈ C
2,α, by the known regularity
theory for the double obstacle problem, we obtain that U ∈ C
loc1,1. Thus the
difference u
1− u
2∈ C
1,1, while the question if u
1∈ C
1,1loc
and u
2∈ C
1,1 locis not
covered by the regularity theory of the double obstacle problem.
20
CHAPTER 3. THE OPTIMAL SWITCHING PROBLEM
Let us sketch the main results in Paper A. Consider system (3.7), and let
Ω
1:= {−∆u
1> 0}, Ω
2:= {−∆u
2> 0}, and Ω
12:= Ω \ Ω
1∪ Ω
2be disjoint
open sets. Then
−∆u
1= ∆ϕ
1> 0, −∆u
2= 0 in Ω
1−∆u
2= ∆ϕ
2> 0, −∆u
2= 0 in Ω
2−∆u
1= 0, −∆u
2= 0 in Ω
12.
(3.9)
In the set Ω \ Ω
1we get −∆u
1= 0, and therefore u
2solves the obstacle
problem
min(−∆u
2, u
2− u
1+ ϕ
2) = 0,
with a C
2,α-obstacle u
1− ϕ
2. Hence u
2∈ C
1,1loc
(Ω \ Ω
1). By a similar argument
we obtain u
1∈ C
1,1loc
(Ω \ Ω
2).
It remains to study the regularity of the solution in a neighborhood of the
set ∂Ω
1∩ ∂Ω
2∩ Ω. Note that ∂Ω
1∩ ∂Ω
2⊂ L , since u
1− u
2+ ϕ
1= 0 in Ω
1and u
2− u
1+ ϕ
2= 0 in Ω
2.
Let us study the regularity of (u
1, u
2) in the interior of
L , denoted by
L
0. We have already observed before that u
2= u
1+ ϕ
1in
L . Hence we
obtain from the third equation in system (3.7) that min(−∆u
1, −∆u
2) =
min(−∆u
1, −∆u
1− ∆ϕ
1) = 0, therefore u
1solves the following equation,
−∆u
1= (∆ϕ
1)
+in
L
0,
(3.10)
and u
2= u
1+ ϕ
1in
L . By the classical regularity theory for elliptic
equa-tions, solutions to equation (3.10) are locally C
2,α, if ∆ϕ
1∈ C
α. Thus in a
neighborhood of the points x ∈ ∂Ω
1∩ ∂Ω
2∩ L
0, the solution (u
1, u
2) ∈ C
2,α.
The main difficulty is to study the regularity of (u
1, u
2) at the points
x
0∈ ∂Ω
1∩ ∂Ω
2∩ ∂L , called ‘meeting’ points.
Let u be a twice continously differentiable function in B
1(x
0), having
Hölder continuous second order derivatives, i. e. u ∈ C
2,α(B
1(x
0)), for some
0 < α < 1. Let p
x0be the second degree Taylor polynomial for u at x
0, then
sup
x∈Br(x0)|u(x) − p
x0(x)| ≤ C
nkD
2
uk
Cαr
2+α,
(3.11)
where C
nis a dimensional constant. Estimate (3.11) can be obtained easily
by using the mean-value property for continously differentiable functions and
that D
2u ∈ C
α.
On the other hand if u is a continuous function in Ω, and for every x
0∈ Ω
there exists a polynomial p
x0such that
sup
x∈Br(x0)|u(x) − p
x0(x)| ≤ M r
3.1. SUMMARY OF PAPER A
21
for a uniform constant M > 0, then u is twice continously differentiable, and
kD
2uk
Cα
≤ c
nM .
Our main result in Paper A is the following theorem.
Theorem 3.1 (Theorem 4 in Paper A). Assume ϕ
1, ϕ
2∈ C
2,α, and
L = L
0.
Then the solution to the system (3.7), (u
1, u
2) is C
2,α-regular on ∂Ω
1
∩ ∂Ω
2∩
∂
L ∩ Ω, in the sense that for every x
0∈ ∂Ω
1∩ ∂Ω
2∩ ∂L ∩ Ω, there exist
second order polynomials p
1 x0, p
2 x0, such that
sup
x∈Br(x0)|u
i(x) − p
ix0(x)| ≤ Cr
2+α(3.13)
where the constant C > 0 depends only on the given data.
