6856 Chap01

August 6, 2008 15:46
9in x 6in
B-640
ch01
Chapter 1
Elementary Aspects of Potential Theory
in Mathematical Physics
1.1. Introduction
Many problems of elementary classical mathematical-physical analysis
are connected with the solution through integral representations of the
Dirichlet, Newmann, and Poisson problems for the Laplacian operator,
including its generalization to the Cauchy problem for wave and diusion
equations. In this chapter, we present the basic and introductory mathematical methods analysis used in obtaining such above-mentioned integral
representations.1
1.2. The Laplace Dierential Operator

and the PoissonDirichlet Potential Problem
Let be an open-bounded set in RN . We dene C 2 () to be the vectorial
space composed by those functions f (x) f (x1 , . . . , xn ) f (r ). (Here we

i
ei denoting the canonical
adopt the physical notation r = N
i=1 x ei with
N
vectorial basis of R ), continuously dierentiable until the second order in
, namely,
2 f (x)
C(),
xk xm
for any (k, m) {1, . . . , n}.
Let us consider the following linear transformation with domain C 2 ()

and range C()
(N ) : C 2 ( C()
f (x) (f )(x) =
n

2
x2i
i=1

f (x1 , . . . , xn ).
(1.1)
We now have the following simple result about the kernel of the above
linear transformation in R3 (holding true for any RN , with N 2).
1
LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

World Scientific Publishing Co. Pte. Ltd.
http://www.worldscibooks.com/physics/6856.html
August 6, 2008 15:46
9in x 6in
B-640
ch01
Lecture Notes in Applied Dierential Equations of Mathematical Physics
Lemma 1.1. The kernel of the Laplacean operator as dened by Eq. (1.1)
has innite dimension in R3 .
2
f
Proof. Let f C 2 (C R) such that 2 w
(w, t) exists and the function
1
2
3
dened below (x = x, x = y, x = z the usual classical notation in
potential theory in the physical space R3 )
Ut (x, y, z) = f(ix cos t + iy sin t + z, t).
(1.2)
We wish to show that

(3) Ut (x, y, z) = 0.
(1.3)
This result is a simple consequence of the following elementary

identities:
2
2
f (w, t)|w=z+ix cos t+iy sin t ,
U
(x,
y,
z)
=
t
2z2
2w
(1.4a)
2
2
Ut (x, y, z) = 2 f(w, t)|w=z+ix cos t+iy sin t ((i cos t)2 ), (1.4b)
2
2
x
w
2
2
Ut (x, y, z) = 2 f(w, t)|w=z+ix cos t+iy sin t (i sin t)2 .
2
2
y
w
(1.4c)
As a consequence

2
(3) Ut (x, y, z) = 2 f (w, t)w=z+ix cos t+iy sin t (1 cos2 t sin2 t) 0.
w
(1.5)
Note that a whole class of harmonic functions H()

in the class of
C 2 ()-functions in the kernel of the Laplacean operator Eq. (1.3) may be
obtained from Eq. (1.3):

dt g(t)(f(w, t)|w=ix cos t+i sin ty+z ),
U (x, y, z) =
a
and for the special function below:

f(w, t) = wn eit ,
( R, n Z),

(1.6)
August 6, 2008 15:46
9in x 6in
B-640
ch01
Elementary Aspects of Potential Theory in Mathematical Physics
we have the harmonic legendre polynomials in the three-dimensional

spherical ball

U (x, y, z) =
eit (z + ix cos t + iy sin t)n dt
dt eit (r cos + ir sin cos cos t + ir sin sin sin t)n
= 2rn ei
dt cos(t)(cos + (sin sin t)n

0
= rn ei Pn, (cos ).
(1.7)
After these preliminary elementary remarks we pass on the discussion

of the famous classical PoissonNewton problem: Let R3 be a bounded
open set of R3 with a boundary being an orientable simply connected C 2 -surface. We search for the solution of the nonhomogeneous linear
problem in through an integral representation (the Laplacean Green
function).
3 U (x, y, z) = f (x, y, z).
(1.8)
(
= ).
Here f (x, y, z) C 1 ()
We have the basic result in answering the PoissonNewton problem.

Theorem 1.1. The Laplacean operator

C 1 ()
3 : C 2 () H()
(1.9)
is inversible, and its inverse has the following integral representation:

f (x , y , z )
1

dx dy dz
U (x, y, z) =
4
(x x )2 + (y y )2 + (z z )2
def

d3r f (r ) G(0) (|r r |).

(1.10)
August 6, 2008 15:46
9in x 6in
B-640
ch01
Proof. Let us re-introduce the classic mathematical-physicist vectorial

notation in our formulas
r = (x, y, z);
r = (x , y , z ).
(1.11)
Let us consider the -approximant solution for the problem

1
()
d3r f (r ) G(0) (|r r |).
U () (r ) =
4
Here the regularized Green function is given by

1
for |r r | > ,
|r r | ,

()
G(0) |r r | =
1
|r r |2
, for |r r | .
3
2
2
(1.12)
(1.13)
Note that the above regularized Green function is a continuous

function in both variables, i.e.
G(0) (|r r |) C( )(since G(0) ( + ) = G(0) ( )).
()
()
We now have the estimate for 0:

1
()

2
sup |U U |(r ) f L2()

r
f L2() ( 2),
()

2
d rG() (|r r |)
3
12
B (0)
1
2
(1.14)
which shows the uniform convergence of the sequence of approximate

solutions to the (well-dened) function U (r ) by the Weierstrass uniform
convergence criterium.
That the integral representation Eq. (1.10) denes a well-dened
function comes straightforwardly from the estimate around our innitesimal
ball B (r ) = {r | ||r r| }.

1

1
r )
3 f (
2
2

d

r
4
dv
f

2
L ()

|r r |
v2
0
B (
r)
1
(4) 2 f L2 2() < .
(1.15)
As a second step to the proof of Theorem 1.1, we point out that the
sequence of functions

1
U () =
d3r f (r ) G() (|r r |)
(1.16)
4

August 6, 2008 15:46
9in x 6in
B-640
ch01
converges uniformly to the (well-dened) function in C()

I=+
3
d r f (r )r
1
4|r r |

.
(1.17)
The proof of the above made claim is a consequence of the following

estimate:

2

|r r |2
(x x )
d3r rx
+
= 0.
2
|r r |3
B (
r)

lim
(1.18)
Let us now pass on the problem of evaluating second derivatives.

In this case we have the Gauss theorem (integration by parts in R3 !)

1
|r r |

1
3

d r f (r ) r
=
|r r |

1
r ) dS(r )
cos(N
f (r )
= O

|r r |

1
.
d3r r f (r )
+
|
r
r |

d3r f (r ) r
(1.19)
r ) is the cosine between the normal eld of the globally oriHere cos(N
ented surface and the point r, argument of the function U (r ). At this
However, Eq. (1.19) has a rigpoint we are using the fact that f C 1 ().
2
orous meaning for f L () (left as an exercise to our diligent reader!).
It is important to remark that at this point of our exposition all these
topologicalgeometrical constraints imposed on the surface by means
of Gauss theorem must be imposed as complementary hypothesis on the
geometrical nature of the Poisson problem.
By deriving a second time the left-hand side of Eq. (1.19) and taking
into account that in the surface integral we always have r = r (we do
not evaluate the solution U (r ) at the boundary points) and obviously
r f (r ) 0, we arrive at the following result (without bothering ourselves
August 6, 2008 15:46
9in x 6in
B-640
ch01
with the 4-overall factor in Eq. (1.10)):

1

r )dS(r )
r U (r ) = O f (r ) r
cos(N
|r r |

1
d3r r f (r ) f (r ) r
+
|r r |

1

= O f (r ) r
cos(N r )dS(r )
|r r |

1
r )dS(r )
f (r ) f (r ) r
+ O
cos(
N
|r r |

1

3
(f (r ) f (r )) r
(1.20)
d r .
|r r |
Now we have that

f (r ) O

r
1
|r r |

= f (r ) O

cos(N r )dS(r )

r
N
ds(r ) = 4f (r )
|r |3
(1.21)
and (with xa = (x, y, z) and a = 1, 2, 3):

3
(xa xa )2 3

(f (r ) f (r )
+
d r

|r r |3
|r r |5
B (
r)

sup

r B (
r)

|Df |(r ) 3

+
B (
r)
d3r
B (
r)
|xa xa |2 |r r | 3

d r
|r r |5
|r r |
|r r |3

0 .
0
(1.22)
Precisely, to obtain this estimate, we need the condition that the data
(including the continuity of the
f (r ) must belong to the space C 1 ()
derivative at the boundary !).
As a general exercise in this section we left to our reader to prove
Theorem 1.1 in the general case of the domain RN (N 3) with the
Green function in RN .

