PDE Notes Miw04!28!08

DRAFT: April 28, 2008
1
1
Lectures on PDEs- APMA- E6301
DRAFT IN PROGRESS
Michael I. Weinstein
April 28, 2008
Contents
1 First order partial dierential equations and the method of characteristics 4
2 The Laplacian - Laplaces and Poissons equations 4
2.1 The Newtonian potential in R
n
. . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Boundary value problems and Greens functions . . . . . . . . . . . . . . . . 4
2.2.1 Greens function for the ball in R
n
. . . . . . . . . . . . . . . . . . . 4
2.2.2 Greens function for the upper half plane in R
2
, R
n
. . . . . . . . . . 4
2.3 Single and Double Layer Potentials and their boundary limits . . . . . . . . 4
2.4 Boundary integral formulation of the Dirichlet and Neumann Problems In-
terior and Exterior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.5 Fredholm Operators in Banach Spaces, Solvability of Fredholm Integral Equa-
tions on C
0
(S) and the Fredholm Alternative . . . . . . . . . . . . . . . . . 4
2.6 Application of Fredholm Integral Operator Theory to the solution of the
Dirichlet and Neumann Problems . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Distributions and Fundamental Solutions 5
4 Introduction to Hilbert Space 7
4.1 Representation of Linear Functionals on H . . . . . . . . . . . . . . . . . . 10
5 Applications of Hilbert Space Theory to the Existence of Solutions of
Poissons Equation 12
5.1 Generalizations of the Weak Dirichlet Problem . . . . . . . . . . . . . . . . 14
5.2 Relation between weak and strong (classical) solutions of elliptic PDE . . . . 15
5.3 Introduction to the Finite Element Method . . . . . . . . . . . . . . . . . . 15
Department of Applied Physics and Applied Mathematics, Columbia University, New York, NY 10027;
miw2103@columbia.edu
2
CONTENTS DRAFT: April 28, 2008
6 Operators on Hilbert spaces 18
6.1 Bounded operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.2 Compact operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
7 Applications of compact operator theory to eigenfunction expansions for
elliptic operators 21
8 Introduction to Variational Methods 24
8.1 Variational Problem 1: Dirichlets Principle . . . . . . . . . . . . . . . . . . . 24
8.2 Variational Problem 2: Smallest eigenvalue of . . . . . . . . . . . . . . . 24
8.3 Direct methods in the calculus of variations . . . . . . . . . . . . . . . . . . 26
8.4 Ideas and general discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
8.5 Compactness and minimizing sequences . . . . . . . . . . . . . . . . . . . . . 27
9 The Heat / Diusion Equation 29
9.1 Remarks on solutions to the heat / diusion equation . . . . . . . . . . . . . 30
9.2 Inhomogeneous Heat Equation - DuHamels Principle . . . . . . . . . . . . . 31
9.3 Heat equation on a bounded domain, . . . . . . . . . . . . . . . . . . . . . 32
9.4 Energy inequalities and heat equation . . . . . . . . . . . . . . . . . . . . . . 33
9.5 The maximum principle for the heat / diusion equation . . . . . . . . . . . 33
9.6 Application: Burgers equation and Shock Waves . . . . . . . . . . . . . . . 33
10 The Wave Equation 33
10.1 One dimensional wave equation . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.2 Three dimensional wave equation . . . . . . . . . . . . . . . . . . . . . . . . 34
10.3 2D wave equation - Hadamards method of descent . . . . . . . . . . . . . . 36
10.4 Principle of causality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
10.5 Inhomogeneous Wave Equation - DuHamels Principle . . . . . . . . . . . . . 37
10.6 Initial boundary value problem for the wave equation . . . . . . . . . . . . . 37
11 The Schrodinger equation 38
11.1 Quantum mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
11.2 Diraction of classical waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
11.3 Free Schr odinger - initial value problem . . . . . . . . . . . . . . . . . . . . . 39
11.4 Free Schr odinger in L
p
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
11.5 Free Schr odinger evolution of a Gaussian wave packet . . . . . . . . . . . . . 41
11.6 The Uncertainty Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
12 Background 43
12.1 Linear Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
12.2 Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
12.3 Local existence theorem for ordinary dierential equations (ODEs) . . . . . 45
12.4 Compactness and convergence results . . . . . . . . . . . . . . . . . . . . . . 46
12.5 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
12.6 Fourier inversion on o and L
2
. . . . . . . . . . . . . . . . . . . . . . . . . . 48
3
1 First order partial dierential equations and the method
of characteristics
Consider the rst order PDE
F(x, u(x), Du(x)) = 0, Du(x) = (
x
1
u, . . . .
xn
u) . (1.1)
Here, F(x, z, p) denotes a smooth real-valued function of 2n + 1 variables x R
n
, p R
n
and z R
1
, with x = (x
1
, . . . , x
n
), p = (p
1
, . . . , p
n
).
Transport equation:
t
u(x, t) + v u(x, t) = 0, x R
n
, t R
1
.
Linear rst order equations: b(x) Du(x) + c(x)u(x) = 0
Quasilinear rst order equations: b(x, u(x)) Du(x) + c(x, u(x)) = 0
Fully nonlinear equations: equations which are nonlinear in Du(x), e.g. The eikonal
equation, [u(x)[
2
= 1.
2 The Laplacian - Laplaces and Poissons equations
2.1 The Newtonian potential in R
n
2.2 Boundary value problems and Greens functions
2.2.1 Greens function for the ball in R
n
2.2.2 Greens function for the upper half plane in R
2
, R
n
2.3 Single and Double Layer Potentials and their boundary limits
2.4 Boundary integral formulation of the Dirichlet and Neumann
Problems Interior and Exterior
2.5 Fredholm Operators in Banach Spaces, Solvability of Fred-
holm Integral Equations on C
0
(S) and the Fredholm Alterna-
tive
2.6 Application of Fredholm Integral Operator Theory to the so-
lution of the Dirichlet and Neumann Problems
4
3 Distributions and Fundamental Solutions
1
Denition 3.1 Let C
0
() or C
c
() denote the set of compactly supported C
functions.
C
0
() = f : there exists K compact subset of , f(x) 0, x K
One often denotes C
0
() by T() or T
2
. Functions in T are often called test functions.
Denition 3.2 A linear functional, T, on T is linear mapping which associates to any
T a number T[], i.e.
T[ + ] = T[] + T[]
for any , scalars and any , T.
Examples: (a) Distributions obtained from locally integrable functions: Let f L
1
loc
be any
locally integrable function
3
. Dene
T
f
[] =
_
R
n
(x)f(x) dx (3.1)
Then, T
f
is a linear functional on T. For simplicity, we will often write f[] instead of T
f
[].
(b) The Dirac delta function:
[] = () (3.2)
denes a linear functional on T.
Continuity of linear functionals on T, notion of distribution
Denition 3.3 Consider an arbitrary sequence of functions
k
in T, which tend to zero
in the sense of the following two conditions: (a)
k
vanish outside a xed compact subset,
K, of . and (b) For any multi-indx , lim
k
k
(x) = 0 uniformly in x K.
A linear functional T is continuous if lim
k
T[
k
] = 0 for any sequence
k
, which
tends to zero in the above sense. A continuous linear functional on T is called a distribution.
The set of distributions on T() is denoted by T
t
().
Examples: T
f
, dened for any locally integrable f and
are examples of distributions.

Dierentiation of distributions:
1
References: Chapter 3, F. John PDEs; Section E of Chapter 0 in G.B. Folland Introduction to PDEs
2
Here, is taken to be an open set. A compact set is a set which is a set which is closed and bounded.
3
f L
1
loc
means that for any compact set C,
_
C
[f(x)[dx < .
5
Notation: A multi-index is a vector of non-negative integers = (
1
, . . . ,
m
) N
m
0
. The
order of a multi-index is given by [[
1
+ +
m
.
f(x) =
1
x
1

m
xm
f(x
1
, . . . , x
m
).
Note that if f, T, then
_

f(x)(x) dx = (1)
[[
_
f(x)
(x) dx, (3.3)

which we can write as
f[] = (1)
[[
f[
]. (3.4)
Equation (3.4) motivates our denition of the derivative of a distribution. Let T T
t
. Then,
T[] = (1)
[[
T[
] (3.5)
Distribution solutions to PDEs
Let L denote the linear partial dierential operator L =
I
a
(x)
. Suppose we have a
PDE
Lu = T, (3.6)
where T T
t
. Given the above denition of derivative of a distribution, we say that the
PDE has a distribution solution u, provided
u[L
t
] = T[], T. (3.7)
Here, L
t
=
I
(1)
[[
( a
(x) (x) ) denotes the adjoint of the dierential operator

L, which satises the generalization of (3.3): for , T
_
L(x) (x) dx =
_
(x) L
t
(x) dx
Fundamental Solutions
Denition 3.4 A fundamental solution of a linear dierential operator L with pole at x is
a distribution solution, u, of Lu =
x
.
Examples: Laplace Operator L = . L
t
= . We have shown that a fundamental
solution for with pole at x is:
(x y) =
1
2
log [x y[, n = 2
=
1
(2 n)
n
[x y[
2n
, n 3
Exercise: In one space dimension, solve
d
2
dx
2
=
x
where
x
denotes the Dirac delta function (distribution).
6
4 Introduction to Hilbert Space
Denition 4.1
45
(a) S is a vector space or linear space: u, v S implies, for all , scalars that u+v
S, plus vector spaces axioms.
(b) S is an inner product space if it is a linear space and there is a function
(, ) : S S R
u, v S (u, v) R (4.1)
with the following properties
(P1) (u, v) = (v, u)
(P2) (u + v, w) = (u, v) + (v, w), , R.
(P3) (u, u) 0 and (u, u) = 0 = u = 0.
Example 1a: S = R
n
. u, v R
n
. (u, v) =
n
i=1
u
i
v
i
.
Example 2a: S = C
(), R
n
bounded open, (f, g) =
_
u(x) v(x) dx.

Denition 4.2 A normed linear space is a vector space equipped with a notion of length,
called the norm, | |, satisfying the following properties
(n1) |u| = 0 implies u = 0
(n2) |u| = [[ |u|, R
(n3) |u + v| |u| +|v|
Given an inner product space, S, there is a natural norm, or measure of size of a vector in S
|u| =
_
(u, u) (4.2)
4
References: We seek to give the required working knowledge of the basic functional analysis, e.g. Hilbert
space, linear operators, spectral theory, . . . , with a view toward its application to PDEs. See, for example,
(a) L.C. Evans, PDEs, Appendix D on Linear Functional Analysis and Appendix E on Measure and Inte-
gration; (b) F. John, PDEs, Springer 4th Ed., Chapter 4, section 5; G.B. Folland, Introduction to PDEs,
Princeton University Press, Chapter 0. A good text on elementary notions in analysis (convergence, uniform
convergence, integration,. . . is the text: W. Rudin, Principles of Mathematical Analysis, McGraw Hill. a
more advanced text is that of E. Lieb and M. Loss, Analysis, Am. Math. Soc.
5
Throughout we will construct spaces of functions, with notions of size and distance within these spaces
dened in terms of integrals. These integrals are assumed throughout to be taken in the Lebesgue sense, an
important extension notion of Riemann integral. Note that if f is integrable in the sense of Riemann, it is
in the sense of Lebesque and the values of these integrals is the same. Appendix E of L.C. Evans PDEs
contains a very terse treatment of measure theory and the Lebesgue integral.
7
Any norm satises the Cauchy Schwarz inequality
[(u, v)[ |u| |v| (4.3)
Example 1b: S = R
n
. u, v R
n
. |u| =
_
k=1
[u
k
[
2
.
Example 2b: S = C
(), R
n
bounded open. |u|
2
=
_ _
[u(x)[
2
dx
_1
2
.
Convergence:
The sequence u
j
j1
in S converges to a point u
S provided
lim
j
|u
j
u
| = 0. (4.4)
In this case we say that the sequence u
j
j1
converges strongly to u
or converges to u
in
norm. Later, we shall introduce the notion of weak convergence.
Cauchy sequence:
The sequence u
j
j1
in S is called a Cauchy sequence if |u
j
u
k
| 0 as j, k .
Completeness:
The normed linear space S is complete if every Cauchy sequence in S converges to some
element of S. A complete normed linear space is called a Banach space. A complete normed
linear space, whose norm derives from an inner product, as in (4.2), is called a Hilbert space.
Completion Theorem: Every normed linear space can be completed. Specically, if S is
a normed linear space, then it is possible to naturally extend S to a possibly larger set,

