Vous êtes sur la page 1sur 13

Greens functions, formulas and representations

Suppose that we want to solve a linear, inhomogeneous equation of the


form
Lu(x) = f(x) + homogeneous boundary conditions. (1)
Here u, f are functions whose domain is , which could be either a bounded
or unbounded domain in any dimension. The boundary conditions we will
consider will be Dirichlet, Neumann, or some mixture.
In this discussion, L will be self-adjoint with respect to the usual L
2
inner product using the imposed homogeneous boundary conditions. In
addition, we will assume that equation (1) admits a unique solution for any
(sufciently smooth) right hand side f. Both assumptions can be relaxed,
by the way, using adjoint- or modied- Greens functions, although these are
not pursued here.
1 The delta function and distributions
There is a great need in differential equations to dene objects that arise
as limits of functions and behave like functions under integration but are
not, properly speaking, functions themselves. These objects are sometimes
called generalized functions or distributions. The most basic one of these is
the so-called -function.
For each > 0, dene the family of ordinary functions

(x) =
1

e
x
2
/
2
. (2)
When is small, the graph of

is essentially just a spike of unit integral at


x = 0. Thus for any continuous and bounded f(x),

f(x)

(x x
0
)dx f(x
0
)

(x x
0
)dx = f(x
0
)
for small . On the other hand, taking the limit 0 inside the integral
makes no sense: the limit of

is not a function at all! To get around this,


we dene the object to act as follows:

f(x)(x x
0
)dx = f(x
0
). (3)
Informally speaking, the -function picks out the value of the function
f(x) at one point.
1
There are functions for higher dimensions also. We dene the n-
dimensional function as

R
n
f(x)(x x
0
)dx
n
= f(x
0
),
where x, x
0
R
n
. Sometimes we write this higher dimensional delta func-
tion as a product of one dimensional ones (x) = (x) (y) . . ..
1.1 A more precise denition
To be concrete about distributions, one needs to talk about how they act
on smooth functions. Note that (3) can be thought of as a linear mapping
from smooth functions f(x) to real numbers by the process of evaluating
them at x
0
. Linear mappings from a vector space to the real numbers are
often called linear functionals.
Now we come to the precise denition. A distribution is a continu-
ous linear functional on the set of innitely differentiable functions with
bounded support (this vector space of functions is denoted C

0
or simply
D). We can write d[] : D R to represent such a map: for any input
function , d[] gives us a number.
The class of distributions includes just plain old functions in the follow-
ing sense. For some integrable function g(x), we dene the corresponding
linear functional to be
g[] =

g(x)(x)dx.
(notice the subtlety of notation: g(x) means the function evaluated at x,
whereas the bracket g[] means multiply by function and integrate). In
fact, all distributions can be approximated by smooth, ordinary functions.
This means that, for any distribution d, there exists a sequence of smooth
functions d
n
(x) D so that
d[] = lim
n

d
n
(x)(x)dx, for all D.
For example, the sequence

that we rst investigated are smooth approx-


imations to the delta function distribution.
We can now dene what it means to integrate a distribution (notated as
if it were a regular function d(x)) simply by

d(x)(x)dx = d[], for any D.


2
This is consistent with the formula (3) since (x) maps a function onto its
value at zero.
Here are a couple examples. A linear combination of two delta func-
tions such as d = 3(x1)+2(x) denes a distribution. The corresponding
linear functional is
d[] = 3(1) + 2(0) =

d(x)(x)dx.
The linear operation
d[] =

(0)
is a linear map (check it!), taking a smooth function and returning the value
of its derivative at zero. A sequence of smooth functions that approximates
this is

(x), since integration by parts gives


lim
0

(x)(x)dx = lim
0

(x)

(x)dx =

(0),
1.2 Distributions as derivatives
One useful aspect of distributions is that they make sense of derivatives of
functions which are non-smooth or unbounded. Suppose that g(x) is an
integrable function, but cannot be differentiated everywhere in its domain.
It can make sense, however, to talk about the integrals involving g

. Though
integration by parts doesnt technically hold in the usual sense, for D
we can dene

(x)(x)dx

g(x)

(x)dx.
Notice that the expression on the right makes perfect sense as a usual in-
tegral. We dene the distributional derivative of g(x) to be the distribution
g

[] so that
g

[]

g(x)

(x)dx, for all D.