N2
N
U (r ) =
dN r f (r )|r r |2N .
(1.23)
(N 2)2( )

August 6, 2008 15:46
9in x 6in
B-640
ch01
Note that the regularized Green function G() is now given by

G() |r, r )
|r r |N 2

=

N
N 2
2N
|r r |
2
2
(N 2)wN
N
2
if |r r | > ,
if |r r | .
(1.24)
In R2 , we can show that (left to our readers)

1
1

d2r g
)
.
U (r ) =
f
(
r
2
|r r |
(1.25)
This case can be seen as coming from the general case Eq. (1.23) after
considering the (distributional) limit
2

2
1
1
(2N )g|
r
r |
e
g|r r |.
lim
(1.26)
=
N 2 2( )2 N 2
2
We consider now the well-known Dirichlet problem: Determine a U (r )
such that
C 2 () C 1 ()

U (r ) = 0,

U (r ) = g(r )
for r ,
(1.27)
a very useful problem in string path integrals for R2 .

By a simple application of the rst Green identity we obtain the
uniqueness property for the Dirichlet problem

2
3
U U +|U | (r )d r = O (U N U )(r )d(r ) = 0.
0
(1.28)
As a consequence,

|U |2 (r )d3r = 0 U (r ) 0
in .
(1.29)
An explicit solution for Eq. (1.27) is easily obtained by considering the

Poincare method of considering a Green function of the structural form
below.
August 6, 2008 15:46
9in x 6in
B-640
ch01
Let
G(r, r ) = H(r, r ) +
1
,
4|r r |
(1.30)
where H(r, r ) is a harmonic function in in relation to the variable r ,

satisfying the following condition:

1

.
(1.31)
H(r, r )|r =
4|r r | r
Then we have the following explicit representation:
Theorem 1.2. We have the integral representation for the Dirichlet

problem

1

g(
r
)
G(
r
,

r
)
dS(
r
)
U (r ) =
O

4
N
(1/2 if r exercise)

1
1

=
)
. (1.32)
g(r )N H(r, r ) +
O
dS(
r
4
4|r r |
Proof. Let us consider the regularized form for the Green function as
proposed by the Poincare ansatz
Here
with
G() (r, r ) = H () (r, r ) + U () (r r ).
(1.33)
r H () (r, r ) = 0,

H () (r, r )
= U () (r r ),

r
(1.34)
1
,
4|
r
r |
()

U (r r ) =

1 1
|r r |2
,
3
4 2
2
if |r r | > ,
(1.35)
if |r r | .
By applying the second Green identity to the pair of functions U (r )

and G() (r, r ), we obtain the relation

U (r ) r U () (r, r ) d3r

= O
U (r ) N G() (r, r )dS(r ).

(1.36)
August 6, 2008 15:46
9in x 6in
B-640
ch01
By noting that we have explicitly the Laplacean action on the regularized function U () :
0,
if |r r | > ,

()

r U (r, r ) =
(1.37)
3
, if |r r | ,
3
4
and by taking the limit of 0, it yields the integral representation
Eq. (1.30).
Our problem now is to obtain an explicit procedure to determine the
Harmonic piece of the Green function, namely, H(r, r ).
A solution was rst given by H. Poincare through the use of the so-called
double-layer potential (r, r ), satisfying the Poincare integral equation in
L2 (, dS):

1
1
1
(r, r )
=
(r, w)
dS(w).

N
4|r r |
2
4 |r r |
(1.38)
Note the somewhat geometrical complexity of the geometrical measure
dS on the surface , when the integral equation above is solved by the
standard Fixed Point Theorems.1,2
After solving Eq. (1.38), Poincare showed that one easily has a formula
for the function H(r, r ):

1
H(r, r ) = O (r, r )N
(1.39a)
dS(r ).
|
4|
r

r
Another point worth to call the readers attention is that it appears more
straightforward to consider the following integral equation for the surface
derivative of the harmonic function U (r ). Since

1
1
1

O
U (r ) U (r )N
dS(r ),
U (r ) =
4 |r r | N
|r r |
(1.39b)
we have

(r ) N
(r ) = U (r ) = (r )
r U
N
r )
(

(r r )
1
=
N (r ) N U (r )
O
4
|r r |3

1
g(r )N N
dS(r ).
|r r |

(1.39c)
August 6, 2008 15:46
10
9in x 6in
B-640
ch01
One can in principle solve the above-written integral equation in

L (, dS(r )) by the Banach xed point theorem at least for small volumes
and for the normal derivative of the harmonic function U (r ) and then using
again the Liouville integral representation Eq. (1.39b) to determine U (r )
in .
A more useful approach for the explicit determination of the harmonic
term H(r, r ) in the PoincareGreen function is by means of the eigenvalue
problem for our Laplacean operator with a Dirichlet condition
2
n (r ) = n n (r )

n (r ) = 0
for r ,
(1.40)
for r ,
or in the equivalent integral form (see Theorem 1.1)

n (r )
d3r
n n (r ) =
4|r r |
with
1
n =
=
n

(1.41)

|n | (r )d r
2n (r )d3r
> 0.
Now it is straightforward to see that

(N n (r ))dS(r )
n (r )
1
d3r G(r, r )n (r ) =
O
n
n
4|r r |

r )) 3
r n (
(
+
H(r, r )
d r .
n
(1.42)
At the point we left as an exercise to our reader to show that the surface
integral and the volume integral cancel out.
By the usual Mercer theorem (if supr |n (r )| M , for n Z+ ), we
have the uniform convergence of the spectral
series dening the Dirichlet
n
< ,
Green Function as a result that limn |n|
2

|n (r )n (r )|
1

2
M
< .
(1.43)
|G(r, r )|
n
|n|2
3
3
nZ
nZ
The same result holds true for the case of the Newmann problem written
below:
U (r ) = 0,
for r ,

(1.44a)
N U (r ) = g(r )
O g(r )dS(r ) 0 ,

August 6, 2008 15:46
9in x 6in
B-640
with

U (r ) = + O
g(r )G(r, r )dS(r ),
ch01
11
(1.44b)
where
G(r, r ) =
n (r )n (r )
n
and with the eigenfunction/eigenvalue problem

n (r ) = n n (r ),
for r ,
N n (r ) = 0,
for r .
(1.44c)
(1.45)
Finally, it is left to our readers as a highly nontrivial exercise to generalize all the above results (their proofs!) for the case of possessing a Riemannian structure: The famous LaplaceBeltrami operator acting on scalar
functions into the domain (, gab (x)), RN , a = x a (a = 1, . . . , n)

1
g U = a ( g g ab b ) U (xa )
(1.46)
g
mathematical object instrumental in quantum geometry of strings and
quantum gravity.4
For instance, for two-dimensional Riemann manifolds, one can always
locally parametrize the Riemann metric into the conformal form in the
trivial topological Riemann surface sector gab (x, y) = (x, y)ab .
An important application in potential theory is the use of the famous
multipole expansion inside the Laplacean Green function (r = r|, r = |r |):