S,
and the norm | | to be dened on all

S, in such a way that the set

S, equipped with the
norm | |. is complete. The completion is constructed by identifying all Cauchy sequences
with points in the extended space

S. If v

S but v / S, then there is a sequence v
k
in S,
such that |v
k
v| 0 as k . We say that the space S is dense in

S.
Example 1c: Let Q denote the set of rational numbers, numbers of the form p/q, where
p and q are integers and q ,= 0. Consider the inner product space S = Q
n
. u, v Q
n
.
|u| =
_
k=1
[u
k
[
2
, the Euclidean norm. While Q
n
equipped with this norm is not
complete, Q
n
has a completion, namely R
n
equipped with the norm |u|.
Example 2c: C
() is not complete with respect to the norm |u|

2
=
_
[u(x)[
2
dx. There
are many important completions of C
() with respect to dierent norms:

(1) The Hilbert Space L
2
(): The completion of C
() with respect to the norm | |

2
is called L
2
(). Since the norm on L
2
derives from an inner product, f, g
_
fg, L
2
is a Hilbert space.
8
(2) Sobolev Spaces - H
k
(), k 0: Dene the inner product of two functions f, g H
1
by
(f, g)
H
1 =
_
f g + fg dx (4.5)
The H
1
Sobolev norm is dened by:
|u|
2
H
1 = (u, u)
H
1 =
_
[u(x)[
2
+ [u(x)[
2
dx (4.6)
More generally, for any integer k 0, dene H
k
() to be the completion of C
()
with respect to the norm
|u|
2
H
k
=
[[k
_
u(x)[
2
dx (4.7)
Note that H
0
() = L
2
().
(3) Sobolev Spaces - H
k
(R
n
), k 0: Replace the bounded open set, , with = R
n
.
First consider C
0
(R
n
) and dene H
k
(R
n
) to be the completion of C
0
(R
n
) with respect
to the H
k
(R
n
) norm:
|u|
2
H
k
=
[[k
_
R
n
[
u(x)[
2
dx (4.8)
Remark 4.1 Sobolev spaces are important in analysis of functions largely due to their role
in relating global (integral) properties and local (pointwise) properties. For example, here is
the most elementary
Sobolev inequality: There is a constant K > 0 such that for any f C
0
(R) we have
[f(x)[
2
K
_ _
[f(x)[
2
+ [f
t
(x)[
2
dx
_
= K |f|
2
H
1 (4.9)
Proof: Exercise
Weak convergence: A sequence u
j
j1
in H converges to a limit u
in H if for all w H
(u
j
, w)
1
(u
, w)
1
(4.10)
Compactness A closed subset, C, of H is compact if any sequence u
j
j1
in C has a
convergent subsequence.
Remark 4.2 Any closed and bounded subset of C
n
is compact. However, in innite dimen-
sional spaces, e.g. L
2
, this is not the case. See section 12.4.
9
4.1 Representation of Linear Functionals on H DRAFT: April 28, 2008
4.1 Representation of Linear Functionals on H
(1) A linear functional on S is a mapping : S R, u (u), such that for all
u, v S and all scalars , we have
(u + v) = (u) + (v).
One sometimes represents the action of on a vector u with notation
(u) = , u).
(2) A bounded linear functional is a linear functional for which there is a constant, M > 0,
such that for any u S
[(u)[ M |u| (4.11)
The set of bounded linear functionals on a Hilbert Space, H, is denoted H
and is called the

dual space of H. The smallest constant, M, for which the inequality (4.11) holds is called
the norm of the linear functional , ||
1
.
Example 1d: For any vector v R
n
, (u) = (u, v) = u v =
n
k=1
u
k
v
k
is a bounded linear
functional with norm ||
(R
n
)
= |v|
R
n. Moreover, it is straightforward to see that given any
linear functional on R
n
, , there is a unique vector v
R
n
such that (u) = (u, v
) for all
u R
n
.
What about bounded linear functionals on Hilbert space? Let H denote any Hilbert
space and let v be a xed vector in H. Then, in analogy with the construction in Example
1d, () = (, v) denes a bounded linear functional on H with norm |v|. The joke is that
every bounded linear functional on a Hilbert space is obtained in this way:
Theorem 4.1 Riesz Representation Theorem: Let H denote a Hilbert space and H
the set of all bounded linear functionals on H. For any H
, there is a unique vector

v
H, such that
(u) = (u, v
), for any u H. (4.12)

Furthermore
6
, ||
1
= |v
|
1
.
The proof of this theorem relies on some elementary geometry of Hilbert space, in particular,
the projection theorem.
Proof of the Riesz Representation Theorem: If (u) = 0 for every u H, then we
take v
= 0. Thus we suppose that (q) ,= 0 for some q H. By linearity, (q (q)

1
) = 1.
The proof can be broken down into a sequence of claims.
Claim 1: Let m = inf |u| : (u) = 1 . The inmum, m, is attained, i.e. there exists
w H such that (w) = 1 and |w| = m.
We begin with the identity (exercise), which holds for all a, b H:
Parallelogram Law :
1
4
|a + b|
2
+
1
4
|a b|
2
=
1
2
|a|
2
+
1
2
|b|
2
, (4.13)
6
H and H
are isometrically isomorphic

10
4.1 Representation of Linear Functionals on H DRAFT: April 28, 2008
which follows by direct computation, using that |a|
2
= (a, a).
Now let w
k
denote a minimizing sequence. Thus, |w
k
| m, as k , (w
k
) = 1. We use
(4.13) to show that w
k
is a Cauchy sequence. Set a = w
k
and b = w
k
. Then, (4.13)
1
4
|w
k
w
j
|
2
=
1
2
|w
k
|
2
+
1
2
|w
j
|
2
w
k
+ w
j
2
1
2
|w
k
|
2
+
1
2
|w
j
|
2
m
2
The latter inequality holds because (
w
k
+w
j
2
) = 1, and therefore
w
k
+w
j
2
m. Now let k, j
tend to innity and we have, since |w
k,j
| m, that |w
k
w
j
| 0, i.e. the sequence w
k
is Cauchy. It therefore has a limit w H.

Claim 2: The minimizer, w, is unique.
Again this follows from the parallelogram law (4.13). If w
A
and w
B
both belong to H with
|w
A
| = |w
B
| = m, we have
w
A
+w
B
2
2
m
2
. Substitution into (4.13) gives:
m
2
+
1
4
|w
A
w
B
|
2
w
A
+ w
B
2
2
+
1
4
|w
A
w
B
|
2
= m
2
.
This gives a contradiction.
Claim 3: w is orthogonal to the null space of , i.e. for all u such that (u) = 0, we have
(w, u) = 0.
If (u) = 0 then
_
w
(w, u)
(u.u)
u
_
= (w) = 1 (4.14)
Since w has minimal norm,
w
(w, u)
(u, u)
u
2
|w|
2
.
Expanding and cancelling terms gives [(w, u)[
2
0 and therefore (w, u) = 0.
Claim 4: w ,= 0.
1 = (w
k
) (w) by continuity of . If w = 0 then (w) = 0, a contradiction.
Claim 5 and Completion of the Proof:
Dene v
= |w|
2
w. Then, for all u H, (u) = (u, v
).
Let u H be arbitrary. Recall that by Claims 3 and 4, w ,= 0 and x : (x) = 0 w.
Now express u as the sum of two orthogonal components, one in the direction of w and one
lying in the null space of .
u = [ u w(u) ] + w(u)
(u, w) = (u w(u) , w) + (u) (w, w)
Note: (u w(u) , w) = 0 since (u w(u)) = 0. Therefore,
(u, w) = (u) (w, w) (u, v
) = (u), v

w
|w|
2
This completes the proof of the Riesz Representation Theorem.
11
5 Applications of Hilbert Space Theory to the Exis-
tence of Solutions of Poissons Equation
We begin with Poissons equation for a function u dened on an open bounded subset of
R
n
:
v = f, v [
= 0 (5.1)
Here, we begin by taking f C(
).
We seek a weak formulation of (5.1). That is, we seek a formulation which is satised by
classical solutions, i.e. u of class C
2
(
) satisfying (5.1). To derive a weak formulation, let

u C
0
() and suppose that v is a classical solution. Now multiply (5.1) by u and integrate
over . Integration by parts, using the Gauss divergence theorem and that u vanishes on
gives
_
uf dx =
_
u (v) =
_
(uv) dx +
_
u v dx
=
_
u v dx. (5.2)
Dene
(u) =
_
u(x)f(x)dx
(u, v)
D
=
_
u(x) v(x) dx
The result (5.2) can be expressed as follows: If v is a classical solution to Poissons equation
with zero (Dirichlet) boundary conditions (v(x) = 0, x ), then for all u C
0
(), we
have (u) = (u, v)
D
. Note, on the other hand, that (u) is well-dened for any u, v L
2
()
and (u, v)
D
is well-dened if u and v are both in L
2
(). With the Riesz Representation
Theorem in mind, it is natural to regard (u, v)
D
as dening an inner product with respect
to which we shall represent the bounded linear functional (u).
Toward the correct formulation, we dene the
Dirichlet norm on C
0
()
|u|
D
=
_
(u, u)
D
. (5.3)
The key to showing that | |
D
is indeed a norm is to observe that if |u|
D
= 0, then u is
identically constant, and because u vanishes on , we must have u 0.
Next, to put our problem in a Hilbert space setting we dene the
Sobolev space H
1
0
(), which encodes zero boundary conditions, to be the completion of
the space C
0
() with respect to the Dirichlet norm, | |
D
.
We can now formulate the Weak Dirichlet Problem:
Let f L
2
(). Find v H
1
0
() such that for all u H
1
0
(), we have
(u, v)
D
=
_
u(x) v(x) dx =
_
u(x) f(x) dx =
f
(u). (5.4)
12
Theorem 5.1 The weak Dirichlet problem has a unique H
1
0
() solution.
Proof: We have set things up so that the result follows directly from the Riesz Representation
Theorem. The main thing to check is that (u) is a bounded linear functional on H
1
0
(),
i.e. there exists a constant M > 0, such that
[(u)[
uf dx
M
f
|u|
D
, (5.5)
for some M
f
, depending on f but independent of u. By the Cauchy Schwarz inequality
7
uf dx
|f|
L
2 |u|
L
2
Therefore, to prove that (u) is a bounded linear functional and therewith the theorem, we
must prove the:
Poincare Inequality: Let be an open connected subset in R
n
, which lies between two
planes, a distance 2a apart
8
. There exists a positive constant, p
, depending on , such that

for any u H
1
0
()
|u|
L
2 p
|u|
D
(5.6)
Note that by the Poincare inequality, (u) is a bounded linear functional on H
1
0
(), with
norm bound M
f
= p
|f|
2
.
Proof of the Poincare inequality: After possibly rotating the region , we may assume
it lies between the two planes x
1
= a and x
1
= a. Note that it suces to prove the
inequality for u belonging to a dense subset of H
1
0
(), u C
0
(). Let x = (x
1
, x
t
), with
x
t
= (x
2
, . . . , x
n
). For xed u(x), we have
u(x
1
, x
t
) =
_
x
1
a
x
1
u(
1
, x
t
) d
1
Squaring and applying the Cauchy-Schwarz inequality, we have
u
2
(x
1
, x
t
) =
_ _
x
1
a
1
x
1
u(
1
, x
t
) d
1
_
2
_ _
x
1
a
1
2
d
1
_

_ _
x
1
a
[
x
1
u(
1
, x
t
)[
2
d
1
_
(x
1
+ a)
_
a
a
[
x
1
u(
1
, x
t
)[
2
d
1
7
_
[fg[ |f|
2
|g|
2
8
Thus we allow domains, more general than bounded domains.
13
5.1 Generalizations of the Weak Dirichlet Problem DRAFT: April 28, 2008
Integration with respect to x
1
gives
_
a
a
[u(x
1
, x
t
)[
2
dx
1
2a
2
_
a
a
[
x
1
u(
1
, x
t
)[
2
d
1
Finally, integation with respect to the remaining n 1 variables, x
t
gives:
_
[u(x
1
, x
t
)[
2
dx
1
d
1
dx
t
2a
2
_
[
x
1
u(
1
, x
t
)[
2
d
1
dx
t
This proves Poincares inequality with p
2a.
5.1 Generalizations of the Weak Dirichlet Problem
We briey discussed how the forgoing discussion can be generalized from the Laplacian, ,
to a treatment of the Dirichlet problem for for more general linear elliptic partial dierential
operators, L:
Lv = f, x , v [
= 0
(a) L = a(x), where a(x) is a smooth function satisfying the strong ellipticity
condition: min
x
a(x) . Dene the Dirichlet inner produce

(u, v)
a
=
_
a(x)u(x) v(x) dx.