The idea of the proof comes from [2], where the authors derive the optimal
regularity for the no-sign obstacle problem. Let u ∈ W
2,p, for p < ∞, if ∆u is
a bounded function, it does not imply that D
2u is bounded, but it does imply
that D
2u ∈ BM O. First, we show that ∆u
1and ∆u
2are bounded functions,
then use the BM O-argument to derive D
2u
i∈ L
∞.
Let 0 ∈ ∂Ω
1∩ ∂Ω
2∩ ∂L
0, be a meeting point. Since ϕ
i∈ C
2,α, we get
∆ϕ
1(0) + ∆ϕ
2(0) = 0. According to (3.9), ∆ϕ
1≥ 0 in Ω
1
and ∆ϕ
2≥ 0 in Ω
2,
hence ∆ϕ
1(0) = ∆ϕ
2(0) = 0.
Employing the BM O-estimates for D
2u
1and D
2u
2, we show that the
polynomials
p
1r(x) := (u
1)
r+ (∇u
1)
r· x +
1
2
x · (D
2u
1)
r· x and
p
2r(x) := (u
2)
r+ (∇u
2)
r· x +
1
2
x · (D
2u
2)
r· x
converge to harmonic polynomials, denoted respectively by p
10and p
20. We also
describe the rate of convergence,
sup
x∈Br(0)|p
ir(rx) − p
i0(rx)| ≤ Cr
2+α,
where C is just a constant. Considering the following rescalings;
v
ri(x) =
u
i
(rx) − p
i 0(rx)
r
2+α,
we study the corresponding system for (v
1r
, v
r2). Taking into account that
∆ϕi(rx)rα
is uniformly bounded for i = 1, 2, we show that rescalings (v
1
r
, v
2r) are
uniformly bounded in the ball B
1, which is equivalent to (3.13).
22
CHAPTER 3. THE OPTIMAL SWITCHING PROBLEM
It follows from Theorem A that the solution to system (3.7) is locally C
1,1if
L = L
0.
In the end of the article we justify our assumption x
0∈ ∂L
0with a
counterexample. By considering a particular system in
R
2, with
L = {0},
we find an explicit solution, which is not C
1,1in any neighborhood of the
origin. This example reveals that the regularity of the solution to the optimal
switching problem depends on the topological properties of the zero loop set.
We saw that at the so-called meeting points |D
2u
i(rx)| is of order r
α. While if
0 is an isolated point of
L , we may have |D
2u(rx)| ≈ − ln r as r → 0+. This
is a new result, and it shows how different the optimal switching problem is
from the obstacle or double obstacle problems.
Let us also mention that Paper C was born from the example in Paper A,
although we gave the summary of Paper C in the previous chapter.
Chapter 4
The biharmonic obstacle problem
Let Ω ⊂
R
nbe a given domain, and ϕ ∈ C
2(Ω), ϕ ≤ 0 on ∂Ω be a given
function, called an obstacle. Consider the problem of minimizing the following
functional
J [u] =
Z
Ω
(∆u(x))
2dx,
(4.1)
over all functions u ∈ W
02,2(Ω), such that u ≥ ϕ. The functional J admits
a unique minimizer, called the solution to the biharmonic obstacle problem
with obstacle ϕ.
A variational argument easily verifies that ∆u is a weakly subharmonic
function. Furthermore, ∆u is a harmonic function in the noncoincidence set
{u > ϕ}. Therefore the solution satisfies the following variational inequality
∆
2u ≥ 0, u ≥ ϕ, ∆
2u · (u − ϕ) = 0.
It has been shown in [6] and [9] that the solution u ∈ W
loc3,2(Ω), ∆u ∈
L
∞loc(Ω), and moreover u ∈ W
loc2,∞(Ω).
Furthermore, in the paper [6], the
authors show that in dimension n = 2 the solution u ∈ C
2(Ω) and that the
free boundary Γ
u:= ∂{u = ϕ} lies on a C
1-curve in a neighbourhood of the
points x
0∈ Γ
u, such that ∆u(x
0) > ∆ϕ(x
0).
The setting of our problem is slightly different from the one in [6] and
[9]. We consider a zero-obstacle problem with general nonzero boundary
con-ditions. We look for a minimizer to the functional (4.1) over the admissible
set
A :=
(u ∈ W
2,2(Ω), u ≥ 0, u = g > 0,
∂u
∂ν
= f on ∂Ω
).