3 r r
1 r r
1
(r )2
= + 3 +
+ .
(1.47a)
|r r |
r
r
2
r2
2r3
A useful explicit expression for the Laplace Green function can be
obtained for the Ball Ba (0) = {r R3 | |r| = r a} in spherical
coordinates

2

f ( , )
a(r2 a2 )

U (r, , ) =
d
sin d
,
4
(a2 + r2 2ar cos )
0
0
(1.47b)
where the composed angle in the above-written formula is dened as

(1.47c)
(, ); (, ) = cos cos + sin sin cos( ).
August 6, 2008 15:46
12
9in x 6in
B-640
ch01
Another point worth remarking is that Eqs. (1.38), (1.39a) and (1.39c)
have potentialities for problems involving weak perturbations of the domain
boundary (0) (0) + (1) , since we have for instance the
explicit perturbation result for the integrand of Eq. (1.38) for a given
parametrization of the perturbed surface
(0)

= W1 (u, v)

(1) (u, v)
W
e1 + W2 (u, v)
e2 + W3 (u, v)
e3 = W
+ W
(1)
(0)
(1)
(0)
(u, v) = (W + W ) (W + W ) = N
(0) + N
(1) +
(0) + W
(1) ) (W
(0) + W
(1) )| 3
|(W
R
(0) W
(1) )
1
1
(r W
=
+
+
(0) W
(1) |
(0) |
(0) |3
|r W
|r W
|r W
) = (0) (r, W (0) ) + (1) (r, W (0) ) W

(1) + .
(r, W
(1.47d)
Rigorous mathematical analysis of such -perturbative expansion is left
to our mathematically oriented readers, including the open problem in
advanced calculus to nd explicitly the Green function for a toroidal surface
and the M
oebius surface band and its weakly random perturbations.

An important mathematical question is the one about the wellposedness properties of the Dirichlet problem in relation to the uniform convergence of the datum. We have the following result in this direction:
Lemma 1.2. If in Eq. (1.27) gn (r ) C 0 () converges uniformly for a
function g(r ) C0 (), then we have the uniform convergence of the solutions (1.32) associated with the sequence data {gn } to the solution associated
with g(r ).
be a compact in such that dist(G,
) = > 0. Since we
Proof. Let G
have the following estimate for the Green function

1

1

(1.47e)
sup
sup |G(r, r )|

r r | 4
r
r 4|

r G;

rG;
and so on,
sup |Un (r ) U (r )|

r G
Area()
||gn g||C 0 () 0.
4
(1.47f)

August 6, 2008 15:46
9in x 6in
B-640
ch01
13
Finally, let us announce the combined problem of PoissonDirichlet in

R and its associated integral representation.
3
Theorem 1.3. Let R3 be a region satisfying the hypothesis of the

the solution of
Gauss theorem. Let g(r ) and f (r ) be elements in C 1 ();
2
1
this problem in C () C ()

U (r ) = f (r ); r ,

U (r ) = g(r ) ,
(1.48)
is (uniquely) given by the integral representation written below:

U (r ) =
1
4

d3r
f (r )
|r r |
+ v(r ).
(1.49)
Here the function v(r ) is the harmonic function related to the boundary
data

v(r )
v(r ) = 0,

r )
3 f (

+
g(
r
)
=
d r
.

|r r |

r
(1.50)
1.3. The Dirichlet Problem in Connected Planar Regions: A

Conformal Transformation Method for Green Functions
in String Theory
Let be a simply connected region in R2 and W R2 another region
with similar topological properties. A coordinate transformation (dieomorphism) of W into
F
W R2
R2
(, )
(x(, ), y(, ))
(1.51)
is called conformal when one has the validity of the conditions in W :
x(, )
y(, )
=
,

x(, ) = y(, ) .

(1.52)
August 6, 2008 15:46
14
9in x 6in
B-640
ch01
In this case one can see that z = x + iy = f 1 ( + i) and f (z) is an

analytic function in . We have the following relation among the functional
expressions of the operator Laplaceans among the two given domains W
and :

2
2

y
x

y)
+
, (1.53)
, U (, ) = x,y U(x,

+i=f (x+iy)
where U ((x, y), (x, y)) U(x,

y).
In the case of the given domain transformation F possesing a unique
extension as a dieomorphism between the domain boundaries 1 = W
and 2 = , we have the following relationships among the Dirichlet
problems in each region:
, U (, ) = 0 (, ) W,
(a)
(1.54)
U (, ) = f (, ) ,
1
1
(x, y) = 0
x,y U
(b)
U
(x, y)
(x, y) ,

= f(x, y) .
(1.55)
Then,
(x, y).
U (, ) = U
(1.56)
It is a deep theorem of B. Riemann in the theory of one-complex

variables that given two simply connected bounded domains where the
boundaries are curved with continuous curvature there is a conformal transformation of the given domains realizing a coordinate (sense preserving)
of the boundaries.3 A proof of such a result can be envisaged along the
following arguments: Firstly, one considers the universal domain B1 (0) =
{(x, y) | x2 + y 2 < 1} as the domain . Secondly, one considers a triangularization T(n) of the region W . It is possible to write explicitly a conformal
transformation of T(n) into B1 (0) by means of the well-known Schwartz
Christoell transformation of a polygon into the disc B1 (0) denoted by
fn (z). One can show now that the family of functions {fn (z)} is a uniformly
bounded (compact) set in the space H() (Holomorphic function in ).
By choosing the family of triangulations Tn in such a way that Tn Tn+1 ,
August 6, 2008 15:46
9in x 6in
B-640
ch01
15
one can expect that fn (z) converges into H() to a function f (z) H()
carrying the conformal transformation of W into B1 (0). (Details of our suggested proof are left to our mathematically oriented readers again.)
At this point we intend to present a Hilbert space approach to construct
explicitly such canonical conformal mapping of the domain W into the disc
B1 (0).
Let H = {F (W, C), f : W C, f is a holomorphic function in W }.
We introduce the following Hilbert space inner product in this space of
holomorphic (analytical) functions in W :

1
dz d
z f (z)g(z).
(1.57)
f, gc =
2i
W
We have the following straightforward result for two elements of the
Hilbert space (H, , c ) obtained by a simple application of the Green
W:
theorem in the plane for any domain W

1
g(z)h(z)
dz =
(g(z) h (z))dz d
z.
(1.58)
2i
=1
W
W
= Br (z0 ) in Eq. (1.58), one obtains
If one considers h(z) = z z0 and W
that

g(z)
r2
2
g(z0 ). (1.59)
g(z)(z z0 )dz = r
=
2i
|zz0 |=r
|zz0 |=r z z0
In other words,
2
|g(z0 )| 2
r

1

g(z) dxdy

2
|zz0 |r
12
12

2 1
2
|g
(z)|
dxdy
dxdy
r2 2
|zz0 |r
|zz0 |r

r ||g||2H(W )
r2
3/2
||g||2H(W ) c||g||2W (D) .
(1.60)
r
By the Riesz theorem for representations of linear functionals in
Hilbert spaces, the bounded (so continuous) linear evaluation functional on
(H(D), , c ) = HD has the following representation:

1
g(z0 ) =
dz d
z R(z, z0 )g(z)
(1.61)
2i
D

August 6, 2008 15:46
16
9in x 6in
B-640
ch01
and
H
R(z, z0 ) =
ek (z)ek (z0 ),
(1.62)
k=1
where {ek (z)} denotes any orthonormal set of holomorphic functions in

HD .
Let f : W B1 (0) be the aforementioned Riemann function mapping
holomorphically the given bounded simple connected-smooth boundary
domain in B1 (0). Since f (z) is univalent, we have that for f (z0 ) = 0 and
f (z0 ) > 0 the following identity holds true:

1
g(z0 )
g(z)
=
for any g HD .
(1.63)
2i W =1 f (z)
f (z0 )
Let Cr = {(f (z))(f (z)) = r < 1} be the curve in the complex plane.
For this special contour, we have that (see Eq. (1.58))