We then introduce the Dirichlet norm is then |u|
1
1
a
. Let H
1
a
denote the completion
of C
0
() with respect to |u|
1
1
a
. Then, we can prove (Exercise using Poincares
inequality!) that (u) = (u, f)
L
2 is a bounded linear functional on H
1
a
. We can then
apply the Riesz Representation Theorem to show that there exists a unique v
H
1
a
,
such that for all u H
1
a
, (u) = (u, v
)
a
.
(b) More general strongly elliptic operators: L =
jk

x
j
a
jk
(x)
x
k
, where A = (a
jk
(x))
is a n by n symmetric and positive denite matrix function, whose minimum eigenvalue,
min
(x) satises min
x
min
(x) > 0, for some . Note that example (a) is a special
case with A =
jk
a(x). Introduce the Dirichlet norm: |u|
1
1
A
and H
1
A
, the completion
of C
0
() with respect to |u|
1
1
A
. Then, (Exercise using Poincares inequality)
that (u) = (u, f)
L
2 is a bounded linear functional on H
1
A
. We can then apply the
Riesz Representation Theorem to show that there exists a unique v
H
1
A
, such that
for all u H
1
A
, (u) = (u, v
)
A
.
(c) Yet more general elliptic operators - elliptic operators with non-symmetric lower order
terms: Gardings inequality and the Lax-Milgram Lemma (representation theorem,
which generalizes Riesz Rep Thm, which is applicable to non-symmetric forms).
14
5.2 Relation between weak and strong (classical) solutions of elliptic PDE DRAFT: April 28, 2008
5.2 Relation between weak and strong (classical) solutions of el-
liptic PDE
For strongly elliptic linear partial dierential operators (see 5.1 one has the following impor-
tant theorem, showing that weak implies strong.
Theorem 5.2 (Elliptic Regularity Theorem) Consider the Dirichlet problem
Lv(x) = f(x), x , v(x) = 0, x , (5.7)
where is an open and bounded subset of R
n
with smooth boundary and f L
2
().
Let v H
1
0
() be a weak solution of Lv = f, where f C
(). Then, v C
and
lim
x, x
v(x) = 0.
The proof is very technical; see L.C. Evans, PDE, section 6.3.
5.3 Introduction to the Finite Element Method
9
The notion of weak solution of the Dirichlet problem for Poissons equation v(x) =
f(x), v(x) = 0, x (see (5.4) ) motivates The Finite Element Method - (FEM), a
method of central importance in the numerical solution of PDEs. We introduce FEM via a
simple example.
Consider the two-point boundary value problem:
2
x
v(x) = f(x), v(0) = v(1) = 0. (5.8)
Above we saw that this can be formulated, in terms of the Weak Dirichlet Problem (5.4),
which we present in modied form as follows.
Dene
| = u : u C
0
[0, 1], u C
1
piecewise
[0, 1] . (5.9)
Here, f C
1
piecewise
[0, 1] if there are nitely many points 0 <
1
< <
r
< 1 such that
f C
1
(
j
,
j+1
), j = 1, . . . , r 1
Our modied form of the weak Dirichlet problem is:
Find u | such that for all v |
(
x
u,
x
v) = (f, v) (5.10)
To obtain the Galerkin or Finite Element Method for the (5.8) we will replace |, by a
nite-dimensional approximation, |
h
, which we now construct.
9
Reference: Numerical solution of partial dierential equations by the nite element method, Claes John-
son, Cambridge 1992
15
5.3 Introduction to the Finite Element Method DRAFT: April 28, 2008
Partition the interval [0, 1] into subintervals:
[0, 1] = [x
0
, x
1
) (x
1
, x
2
) (x
M1
, x
M
) (x
M
, x
M+1
] =
M+1
j=1
I
j
0 = x
0
< x
1
< x
2
< < x
M1
< x
M
< x
M+1
= 1
The interval lengths in the partition are h
j
= x
j+1
x
j
. If the points, x
j
, are equally spaced
than h
j
= h = (M + 1)
1
.
Denition of |
h
:
|
h
= u : u linear on I
j
, j = 1, . . . , M + 1, v continuous, u(0) = u(1) = 0 (5.11)
Proposition: |
h
has dimension M and has the basis (spanning set which is linearly inde-
pendent) consisting of the triangular tent functions dened by:
j
|
h
j
(x
i
) =
ij
, (1, i = j; 0, i ,= j)
j = 1, 2, . . . , M
proof: Let u |
h
. Then, u(x) =
M
j=1
j
(x), where
j
= u(x
j
). Therefore, the functions
i
(x), i = 1, . . . , M span |
h
. It is simple to check they are linearly independent.
The Galerkin / Finite Element Method
Find u
h
|
h
such that
(
x
v
h
,
x
u) = (f, u), for all u |
h
or equivalently
(
x
v
h
,
x
j
) = (f,
j
), for all j = 1, . . . , M (5.12)
Exercise: Using the ideas from our discussion of the Dirichlets principle, show that (5.12)
is equivalent to nding u
h
|
h
, such that
I[v
h
] = min
u|
h
I[u] (5.13)
Implementation of the nite element method
Since v
h
|
h
, we have that v
h
(x) =
M
k=1
k
(x), where
k
= v
h
(x
k
) is to be determined.
Substitution into (5.12) gives
_
M
k=1
k
,
x
j
_
= (f,
j
), j = 1, . . . , M.
Equivalently,
A = b, (5.14)
16
5.3 Introduction to the Finite Element Method DRAFT: April 28, 2008
where A = (a
ij
) denotes the stiness matrix and b = (b
i
) denotes the load vector given
by:
a
ij
= (
x
i
,
x
j
) , i, j = 1, . . . , M
b
i
= (f,
i
).
Clearly, A is symmetric.
Computation of and properties of the stiness matrix
(1) Note that (
x
i
,
x
j
) = 0, [i j[ > 1. Therefore, A is a tridiagonal matrix. The
diagonal and o-diagonal elements are given by:
a
ii
= (
x
i
,
x
i
) = h
1
i
+ h
1
i+1
, a
i,i+1
= a
i+1,i
= (
x
i
,
x
i+1
) = h
1
i
(2) A is symmetric and positive denite. Symmetry was noted above. Positive deniteness
is seen as follows: For any R
M
T
A =
M
i,j=1
i
(
x
i
,
x
j
)
j
=
_
M
i=1
i
,
M
j=1
j
_
= |v|
2
0.
where v =
M
i=1
i
. Moreover,
T
A = 0 only if = 0 because
i
: i = 1, . . . , M
is a linearly independent set.
Theorem: The nite element method reduces to the solution of the system of M nonhomo-
geneous linear algebraic equations A = b, (5.14), with stiness matrix, A, and load vector,
b. A is symmetric and positive denite. It is therefore invertible and the system can be
solved for any load vector b.
Special case: Uniform partition of [0, 1]: Here, h
j
= x
j+1
x
j
= h = (M +1)
1
. Therefore,
a
ii
= 2/h and a
i,i+1
= a
i+1,i
= 1/h. Thus,
A =
1
h
_
_
_
_
_
_
_
_
2 1 0 0 . . . 0
1 2 1 0 . . . 0
0 1 2 1 . . . 0
. . . . . . . .
0 . 0 1 2 1
0 0 . . . 0 1 2
_
_
_
_
_
_
_
_
(5.15)
17
6 Operators on Hilbert spaces
The setting: complex Hilbert space with inner product, (, )
1
.
6.1 Bounded operators
Denition 6.1 Bounded linear transformation or bounded linear operator A linear
transformation on H is bounded if there is a constant, C 0, such that for all x H,
|T(x)| C|x| (6.1)
The smallest constant, C, for which (6.1) holds, is called the norm of T and is denoted by
|T|,
|T| = sup
x,=0
|T(x)|
|x|
= sup
|x|=1
|T(x)| (6.2)
We sometime write |T|
/(1,1)
to explicitly denote the operator norm of T, which maps H to
H, and more generally write,
|T|
/(1
1
,1
2
)
= sup
|x|
H
1
=1
|T(x)|
1
2
when speaking of T as an operator between dierent spaces H
1
and H
2
.
Denition 6.2 (Self-adjoint operator) T : H H is self-adjoint if
(Tx, y) = (x, Ty), for all x, y H (6.3)
Example 1: H = C
n
, (x, y)
C
n =
j
x
j
y
j
. Linear transformation on C
n
: T(x) = Tx,
T = (t
ij
), n by n matrix. T is self-adjoint if T = (t
ij
) = (t
ji
) = T
.
T = T
= |T| = max [[ : (T) ,

where (T) denotes the set of eigenvalues, the spectrum, of the matrix T.
Example 2: H = L
2
(R
n
). Fix R
n
. Dene T
f = f(x + ). By change of variables

|T
f| = |f|. Thus, T
is unitary on L
2
; |T
| = 1.
Example 3: H = L
2
(R
n
). Dene T[f]() =

f() =
_
e
2ix
f(x) dx, the Fourier transform
of f; see section 12.5. The Plancherel Theorem states: |T[f]| = |f|. Thus, T is unitary
on L
2
; |T| = 1.
Example 4: H = the set of 1 periodic L
2
functions:
L
2
(S
1
) = f : f(x + 1) = f(x), x R,
_
1
0
[f[
2
<
Dene, for n Z, the Fourier coecients
T[f](n) =

f(n) = (f,
n
) =
_
1
0
f(z)e
2inz
dz
18
6.2 Compact operators DRAFT: April 28, 2008
Parseval - Plancherel Theorem states
|f|
2
L
2
(S
1
)
=
_
1
0
[f(z)[
2
dz =
nZ
[
f(n)[
2
Thus, the mapping T : L
2
(S
1
) l
2
(Z), where l
2
(Z) is the set of sequences, which are square
summable, is unitary in L
2
; |T|
/(L
2
(S
1
);l
2
(Z))
= 1.
Denition 6.3 An operator T : H H has nite rank if its range is nite dimensional.
Example 5: Let H, and ||
1
= 1. Dene the projection operator
P
f = (f, )
1