The minimizer u exists, and it is unique. The minimizer is called the solution
to the biharmonic obstacle problem. As in the obstacle problem, we denote
the free boundary by
Γ
u:= ∂Ω
u∩ Ω, where Ω
u:= {u > 0}.
23
24
CHAPTER 4. THE BIHARMONIC OBSTACLE PROBLEM
There are several important questions regarding the biharmonic obstacle
problem that remain open. For example, the optimal regularity of the solution,
the characterization of blow-ups at free boundary points, etc..
In Paper B we focus on the regularity of the free boundary for an
n-dimensional biharmonic obstacle problem, assuming that the solution is close
to the one-dimensional solution
16(x
n)
3+. We saw in Chapter 1 that the solution
to the obstacle problem at a regular point has a unique blow-up of the form
1 2
(x
+
n
)
2in some coordinate sytem. A flatness improvement argument shows
that if the solution to the obstacle problem is close to the halfspace solution
1 2
(x
+
n
)
2, then the free boundary is locally a graph of a C
1,α-function. We expect
to obtain a similar result for the biharmonic obstacle problem assuming that
the solution is close to the one-dimensional solution c(x
+n
)
3. We choose c = 1/6
so that ∇∆
16x
3n
= e
n, but we could have as well as chosen a different positive
constant.
4.1
Summary of Paper B
In [1], using flatness improvement argument, the authors show that the free
boundary in the p-harmonic obstacle problem is a C
1,αgraph in a
neighbor-hood of the points where the solution is almost one-dimensional. We apply
the same technique in order to study the regularity of the free boundary in
the biharmonic obstacle problem.
In the first step we study one-dimensional solutions.
Assume that u
0solves the biharmonic obstacle problem with zero obstacle in an open
inter-val (−a, a) ⊂
R, and let the origin be a free boundary point. By a direct
computation we see that u
0is one of the following functions; cx
3+, cx
3−or cx
3,
where c is a positive constant. If u(x) = u
0(x
n) = c(x
n)
3+, for x ∈
R
n, then
the noncoincidence set for u is the halfspace {x ∈
R
n, x
n> 0}, and the free
boundary is the plane {x
n= 0}.
Let us choose c =
16, and introduce the class
B
%κ
(ε) of solutions to the
biharmonic obstacle problem, that are close to the one-dimensional solution
1 6
(x
+ n
)
3.
Definition 4.1. Let u ≥ 0 be the solution to the biharmonic obstacle problem
in a domain Ω ⊂
R
n, B
2
⊂⊂ Ω and assume that 0 ∈ Γ
uis a free boundary
point. We say that u ∈
B
%κ
(ε), if the following assumptions are satisfied:
1. u is almost one dimensional, that is
k∇
0uk
W2,2(B 2)≤ ε,
4.1. SUMMARY OF PAPER B
25
2. The set Ω
u:= {u > 0} is a non-tangentially accessible (NTA) domain
with constants r
0= M
−1= %.
3. There exists 2 > t > 0, such that u = 0 in B
2∩ {x
n< −t}.
4. We have the following normalization
kD
3uk
L2(B 1)=
1
6
D
3(x
n)
3+ L2(B 1)=
|B
1|
1 22
12:= ω
n,
(4.2)
and we also assume that
kD
3uk
L2(B 2)< κ,
(4.3)
where κ >
16D
3(x
n)
3+ L2(B 2)= 2
n2ω
n.
Assumption 1 states that the solution is flat in the x
n-direction. This can
be compared to the flatness assumption in [1].
We show that the precise value of the parameter t in assumption 3 is not
very important. The normalization in point 4 corresponds to our choice of
c =
16, and it can always be achieved by a renormalization.
For the biharmonic obstacle problem there are no optimal growth or
non-degeneracy properties known, that could help us avoid additional assumptions
on the set Ω
u. Hence we make assumption 2 in the definition of
B
%κ(ε). The
NTA-domain assumption is not very strong, and it holds for a wide class of
domains. However, it is not easy to verify directly that the noncoincidence set
is an NTA-domain.
Evidently
16(x
n)
3+∈ B
κ%(ε), for any ε > 0 and % > 0. Our first step is to
show that if u ∈
B
%κ
(ε), with ε > 0 small, then u ≈
16(x
n)
3+