1
g(z)
1
dz =
g(z)f(z) dz
2i Cr f (z)
2ir2 Cr

1
g(z)f (z) dxdy.
(1.64)
= 2
r
(W \f 1 (Br (0))
Note that we have implicitly assumed that the domain W is a homotopical
deformation of its boundary 1 !
As a consequence (by considering the limit of r 1)

1
g(z0 ) = f (z0 )
g(z) f (z) dxdy .
(1.65)
W
In view of the uniqueness of the Reproducing Kernel Eq. (1.57), we have
the explicit representation for the conformal transformation (f (z0 ) = 0 and
f (z0 ) > 0)

f (z) =
R(z0 , z0 )
12
R(, z0 )d
.
(1.66)
z0
For instance, one could consider the GrahamSchmidt orthogonalization

process applied to the functions {z n , zn }nZ+ and obtain Eq. (1.66) as an
innite sum of holomorphic functions, or use triangulations of the domain D
in order to evaluate the functions en (z) in Eq. (1.62).
August 6, 2008 15:46
9in x 6in
B-640
ch01
17
Finally let us devise some useful formulae for the Dirichlet problem in
planar smooth regions, useful in string path integral theory.
Let U0 f 1 : B1 (0) R be the associated harmonic function on the
disc B1 (0) related to the Dirichlet problem in the region W . Note that by
construction f (z0 ) = 0 and f (z0 ) > 0 for a given z0 W . As a consequence
of the mean value for harmonic function, we have that
2
1
U (z0 ) = (U0 f 1 )(0) =
(U0 f 1 )(cos , sin )d
2 0

f (z)
1
dz.
(1.67)
U (z) z0
=
2i W
fz0 (z)
From
Eq. (1.67) we can write explicitly the Green functions G(z, z0 ) =
(z) fz0 (z) for the given canonical regions. For instance, fz0 (z) = R(z
fz 0
z0 ) (R2 z0 z) applies BR (0) conformally into B1 (0), leading to the Poisson
formulae in the circle of radius R

2
R2 r 2
1
d U (Rei )
U (r, ) =
.
(1.68)
2 0
R2 + r2 2Rr cos( )

The function fz0 (z) = (zz0 ) (z+z0 ) applies conformally the right-half+
= {z = x + iy, x 0, < y < } into B1 (0) and fz 0 (z0 ) = 0.
plane
As a consequence, one has the integral representation for this region

x + U (it) dt
(1.69)
U (x, y) =
(t y)2 + x2

The function fz0 (z) = (z 2 z02 ) (z 2 (
z0 )2 ) solves the Dirichlet problem
in the rst quadrant plane, leading to the integral representation below:

t U (t) dt
4xy
U (x, y) =
t4 2t2 (x2 y 2 ) + (x2 y )2

0

t U (it) dt
+
.
(1.70a)
t4 + 2t2 (x2 y 2 ) + (x2 y 2 )
0
In general
G(z, z ) =
fz0 (z ) fz0 (z)

1 fz0 (z) fz0 (z )
(1.70b)
with f (z0 ) = 0 is the explicit Green function associated with the Dirichlet
problem in the planar region W .
August 6, 2008 15:46
18
9in x 6in
B-640
ch01
Just for completeness let us write the Green function of the Dirichlet
2
in polar coordinates
problem in the SR

1
G((, ), ( , )) =
log[2 + 2 2 cos( )
4

2 2

+ log R2 +
2
cos(
)
. (1.70c)
R2
Here

, G((, ), ( , ) = 0,

1

G((R, ), (R, )) = +
g|r r |

r =Rei

r =Rei
(1.70d)
1.4. Hilbert Spaces Methods in the Poisson Problem

Let us pass now to the formulation of the Poisson and Dirichlet problems in
a compact set of RN . This formulation is
the Hilbert spaces L2 (), with
known in the mathematical literature as the Hilbert space weak formulation
of the problem.
of the vector space Cc1 () in relation to the
Let H01 () be the completion

d4 x f g; the weak formulation of Poisson
inner product f, gH 1 =

problem reads as of
U, H 1 =
dN x U (x) ()(x)
dN x(f )(x)
= f, L2
H01 (),
(1.71)
where one must determine a U (x) H01 () for a given f L2 (). That
such (unique) U (x) exists is readily seen from the fact that the functional
linear below is bounded in H01 ():
Lf : H01 ( C)

dN x(f )(x)
= Lf ()
(1.72)
as a consequence of the Friedrichs inequality

N 1
N
2
N
2
d x|U | (x) diam()

d x|U | (x) .

||U ||2L2 ()
C() ||U ||2H 1 ()

0

(1.73a)
(1.73b)
August 6, 2008 15:46
9in x 6in
B-640
ch01
19
By the Riesz representation theorem, there is a unique element U (x)

such that it satises Eq. (1.72) and has the explicit expansion
H01 (),
U (x)
H01 ()
Lf (en )en (x)
dN y f(y)en (y) en (x)

d y

en (x)en (y) f (y) =

dN y G(x, y)f(y).
(1.74)
Here, en (x) is an arbitrary orthonormal basis in H01 (), which can be

constructed from the application of the GrahmSchmidt orthogonalization
process to any linear independent set of Cc (), for instance, the set of the
polynomials {x|n| , x|m| } with x RN .
This method of determining weak solutions is easily generalized to the
Poisson problem with variable coecients the LaxMilgram theorem, a
useful result in covariant euclidean path integrals in string theory.
a compact set of RN .
Let aij (x) and U (x) be functions in C(), with
Let us consider the following Hilbert space:
L2{a} () = topological closure of

N
)(x) <
f C0 ()
dN x aij (x) i f
d
x
V
(x)(f
f
f
+
x
xj
with aij (x) = aji (x), aij (x)i j a0 ||2

and V (x) > V0 > 0 .
(1.75)
We left as an exercise to our readers to mimic with some improvements

the previous study to prove the existence and uniqueness of the Poisson
problem in L2{a} () for f L2{a} () H01 () and (x) H01 ()

d x aij (x) i U (x) j (x) +

dN x V (x)U (x)(x)
x
x

=
dN x f (x)(x).
(1.76)
It is another nontrivial problem to nd conditions on the data function

f (x) (see Theorem 1.1) in order to have the weak solution U (x) to belong
August 6, 2008 15:46
20
9in x 6in
B-640
ch01
which will lead to the strong solution of the

to the space C 2 () C 1 (),
Poisson problem in

N
(x)
U
(x)
+ V (x)U (x) = f (x) x .
a
ij
xj
xi
i,j=1
(1.77)
Another important generalization of the Poisson problem is the class

of semilinear Poisson problems dened by Lipschitzian nonlinearities as
written below for R3 :
(U )(x) + F (U (x)) + V (x)U (x) = f (x).
(1.78)
Here f (x) L2 () and F (z) are real Lipschitzian functions (i.e. (z1 , z2 )
R R |F (z1 ) F (z2 )| C|z1 z2 | for a uniform bound c).
In order to show the existence and uniqueness through a construction
technique, we rewrite Eq. (1.78) into the form of an integral equation (see
Theorem 1.1)

F (U (x )) + V (x )U (x ) f (x ) 3
1
d x = (T U )(x).
U (x) =
4
|x x |
(1.79)
One can see easily that T denes an application with domain L2 () and
range in L2 (). Besides, T is a contraction application in L2 () for small
domains, as we can see from the following estimate:
||T u T v||2L2 ()

1
1
2
2
2
3
||u
v||
+
||V
||
)
d
x
(c
L2 ()
L2 ()
16 2
|x x |2
C 2 + ||V ||2L2 ()
16 2
(4 diam())||u v||2L2 () .
(1.80)
As a consequence of the Banach Fixed Point Theorem, we have that the

sequence
Un = T (Un+1 )
(1.81)
converges strongly to the solution in L2 () of Eq. (1.79).