An important example is the family of projection operators, P
j
, dened by
P
j
f(x) =

f(j) e
2ijx
for any f L
2
(S
1
). Note that the family of operators P
j
are orthogonal projections. That
is, P
j
P
j
= P
2
j
= P
j
and P
j
P
k
= 0 if j ,= k.
Example 6: Let T
(N)
= span e
2ijx
: [j[ N . Let P
(N)
=
[j[N
P
j
; see the previous
example. The operator
d
dx
P
(N)
maps T
(N)
to itself. What is its matrix representation with
the respect to the basis e
2ijx
: [j[ N ?
6.2 Compact operators
This would be a good time to review the background material on compactness and conver-
gence in section 12.4.
Denition 6.4 Let T : H
1
H
2
denote a bounded linear transformation between Hilbert
spaces. T is compact if for any bounded sequence, u
k
in H
1
, the sequence T(u
k
) has a
convergent subsequence in H
2
. That is, if for all k, |u
k
| M, then there exists a subsequence
T(u
k
j
) and a v
H
2
such that |T(u
k
j
) v
|
1
2
0 as j .
Example 6.1 Finite rank operators are compact. The proof is an exercise, using the com-
pactness of closed and bounded subsets of C
n
.
The importance of nite rank operators in the general theory is that any compact operator
in a separable Hilbert space can be approximated arbitrarily well by nite rank operators.
Theorem 6.1 If H is a separable Hilbert space and T is compact, then T is the norm
limit of operators of nite rank. That is, there exists T
j
compact and nite rank such that
|T
j
T|
/(1,1)
0.
A very important class of operators is the class of Hilbert Schmidt operators:
19
6.2 Compact operators DRAFT: April 28, 2008
Denition 6.5 K(x, y) : S S R is a called a Hilbert-Schmidt kernel if
_ _
SS
[K(x, y)[
2
dx dy <
We use K(x, y) to generate an operator on the Hilbert space L
2
(S).
T
K
[f](x) =
_
S
K(x, y) f(y) dy (6.4)
Theorem 6.2 If K(x, y) is a Hilbert-Schmidt kernel, then T
K
is a compact operator.
Proof: Let
j
(x)
j1
denote an orthonormal basis for L
2
(S). Then,
jk
(x, y) =
j
(x)
k
(y)
is an orthonormal basis for L
2
(SS). Dene K
N
(x, y) to be the partial sum approximation
to K(x, y) given by:
K
N
(x, y) =
j+kN
a
jk
jk
(x, y) (6.5)
Here, a
jk
=
_
S
_
S
K(x, y)
j
(x) dx
k
(y)dy is the (j, k) Fourier coecient of K. Then,
|T
K
T
K
N
| |K K
N
|
L
2
(SS)
0, N
Denition 6.6 Let T denote a bounded operator on H. The resolvent set of T, denoted
(T), given by
(T) = z C : (T zI)
1
exists and is a bounded operator on H . (6.6)
The spectrum of T, denoted by (T), is the complement of (T):
(T) = C (T)
Theorem 6.3 (Variant of the Riesz-Schauder Theorem)
Let T denote a compact operator on H.
(1) (T) is discrete
(2) If H is innite dimensional, then (T) is innite and consists of eigenvalues accumu-
lating at zero. Thus, 0 (T).
(3) The eigenvalues of T have nite multiplicity. If ,= 0 is an eigenvalue of T, then
dimx : Tx = x is nite.
The following result generalizes the well-known fact that Hermitian symmetric or self-adjoint
n n matrices have a complete set of eigenvectors that span C
n
.
Theorem 6.4 (Hilbert Schmidt Theorem) Let T : H H be a compact and self-
adjoint operator. Then, there exists an orthonormal basis for H consisting of eigenvectors
of T. That is, there exists x
j
j1
, Tx
j
=
j
x
j
, (x
j
, x
k
) =
jk
, such that for any x H
_
_
_
_
_
x
N
j=1
(x, x
j
)x
j
_
_
_
_
_
1
0
20
7 Applications of compact operator theory to eigen-
function expansions for elliptic operators
By Theorem 5.1, the Dirichlet problem for Poissons equation
v = f, x ; u [
= 0
has a weak H
1
0
() solution, i.e. There exists a unique v
f
H
1
0
() such that
_
u v
f
dx =
_
uf dx, u H
1
0
() (7.1)
The solution operator T : f v
f
, L
2
() H
1
0
() is a bounded operator. Think of this
mapping as
(
D
)
1
: f (
D
)
1
f,
where
D
denotes the Dirichlet Laplacian, the Laplace operator acting in the space of
functions with zero boundary conditions.
We next view T = (
D
)
1
as an operator from L
2
() to itself and can estimate its
norm.
Setting u = v
f
in the weak formulation (7.1) we obtain
_
[v
f
[
2
dx =
_
v
f
fdx.
By the Cauchy Schwarz inequality,
_
[v
f
[
2
dx
__
[v
f
[
2
dx
_1
2
__
[f[
2
dx
_1
2
.
By the Poincare inequality (5.6), (|g|
L
2 p
|g|
L
2, for g H
1
0
()), we have
p
2
|v
f
|
2
L
2
_
[v
f
[
2
dx |v
f
|
L
2 |f|
L
2
In other words,
|Tf|
L
2 p
2
|f|
L
2
()
; (7.2)
T is bounded on L
2
and |T| p
2
.
We now claim that T is compact and self-adjoint.
Theorem 7.1 The solution operator T : f Tf = v
f
, the solution of the Dirichlet problem
for Poissons equation is compact and self-adjoint operator for L
2
() to L
2
(). Therefore,
by the Hilbert-Schmidt Theorem 6.4 there exists an orthonormal set of eigenfunctions of T
which is complete in L
2
().
21
Proof: To show that T is compact, we must show that if f
j
is a bounded sequence in
L
2
(), then Tf
j
has a convergent subsequence. The key tool here is the Rellich compactness
Lemma, Theorem 12.11
10
Suppose that f
j
is a sequence for which there is a constant M such that |f
j
|
L
2 M.
By (7.2)
|Tf
j
|
D
p
2
|f|
L
2
()
p
2
M (7.3)
Thus, Tf
j
is a bounded sequence in H
1
0
(). By Theorem 12.11, this sequence has an L
2
()
convergent subsequence, Tf
j
k
.
Next, we verify self-adjointness of T. By the weak formulation of the Dirichlet problem
and the denition of T:
_
Tf u dx =
_
f u dx, u H
1
0
() (7.4)
Setting u = Tg, where g L
2
(), we have
_
Tf Tg dx =
_
f Tg dx, (7.5)
Similarly,
_
Tg Tf dx =
_
g Tf dx, u H
1
0
() (7.6)
Therefore,
_
Tf g dx =
_
f Tg dx. (7.7)
That is, (Tf, g)
L
2 = (f, Tg)
L
2; T is self-adjoint. The existence of a complete set of eigen-
functions of T is now a consequence of the Hilbert Schmidt Theorem, Theorem 6.4.
Corollary 7.1 L
2
() has an orthonormal basis of eigenfunctions of the , satisfying
Dirichlet boundary conditions.
Proof: Let
j
H
1
0
() denote the orthonormal sequence of eigenfunctions of T. By the
denition of T
(T
j
, u)
D
= (
j
, u)
L
2
(
j
j
, u)
D
= (
j
, u)
L
2

j
(
j
, u)
D
= (
j
, u)
L
2,
j
real
_

j
u dx =
_
1
j
u dx
Therefore, the eigenfunctions of T are weak solutions of
j
=
j
j
,
j
=
1
j
,
j

H
1
0
(). Interior elliptic regularity results ensures that these are, in fact, C
(). It can
10
Theorem: ( Rellich compactness Lemma) Suppose u
j
is a sequence in H
1
0
() such that |u
j
|
H
1
0
()
C.
Then, there exists a subsequence u
j
k
and an element u
H
1
0
such that |u
j
k
u
|
L
2 0 as j
k
.
22
also be shown, using boundary elliptic regularity results that if is suciently smooth,
then lim
x
j
(x) = 0
11
.
Moreover, setting u =
j
above yields positivity of the eigenvalues
j
=
1
j
=
_
[
j
[
2
dx.
FInally, since
j
0 as j ,
j
.
11
See L.C. Evans, PDE, section 6.3 for a discussion of interior and boundary elliptic regularity
23
8 Introduction to Variational Methods
In these notes we introduce Variational Methods or The Calculus of Variations via two
classical problems:
(1) Dirichlets principle, which characterizes the solution of the Poisson equation as
the solution of an optimization (minimization) problem
12
.
(2) The Rayleigh-Ritz, characterization of the smallest eigenvalue of the Dirichlet prob-
lem for the Laplacian as the solution of a minimization problem
13
8.1 Variational Problem 1: Dirichlets Principle
We introduce a class of admissible functions, /
/ w C
2
(
U) : w = g on U (8.1)
For w /, dene the Dirichlet energy:
I[w] =
_
U
1
2
[w(x)[
2
w(x)f(x)dx
(8.2)
Theorem 8.1 (Dirichlets principle
14
)
(1) If u / and u = f in U, then
I[u] = min
w,
I[w]. (8.3)
(2) Conversely, if u / and I[u] = min
w,
I[w], then u = f and u = g for x U.
We have studied, by Hilbert Space - functional analytic methods, the solution of the Dirichlet
problem for Poissons equation, for the case g = 0. The above Theorem gives an alternative
characterization of the solution.
8.2 Variational Problem 2: Smallest eigenvalue of
Poincare inequality
15
: Let denote a domain in R
n
which is open, connected and can be
bounded between two planes. There exists a positive constant, p
, depending on the domain

, such that for any u H
1
0
()
_
[u[
2
dx p
[u[
2
dx. (8.4)
12
Section 2.2.5 of L.C. Evans PDEs, pages 42-43
13
Section 6.5.1,Theorem 2, We follow an approach to Theorem 2, using the methods of Chapter 8 of L.C.
Evans, PDEs.
14
Theorem 17, page 42 of L.C. Evans PDEs
15
We established Poincares inequality in our study of weak solutions to Poissons equation.
24
8.2 Variational Problem 2: Smallest eigenvalue of DRAFT: April 28, 2008
We proved, in particular, that if is bounded between two planes a distance 2a apart, then
p
can be taken to be 2a
2
.
Question: What is the smallest constant, p
for which the Poincares inequality holds?

To answer this question, we need to compute
1
p
inf
0,=uH
1
0
()
R[u]
= inf
0,=uH
1
0
()
_
[u[
2
dx
_
[u[
2
dx
(8.5)
Remark 8.1 As in Dirichlets principle, we have an admissible class of functions,, / =
H
1
0
() in this case, and a functional R[u], dened on / which we seek to minimize.
Suppose the minimium in (8.5) is attained a function u
H
1
0
(). Let C
0
() be xed.
Then, we have for any , u
+ H
1
0
() and
R[u
+ ] R[u
]
Thus the function r() = R[u
+ ] is minimized at = 0. In particular, we must have

d
d
R[u
+ ] [
=0
= 0. (8.6)
Equation (8.6) can be rewritten equivalently as
_
u dx R(u
)
_
u = 0
Thus, for any C
0
()
_
( u R(u
) u ) dx = 0 (8.7)
In other words, (u
, R(u
)) is a weak solution of the eigenvalue problem

= , = 0, x . (8.8)
u
is an eigenfunction of with corresponding eigenvalue R(u
).
We have previously seen that the eigenvalue problem (8.8) has an innite sequence of
Dirichlet eigenvalues
j
, with
0
<
1

2
. . . , and corresponding eigenfunctions
j

H
1
0
()
16
. Note that R(
j
) =
j
and therefore we have
1
= R(u
). Borrowing from
terminology in quantum physics,
1
() is sometimes called the ground state energy and u
the ground state eigenfunction. We therefore have the following equivalence

