Lecture One: Harmonic Functions and The Harnack Inequality

Lecture One: Harmonic Functions and the Harnack Inequality
1 The Laplacian
Let Ω be an open subset of Rn , and let u : Ω → R be a smooth function. We define the
Laplacian by
n
� ∂2u
�u = . (1)
i=1
∂x2i
The equation
�u = 0 (2)
is called the Laplace equation, and functions which satisfy it are said to be harmonic.
Harmonic functions turn out to be very important, and much of this course will be devoted
to their study. Also of interest are functions with nonnegative or nonpositive laplacian.
These are termed sub and superharmonic respectively.
2 The Maximum principle

The maximum principle is simply the statement that the gradient of a function at a max
¯r (x) with
imum is zero. Formally, if u is a twice differentiable function on a closed ball B
a maximum at x, then
�u = 0 at x and �u ≤ 0 at x. (3)
The one dimensional case should be familiar, and proofs of other cases are analogous.
3 Dirichlet Energy
Recall that the Dirichlet energy of a function v : Ω → R is given by Ω |�v|2 . We will show
�
that harmonic functions correspond to critical points of Dirichlet energy. For fixed v and
for any φ ∈ C0∞ (Ω) (i.e. φ an infinitely differentiable real valued function on Ω) we define
�
Eφ (t) = |�(v + tϕ)|2 . (4)
Ω
1
Now compute
� � �
Eφ (t) = |�v|2 + t2 |�φ|2 + 2t �v · �φ,
Ω Ω Ω
so �
d
|t=0 Eφ (t) = 2 �v · �φ. (5)
dt Ω
�
Since φ = 0 on ∂Ω, it is clear that ∂Ω φ�v · dS = 0. Applying Stokes’ theorem to this
gives
� �
�v · �φ = − φ�v,
Ω Ω
which we apply to (5) to get

�
d
|t=0 Eφ (t) = −2 φ�v. (6)
dt Ω
d
Thus, if �v = 0, then dt |t=0 Eφ (t) = 0 for all φ. The converse is also true but we will not
prove it here. These results give our correspondence. The following proposition makes it
more explicit.
Proposition 3.1 Let Ω ⊂ Rn be open. If u is a harmonic function on Ω then

� �
2
|�u| ≤ |�v|2 , (7)
Ω Ω
for all functions v satisfying u = v on ∂Ω. In other words harmonic functions have the
smallest Dirichlet energy for their boundary values.
�
Proof is by calculation. Clearly ∂Ω (v − u)�(v + u) · dS = 0. Therefore, by Stokes’
theorem,
� �
�(v − u) · �(v + u) = − (v − u)�(v + u). (8)
Ω Ω
Similarily � �
�(v − u) · �(v − u) = − (v − u)�(v − u). (9)
Ω Ω
Apply these to get
� �
|�v|2 − |�u|2 = �(v − u) · �(v + u)
Ω Ω
�
= − (v − u)�(v + u)
�Ω
= − (v − u)�(v − u)
� Ω
= |�(v − u)|2
Ω
≥ 0
as required.
4 The Mean Value Property

The following property of harmonic properties turns out to be very useful
Theorem 4.1 Let x ∈ Rn and take Br0 (x) a ball around x. If u is a harmonic function
on Br0 (x) then
�
1
u(x) = u, (10)
vol ∂Br (x) ∂Br (x)
and
�
1
u(x) = u (11)
vol Br (x) Br (x)
for all 0 < r ≤ r0 .
Proof First note that
� �
d 1 1 ∂u
u= . (12)
dr vol ∂Br (x) ∂Br (x) vol ∂Br (x) ∂Br (x) ∂n
This is simply saying that the derivative of the average of u is the average of the outward
normal derivative. Using this we can calculate
� �
d 1 1 ∂u
u =
dr vol ∂Br (x) ∂Br (x) vol ∂Br (x) ∂Br (x) ∂n
�
1
= �u · dS
�
1
= �u
vol ∂Br (x) Br (x)
= 0
3
by Stokes’ theorem. Thus
�
1
u
is constant on 0 < r ≤ r0 . This gives

� �
1 1
u(x) = lim u= u
s→0 vol ∂Bs (x) ∂Bs (x) vol ∂Br (x) ∂Br (x)
as required. For the second statement calculate

� � �� r
�
u = u ds
Br (x) 0 ∂Bs (x)
� r
= u(x) vol ∂Bs (x)ds
0
= u(x) vol Br (x).
By a similar argument we can also show

� �
1 1
u(x) ≤ u and u(x) ≤ u (13)
vol ∂Br (x) ∂Br (x) vol Br (x) Br (x)
for subharmonic u, or
� �
1 1
u(x) ≥ u and u(x) ≥ u (14)
vol ∂Br (x) ∂Br (x) vol Br (x) Br (x)
for superharmonic u. One consequence of the mean value property is the following.
Corollary 4.2 If x ∈ Rn and u is a harmonic function on Br (x) then u takes both it’s
maximum and it’s minimum value on the boundary ∂Br (x).
Proof If u has no interior maximum then it’s maximum must be on the boundary, and
we’re done. Else take y an interior maximum, and set c = u(y). By the maximum principle
c is the average of u over any sphere surrounding y. Since y is a maximum we also have
u ≤ c on each of these spheres, and we conclude that u = c on each sphere. Now take a
sphere that intersects with the boundary, and u takes it’s maximum on this intersection.
The argument for the minimum is similar.
By a very similar argument we can extend this result to shapes other than spheres, show
that subharmonic functions take their maximum on the boundary, and show that super
harmonic functions take they’re minimum on the boundary.
4
5 Harnack Inequality
Another useful property of harmonic functions is the Harnack inequality.
Theorem 5.1 Let B2r (0) be an open ball in Rn . There is a constant C depending only on
the dimension n such that
sup u ≤ C inf u. (15)

Br (0) Br (0)
for all functions u that are nonnegative and harmonic on B2r (0).
Proof Pick x, y ∈ Br (0). We must show that u(x) ≤ Cu(y). Let d ≤ 2r be the distance
between x and y, and pick w and z one and two thirds of the way from x to y respectively.
Note that u is positive and harmonic on Br (x) and that Br/3 (w) ⊂ Br (x). By the mean
value property we have
�
1
u(w) = u
vol Br/3 (w) Br/3 (w)
3n
�
= u
vol Br (x) Br/3 (w)
3n
�
≤ u
vol Br (x) Br (x)
≤ 3n u(x).
By a similar calculation we compare w, z and z, y to get
u(x) ≤ 33n u(y) (16)

as required.
This is a very powerful result about harmonic functions, with several consequences. For
example, if we have u ≥ 0 and �u = 0 on B2r (0) and inf Br (0) u = 0, then u is identically
0 on Br (0). In fact, by modifying the above argument, we can obtain a similar result for
any radius s < 2r.

Lecture One: Harmonic Functions and The Harnack Inequality

Uploaded by

Copyright:

Available Formats

Lecture One: Harmonic Functions and The Harnack Inequality

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture One: Harmonic Functions and The Harnack Inequality

Uploaded by

Copyright:

Available Formats

Lecture One: Harmonic Functions and the Harnack Inequality

2 The Maximum principle

which we apply to (5) to get

Proposition 3.1 Let Ω ⊂ Rn be open. If u is a harmonic function on Ω then

4 The Mean Value Property

is constant on 0 < r ≤ r0 . This gives

as required. For the second statement calculate

By a similar argument we can also show

sup u ≤ C inf u. (15)

By a similar calculation we compare w, z and z, y to get

u(x) ≤ 33n u(y) (16)

You might also like