The reader should repeat the above-made analysis in the general case
RN (see Eq. (1.23)).
August 6, 2008 15:46
9in x 6in
B-640
ch01
21
1.5. The Abstract Formulation of the Poisson Problem

Let A be a linear operator dened in a Hilbert space H = (H, ,) (A : H H,
A(U + v) = A(u) + A(v)) such that its domain D(A) forms a (nonclosed) vectorial subspace dense in H. Note that D(A) has the explicit
analytical representation

(1.82)
D(A) = u H | Au H Au, AuH < .
Let C such that there is a vector U H, such that AU = U .
Such vector is called the eigenvector of A associated with the eigenvalue .
The set of all these eigenvectors {U } makes an invariant subspace for the
operator A.
We have the following results in operator theory in separable Hilbert
spaces, useful to the Poisson problem solution.
Theorem 1.4. If the operator A satises the symmetric condition (so A
is called a symmetric operator) Au, vH = (u, Av)H (u, v D(A)), then all
eigenvalues are real numbers.
Proof. Let AU = U . As a consequence of the symmetry property of
the operator A, we have that
||2 .
AU , U H = U , U = ||U ||2 = U , AU = ||U
(1.83)
Theorem 1.5. If A is a symmetric operator, then dierent eigenvectors

corresponding to dierent eigenvalues are orthogonals. So A has at most a
countable number of eigenvalues in the separable H.
Proof. Let AU1 = 1 U1 and AU2 = 2 U2 (1 = 2 ). Thus we have the
identity
0 = AU1 , U2 H U1 , AU2 H = (1 2 )U1 , U2 H ,
(1.84)
which means the result searched

U1 , U2 H = 0.
(1.85)
We have our basic result in the theory of the Poisson problem in Hilbert
spaces.
August 6, 2008 15:46
22
9in x 6in
B-640
ch01
Theorem 1.6. (DirichletRiemann Variational formulation of Poisson

problems.) Let A be a positive-denite symmetric operator in a given separable Hilbert space H = (H, , ). (Exists c R+ such that AU, U H
c(U, U )H for any U D(A).) Let the functional given below for any
F Range(A) be
F(A) () = A, H f, H .
(1.86)
Then, the functional equation (1.86) has a minimum value at U D(A)

such that
A U = f,
(1.87)
the converse still holds true.

Let us exemplify the above result for the Poisson problem in bounded
open sets in RN with a Riemannian structure {gab (xa ); a, b = 1, . . . , n}
(manifolds charts).
Thus we consider the following covariant Hilbert space associated with
the given metric structures

dN x g(x)f (x)f (x) ,
L2g () = closure of f Cc () | ,
(1.88a)
1
H0,g
() = closure of

f C01 () | , =
dN x g g ab a f b f ,
(1.88b)

p
H0,g
() = closure of f C0p () | , H 0

=
dN x g(a1 . . . a(p/2) f )(x)
{g a1 ,a(p/2+1) . . . g a(p/2) ap }(x)(a(p/2+1) . . . ap f )(x). (1.88c)

1
(), we consider the functional
In the covariant Hilbert space H0,g
(positive-denite) associated to the LaplaceBeltrami operator g =
1g a ( g ab b ) acting on real functions

Fq () =
dN x

g (g ) (x)

dN x
g f (x)(x).

(1.89)
August 6, 2008 15:46
9in x 6in
B-640
ch01
23
1
H0,g
The minimizing U
() of the above-written functional is the
solution of the covariant Poisson problem in the open domain W :
)(x) = f (x) for x W.

(g U
(1.90)
Let us now analyze the eigenvalue problem in the Dirichlet problem in

the Hilbert space L2 ().
We can accomplish such studies by considering the problem rewritten
into the integral operator form
1
4
U (r ) =
d3r
U (r )
= (T U )(F ).
|r r |
(1.91)
First, let us see that the integral operator T has as its domain C 1 ()
L () and as its range the functional space C 2 (). However, it can be
extended to the whole space L2 () by the HahnBanach theorem
2
T : L2 () L2 ()
(1.92)
since we have the estimate

|(T U )(r )|

1
4
d3r
1
|r r |2
12
12
d3r U 2 (r )
12
1
diam() ||U ||L2 () .
3
(1.93)
We can see that T is a symmetric bounded compact operator with a set

of eigenvalues |r | 4
3 diam() and n 0 for n .
1.6. Potential Theory for the Wave Equation in R3
Kirchho Potentials (Spherical Means)
Let us start this section by dening the PoissonKirchho potential associated with the given C 2 () functions. The rst potential associated with a
certain f (r ) C 2 () is dened by the spherical means (r = (x, y, z) )
U
(1)
t
(r, t, [f ]) =
4

3 (
Sct
r)
f (x + ct, y + ct, z + ct)d.

(1.94)
August 6, 2008 15:46
24
9in x 6in
B-640
ch01
3
3
Here, Sct
(r ) = {Bct
(r )} is the spherical surface centered at the point
r = (x, y, z) with radius |r r| = ct. The spherical parameter , , and
are dened as usual as
= cos sin ,
= sin sin ,
(1.95)
= cos ,
d = sin d d.
The second potential is dened in an analogous way as

1 1 (1)
U (r, t, [g]) .
U (2) (r, t, [g]) =
4 t 2
(1.96)
We have the following elementary initial-value properties for the above

wave potentials, which are the functions dened in the innite domain
R3 R+ :
U (1) (r, 0, [f ]) = 0,
(1)
U (r, 0, [f ]) = f (x, y, z) lim
t
t0+
t
4

3 (
Sct
r)
= f (r ),
(1.97)
U (2) (r, 0, [g]) = g(r ),
U2 (r, 0, [g]) = 0
(exercise).
t
The main dierentiability property of the above-written wave potentials
is the following:
r U
(1)
t
(r, t, [f ]) =
4
d
0

sin dr f (x + ct, y + ct, z + ct) .
(1.98)
We should now evaluate the second-time derivative of the potential
(
x = x + ct, y = y + ct, z = z + ct)
U (1) (r, t, [f ])
t
t
U (1) (r, t, [f ])
+
=
t
4
2
+ c
+ c
c
f (
x, y, z)d
3 (
x
y
z
Sct
r)

August 6, 2008 15:46
9in x 6in
B-640
ch01
25
1
U (1) (r, t, [f ])
+
t
4ct

n

+
+
f (
x, y, z)dA
3 (
x
y
z
Sct
r)
(Gauss theorem)
1
U (1) (r, t, [f ])
+
=
t
4ct
U (1) (r, t, [f ])
1
=
+
t
4ct
since

3 (
Bct
r)

( f )(
x, y, z)d
x d
y d
z

ct
d
0

d A(f )(
x, y, z)
2
3 (
Sct
r)
d
x d
y d
z = d d2 A,
0 ct.
(1.99)
(1.100)
Since we have the usual Leibnitz rule for derivatives inside integrals

ct
d
f (x) dx = f (ct) c.
(1.101)
dt
0
We have the result for evaluation of the second-time derivative of Eq. (1.94):
1
2 (1)
U (1) (r, t, [f ]) 1
2 U (1) (r, t, [f ])
U
(
r
,
t,
[f
])
=
2
t
t
t
t

1
2
+
c
d A(f )(
x, y, z)
3 (
4ct
Sct
r)
1
4ct2

ct
d
0

d A(f )(
x, y, z) .
2
(1.102)
S3 (
r)
Using Eq. (1.99) again to simplify Eq. (1.102), yields

1
2 U (1) (r, t, [f ])
2
=
d
A(f
)(
x
,
y
,
z
)
2t
4t S3 (r )

