Fall 2023: Gravitation and Cosmology (PH 6450 / EP 4350)

Lecture 2: Newton’s Gravitation Reloaded

August 7, 2024

1 Poisson and Laplace equations: Solution in a bounded region

In lecture 1, we deduced that gravitational field is specified by a single scalar1 quantity, namely the
Newton scalar potential Φ(x) which obeys Poisson’s equation,

∇2 Φ(x) = 4πGN ρ(x), (1)

just as the electrostatic potential does. In regions outside mass sources (ρ(x) = 0), the potential obeys
Laplace equation,
∇2 Φ(x) = 0.
We also recalled the Uniqueness Theorem concerning the kind of boundary data one needs to supply
to uniquely solve or determine the gravitational Newton’s scalar potential in a bounded region (say Γ).
According to the Uniqueness Theorem there are two distinct boundary data which determine the
solution to Poisson’s equation in region Γ uniquely, to wit,
• Dirichlet data i.e. Φ has to be specified on the boundary (say ∂Γ) or alternatively,
∂Φ
• Neumann data, i.e. the normal derivative of the potential ∂n needs to be specified at all points on
the boundary ∂Γ
Let’s try to write down the solution to Poisson equation (1) using the Coulomb Green function, G(x, x0 ) =
1 0
4π|x−x0 | where x, x are points in Γ. Evidently this satisfies,

∇2 G(x, x0 ) = ∇02 G(x, x0 ) = −δ 3 (x − x0 ).

Next we utilize the Green identity,

ˆ   ‹
2
d3 x0 ϕ ∇02 ψ − ψ∇0 ϕ = dS(x0 ) ϕ n̂ · ∇0 ψ − ψ n̂ · ∇0 ϕ


Γ ∂Γ

and substitute ϕ = Φ(x0 ) and ψ(x0 ) = G(x, x0 ). We obtain,

 
 
ˆ ‹  
d x Φ(x0 ) ∇02 G(x, x0 ) −G(x, x0 ) ∇02 Φ(x0 )  =
3 0
dS(x0 ) Φ(x0 ) n̂ · ∇0 G(x, x0 ) −G(x, x0 ) n̂ · ∇0 Φ(x0 ) ,
   
Γ | {z } | {z } ∂Γ  | {z } | {z } 
=−δ 3 (x−x0 ) =4πGN ρ(x0 ) 1
=− 4π x0 −x = ∂Φ
∂n
|x−x0 |3

which, after some simplifications, yields

ˆ 3 0 ‹ ‹
d x ρ(x0 ) dΩ(x0 , x) 0 1 dS(x0 ) ∂Φ(x0 )
Φ(x) = −GN 0
+ Φ(x ) + . (2)
Γ |x − x | ∂Γ 4π 4π ∂Γ |x − x0 | ∂n
1
scalar under rotations and Galilean boosts

1
Here
dS(x0 ) n̂ · (x0 − x)
dΩ(x0 , x) ≡
|x − x0 |3
is the solid angle subtended by the directed area element dS(x0 ) n̂ on the boundary to the interior point
x. The equation (2) represents the solution of the scalar potential at any point x inside the region Γ.
But this solution is of the Cauchy form, in the sense that in the RHS one needs both the information
about the potential and its normal derivative on the boundary (second and third terms in the RHS of
(2))! On the other hand the uniqueness theorem states that one needs either the potential or its normal
derivative, but not both to determine the potential. So this presents a puzzle: if we only need either
Dirichlet data or Neumann data to unique determine the scalar potential, why does the (more) general
form of the solution (2) appear to require Cauchy data (i.e. both Dirichlet and Neumann data). The
resolution of this puzzle is left as a homework exercise.