Notice that g does not even need to be a function for this denition to make
sense. If g is merely a distribution, then its derivative is the rule
g

[] g[

].
3
For example, if H(x) is the step function which is zero when x < 0 and
one when x > 0, the distributional derivative H

[] of H is the rule
H

[] =

H(x)

(x)dx =

(x)dx = (0) =

(x)(x)dx.
Therefore the delta function is the distributional derivative of the unit step
function.
In higher dimensions, one can make similar denitions of distributional
derivatives by using Greens identities. For some smooth function u(x) and
D, one has

R
n
(u)dx =

R
n
udx,
since the boundary terms in the Green identity vanish. This motivates a
denition of the distributional Laplacian for non-smooth functions, which is
a distribution with linear functional
(u)[] =

R
n
udx.
As an example, consider the statement u(x) = (x). This means that
the distributional Laplacian of u is

R
n
udx = [] = (0), (4)
for every D. Suppose rst we choose a point x
0
= 0, and let be any
function whose support is in a small neighborhood B of x
0
. Then suppos-
ing u is a regular smooth function near x
0
,

B
udx =

B
u = (0) = 0, (5)
by using (4). Since this is true for any choice of with = 0 outside B, this
means that
u(x
0
) = 0, x
0
= 0. (6)
Conversely, take so that = 1 on some neighborhood B containing the
origin. Using (4),
1 = (0) =

R
n
udx =

R/B
udx, (7)
4
since the integrand is zero on B. We can now use Greens identity on the
remaining integral, since u is a perfectly nice function away fromthe origin.
By virtue of (6), and the fact that = 0 on B, we are left with

B
u ndx = 1. (8)
One way of thinking of this statement is that the vector eld u has a unit
source at the origin. The two conditions (6),(8) will be needed in the next
section to nd such a function u(x).
2 Greens functions
Recall that differential operators often have inverses that are integral oper-
ators. So for equation (1), we might expect a solution of the form
u(x) =

G(x; x
0
)f(x
0
)dx
0
(9)
If such a representation exists, the kernel of this integral operator G(x; x
0
)
is called the Greens function.
It is useful to give a physical interpretation of (9). We think of u(x) as the
response at x to the inuence given by f(x). For example, if the problem
involved elasticity, u might be the displacement subject to an external force
f. If this were a heat equation, u might be the temperature subject to a heat
source described by f. The integral can be though of as the sum over the
inuences created by sources at each x
0
. For this reason, G is sometimes
called the inuence function.
Is there only one such function G so that (9) holds? Suppose that there
is another function G
2
so that
u(x) =

G
2
(x; x
0
)f(x
0
)dx
0
(10)
Subtracting (9) and (10) gives
0 =

G(x; x
0
) G
2
(x; x
0
)

f(x
0
)dx
0
, (11)
for every function f, which means that G G
2
(as a function of x
0
) is or-
thogonal to every function, so that G G
2
= 0. Thus the Greens function
is unique.
5
2.1 Relationship to the delta function
Part of the problem with the denition (9) is that it doesnt tell us how to
construct G. It is useful to imagine what happens when f is a point source
at x
i
, i.e. f(x) = (x x
i
). Plugging into (9) we learn that the solution
to
Lu(x) = (x x
i
) + homogeneous boundary conditions (12)
is just
u(x) =

G(x; x
0
)(x
0
x
i
)dx
0
= G(x; x
i
). (13)
In other words, we nd that the Greens function G(x; x
0
) formally satises
L
x
G(x; x
0
) = (x x
0
) (14)
(the subscript means that the linear operator acts on x, not x
0
). This equa-
tion says that G(x; x
0
) is the inuence felt at point x due to a source at point
x
0
.
Equation (14) is a more useful way of dening G since we can in many
cases solve this almost homogeneous equation, either by direct integra-
tion or using Fourier techniques. In particular, (14) can be rewritten as two
conditions
L
x
G(x, x
0
) = 0, when x = x
0
(15)
and integrating (formally) both sizes of L
x
G(x; x
0
) = (x x
0
)