1
2
2
=c t
d (f )(
x, y, z) = 2 r U (1) (r, t, [f ]).
3 (
c
Sct
r)
(1.103)
As a result of the above dierential calculus evaluations we have the
analogous of Theorem 1.1 (the Cauchy problem for the wave equation).
August 6, 2008 15:46
26
9in x 6in
B-640
ch01
Theorem 1.7. The solution for the Cauchy (initial values) problem for
the wave equation with smooth C 2 () data (including an external forcing
F (r, t), continuous in the t-variable), is given by the wave potentials
2
U (r, t)
1
= 2 r U (r, t) + F (r, t),
c
t
(1.104)
U
(
r
,
0)
=
f
(
r
), r ,
Ut (r, 0) = g(r ), r ,
U (r, t) = U (1) (r, t, [f ]) + U (2) (r, t, [g])

F (r , ct |r r |) 3
1
d
+

r
.
4
|r r |
|
r
r |ct
In the case of R2 we have the Poisson wave potentials:

1
f (r )
2

U (r, t) =
d r
t 2 |rr |<ct c2 t2 |r r |2

1
g(r )
2

+
d r
2 |rr |<ct c2 t2 |r r |2

t
1
F (r , c(t t )) 2

+
dt
d r .
2 0
(t )2 |r r |2
|
r
r |<ct
(1.105)
(1.106)
Let us give a proof for the problem uniqueness for the Cauchy initial
value Eq. (1.104).
In order to show such a result, let us consider the energy wave (functional) inside a spherical ball of radius R:

2

Ut + c2 |U |2 (r, t)d3r.
(1.107)
ER (t) =
BR (0)
Let us analyze its time derivative

dER (t)
lim
=
2Ut Utt + 2c2 t U U (r, t)
R
dt
BR (0)

2
=2
Ut N U dS
Ut (Utt c U ) + O
3 (0)
BR

= 2 lim O
R
3 (0)
SR

Ut N U dS
3 (0)
SR
= 0,

(1.108)
August 6, 2008 15:46
9in x 6in
B-640
ch01
27
if one imposes a sort of HelmholtzSommerfeld radiation condition at

innity (the vanishing of the surfaces term in Eq. (1.108). As a consequence,
if f (r) = g(r) 0 (vanishing of the initial conditions), we have that
E(t) = E(0)

2
= lim c O
R

U U (r, 0) + lim O

Ut N U dS
3 (0)
SR
BR (0)
= 0.
(1.109)
1.7. The Dirichlet Problem for the Diusion Equation
Seminar Exercises
Let us consider the following Dirichlet problem for the diusion equation in
a given domain R3 with a nontrivial time-dependence on the boundary
(an interior problem):
U (r, t) = 2
U(r,t) ,
a
t
(1.110)
U (r, 0) = f (r ) C (),

U (r, t) = g(r, t) C ().

Let us introduce the double-layer like Poisson potential for a not-yet
determined density (r, t)
U
(1)
1
(r, t) =
4
=

d O
1
8a2

2r
(r , )

4a
(t)
e
dS(r )
(t ) N

t
d O

0

r
(r , ) 4a2 (t)
(r )dS(r ) .
e
cos(
N
(t )2
(1.111)
We observe that U (1) (r, t) satises Eq. (1.110) and vanishes at t = 0

(exercise).
The solution for the Dirichlet-like problem Eq. (1.110) will be determined from the ansatz

U (r, t) = U
(1)
(r, t) +
U (2) (
r,t)
d r f (r )e
(2a2 t) 2

3


2
r
r )
(
4a2 t
(1.112)
August 6, 2008 15:46
28
9in x 6in
B-640
ch01
It is clear that the formal solution written above satises the boundary
condition in Eq. (1.110) if one can determine the density function (r, t)
from the integral equation coming from the imposition of the boundary
condition as given by Eq. (1.110) (exercise):

U (r, t) = g(r, t) = U (2) (r, t)

+ (r, t) +
1
8a2

cos(N r )dS(r ) .

d O
2
(r , t) 4a2r(t)
e
r
(t )2
(1.113)

One can show that for small densities (r, t) (r, t) with 1 ,
one can solve Eq. (1.113) by means of the xed point theorem of Banach
(exercise) and yielding the solution as a power series in (the size of the
area of ).
1.8. The Potential Theory in Distributional Spaces The

GelfandChilov Method
Seminar Exercises
In this brief section, we intend to show the general method of Kirchho
Poisson potentials in the context of Schwartz distribution theory in S (RN ).
Let us start this section by considering the problem of determining the
fundamental solution of an arbitrary elliptic dierential operator of order
m in the whole space RN with constant coecients

a Dx
U (x) = (N ) (x).
(1.114)
||m
Here U (x) shall belong to S (RN ). In the distributional sense2

U (x), (x) =

||m
t
a Dx

(x)
(1.115)
x=0
Let us consider the analytical regularization of the Dirac distribution in

the right-hand side of Eq. (1.114) (the well-known analytical regularization
August 6, 2008 15:46
9in x 6in
B-640
ch01
29
scheme in quantum eld theory) leading to a usual function-theoretic partial

dierential equation expressed below:

2|x|

a Dx U () (x) =
n +n
2
||m

1

=
|x r | dS(r ) .

n1
(n)
n 2 +n
S1 (0)
2
(1.116)
At this point one can consider the following analytical regularized associated problem:

|r r |
(1.117)
(L U )(r ) =
a Dr v () (r ) =
+n
n1
2
n
2
||m
Due to the linearity of Eq. (1.116), we should search for a solution of this
analytical regularized equation in the linear-superposing form of a spherical
mean (Radon transforms)

()
d(r ) v () (r, r ).
(1.118)
U (r ) =
(n)
S1
(0)
The solution of Eq. (1.117) reduces to the solution of an ordinary nonhomogeneous dierential equation in R after introducing the variable
d
, where r = {xa }a=1,...,n and r S1n (0) r =
change x a = a d
a
1 2
{ }a=1,...,n [( ) + + ( n )2 = 1)] with < < +.
It yields the following simple result:

d
||
(1.119)
L a
v () =

n1
d
n 2 +n
2
The solution of Eq. (1.119) is easily written down in the distributional
sense in S (R):

+
1
()

v () =
G(, )| | d .
(1.120)

n1
n 2 +n
2
Here, G(, ) is the Green function of the ordinary dierential operator in
Eq. (1.119), obtained through Fourier transforms in S (R), namely,

d
L a
(1.121)
G(, ) = (1) ( ).
d
August 6, 2008 15:46
30
9in x 6in
B-640
ch01
Now it can be shown case by case that the limit of n in our

already-obtained distribution-regularized solutions converges in the distributional sense to the expected S (RN )-distributional solution of Eq. (1.114).
As an exercise, the reader can nd the distributional solution of the
-power Laplacean through the analytical regularized nonhomogeneous
ordinary dierential equation. (Here C, and 0 < Real() < n).

2
d
d2
||

v () =
L a
= ((1 )2 + (2 )2 + (3 )2 ) 2
d
d
3 +n
2
(1.122)
The full solution is given by the famous RieszPoisson potential in RN
(compare with Eq. (1.41):
() 2 U (x) = f (x),

(n)
f (x )
dN x
.
U (x) = 2 (n/2)
|x x |n
2 2
RN
(1.123)
References
1. C. Hilbert, Methods of Mathematics Physics, Vol. I, II (Interscience
Publishers, Inc., New York, 1978).
2. B. Simon, J. Math. Phys. 12, 140 (1971).
3. M. Lavrentiev and B. Chabat, Methodes de la theorie des fonctions dune
variable complexe, Edition

MIR (Moscou, 1976).
4. L. C. L. Botelho, Methods of Bosonic and Fermionic Path Integrals Representations: Continuum Random Geometry in Quantum Field Theory (Nova
Science Publisher, NY, USA, 2008).
5. C. Luiz and L. Botelho, Il Nuovo Cimento 112A, 1615 (1999).
6. S. Weinberg, Gravitation and Cosmology (John Wiley & Sons Inc., 1992).
7. P. G. Bergmann, Introduction to the Theory of Relativity (Prentice-Hall, Inc.,
England, Clis, NJ, USA).
8. V. Fock, The Theory on Space, Time and Gravitation, 2nd revised edn.
(Pergamon Press, 1964).
9. S. Sternberg, Lectures on Dierential Geometry (Chelsea Publishing
Company, NY, 1983), Theorem 4.4, p. 63.
10. S. W. Hawking, General Relativity: An Einstein Centenary, Survey
(Cambridge University Press, 1979).
11. H. Flanders, Dierential Forms (Academic Press, 1963).
12. J. Milnor, Topology from the Dierentiable Viewpoint (University Press of
Virginia, 1965).
August 6, 2008 15:46
9in x 6in
B-640
ch01
31
Appendix A. Light Deection on de-Sitter Space