16
This we shown by proving that the mapping ()
1
: f u
f
= ()
1
f which maps f L
2
() to
the weak H
1
0
() solution of the Dirichlet problem is a compact linear operator from L
2
() to itself. We then
applied the general spectral theorem for compact self-adjoint operators in Hilbert space.
25
8.3 Direct methods in the calculus of variations DRAFT: April 28, 2008
Theorem 8.2 (a)
0
() > 0 is the smallest Dirichlet eigenvalue of the Laplacian, .
(b)
0
() = min R(u), the minimum of the Rayleigh quotient over u H
1
0
().
(c) (
0
())
1
is the best (smallest) constant for which the Poincare inequality holds, i.e.
for any u H
1
0
()
_
[u(x)[
2
dx
1
0
()
_
[u(x)[
2
dx. (8.9)
8.3 Direct methods in the calculus of variations
In both variational problems, we assumed that the minimum of of the Dirichlet energy,
(8.3), and the Rayleigh quotient, R(u) in (8.5) are attained at an admissible function. The
question of whether this is in fact the case is quite subtle. In both problems we have the
following set up:
(a) a functional J[u] dened on an admissible class of functions, / (J[u] = I[u] in varia-
tional problem 1 and J[u] = R[u] in variational problem 2).
(b) J[u] > for u /.
Thus, inf
u,
J[u] = J
exists. In particular, there is a minimizing sequence of functions

u
j
/ such that J[u
j
] J
.
Exercise 1: Prove that for both Dirichlets and Rayleigh-Ritz principles that J[u] has a
lower bound, when u varies over the appropriate admissible set of functions, /.
THE QUESTION: Does the minimizing sequence u
j
converge to some u
which is (a)
admissible, i.e. u
/ and (b) a minimizer, i.e. J[u
] = lim
j
J[u
j
] = J
?
The property that u
j
is a minimizing sequence can often be used to conclude some
kind of
(c) uniform upper bound on norms of the functions u
j
.
The goal is to deduce from (a), (b) and (c) that
(d) the sequence of u
j
, or possibly some subsequence of it, has good convergence prop-
erties to a limiting admissible function u
/, for which J[u
] = j
.
Assuming this and if we set aside the issue of whether minimizers are C
2
functions
17
, the
arguments of the previous section apply to give variational characterizations to solutions of
PDEs.
For the remainder of these notes, we shall focus primarily on the solution of Variational
Problem 2: Lowest eigenvalue of and Poincares inequality.
17
This is an important and very technical issue. See the discussion in L.C. Evans PDEs, Chapter 8.3
26
8.4 Ideas and general discussion DRAFT: April 28, 2008
8.4 Ideas and general discussion
We now turn to the following questions:
(Q1) How do we deduce a uniform bound on norms of the function u
j
from (a) and (b)?,
and
(Q2) How do we show that, perhaps by passing to some subsequence of the u
t
j
s that we
can nd a sequence which converges to a function u
in the admissible set, at which the

functional attains its minimum.
The approach we take to the existence of solutions of a PDE via the proof that a functional
is minimized is called the direct method in the calculus of variations
18
.
Exercise 2: Concerning (Q1), prove that if u
j
is a minimizing sequence for either of the
variational problems in section 8, then there is a constant C, such that |u
j
|
H
1
0
()
C for
all j.
Concerning (Q2), we are interested in when bounds on a sequence of functions imply the
existence of a convergent subsequence. This is the subject of compactness; see section 12.4.
8.5 Compactness and minimizing sequences
Exercise 5: Prove that minimizing sequences for the variational problems in section 8 have
subsequences which converge weakly in H
1
.
To show that subsequence of any minimizing sequence for the problems of section 8
converge to an admissible minimizer, we use the following key properties:
Property A Weak compactness of minimizing sequences: There is a subsequence u
j
such u
j

u
H
1
0
() weakly.
Property B (Strong) compactness in L
2
() of minimizing sequences: There is a subsequence for
which |u
j
u
|
2
0.
Property C Weak lower semicontinuity of a functional with respect to weak convergence
19
:
_
[u
[
2
dx liminf
j
_
[u
j
[
2
dx (8.10)
Heres how Properties A, B and C are used to established the minimum is attained.
Consider variational problem 2 to minimize R(u). Recall that J
= 1/p
> 0 denotes the

positive inmum. By scaling, we can assume that u
j
is such that |u
j
|
2
= 1. Moreover,
it is clear since R(u
j
) approaches its positive inmum, that |u
j
|
2
is bounded. Thus u
j
is a bounded sequence in H
1
0
(). By weak compactness, theres a subsequence which well,
for simplicity call u
j
, which converges weakly to some u
H
1
0
(). Furthermore, Rellichs
Lemma implies there is a subsequence, which well call once again u
j
which converges strongly
to some u
. (Note: the weak limit is unique.). By strong convergence in L

2
we
[ |u
j
|
2
|u
|
2
[ |u
j
u
|
2
0
18
An excellent introductory reference is the book of Gelfand and Fomin: Calculus of Variations
19
This is a special case of the lower semi-continuity of any convex functional with respect to weak conver-
gence. Compare also, with Fatous Lemma from basic real analysis.
27
8.5 Compactness and minimizing sequences DRAFT: April 28, 2008
Thus, u
is admissible, i.e. u
H
1
0
() and |u
|
2
= 1.
By weak lower semicontinuity and the fact that u
j
is a minimizing sequence
J
= lim
j
R[u
j
] = liminf
j
R[u
j
] R[u
].
But u
is admissible, so R[u
] J
and the minimum is attained R[u
] = j
.
Whats left is the Justication of Properties A, B and C:
Property A follows from the above theorem on weak compactness.
Property B is a consequence of the following theorem, which was at the heart of the Hilbert
space approach to the Dirichlet problem and eigenfunction expansions.
Proof of Property C - weak lower semicontinuity: To prove (8.10) we begin with the
following observation:
_
[u
k
[
2
[u
[
2
=
_
[(u
k
u
)[
2
+ 2
_
(u
k
u
), (8.11)
which is proved by simply expanding (u
k
u
) (u
k
u
). Dropping the positive term

on the right hand side gives
_
[u
k
[
2
[u
[
2
2
_
(u
k
u
), (8.12)
Letting k tend to innity and using that u
k
tends weakly to u
in H
1
we see that the term
on the right hand side approaches zero as k . This implies
_
[u
[
2
liminf
k
_
[u
k
[
2
.
28
9 The Heat / Diusion Equation
Initial value problem
t
u = u
u(x, 0) = f(x)
Solution by Fourier transform
t
u(, t) = 4
2
[[
2
u(, t), u(, 0) =

f()
u(, t) = e
4
2
[[
2
t

f()
u(x, t) =
_
e
2ix
e
4
2
[[
2
t

f() d
=
_
f(y) dy
_
e
2i(xy)
e
4
2
[[
2
t
d
=
_
K
t
(x y) f(y) dy, K
t
(z) =
1
(4t)
n
2
e
z
2
/4t
Theorem 9.1 Let f be bounded and continuous on R
n
. For t > 0, dene
u(x, t) =
_
R
n
K
t
(x y) f(y) dy e
t
f. (9.1)
Then,
(a) u C
for any t > 0,

(b) u(x, t) satises the heat equation, and
(c) lim
t0
u(x, t) = f(x).
K
t
(x) is called the fundamental solution for the operator L =
t
, i.e.
(
t
) K
t
(x) = 0, t > 0 lim
t0
K
t
(x) =
0
,
where
0
denotes the Dirac delta distribution.
Proof: The explicit construction guarantees that (a) and (b) hold. To prove part (c) we
note from the following properties of K
t
(x).
(K1) K
t
(x) > 0, t > 0
(K2)
_
K
t
(x) dx = 1, t > 0.
(K3) Let > 0 be xed. Then,
lim
0<t0
_
[xy[
K
t
(y) dy = 0,
uniformly in x.
29
9.1 Remarks on solutions to the heat / diusion equation DRAFT: April 28, 2008
We need to show u(x, t) f() 0 as (x, t) (, 0).
[ u(x, t) f() [ =
_
R
n
K
t
(x y) ( f(y) f() ) dy
=
_
R
n
K
t
(x y) [f(y) f()[ dy, (K
t
> 0)
=
_
[xy[
K
t
(x y) [f(y) f()[ dy +
_
[xy[>
K
t
(x y) [f(y) f()[ dy
= I
1
+ I
2
Here, we have used property (K1). Estimation of I
1
: Here, [xy[ : Let > 0 be arbitrary.
Using the continuity of f, we can choose = () so that if [y[ < 2 then f(y)f()[ < .
By taking [x [ < , we have [y [ [y x[ +[x [ 2. Thus,
I
1

_
[xy[
K
t
(x y) dy
_
R
n
K
t
(x y) dy = ,
where we have used property (K2).
Estimation of I
2
: Here, [x y[ .
I
2
2|f|
L
_
[xy[()
K
t
(x y) dy, (9.2)
which tends to zero, as t 0 uniformly in x, by property (K3).
Thus we have shown
[u(x, t) f()[ + o(1)
as t 0, where is arbitrary. This proves the theorem.
9.1 Remarks on solutions to the heat / diusion equation
Let u(x, t) = e
t
f, the solution of the heat equation with initial data f(x).
(1) Instantaneous smoothing and innite propagation speed f 0, bounded and of com-
pact support = for all t > 0, e
t
f C
, e
t
f > 0 for all x.
Contrast this with the transport (wave) equation (
t
+c
x
)u(x, t) = 0 with initial data
u(x, 0) = f(x). The solution,u(x, t) = f(x ct) is clearly no smoother than the initial
data.
(2) f 0, u(x, 0) = f L
1
=
_
u(x, t)dx =
_
u(x, 0)dx, t > 0.
(3) |e
t
f|
L
2 |f|
L
2, t 0. More precisely,
d
dt
_
[u(x, t)[
2
dx = 2
_
[u(x, t)[
2
dx (9.3)
(4) [u(x, t)[ C t
n
2
|f|
L
1, t > 0.
30
9.2 Inhomogeneous Heat Equation - DuHamels Principle DRAFT: April 28, 2008
(5) inf f e
t
f sup f, t > 0.
(6) The basic properies of e
t
can be proved for e
Lt
, where L is a strictly elliptic diver-
gence form operator:
L =
j,k
x
j
a
jk
(x)