2 Gravitational Potential Energy and the Energy stored in the gravi-

tational field
Newton’s gravitation is not a field theory, but an action-at-a-distance theory, which means the gravi-
tational energy is only present in/possessed by the masses and no energy is stored in the empty space
outside the sources. However in relativistic theory of gravitation, one will eventually have to introduce
a energy-momentum carrying entity “gravitational field” in the region surrounding (and inside) massive
bodies. By a heuristic argument, one can arrive at the energy (density) contained in static gravitational
field from Newton’s theory despite the fact that Newton’s theory is not a field theory. Here we “de-
rive” this energy stored in the region of space around massive bodies. To begin with, we recall that the
gravitational potential energy of a pair of (point) gravitational masses mi and mj located at xi and xj
respectively is,
mi mj
Uij = −GN .
|xi − xj |
This is a negative definite quantity. Now when we have a collection of such point masses then we need to
sum over all such pairs, X
U= Uij
i,j;i>j
where we are summing over all masses labeled by indices i and j, but we maintain the condition, i > j
so that we count a pair only once. If we remove this restriction and sum over all i and j then we are
counting each pair twice, so we need to put a half,
X1
U = Uij
i,j
2
1 X mi mj
= − GN .
2 i,j
|xi − xj |

Instead of a point mass distribution, if we have a continuous distribution of mass, this turns into
ˆ
1 ρ(x)ρ(x0 )
U = − GN d3 x d3 x0 .
2 |x − x0 |
Note that since every quantity inside the integral sign is positive, the potential energy is negative definite
(because this is bound system held together by gravitational attraction and it will take positive work
done to rip it apart). Now recall that the gravitational potential at x is given by,
ˆ
ρ(x0 )
Φ(x) = −GN d3 x0 ,
|x − x0 |

2
so the energy can be rewritten as, ˆ
1
U= d3 x ρ(x) Φ(x). (3)
2
This is a negative definite quantity as well since the starting point as negative. To extract the energy
stored in the region outside the massive objects or sources, which is positive definite, let us split up this
energy expression in the following manner2 ,
ˆ ˆ
1
 
3 3
U = d x ρ(x) Φ(x) + − d x ρ(x) Φ(x) .
2
| {z } | {z }
Umatter Uf ield

Next, we use the Poisson equation in the second term to obtain,

ˆ ˆ
1 3 1  
Uf ield = − d x ρ(x) Φ(x) = − d3 x ∇2 Φ Φ
2 8πG
ˆ
1 h i
= − d3 x ∇ · (Φ∇Φ) − (∇Φ)2
8πG
‹ ˆ
1 1
= − dS · Φ∇Φ + d3 x (∇Φ)2 .
8πG ∞ 8πG
where we have used the Gauss divergence theorem to convert the first term into a surface integral at
(spatial) infinity. Now assuming the mass distribution is localized (i.e. does not extend all the way to
1
infinity), one has, Φ ∼ |x| , and so the integrand of the surface integral term falls off as Φ∇Φ ∼ |x|1 3 , while
the elemental surface area grows as dS ∼ |x|2 . As a result the surface integral at infinity vanishes and we
get, ˆ ˆ
1 1
Uf ield = − 3
d x ρ(x) Φ(x) = d3 x (∇Φ)2 .
2 8πG
Since this is a volume integral, we can write down an energy density of the form,
1 1 2
uf ield = (∇Φ)2 = g .
8πG 8πG
This result holds even at regions where there is no matter, i.e. ρ = 0, but a non-vanishing gravitational
field, g. So that is why it is natural to think of this as the density of energy contained in the region
external to the massive bodies i.e. the gravitational fields in the relativistic theory (this explains the
subscript, “f ield”).