A.1. The Light Deection
The most general covariant second-order equation for the gravitation eld
generated by a given (covariant) enegymatter distribution on the space
time is given by the famous Einstein eld equation with a cosmological
constant with dimension (length)2 (Ref. 5), namely,

1
(A.1)
R (g) g R + g (x) = 8GT (x),
2
where x belongs to a spacetime local chart.
It is well known that studies on the light deection by a gravitational
eld generated by a massive point-particle with a pure time like geodesic
trajectory (a rest particle sun for a three-dimensional spatial spacetime
section observer!) is always carried out by considering 0.6,7
Our purpose in this appendix is to understand the light deection
phenomena in the presence of a nonvanishing cosmological term in Einstein
equation (A.1), at least on a formal mathematical level of solving trajectory
motion equations.
Let us, thus, look for a static spherically symmetric solution of Eq. (A.1)
in the standard isotropic form6,7
(ds)2 = B(r)(dt)2 A(r)(dr)2 r2 [(d)2 + sin2 (d)2 ].
(A.2)
In the spacetime region r = + | x |2 > 0, where the matterenergy

tensor vanishes identically, we have that the Einstein equation takes the
following form:
R (g)(x) = (g (x)).
(A.3)
In the above cited region, the Ricci tensor is given by
B(r)
0
0
0
0
A(r)
0
0
g =
2
0
0
0
r
2
2
0
0
0
r sin
Rtt
0
=
0
0
0
Rrr
0
0
0
0
R
0
0
0
.
0
R

(A.4)
August 6, 2008 15:46
32
9in x 6in
B-640
ch01
We have, thus, the following set of ordinary dierential equations in place

of Einstein partial dierential Eq. (A.1)

1 B
A
B
B
1 B
+
+
Rtt =
= B,
(A.5)
2A 4 A
A
B
r A

1 B
A
B
B
1 A
+
Rrr
= A,
(A.6)
2B
4 B
A
B
r A

B
r
A
1
(A.7)
R = 1 +
+
+ = r2 ,
2A
A
B
A
R = sin2 R = (sin2 )r2 .
(A.8)
At this point we note that

Rrr
Rtt
+
= 0,
B(r) A(r)
(A.9)
or equivalently
A(r) =
,
B(r)
(A.10)
where is an integration constant.

2
Since R = 1 + r B + B
= r , we get the following expression for
the B(r) function:
B(r) =
r2
++ ,
3
r
(A.11)
with denoting another integration constant.

In the literature situation,6,7 one always considers the case = 0 in
a pure classical mathematical vacuum situation context, the so-called deSitter vacuum pure gravity. However, in our case it becomes physical to
consider that our solution depends analytically on the cosmological constant. In other words, if the parameter 0 in our solution, it must
converge to the usual Schwarzschild solution with a mass singularity at the
origin r = 0, that is our boundary condition hypothesis imposed on our
solution.

August 6, 2008 15:46
9in x 6in
B-640
ch01
33
As a consequence, one gets our proposed Schwarzschildde-Sitter

solution
2
r
2M G
2
+1
(ds) =
(dt2 )
3
r

r2
2M G
+1
3
r
1
(dr)2 r2 [(d)2 + (sin2 )(d)2 ].
(A.12)
At this point let us comment that for the spacetime region exterior to
(1/3)
, the eld gravitation approximation leads
the spatial sphere r > ( 3mG
)
to the antigravity (a repulsion gravity force) if < 0; So, explain from
this Einstein Gravitation theory of ours the famous Huble accelerating
Universe expansion.
In what follows we are going to consider a nonvanishing < 0 and study
the path of a light ray on such negative cosmological constant Einstein
manifold Eq. (A.12).
We have the following null-geodesic equation for light propagating in
= /2 plane (Einstein hypothesis) for light propagation in the presence
of the sun (Sec. A.2 for the related formulae):
0 = B(r)
1
B(r)
dr
dt
2
r2
d
dt
2
.
At this point we note that

d
B(r)J
,
=
dt
r2
(A.13)
(A.14)
where J is an integration constant.

After substituting Eq. (A.14) into Eq. (A.13) we have the following
dierential equation for the light trajectory as a function of the deection
angle
2

dr d
J 2 B 3 (r)
+
B 2 (r) = 0,
(A.15)
d dt
r2
which is exactly integrable:
'
d =
r2
dr
1
J2
.
B(r)
r2

(A.16)
August 6, 2008 15:46
34
9in x 6in
B-640
ch01
By
supposing a deection point rm where
rm / B(rm ), we get the deection angle

1 =
rm
dr
B(rm )
r2 [ r 2
m
B(r) (1/2)
r2 ]
=
0
1
rm
dr
dt
= 0 and, thus, J =
dU
,
2 U 2 ) 2M G(U 3 U 3 )](1/2)
[(Um
m
(A.17)
which is exactly the one given in the pure ( = 0) Schwarzschild famous
case. However, if one supposes that there is no deection (a continuous
monotone trajectory r = r()!), the total deection angle now depends on
the cosmological constant and is given formally by the expression below:
rm
1
1
= 1 . (A.18)
+
'
dr
2 =
3J 2
r2 12 + 2mG
r 3 ( 3J 2 )
r
r3
1 + 2MGr
As a general conclusion of our note we claim that the usual lightdeection experimented test does not make any dierence between the usual
noncosmological Schwarzschild case and our case Eq. (A.12), and, thus,
it should not be considered as a denitive physical support for Einstein
General Relativity without cosmological constant.
A.2. The Trajectory Motion Equations
The body trajectory (t(p), r(p), (p), (p)) in the presence of the gravitational eld generated by the metric Eqs. (A.2)(A.10) is described by the
following geodesic equations:

B dr
dt
d2 t
+
= 0,
(A.19)
d2 p
B dp
dp
A
d2 r
+
d2 p 2A
dr
dp
d
dp
r sin2
d 2 d dr
+
sin cos
d2 p r dr dp
d
dp
d
dp
B
+
2A
dt
dp
2
= 0,
(A.20)
2
= 0,
d2 2 d dr
d d
+
+ 2cotg()
= 0.
d2 p
r dp dp
dp dp
(A.21)
(A.22)
August 6, 2008 15:46
9in x 6in
B-640
ch01
35
At this point we remark that by multiplying Eq. (A.19) by B(r(p)),

it reduces to the exact integral form relating the Einstein proper-time
(physical evolution parameter) p with the geometrical-dependent coordinate Newtonian time t:
1
dt
=
.
dp
B(r)
(A.23)
We remark either that Eq. (A.22) can be rewritten in the form

d
d
2
+ nr + 2 n sin = 0,
(A.24)
n
dp
dp
which reduces to the following form:

d 2
2
r (p) sin ((p)) = J,
dp
(A.25)
where J is an integration constant.

By substituting Eqs. (A.23) and (A.25) into Eqs. (A.20) and (A.21) we
obtain the full set of equations describing the body trajectory in relation
to the (r, ) variables
B
d2 r
d2 p 2B
dr
dp
J 2B
B
= 0,
+
r3
2B
(A.26)
d2 2 d dr
cos J 2
+
= 0.
d2 p r dr dp sin3 r4
(A.27)
rB
d
dp
For Einstein hypothesis of light propagation on the plane = /2,

Eq. (A.27) vanishes and Eq. (A.26) takes the form
d2 r
B
2
d p 2B
dr
dp
J 2B
B
=0
+
3
r
2B
(A.28)
or in a more manageable alternative form after multiplying Eq. (A.28) by

2 dr
B ( dp ) and by using Eq. (A.23) for exchanging the geometrical parameter p
by the time manifold coordinate

dr
dt
1
J2
1
+
+ E = 0,
B3
r2
B
where E denotes another integration constant.