x
k
, (9.4)
where
j,k
a
jk
(x)
j
k
|[
2
, where > 0 is independent of x and R
n
is arbitrary.
Theorem 9.2 ( Nash (1958) and Aronson (1967) )
e
Lt
f =
_
K
L
t
(x, y) f(y) dy
and there exist constants a, a
t
, b, b
t
> 0 depending on and such that for all x, y, R
n
,
K
L
t
satises Gaussian upper and lower bounds:
a
t
t
n
2
exp(b
t
[x y[
2
t
) K
L
t
(x, y) a t
n
2
exp(b
[x y[
2
t
)
9.2 Inhomogeneous Heat Equation - DuHamels Principle
Consider the scalar ODE
du
dt
= au + f(t), u(0) = u
0
where a denotes a constant. Multiplying by the integrating factor e
at
, the resulting equation
can be rewritten as
d
dt
(e
at
u(t)) = e
at
f(t)
which we can now integrate and conclude:
u(t) = e
at
u
0
+
_
t
0
e
a(ts)
f(s) ds
The formula generalizes immediately to the case u(t) is the solution of the system of
ordinary dierential equations:
du
dt
= Au + f(t), u(0) = u
0
where A is a constant n n matrix and f(t) is a n 1 vector function of t:
u(t) = e
At
u
0
+
_
t
0
e
A(ts)
f(s) ds (9.5)
Under suitable hypotheses this can be generalized to the case where A is an operator
acting on a Hilbert space and f(t) takes values in a Hilbert space
20
.
20
A. Pazy, Semigroups of linear operators and application to partial dierential equations, Springer 1983
31
9.3 Heat equation on a bounded domain, DRAFT: April 28, 2008
Equation (9.5) is often called DuHamels formula or DuHamels principle. Applying it
with A = , we obtain that the solution to the inhomogeneous heat equation:
t
u(t, x) = u(t, x) + f(t, x), u(0, x) = u
0
(x)
is given by
u(t, x) = K
t
u
0
+
_
t
0
K
ts
f(s) ds, (9.6)
where e
t
g = K
t
g and K
t
denotes the heat-kernel
K
t
(z) =
1
(4t)
n
2
e
|z|
2
4t
(9.7)
9.3 Heat equation on a bounded domain,
t
u = u, (t, x) (0, )
u(t = 0, x) = f(x)
u(t, x) = 0, x , t > 0 (9.8)
Solution: From section 7, we have that L
2
() is spanned by the complete orthonormal set
of eigenfunctions of , F
j
(x), j = 0, 1, 2, . . . with eigenvalues
j
> 0:
F
j
(x) =
j
F
j
(x), x
F
j
(x) = 0, x
_
F
j
(x) F
k
(x) dx =
jk
We write the solution u(x, t) as a weighted sum of solutions of the form e
j
t
F
j
(x):
u(x, t) =
j=0
a
j
F
j
(x) e
j
t
(9.9)
The formal sum (9.9), by construction, solves the heat equation and satises the Dirichlet
boundary condition. To complete the solution, it suces to choose the constants a
j
so that
the initial condition u(0, x) = f(x) is satised:
f(x) = u(0, x) =
j=0
a
j
F
j
(x)
Since the eigenfunctions F
j
form a complete orthonormal set we have
a
j
= (f, F
j
)
L
2 (9.10)
Therefore,
u(x, t) =
j=0
(f, F
j
)
L
2 F
j
(x) e
j
t
(9.11)
32
9.4 Energy inequalities and heat equation DRAFT: April 28, 2008
It is clear from (9.11) and the fact that
0
<
1
< . . . , that each term in the expansion
of u(x, t) decays exponentially in t. The dominant behavior of u(t, x) for large t, if a
0
,= 0,
is given by
u(t, x) (f, F
0
)
L
2 e
0
t
, (9.12)
where
0
denotes the smallest eigenvalue of the
D
, the Laplacian on with Dirichlet
(zero) boundary conditions.
9.4 Energy inequalities and heat equation
Another, more general, approach to time-decay estimates is based on energy inequalities or
energy estimates.
Multiplication of the heat equation by u(t, x) and integration over gives
1
2
d
dt
_
u
2
(t, x) dx =
_
[u(t, x)[
2
dx (9.13)
By Theorem 8.2 we have the sharp form of Poincares inequality, valid on H
1
0
():
_
[u(x)[
2
dx
1
0
()
_
[u(x)[
2
dx. (9.14)
Therefore, by (9.13)
d
dt
_
u
2
(t, x) dx 2
0
()
_
[u(t, x)[
2
dx (9.15)
Thus,
_
u
2
(t, x) dx e
2
0
()t
_
u
2
(0, x) dx
|u(t)|
L
2 e
0
t
|u
0
|
L
2, (9.16)
consistent with the asymptotic result (9.12).
It is simple to extend this approach to parabolic partial dierential equation, in which,
is replaced by a second order strictly ellptic operator.
9.5 The maximum principle for the heat / diusion equation
9.6 Application: Burgers equation and Shock Waves
10 The Wave Equation
10.1 One dimensional wave equation
2
t
u = c
2
2
x
(10.1)
Characteristic variables:
= x + ct, = x ct (10.2)
33
10.2 Three dimensional wave equation DRAFT: April 28, 2008
Wave equation is equivalent to
u(, ) = 0.
General Solution
u(x, t) = F(x + ct) + G(x ct) (10.3)
consisting of right and left going waves.
Solution to the initial value problem with data:
u(x, 0) = f(x),
t
u(x, 0) = g(x) (10.4)
DAlemberts solution
u(x, t) =
1
2
( f(x + ct) + f(x ct) ) +
1
2c
_
x+ct
xct
g() d (10.5)
Remarks: Domain of Dependence, Range of Inuence, Forward and Backward light cones
in (x, t)
10.2 Three dimensional wave equation
2
t
u = c
2
u (10.6)
with initial data u(x, 0) = f(x) and
t
u(x, 0) = g(x). Introduce the spherical mean of a
continuous function, h on a ball of radius r about a point x:
M
h
(x, r) =
1
n
r
n1
_
[yx[=r
h(y) dS
y
=
1
n
_
[[=1
h(x + r) dS
(10.7)
The latter expression for M
h
permits extension of M
h
to all r R as an even function. Note
that if h is continuous then
lim
r0
M
h
(x, r) = h(x). (10.8)
Euler-Poisson-Darboux
1
r
n1

r
r
n1
r
M
h
(x, r) =
x
M
h
(x, r) (10.9)
The operator on the LHS, acting of M
h
, is the radial part of the n-dimensional Laplace
operator.
Now consider the spherical mean of a solution of the wave equation, M
u(,t)
(x, r). We
have, using (10.9), that
2
t
M
u(,t)
(x, r) = c
2
r
M
u(,t)
(x, r) (10.10)
Remarkably, in the three space dimensions, n = 3, this implies that r M
u(,t)
(x, r) satises
the one dimensional wave equation!!
2
t
_
r M
u(,t)
(x, r)
_
= c
2
2
r
_
r M
u(,t)
(x, r)
_
34
10.2 Three dimensional wave equation DRAFT: April 28, 2008
with initial conditions
r M
u(,0)
(x, r) = r M
f
(x, r),
t
M
u(,0)
(x, r) = r M
g
(x, r) (10.11)
DAlemberts formula (10.5) applies and yields an explicit formula for rM
u(,t)
(x, r). By
(10.8) lim
r0
M
u(,t)
(x, r) = u(x, t), and we obtain an expression for u(x, t) in terms of the
data f and g.
The result is
Theorem 10.1 Consider the three-dimensional wave equation with initial data u(x, 0) =
f(x) and
t
u(x, 0) = g(x). The solution is given by:
u(x, t) = tM
g
(x, ct) +
t
( t M
f
(x, ct) )
=
t
4
_
[[=1
g(x + ct)dS
+

t
_
t
4
_
[[=1
f(x + ct)dS
_
=
1
4c
2
t
_
[xy[=ct
g(y) dS
y
+

t
_
1
4c
2
t
_
[xy[=ct
f(y) dS
y
_
(10.12)
This is Kirchos formula; see Evans, page 73.
Remarks on the solution of the three-dimensional wave equation:
Loss of pointwise smoothness: In contrast to one-dimensional wave equation, solu-
tion is not as smooth as data in a pointwise sense. This is evident upon dierentiating
out the expression for u(x, t) on the second line of (10.12):
u(x, t) =
t
4
_
[[=1
g(x + ct)dS
+
1
4
_
[[=1
f(x + ct)dS
+
ct
4
_
[[=1

x
f(x + ct)dS
(10.13)
Conservation of energy and preservation of L
2
smoothness: However, there is
a preservation of L
2
type regularity
d
dt
_
(
t
u)
2
+ c
2
[u[
2
dx = 0 (10.14)
Diractive spreading and time decay estimate: f, g of compact support =
[u(x, t)[ = O(t
1
). This can be seen as follows. Rewrite (10.13) using the change of
variables y = x + ct. Thus, dS
y
= c
2
t
2
dS
. We obtain:
u(x, t) =
1
4c
2
t
2
_
[xy[=ct
[ tg(y) + f(y) + (y x)
y
f(y) ] dS
y
(10.15)
Assume f and g are supported within B
= x : [x[ < . Each integral is the form

_
A
h, where h is supported in a ball of radius and A = y : [x y[ = ct. Clearly,
_
A
h
_
A
h 1
h,=0
|h|
_
A
1
h,=0
= |h|
meas(A h ,= 0)
|h|
meas(A B
) |h|
4
2
35
10.3 2D wave equation - Hadamards method of descent DRAFT: April 28, 2008
Applying this bound to each term in (10.15) yields the time-decaying upper bound:
[u(x, t)[ K
2
t
1
( |g|
+|f|
+|Df|
) (10.16)
Huygens principle: Sharp arrival and departure of signals in odd space dimensions,
n = 1, 3.
10.3 2D wave equation - Hadamards method of descent
Wave equation in R
2
:
2
t
u = (
2
x
1
+
2
x
2
)u
View solution to the two-dimensional initial value problem as a solution to the three-
dimensional initial value problem with special data which is constant in x
3
:
u(x
1
, x
2
, x
3
, 0) = f(x
1
, x
2
),
t
u(x
1
, x
2
, x
3
, 0) = g(x
1
, x
2
) (10.17)
Note then that
x
3
u(x
1
, x
2
, x
3
, t) solves the three-dimensional wave equation with zero initial
data. Hence,
x
3
u(, t) = 0 for all t ,= 0 and u(x
1
, x
2
, x
3
, t) = u(x
1
, x
2
, 0, t).
Evaluating Kirchos formula (10.12) we have
u(x
1
, x
2
, 0, t) =
1
4c
2
t
_
(y
1
x
1
)
2
+(y
2
x
2
)
2
+y
2
3
=c
2
t
2
g(y
1
, y
2
)dS
y
+

t
_
1
4c
2
t
_
(y
1
x
1
)
2
+(y
2
x
2
)
2
+y
2
3
=c
2
t
2
f(y
1
, y
2
)dS
y
_
We now parametrize the upper / lower hemispheres of the sphere via y
3
(y
1
, y
2
) =
c
2
t
2
r
2
,
where r
2
= (y
1
x
1
)
2
+ (y
2
x
2
)
2
. Then,
dS
y
=
1 +
_
y
3
y
1
_
2
+
_
y
3
y
2
_
2
dy
1
dy
2
=
ct
c
2
t
2
r
2
dy
1
dy
2
Taking into account contributions from both upper and lower hemispheres, we obtain:
Theorem 10.2 Solution of IVP for wave equation in 2 dimensions
u(x
1
, x
2
, t) =
1
2c
_
r<ct
g(y
1
, y
2
)
c
2
t
2
r
2
dy
1
dy
2
+

t
1
2c
_
r<ct
f(y
1
, y
2
)
c
2
t
2
r
2
dy
1
dy
2
See, for example, Evans, page 80.
10.4 Principle of causality
Computation of energy ow across light cone - See, e.g. Evans Theorem 6, page 84.
36
10.5 Inhomogeneous Wave Equation - DuHamels Principle DRAFT: April 28, 2008
10.5 Inhomogeneous Wave Equation - DuHamels Principle
10.6 Initial boundary value problem for the wave equation
Example 1 :
1-d wave equation dened on x > h(t) where Cauchy data, u(x, 0),
t
u(x, 0) is given for
t = 0, x h(0) = 0 and along the curve (moving boundary), i.e. u(h(t), t) prescribed.
Example 2:
Example 1 with h(t) = c
0
t, c > c
0
. and u(c
0
t, t) = cos(t). Obtain Doppler eect.
37
11 The Schr odinger equation
In this section we introduce the Schrodinger equation in two ways. First, we mention how
it arises in the fundamental description of quantum atomic phenomena. We then show its
role in the description of diraction of classical waves.
11.1 Quantum mechanics
The hydrogen atom: one proton and one electron of mass m and charge e. The state of the
atom is given by a function (x, t), complex-valued, dened for all x R
3
and t R. is
often called the wave function.
Let R
3
. [(x, t)[
2
dx is a probability measure with the interpretation
_
D
[(x, t)[
2
dx = Probability (electron at time t)
Thus, we require
_
R
3
[(x, t)[
2
dx = Probability
_
electron R
3
at time t
_
= 1
Given an initial wave function,
0
i
t
= H
H =

2
2m
+ V (x) (11.1)
Here, denotes Plancks constant divided by 2. The operator H is called a Schr odinger
operator with potential V , a real-valued function determined by the nucleus. For the special
case of the hydrogen atom
H =

2
2m

e
2
r
, r = [x[ (11.2)
The free electron (unbound to any nucleus) is governed by the free Schrodinger equation
(V 0):
i
t
=

2
2m
(11.3)
11.2 Diraction of classical waves
Propagation of waves in a homogeneous medium, described by the classical wave equation
in dimensions n 2 gives rise to the phenomenon of diractive spreading of the wave-eld.
By considering a class of initial value problems with so-called wave packet initial conditions,
we demonstrate how on short time scales waves propagate along classical straight-lined
rays. However, on long (but nite) time scales waves propagate according to the (free)
Schr odinger equation.
38
11.3 Free Schr odinger - initial value problem DRAFT: April 28, 2008
Consider the initial value problem for the wave equation:
2
t
u(x, t) = c
2
u(x, t)
u(x, 0),
t
u(x, 0) given
Fourier representation of the solution
u(x, t) =
_
R
n
e
2ix
u(, t) d
u(, t) = cos(2c[[t) u(, 0) +
sin(2c[[t)
2c[[

t
u(, 0)
Remark 11.1 It is an interesting (optional) exercise to deduce the nite propagation speed
of signals from the Fourier representation. Note: The Fourier representation of the solu-
tion does not make explicit central features of the solution such as nite propagation speed,
domain of dependence. . . A proof of nite propagation speed, starting with the Fourier rep-
resentation, requires arguments closely related to a Paley-Wiener Theorem from complex
variables. Reference: F. John - PDEs, 4
th
edition, Springer, 1982, Chapter 5, Section 2.
Multiscale (wave packet) initial data - an example
u(x, 0) = e
2i
x
1
f(x
1
, x
),
t
u(x, 0) = 0, (11.4)
where x
= (x
2
, . . . , x
n
). We call x
1
the longitudinal variable and x
the transverse variables.