3 Multipole Expansion of the Newton scalar potential

Newton proved that for a perfectly spherical massive body of mass M , the gravitational potential (and
hence the gravitational field) at any point outside the sphere is identical to the expression of potential
(and field) for a point mass sitting at the center of the sphere. However for non-spherical sources/objects
the gravitational potential or field does look like that of point mass and instead there are corrections
due to finite size/extension of the massive source- all of those corrections are captured by the Multipole
expansion of the potential. The gravitational potential at the point x due to the extended source is
given by, ˆ
ρ(x0 )
Φ(x) = −GN d3 x0 . (4)
|x − x0 |
2
Right now there is no legit justification of this way of splitting up the energy into a contribution Uf ield purely located in
the (static) gravitational field region, outside the masses and a contribution, Umatter which is localized in the mass distribution
and this is indeed totally ad hoc. However when we get to Fierz-Pauli theory we will see that indeed in the non-relativistic
approximation, Uf ield will turn out to be the energy contained in the gravitational field outside matter-energy sources.

3
The origin of coordinate system is fixed to some point within the source, let’s say center of mass. i.e.
x0CM = 0. Now if the point x is far from the source, the vector x is very large when compared to a
point within the source, say x0 . So we Taylor expand |x−x
1 0
0 | around x = 0, i.e. the center of mass of the

extended source:
! !
1 1 ∂ 1 1 ∂ ∂ 1
= + x0i + x0i x0j
|x − x0 | |x| ∂x |x − x0 |
0i
x0 =0 2! ∂x ∂x |x − x0 |
0i 0j
x0 =0
!
1 ∂ ∂ ∂ 1
+ x0i x0j x0k + ...
3! ∂x0i ∂x0j ∂x0j |x − x0 | x0 =0
! !
1 xi 1 −δ ij xi xj
= + x0i + x0i x0j + 3
|x| |x|3 2! |x|3 |x|5
!
1 δ ij xk + δ jk xi + δ ki xj xi xj xk
+ x0i x0j x0k −3 + 15 + ....
3! |x|5 |x|7

Plugging this expansion in the formula for the potential (4), we arrive at the expansion,
ˆ
ρ(x0 )
Φ(x) = −GN d3 x0
|x − x0 |
ˆ  i 
δ ij xi xj
  
3 0 0 1 0i x 1 0i 0j
= −GN d x ρ(x ) +x + x x − 3 +3
|x| |x|3 2! |x| |x|5
ij k jk i ki j
xi xj xk
  
1 0i 0j 0k δ x +δ x +δ x
+ x x x −3 + 3.5 + ...
3! |x|5 |x|7
´ 3 0 ˆ
d x ρ(x0 ) xi xi xj
ˆ 
3 0 0 0i 1
d3 x0 ρ(x0 ) −δ ij x02 + 3x0i x0j


= −GN − GN 3
d x ρ(x ) x − G N 5
|x| |x| 2! |x|
ij k jk i ki j
ˆ
xi xj xk

1 δ x +δ x +δ x
− GN −3 + 3.5 d3 x0 ρ(x0 ) x0i x0j x0k + . . .
3! |x|5 |x|7
i j i j k
M x·D 1 ij x x 1 ijk x x x
= −GN − GN − G N Q − G N Q + ... (5)
|x| |x|3 2! |x|5 3! |x|7

where,
ˆ
D i
= d3 x0 ρ(x0 ) x0i ,
ˆ  
Qij
= d3 x0 ρ(x0 ) 3x0i x0j − x02 δ ij ,
ˆ  
Qijk = d3 x0 ρ(x0 ) 15x0i x0j x0k − x02 x0i δ jk − x02 x0j δ ki − x02 x0k δ ij .

The first term is called a monopole contribution and is equal to the potential in the case if the entire
mass is concentrated at a point, namely the origin which is the center of mass in our case. The second
term is called the dipole term since it is proportional to the (dipole) moment of the mass, Di (The dipole
moment is a vector quantity). About the center of mass, Di = 0 by definition of the center of mass.
So the dipole contribution vanishes. The third term is non-vanishing in the CM frame and is called a
quadrupole term since it is proportional to the quadrupole moment of mass, Qij . (The Quadrupole
moment is a symmetric rank 2 tensor). This quadrupole term is the first nonvanishing correction to the
leading term. Similarly the fourth term is the octupole term (octupole moment is a 3-tensor, symmetric
in all three indices) and so on.