B
L
x
G(x, x
0
)dx = 1, for any ball B centered at x
0
. (16)
Equation (15) is a homogeneous equation with a hole in the domain at
x
0
. Equation (16) is called the normalization condition, and it is used to get
the size of the singularity at x
0
correct. In one dimension, this condition
takes on a slightly different form (see below).
In addition to (15-16), G must also satisfy the same type of homoge-
neous boundary conditions that the solution u does in the original prob-
lem. The reason for this is straightforward. Take, for example, the case of
a homogeneous Dirichlet boundary condition u = 0 for x . For any
point x on the boundary, it must be the case that

G(x; x
0
)f(x
0
)dx
0
= 0. (17)
Since this must be true for any choice of f, it follows that G(x; x
0
) = 0 for
boundary points x (note that x
0
is treated as a constant in this respect, and
can be any point in the domain).
6
2.2 Jump conditions for ODEs
Instead of the integral condition (16), Greens functions in one dimension
(for ordinary differential equations) will satisfy pointwise conditions for
their behavior when x x
0
. Suppose one has a n-th order linear equation
of the form
u
(n)
(x) + F(u
(n1)
(x), u
(n2)
(x), . . .) = f(x),
where F is some expression involving lower order derivatives. The Greens
function G(x; x
0
) formally satises
G
(n)
+ F(G
(n1)
, G
(n2)
, . . .) = (x x
0
),
where G
n
=
n
/x
n
. Formal integration of both sides gives
G
n1
= H(x x
0
) + some continuous function
This means that something special happens at x
0
: the n 1-th derivative is
not continuous, but suffers a discontinuous jump there. Integrating again
shows that G
n2
is continuous.
These facts can be summarized as jump conditions at x
0
:
lim
xx
+
0

n1
G
x
n1
lim
xx

n1
G
x
n1
= 1, lim
xx
+
0

n2
G
x
n2
lim
xx

n2
G
x
n2
= 0. (18)
2.3 Three standard examples
We now specialize to the case of L = .
One dimension. Suppose u : R R solves the (ordinary differential equa-
tion)
u
xx
= f, u(0) = 0 = u(L). (19)
The corresponding Greens function will solve
G
xx
(x; x
0
) = 0 for x = x
0
, G(0, x
0
) = 0 = G(L, x
0
), (20)
The jump conditions are
lim
xx
+
0
G
x
(x; x
0
) lim
xx

0
G
x
(x; x
0
) = 1, lim
xx
+
0
G(x; x
0
) lim
xx

0
G(x; x
0
) = 0.
The piecewise solution to (20) is
G(x; x
0
) =

c
1
x x < x
0
c
2
(x L) x > x
0
(21)
7
Imposing jump conditions gives
c
1
x
0
= c
2
(x
0
L), c
2
c
1
= 1,
so that c
1
= (x
0
L)/L and c
2
= x
0
/L. It follows that the solution to (19)
can be written using G as
u(x) =
1
L

x
0
x
0
(x L)f(x
0
)dx
0
+

L
x
x(x
0
L)f(x
0
)dx
0

.
Three dimensions. Now suppose u : R
3
R solves
u = f, lim
|x|
u(x) = 0. (22)
In this case the homogeneous boundary condition is actually a far-eld
condition. The corresponding Greens function therefore must solve (15),

x
G(x; x
0
) = 0, x = x
0
, lim
|x|
G(x, x
0
) = 0. (23)
The normalization condition (16) is the same as (8),

x
G(x; x
0
) ndx = 1. (24)
for any ball B centered on x
0
. Let us observe the following: if we rotate
the Greens function about x
0
, it still will solve (23-24) since the Laplace
operator is invariant under rotation. Thus G only depends on the distance
between x and x
0
. We can write G = G(r), r = |x x
0
|, where in spherical
coordinates (23) is
1
r
2
(r
2
G

(r))