We take << 1 so that this data is oscillatory (with spatial scale ) with slowly varying
envelope (with spatial scale of order one).
Theorem 11.1 The solution u(x, t) is a superposition of terms of the following multiple
scale type
u(x, t) e
2i
x
1
ct
A(x
1
ct, t, x
) + O(
2
t),
giving a good approximation of the solution for times t o(
2
). The pre-factor e
2i
x
1
ct
is a
plane wave with rapid spatial variation propagating in the x
1
direction at speed c. This plane
wave is modulated by an amplitude function, which is more slowly varying. A(x
1
ct; T, x
)
satises the free Schrodinger equation
i
T
A(; T, x
) =
A(; T, x
) (11.5)
governing the diraction of waves and spreading over energy on time scales t = o(
2
). Here,
denotes the Laplacian with respect to the transverse variables, x
.
11.3 Free Schr odinger - initial value problem
Initial Value Problem
i
t
u = u, u(x, 0) = f(x) (11.6)
39
p
Unlike the heat equation,
t
u = u, which has an exponentially decaying Gaussian fun-
damental solution, the fundamental solution of the Schrodinger equation is a oscillatory
Gaussian with no spatial decay. For this reason, the derivation of the solution to the initial
value problem is more subtle. One approach is to regularize the Schrodinger equation by
adding a small ( > 0) diusive term, which we then take to zero ( 0).
Regularized initial value problem Take > 0.
i
t
u
= (1 i) u
, u
(x, 0) = f(x) (11.7)

i
t
u
= 4
2
(1 i)[[
2
u
u
(, t) = e
4
2
(i+)[[
2
t

f() (11.8)
u
(x, t) =
_
K
t
(x y) f(y) dy
K
t
(x) =
_
e
2ix
e
4
2
(i+)[[
2
t
d
= (4(i + )t)
n/2
e
|x|
2
4t(i+)
For f L
1
, we can pass to the limit 0
+
and dene the free Schr odinger evolution by:
u(x, t) = e
it
f =
_
K
t
(x y) f(y) dy
K
t
(x) = (4it)
n/2
e
i
|x|
2
4t
(11.9)
For f L
2
, we can use that the Fourier transform is dened (and unitary) on all L
2
to
dene, by (11.8),
u(x, t) =
_
e
4
2
i[[
2
t

f()
_
(x, t) = e
it
f (11.10)
p
e
it
f for f L
1
(R
n
): In this case,
[u(x, t)[ =
_
K
t
(x y) f(y) dy

_
[f[ dy
Therefore, if f L
1
(R
n
), then e
it
f L
(R
n
) for t ,= 0 and
| e
it
f |
L
[4t[
n
2
[ |f|
L
1 (11.11)
e
it
f for f L
2
(R
n
): In this case
_
[u(x, t)[
2
dx =
_
[ u(, t)[
2
d =
_
[e
4
2
i[[
2
t

f()[
2
d =
_
[f()[
2
d
40
11.5 Free Schr odinger evolution of a Gaussian wave packet DRAFT: April 28, 2008
Thus, if f L
2
(R
n
), then e
it
f L
2
(R
n
) and
|e
it
f|
L
2 = |f|
L
2 (11.12)
Extension to L
p
: Suppose f L
p
(R
n
) with 1 p 2. Using a theorem of M. Riesz on
interpolation of linear operators
21
, one can show:
Theorem 11.2 Let 1 p 2 and 2 q , where p
1
+ q
1
= 1. If f L
p
(R
n
), then
for t ,= 0 e
it
f L
q
(R
n
) and
|e
it
f|
L
q [4t[
(
n
2
n
q
)
|f|
L
p (11.13)
11.5 Free Schr odinger evolution of a Gaussian wave packet
Consider the evolution of a Gaussian wave-packet in one-space dimension. Let
0
= 2
0
.
i
t
u =
2
x
u
u(x, 0) = e
i
0
x
e
x
2
2L
2
= g
L,
0
(x) (11.14)
Thus u(x, 0) is an oscillatory and localized initial condition with carrier oscillation period
1
0
or frequency
0
. Its evolution has an elegant and illustrative form:
Theorem 11.3
u(x, t) = e
it
g
L,
0
=
e
i
0
(x
0
t)
_
1 +
2it
L
2
_1
2
e
(x2
0
t)
2
2L
2
(1+
2it
L
2
)
(11.15)
Proof: The Fourier representation of the solution is:
u(x, t) =
_
e
2ix
g
L,2
0
() d (11.16)
Note: g
L,2
0
() = g
L,0
(
0
) = (2)
1
2
L e
2
2
L
2
(
0
)
2
. Substitution into (11.16) and
grinding away with such tools as completing the square yields the result.
Remarks:
Phase propagates with velocity
0
, the phase velocity
Energy [u(x, t)[
2
propagates with velocity 2
0
, the group velocity
Solution disperses (spreads and decays) to zero as t . This is seen from the general
estimate (11.11) as well as the explicit solution (11.15).
However, solution does not decay in L
2
. The Schr odinger evolution is unitary in L
2
;
see (11.12).
Concentrated (sharp) initial conditions (L small) disperse more quickly than spread
out initial conditions (L large). The time scale of spreading is t L
2
.
21
See Chapter 5 of Introduction to Fourier Analysis on Euclidean Spaces, E.M. Stein and G. Weiss, Prince-
ton 1971
41
11.6 The Uncertainty Principle DRAFT: April 28, 2008
11.6 The Uncertainty Principle
Recall that [(x, t)[
2
has the interpretation of a probability density for a quantum particle
to be at position x at time t. [
(, t)[
2
has the interpretation of a probability density for
a quantum particle to be at momentum at time t.The expected value of an observable or
operator , A, is formally given by
22
A) = (, A) =
_

A (11.17)
Examples
(i) X), the average position =
_
x[(x, t)[
2
dx.
(ii) ), the average momentum =
_
[
(, t)[
2
d.
(iii) [X[
2
), the variance or uncertainty in position =
_
[x[
2
[(, t)[
2
dx.
(iv) [[
2
), the variance or uncertainty in momentum =
_
[[
2
[
(, t)[
2
d.
Theorem 11.4 (Uncertainty Inequality) Suppose xf and f are in L
2
(R
n
). Then,
_
[f[
2
2
n
__
[xf[
2
_1
2
__
[f[
2
_1
2
or equivalently
_
[f[
2
4
n
__
[xf[
2
_1
2
__
[

f[
2
_1
2
(11.18)
Exercise: (a) Prove the uncertainty inequality, using the pointwise identity
x [f[
2
= (x[f[
2
) n [f[
2
,
(b) Prove that the inequality (11.18) is sharp in the sense equality is attained for the Gaussian
f(x) = exp([x[
2
/2).
Applying the uncertainty inequality (11.18) to a solution of the Schr odinger equation,
with initial condition |(, 0)|
L
2 = 1 and we have,
1 =
_
[(x, 0)[
2
dx =
_
[(x, t)[
2
dx
4
n
__
[x(x, t)[
2
_1
2
__
[

(, t)[
2
_1
2
(11.19)
The latter, can be written as
n
4

_
[X[
2
)(t)
_
[[
2
)(t) (11.20)
and is called Heisenbergs uncertainty principle.
22
We proceed formally, without any serious attention to operator domains etc. For a fully rigorous treat-
ment, see M. Reed and B. Simon, Modern Methods of Mathematical Physics - Volumes 1-4
42
12 Background
12.1 Linear Algebra
Solvability of linear systems: Ax =
b, where A denote an mn matrix.

Theorem 12.1 (i) The system of m equations in n unknowns: Ax =

b is solvable if and
only if z,
b) z y = 0 for all z such that A

T
z = 0. That is, a solution exists if and only
if

b is orthogonal to the null space of the transpose of A.
(ii) Thus, either Ax =
b has a unique solution or it has it has an nr(A) parameter family

of solutions, where r(A) denotes the rank of the matrix A.
The eigenvalue problem: Ax = x
Diagonalization: Let
j
, j = 1, . . . , n denote the n eigenvalues of A, a nn (square) matrix.
This list of eigenvalues may have repetitions. If A has n linearly independent eigenvectors
v
1
, . . . , v
n
, then we say that A has a complete set of eigenvectors. Denote by V the n n
matrix whose j
th
column in v
j
. Then, AV = V , where = diag(
1
, . . . ,
n
), is the di-
agonal matrix with eigenvalues along the diagonal. Note that A = V V
1
or = V
1
AV .
We say A is diagonalized by V .
Symmetric matrices: The eigenvalues of a symmetric matrix are real. Moreover, for any
symmetric matrix A, there exists a complete set of eigenvectors v
1
, . . . , v
n
, which form an
orthonormal set, i.e. v
i
, v
j
) =
ij
. In other words, the matrix V whose columns are the
eigenvectors of A satises V
T
V = V V
T
, i.e. V is an orthogonal matrix.
12.2 Calculus
Let denote an open subset of R
n
and denote its boundary, assumed smooth.
Denition 12.1 C
k
() is the set of all functions f(x), dened on , whose derivatives up
to order k are continuous functions on .
Denition 12.2 A C
k
vector eld on R
n
is a vector function u : R
n
R
n
, mapping x R
n
to u(x) R
n
. Furthermore, the components of u
j
, j = 1, . . . , n of u are each C
k
(R
n
). If
k 1, then the Jacobian matrix of the vector eld D
x
u(x) is well-dened. It is the matrix
(
x
i
u
j
(x) ), 1 i, j n.
Denition 12.3 Given a C
1
vector eld on R
n
, we can dene its divergence:
div u = u =
n
i=1
x
i
u
i
(x) (12.1)
The following profound generalizations of the fundamental theorem of calculus and the
integration by parts formulae play a very important role in the study of PDEs.
43
12.2 Calculus DRAFT: April 28, 2008
Theorem 12.2 Gauss-Green Theorem
Let u be a C
1
vector eld dened on , a subset of R
n
with smooth boundary. Then,
_
u dx =
_
u n dS, (12.2)
where n denotes the unit normal, pointing outward from the region .
Applying (12.2) to the vector eld V (x)
ij
, where V is a scalar C
1
function, we obtain:
_
x
i
V (x) dx =
_
V n
i
dS (12.3)
Substitution of U(x) V (x) for V (x) in (12.3) we obtain:
Theorem 12.3 Integration by parts
Let U(x) and V (x) denote scalar C
1
functions
_
x
i
U V dx =
_
U
x
i
V dx +
_
UV n
i
dS (12.4)
Theorem 12.4 Change of variables
Suppose f : y f(y) is a continuous function on the region R
n
. Let
0
denote another
region and F : x y = F(x) a mapping from
0
to , such that (i) F C
1
and (ii) F is
one to one and onto. Then,
_
f(y) dy =
_
0
=F
1
()
f(F(x)) [ det D
x
F(x) [ dx (12.5)
Theorem 12.5 Implicit Function Theorem
Consider a function F : R
n
x
R
m
R
m
, (x, ) F(x, ), where F is suciently smooth.
We wish to understand the solution set of F(x, y) = 0. Suppose
(i) Solvability at a point (x
0
,
0
): F(x
0
,
0
) = 0.
(ii) Non-vanishing of the Jacobian determinant: Let D
F(x, ) denote the m by m Jaco-

bian matrix of rst partials:
k
F
l
(x, ), k, l = 1, . . . , n. We assume that D
(x
0
,
0
) is
nonsingular, i.e. det D
F(x
0
,
0
) ,= 0
Then, dened in an open neighborhood of x
0
R
n
, U, is a smooth function, = g(x), such
that g(x
0
) =
0
and
F(x, g(x)) = 0, for all x U. (12.6)
Therefore, the equation F(x, ) = 0, locally denes a surface, which can be represented as a
graph of a function = g(x).
Theorem 12.6 The Inverse Function Theorem
Assume f : R
m
R
m
is a smooth function and such that
(i) y
0
= f(x
0
)
(ii) The n by n Jacobian matrix, D
x
f(x
0
) is non-singular, i.e. det D
x
f(x
0
) ,= 0.
Then, dened in a neighborhood U of y
0
, there is function g : R
m
R
m
, such that x
0
= g(y
0
)
and for all y U, f(g(y)) = y.
44
12.3 Local existence theorem for ordinary dierential equations (ODEs) DRAFT: April 28, 2008
Proof of the Inverse Function Theorem: Apply the implicit function theorem to the
function F : R
n
R
m
R
m
dened by F(x, y) f(x) y and the equation F(x, y) = 0.
References: See for example,
(1) R.C. Buck, Advanced Calculus, McGraw-Hill
(2) Appendix C of L.C. Evans, Partial Dierential Equations, AMS - Graduate Studies in
Mathematics, Volume 19
12.3 Local existence theorem for ordinary dierential equations
(ODEs)
Consider the system of n ordinary dierential equations:
dx(t)
dt
= f(x(t), t), x(t
0
) = , (12.7)
where f is a smooth (suciently dierentiable) vector function R
n
R
1
R
n
(vector eld on
R
n
). The condition on x(t) at t = t
0
is called an initial condition. The dierential equation,
together with the initial condition is called the initial value problem or IVP.
When it exists, the solution of the IVP is a function of t as well the initial condition, .
The following theorem ensures the existence of a local solution of the dierential equation,
(t; ), which depends continuously on the initial condition, .
Theorem 12.7 There exists T, which depends on and the details of f, and a function
(t; ), dened and smooth for [t t
0
[ < T
, such that for [t t