(A.29)
August 6, 2008 15:46
36
9in x 6in
B-640
ch01
2
By writing r as a function of and using Eq. (A.25)
get our nal trajectory equation
d r 2
dt B

(1/2)
1
dr
B
BE
2
= r
2 2
,
d
J2
r
J
which leads to the body trajectory geometric form

dr
.
=
1
2
(1/2)
r B
[ J 2 B JE2 r12 ](1/2)
3
= J! , we
(A.30)
(A.31)
Note that for light propagation the integration constant E always

vanishes, a result used in the text by means of Eq. (A.16).
A.3. On the Topology of the Euclidean SpaceTime
One of the most interesting aspects of Einstein gravitation theory is the
question of the nonexistence of holes in the spacetime C 2 -manifold from
the view point of a mathematical observer situated on the Euclidean space
R9 associated with the minimal Whitney imbedding theorem of M on
Euclidean spaces.9
In order to conjecture the validity of such a topological spacetime
property, let us suppose that M is a C 2 -manifold, and the analytically continued (Euclidean) matter distribution tensor generating the (Euclidean)
gravitation eld on M allows a well-dened Euclidean metric tensor
(solution of Euclidean Einstein equation).10
At this point we note that M must be always orientable in order to
have a well-dened theory of integration on M and, thus, the validity of
the rule of integration by parts: Stokes theorem is always needed in order
to construct matter tensor energy momentum. Since Euclidean Einsteins
equations say simply that the sum of sectional curvatures is a measure
of the (classical) matter energy density generating gravity, which must be
always considered positive, it will be natural to expect the positivity of the
Euclidean. EnergyMomentum of the matter content leads to the result
that the associated sectional curvatures are positive individually. Since M
is even-dimensional (four), the famous Synges theorem8 leads to the result
that M is simply connected (note that this topological property is obviously independent of the metric structure being Lorentzian or Euclidean!)
and as a direct consequence of this result, any physical geodesic (particles
August 6, 2008 15:46
9in x 6in
B-640
ch01
37
trajectory) on M can be topologically deformed to a point, and, thus, M

does not possess holes from the point of view of the Whitney imbedding
extrinsic minimal space R9 .
Finally, let us argue that the existence of a (symmetric) energy
momentum tensor on M is associated with the General Relativity
description of the spacetime manifold M by means of charts (the Physics
is invariant under the action of the dieomorphism group of M ), which
by its turn leads to the existence of the matter energymomentum tensor
by means of Noether theorem (a metric-independent result) applied to the
matter distribution Lagrangean (a scalar function dened on the tangent
bundle of M ).
As a consequence, let us conjecture again that the introduction of a
cosmological term on Einstein equation spoils the physical results presented
on the whole topology and the physical requirement of positivity of the
matterenergy universe moments tensor, given, thus, a plausible topological
argument for the vanishing of the cosmological constant at the level of the
globaltopological aspects of the spacetime manifold.
Finally, let us show the mathematical formulae associated with our ideas
and conjectures written above.
Let e0 , . . . , e3 be an orthonormal frame at a point of M (Euclidean). It
is well known that the Ricci quadratic form can be expressed in terms of
sectional curvatures

K(ei ej ),
(A.32)
Ric (ei , ei ) =
j=i
and the Einstein tensor is dened by

1
Gij = Rij gij R.
2
(A.33)
Since the Einstein equation reads in terms of quadratic forms associated

with the sectional curvatures as

G(ep , ep ) = +
K(e
(A.34)
p ) = T (ep , ep ),
with Tij being the matter energy tensor and e
p the basis two-plane
orthogonal to ep , one can in principle write the sectional curvatures
K(ep eq ) in terms of the quadratic energymomentum sectional curvatures
Tij (ep , eq ) at least for short-time cylindrical geometrodynamical space
time congurations as expected in a Quantum theory of gravitation (see
August 6, 2008 15:46
38
9in x 6in
B-640
ch01
Ref. 5). For the two-dimensional case this assertive is straightforward as

one can see from the relations below:
G(e0 , e0 ) = K(e1 e1 ),
(A.35)
G(e1 , e1 ) = K(e0 e0 ).
(A.36)
As a consequence, one should conjecture that the positivity of the

energymomentum tensor T (ep , eq ) leads to the individual positivity of the
sectional curvature set K(er , es ) on the basis of Eq. (A.34), namely,

G(ep , eq ) = T (ep , eq ) = Ric (ep , eq ) pq
K(ei ej ) .
(A.37)
i,j,i=j


6856 Chap01

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

6856 Chap01

Transféré par

Droits d'auteur :

Formats disponibles

August 6, 2008 15:46

1.2. The Laplace Dierential Operator

for any (k, m) {1, . . . , n}.

Let us consider the following linear transformation with domain C 2 ()

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

August 6, 2008 15:46

Lecture Notes in Applied Dierential Equations of Mathematical Physics

Ut (x, y, z) = f(ix cos t + iy sin t + z, t).

We wish to show that

This result is a simple consequence of the following elementary

Note that a whole class of harmonic functions H()

dt g(t)(f(w, t)|w=ix cos t+i sin ty+z ),

and for the special function below:

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

August 6, 2008 15:46

Elementary Aspects of Potential Theory in Mathematical Physics

we have the harmonic legendre polynomials in the three-dimensional

dt eit (r cos + ir sin cos cos t + ir sin sin sin t)n

dt cos(t)(cos + (sin sin t)n

After these preliminary elementary remarks we pass on the discussion

is inversible, and its inverse has the following integral representation:

d3r f (r ) G(0) (|r r |).

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

August 6, 2008 15:46

Lecture Notes in Applied Dierential Equations of Mathematical Physics

Proof. Let us re-introduce the classic mathematical-physicist vectorial

Let us consider the -approximant solution for the problem

Here the regularized Green function is given by

for |r r | > ,

Note that the above regularized Green function is a continuous

We now have the estimate for 0:

which shows the uniform convergence of the sequence of approximate

(4) 2 f L2 2() < .

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

August 6, 2008 15:46

Elementary Aspects of Potential Theory in Mathematical Physics

converges uniformly to the (well-dened) function in C()

The proof of the above made claim is a consequence of the following

Let us now pass on the problem of evaluating second derivatives.

August 6, 2008 15:46

Lecture Notes in Applied Dierential Equations of Mathematical Physics

with the 4-overall factor in Eq. (1.10)):

Now we have that

and (with xa = (x, y, z) and a = 1, 2, 3):

|xa xa |2 |r r | 3

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

August 6, 2008 15:46

Elementary Aspects of Potential Theory in Mathematical Physics

Note that the regularized Green function G() is now given by

In R2 , we can show that (left to our readers)

a very useful problem in string path integrals for R2 .

An explicit solution for Eq. (1.27) is easily obtained by considering the

August 6, 2008 15:46

Lecture Notes in Applied Dierential Equations of Mathematical Physics

where H(r, r ) is a harmonic function in in relation to the variable r ,

Theorem 1.2. We have the integral representation for the Dirichlet

G() (r, r ) = H () (r, r ) + U () (r r ).

By applying the second Green identity to the pair of functions U (r )

U (r ) N G() (r, r )dS(r ).

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

August 6, 2008 15:46

Elementary Aspects of Potential Theory in Mathematical Physics

LECTURE NOTES IN APPLIED DIFFERENTIAL EQUATIONS OF MATHEMATICAL PHYSICS

(4) 2 f L2 2() < .

U (r ) N G() (r, r )dS(r ).

inner product f, gH 1 =