= 0 if r = 0, lim
r
G(r) = 0. (25)
This is easily integrated twice to give the general solution
G =
c
1
r
+ c
2
, (26)
where c
2
= 0 by using the far-eld condition in (25). The normalization
condition (24) determines c
1
. Letting B be the unit ball centered at x
0
(whose surface area is 4),
1 =

x
G(x; x
0
) ndx =

B
c
1
r
2
dx = 4c
1
,
8
so that c
1
= 1/4. Thus the Greens function is G(x; x
0
) = 1/(4|xx
0
|),
and the solution to (22) is
u(x) =

R
3
f(x
0
)
4|x x
0
|
dx
3
0
.
Two dimensions. In this case we want to solve
u = f, lim
r

u(r, ) u
r
(r, )r ln r

= 0. (27)
The far eld condition looks very strange at rst glance. The reason for this
is that solutions to u = f will in general NOT decay to zero at large |x|,
but will grow logarithmically. Again we look for a Greens function of the
form G = G(|x x
0
|) = G(r), so that
1
r
(rG

(r))

= 0 if r = 0, lim
r

GG
r
r ln r

= 0. (28)
The general solution is
G = c
1
ln r + c
2
, (29)
where c
2
= 0 by using the far-eld condition in (28). The normalization
condition (24) gives (where B is the unit disk)
1 =

x
G(x; x
0
) ndx =

B
c
1
dx = 2c
1
,
so that c
1
= 1/2. Thus the Greens function is G(x; x
0
) = ln |x x
0
|/2,
and the solution to (27) is
u(x) =

R
2
ln |x x
0
|f(x
0
)
2
dx
2
0
.
3 Inhomogeneous boundary conditions
Remarkably, the Greens function can be used for problems with inhomo-
geneous boundary conditions even though the Greens function itself satises
homogeneous boundary conditions. To obtain a representation formula for the
solution, we will need a Greens formula particular to the linear operator
in question.
9
3.1 Greens formulas
If a differential operator L is self-adjoint with respect to the usual L
2
inner
product, then for all functions u, v satisfying the homogeneous boundary
conditions in problem (1),

(Lv)udx

(Lu)v dx = 0. (30)
What if u, v dont necessarily satisfy homogeneous boundary conditions?
Then something like (30) would still be true, but terms involving boundary
values of u, v would appear:

(Lv)udx

(Lu)v dx = boundary terms involving u and v. (31)


What this formula actually looks like depends on the linear operator in
question and is known as the Greens formula for the linear operator L.
3.2 Symmetry (reciprocity) of the Greens function
Occasionally we need to utilize a special property of Greens functions of
self-adjoint operators often called reciprocity. This states that G(x; x
0
) =
G(x
0
; x), which means that the associated integral operator is self-adjoint.
To show this, let v(x) = G(x; x
1
) and u(x) = G(x; x
2
). Then because of
(12), we have Lv = (x x
1
) and Lu = (x x
2
). Plugging these into (30)
gives

(x x
1
)G(x; x
2
)dx

G(x; x
1
)(x x
2
)dx = 0 (32)
Using the basic property of the delta function, this simplies to
G(x
1
; x
2
) G(x
2
; x
1
) = 0 (33)
which is what we wanted to show.
A related fact has to do with interchanging partial derivatives of G.
Take the one-dimensional case, and observe

x
G(x, x
0
) = lim
h0
(G(x + h, x
0
) G(x, x
0
))/h = lim
h0
(G(x
0
, x + h) G(x
0
, x))/h
=
x
0
G(x
0
, x).
(34)
In other words, reciprocity implies that we can interchange the partial deriva-
tives with respect to x and x
0
provided we also interchange the arguments
x and x
0
.
10
3.3 Greens formula representation for inhomogeneous bound-
ary conditions
Lets now use (31) to solve a problem of the form
Lu(x) = f(x) + inhomogeneous boundary conditions. (35)
We do this by setting v(x) = G(x; x
0
) in the Greens formula (31), giving

(LG)udx

(Lu)Gdx = boundary terms. (36)


Because LG = (x x
0
) and Lu = f, we have

(x x
0
)u(x)dx

f(x)G(x; x
0
)dx = boundary terms (37)
We can collapse the integral involving the function, leading to
u(x
0
) =