0
[ < T
the function (t; )

satises the ODE and initial condition (12.7). Moreover, for any 0 < T < T
there is a
positive constant, depending on T, r(T), such that
max
[t[T
| (t; ) (t; ) | r(T) | | (12.8)
Here |y| denote the norm of a vector in R
n
, e.g. the Euclidean norm: |y|
2
=
n
j=1
[y
j
[
2
.
The upper bound, (12.8), on the deviation between trajectories quanties the continuous
dependence of the solution with respect to initial conditions. In particular, on any xed
closed interval of existence, as two initial conditions are brought close to each other, the
corresponding solution trajectories become uniformly close.
References: See, for example,
(1) V.I. Arnold, Ordinary Dierential Equations, MIT Press, 1980
(2) M.W. Hirsch and S. Smale, Dierential Equations, Dynamical Systems, and Linear
Algebra, Academic Press, 1974; Chapter 8
(3) E.A. Coddington and N. Levinson, Theory of Ordinary Dierential Equations, McGraw-
Hill, 1955
45
12.4 Compactness and convergence results DRAFT: April 28, 2008
12.4 Compactness and convergence results
The simplest compactnes result, and the one to which many more sophisticated results
reduce, is the classical theorem of analysis of Bolzano and Weierstrass. First, recall that a
subset, X, of R
n
is compact if any sequence in x
n
X has a convergent subsequence, i.e.
there exists a subsequence x
n
k
X and x
X such that |x
n
k
x
| 0 as n
k
.
Theorem 12.8 (Bolzano and Weierstrass) If X R
n
is closed
23
and bounded
24
, then X
is compact.
Exercise 3: Note that this theorem does not hold if R
n
is replaced by an innite
dimensional space. Let X = f L
2
(R
n
) : |f|
2
1 .
(a) Show that X is closed and bounded.
(b) Fix any function f
1
X with |f
1
|
2
= 1. Dene f
m
(x) = m
n/2
f
1
(mx). Show that
|f
m
|
L
2 = 1 and that m
n
f
2
1
(mx)
0
, the Dirac delta distribution, as m . Thus,
no subsequence of f
m
converges to a limiting f
L
2
as m ; the Dirac delta
distribution is not even a function!
Exercise 4: Let X be as in Exercise 3 above and for any m = 0, 1, 2 . . . set g
m
(x) =
f
1
(x m). The sequence g
m
lies in the closed and bounded subset of L
2
, X. Show that
g
m
does not have a subsequence which converges to an L
2
function.
It is therefore clear from the preceding two exercises that boundedness of a se-
quence of functions itself is not sucient to ensure compactness in the strong sense.
That is, if x
n
is a bounded sequence (for all n 0, |x
n
| C, this does not imply the
existence of a subsequence x
n
k
which converges in norm.
A second classical result, whose proof uses the Bolzano-Weierstrass theorem shows how (i)
boundedness and (ii) equicontinuity of a sequence implies compactness.
Denition 12.4 Let C(R
n
) denote the set of continuous real-valued functions on a R
n
and
F denote a subset of C(R
n
). The set F is equicontinuous if for any > 0, there is a
= (; F) such that |x y|
R
n < implies [f(x) f(y)[ < for all f F.
Theorem 12.9 (Ascoli-Arzela Theorem) If F C(R
n
) is a family of functions which is
uniformly bounded
25
and equicontinuous, then there is a sequence in F which converges
uniformly on any compact subset of X to a function in C(X).
Note that although closed and bounded subsets in a Hilbert space H are not generally
compact, THEY ARE HOWEVER WEAKLY COMPACT:
23
X contains all its limit points
24
There exists r
X
> 0 such that X lies within the ball of radius r
X
about the origin
25
There exists M
F
> 0, such that [f(x)[ M
F
for all x R
n
and f F
46
12.5 Fourier Transform DRAFT: April 28, 2008
Theorem 12.10
26
(Weak compactness - special case of Banach -Alagolou ) If x
n
is a
bounded sequence in a Hilbert space H, then there is a subsequence x
n
k
and x
H such
that (x
n
k
, y)
1
(x
, y)
1
for all y H.
A compactness result, rooted in the Ascoli-Arzela Theorem 12.10 and at the heart of
showing that (
D
)
1
is a self-adjoint compact operator is
Theorem 12.11 ( Rellich compactness Lemma)
27
Suppose u
j
is a sequence in H
1
0
() such
that |u
j
|
H
1
0
()
C. Then, there exists a subsequence u
j
k
and an element u
H
1
0
such that
|u
j
k
u
|
L
2 0 as j
k
.
12.5 Fourier Transform
Denition: Schwartz class, o(R
n
), is dened to be the space of all functions which are
C
and which, together with all their derivatives, decay faster than any polynomial rate.
Specically, if f o, then for any , N
n
0
, there exists a constant C
,
such that
sup
xR
n
[x
x
f(x)[ C
,
For f o dene the Fourier transform,

f or Tf by
f() = Tf() =
_
e
2ix
f(x) dx (12.9)
Proposition 12.1 Assume f o.
(a)

f C
and

f() =
_
(2ix)
f(x)
()
(b)

f() = (2i)

f()
(c)

f o. Thus, the Fourier transform maps o to o.
Theorem 12.12 (Riemann-Lebesgue Lemma)
f L
1
(R
n
) = lim

f() = 0.
Proof: Approximate by step functions, for which the result can be checked. Note: no rate
of decay of the Fourier transform is implied by f L
1
.
Theorem 12.13 Let f(x) e
a[x[
2
, 1a > 0. Then,
f() = a
n
2
e
||
2
a
(12.10)
The formula (12.10) also extends to 1a = 0, a ,= 0.
Proof: Write out denition of

f. Note that the computation factors into computing the
Fourier transform of n independent one dimensional Gaussians. For the one-dimensional
Gaussian, complete the square in the exponent, deform the contour using analyticity of the
integrand, and nally use that
_
R
e
y
2
dy = 1.
26
See Theorem 3, Appendix D, page 639 of L.C. Evans PDEs
27
This is a special case of the more general compactness theorem of Rellich and Kondrachov. For a proof
see L.C. Evans, PDEs; see section 5.2, Theorem 1 on page 272. Its proof relies on the Arzela-Ascoli theorem,
given above.
47
2
2
Denition 12.5 For g o dene g
g(x)
_
e
2ix
g() d = g(x)
Proposition 12.2
_

f g dx =
_
f g dx (12.11)
Proof: Interchange orders of integration (Fubinis Theorem)
Remark 12.1 Equation (12.11) is used to dene the Fourier transform of a distribution.
Namely, if T is a distribution, then the distribution

T is dened to be that distribution whose
action on C
0
(R
n
) functions is:
T() = T[
]
Thus, for example we can compute the Fourier transform of the delta function as follows:
x
[] =
x
[
] =

(x)
=
_
R
n
e
2ixy
f(y)dy.
Thus, we identify the

x
with the L
1
loc
function e
2ixy
, i.e.

x
= e
2ixy
.
Theorem 12.14 Fourier inversion formula Assume f o. Then,
f o = g(x) = f(x), where g() =

f() (12.12)
Proof: We shall prove Fourier inversion in the following sense.
lim
0
_
e
2
[[
2
e
2ix

f() d = f(x)
For any > 0 dene
() = e
2x
2
[[
2
whose Fourier transform is
(y) =
1
n
e
|xy|
2
2
=
1
n
g
_
x y
_
g
(x y)
Now,
_
e
2
[[
2
e
2ix

f() d =
_

()

f() d
=
_

(y) f(y) dy
=
_
g
(x y) f(y) dy f(x)
as 0 because g
is an approximation of the identity

28
.
28
Approximation of the identity: Let K(x) > 0 and
_
K(x)dx = 1. Dene K
N
(x) = N
n
K(Nx), N 1.
48
2
Theorem 12.15 Plancherel Theorem
f o =
_
[f(x)[
2
dx =
_
[
f()[
2
d
That is the Fourier transform preserves the L
2
norm on o (|
f|
2
= |f|
2
).
Corollary 12.1 The Fourier transform can be extended to a unitary operator dened on all
L
2
such |
f|
2
= |f|
2
.
Proof: o is dense in L
2
. If f L
2
, there exists a sequence f
j
o such that |f
j
f|
2
0.
Dene

f = lim
j

f
j
.
Let f be a bounded and continuous function on R
n
and consider the convolution
K
N
f(x) =
_
K
N
(x y) f(y) dy (12.13)
Prove: K
N
f(x) f(x) uniformly on any compact subset, C, of R
n
as N , i.e.
max
xC
[ [K
N
f](x) f(x)[ 0.
49

PDE Notes Miw04!28!08

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

PDE Notes Miw04!28!08

Enviado por

Direitos autorais:

Formatos disponíveis

DRAFT: April 28, 2008

are examples of distributions.

(x) dx, (3.3)

(x) (x) ) denotes the adjoint of the dierential operator

u(x) v(x) dx.

() is not complete with respect to the norm |u|

() with respect to dierent norms:

() with respect to the norm | |

and is called the

the set of all bounded linear functionals on H. For any H

, there is a unique vector

), for any u H. (4.12)

= 0. Thus we suppose that (q) ,= 0 for some q H. By linearity, (q (q)

are isometrically isomorphic

is Cauchy. It therefore has a limit w H.

) satisfying (5.1). To derive a weak formulation, let

, depending on , such that

a(x) . Dene the Dirichlet inner produce

a(x)u(x) v(x) dx.

= |T| = max [[ : (T) ,

f = f(x + ). By change of variables

, depending on the domain

for which the Poincares inequality holds?

+ ] is minimized at = 0. In particular, we must have

)) is a weak solution of the eigenvalue problem

is an eigenfunction of with corresponding eigenvalue R(u

the ground state eigenfunction. We therefore have the following equivalence

exists. In particular, there is a minimizing sequence of functions

/ and (b) a minimizer, i.e. J[u

/, for which J[u

in the admissible set, at which the

> 0 denotes the

. (Note: the weak limit is unique.). By strong convergence in L

and the minimum is attained R[u

). Dropping the positive term

for any t > 0,

= x : [x[ < . Each integral is the form

the transverse variables.

denotes the Laplacian with respect to the transverse variables, x

(x, 0) = f(x) (11.7)

b, where A denote an mn matrix.

b) z y = 0 for all z such that A

b has a unique solution or it has it has an nr(A) parameter family

F(x, ) denote the m by m Jaco-

, such that for [t t

the function (t; )

is an approximation of the identity

Você também pode gostar