G(x; x
0
)f(x)dx + boundary terms (38)
Provided we can evaluate everything on the right hand side, this gives a
formula for the solution which has two components: one which accounts
for the inhomogeneous term in the equation, and another (boundary inte-
gral) which accounts for the inhomogeneous boundary conditions.
3.4 Greens formula representation for the Laplace equation
The discussion above was meant to sketch the structure of the Greens func-
tion solution for a general linear equation with general inhomogeneous
boundary conditions. Now lets see exactly what this looks like for the
particular case L = .
One dimension. In this case, the Greens formula (31) is nothing more than
integration by parts twice (i.e. Greens second identity in 1-D), which for
L = d
2
/dx
2
and = [0, L] reads

L
0
uv

vu

dx = [uv

vu

]
L
0
. (39)
If we want to solve
u
xx
= f, u(0) = A, u(L) = B, (40)
11
we use the Greens function (21) appropriate for homogeneous boundary
conditions. Plugging v(x) = G(x; x
0
) into (39), we get

L
0
u(x)G
xx
(x; x
0
) G(x; x
0
)u

(x)dx = [u(x)G
x
(x; x
0
) G(x; x
0
)u

(x)]
x=L
x=0
.
(41)
Using u
xx
= f, G
xx
(x; x
0
) = (xx
0
) and noting that G = 0 on the bound-
aries, this can be written
u(x
0
) =

L
0
G(x; x
0
)f(x)dx + [u(x)G
x
(x; x
0
)]
x=L
x=0
. (42)
To make this look like the original formula (9), we can interchange x and
x
0
notationally and use the reciprocity identities, giving
u(x) =

L
0
G(x; x
0
)f(x
0
)dx
0
+ BG
x
0
(x; L) AG
x
0
(x; 0). (43)
Note this step was for aesthetic purposes only; formula (42) is sufcient to
solve the problem, but (43) is easier to interpret. For example, the inuence
of the right hand boundary condition at a point x is G
x
0
(x; L).
Many dimensions. In this case, the Greens formula (31) is nothing more
than Greens identity

uv vudx =

uv n vu ndx. (44)
Suppose we wish to solve the problem with the inhomogeneous boundary
condition
u = f in , u = h on ,
Let G be the Greens function that solves G = (x x
0
) with homoge-
neous, Dirichlet boundary conditions. Then substituting v(x) = G(x, x
0
)
in (44), we have (being careful to keep x as the variable of integration)

u(x)(xx
0
)G(x; x
0
)f(x) dx =

u(x)
x
G(x; x
0
) n(x)G(x; x
0
)u(x) n(x) dx.
(45)
The notation n(x) reminds the reader the unit normal vector is located at x,
not x
0
. Using the denition of the -function and the fact that G is zero on
the domain boundary, this simplies to
u(x
0
) =

G(x; x
0
)f(x)dx +

h(x)
x
G(x; x
0
) n(x) dx. (46)
12
Again, we can make this look like (9) by interchanging x and x
0
and using
reciprocity:
u(x) =

G(x; x
0
)f(x
0
)dx
0
+

h(x
0
)
x
0
G(x; x
0
) n(x
0
) dx
0
. (47)
The rst term is just the solution we expect for homogeneous boundary
conditions. The second term is more surprising: it is the derivative of G that
goes into the formula to account for the inhomogeneous Dirichlet bound-
ary condition.
3.5 Greens formula representation using the free space Greens
function
Now lets repeat the calculation of the previous section using the free space
Greens function in R
3
instead of the one particular to the domain . Insert-
ing v(x) = 1/(4|x x
0
|) in (44) and going through the same operations,
we arrive at
u(x) =

G(x, x
0
)f(x
0
)dx
n
0
+

u(x
0
)
x
0

1
4|x x
0
|

n +

1
4|x x
0
|

u(x
0
) ndx
n1
0
.
(48)
Notice that the right hand side uses both Dirichlet and Neumann data on
the boundary. Usually both of these are not known, so (48) does not solve
the problem directly. On the other hand, (48) is useful as a formula to con-
nect the values of solutions and their normal derivatives on the boundary.
Exercise. Show

x
0

1
4|x x
0
|

=
x x
0
4|x x
0
|
3
.
13

Vous aimerez peut-être aussi