On the Unification of Electromagnetism and Gravity

by Raymond Côté

Abstract

In this paper we will extend Maxwell’s and Lorentz’s equations of

electromagnetism to five (4+1) dimensions in a process modeled after Kaluza-
Klein theory. Then we will extract a set of equations which are formally identical
to those of Newton’s theory of gravity.

Introduction

Because many readers are unfamiliar with tensor analysis I will begin this
paper with a very brief and non-rigorous exposition. Since we will not need the
machinery of coordinate transforms or curved spacetime in this article I will limit
my comments essentially to notation.
Most readers will be familiar with the concepts of scalars and vectors.
Tensors represent a generalization of these. An intuitive way to look at tensors is
to view them as indexed quantities. We say that a scalar is a tensor of rank zero.
This means the quantity has zero indices. In other words it’s just a number or
function. A vector is a tensor of rank one. This means that a vector is a quantity
with one index. The index runs from 1 to n where n is the dimension of the space
under consideration. Often the index is renumbered from 0 to n-1, and this is in
fact what we will do in this paper. As an example we can consider the position
a 0 1
vector in 3+1 spacetime, x = (t , x, y, z ). Note that x = t , x = x, etc.
The index need not be in the upper position. We can have vectors with a
subscript instead of a superscript, e.g. xa = (−t , x, y, z ). Note that the sign of t
changed, whereas the signs of x, y, and z didn’t. This is because of the form of
the so called metric tensor. This is a tensor of rank two and is used to define
distances and raise and lower indices. This is done as follows:

n −1

xa = ∑ g
b

b =0
ab x,

where g ab
is the metric tensor.
An important bit of notation should be introduced here because it will save
a lot of typing. If an index is repeated in both the superscript and subscript
positions then we sum over that index. Thus we can write:
x =g x
b
a ab

which means exactly the same thing as our previous equation.

ab
g g = δ c , where δ
a a
Our metric tensor is defined in such a way that c
is zero if
bc
a≠c and one if a=c.
We have one final piece of notation to introduce, that of partial
a
y
a
y
c
differentiation. The partial derivative of with respect to x is ,c
. Note that
the superscript of x has become a subscript. Also note the use of a comma to
a
denote differentiation. Strictly speaking y ,c
is not a tensor of rank two. However
one could define a process called covariant differentiation that would create
tensor by differentiation. For our purposes we will not need to make this
distinction.
In order to connect all this with more familiar mathematics in three
dimensions note that ϕ is the gradient of ϕ . The divergence of a vector K, is
,a
a
K ,a
. Due to our summation convention we sum over the index a which in effect
cancels out the two indices so that we end up with a tensor of rank zero, which is
also a scalar. This is in accordance with our knowledge that the divergence of a
vector is a scalar. Likewise the gradient of a scalar is a vector. The curl of a
vector is somewhat more complicated but we have the result
curl K a = K b ,a − K a ,b . Though we are used to thinking of the curl of a vector as a
vector it is actually a tensor of rank two.

Tensor Formulation of Maxwell’s Equations

James Clerk Maxwell derived a set of eight equations to describe the

electromagnetic field. These are augmented by Lorentz’s three equations of
motion for a particle in an electromagnetic field. Using standard vector notation
we can make this more compact by having four electromagnetic equations and
one motional equation. Tensor notation allows an even more compact form by
having just two electromagnetic equations and one motion equation.
ab
This is effected by introducing the Faraday tensor, F . Note that we are
assuming four dimensions here. Our full equations are then:

ab a
J
1) F ,b =
2) F + F + F
ab , c ca ,b bc , a
=0
2 a
d x =q v ab
3)m
dτ
F 2 b
 =( ρ , J , J , J ) is the current density four-vector, q is charge, τ is
a x y z
where J
proper time, and vb is velocity. For simplicity I have assumed that the speed of
light is 1.
One way to look at the Faraday tensor is as a square matrix:

 0
 E x E y E z 
ab − Ex 0 Bz − B y 
F =  − E y − Bz 0 Bx  .
 
− E B y Bx
− 0 
 z 

As a quick example set a=0 in equation (1). This gives,

∂ Ex ∂ Ey ∂ Ez
+ + = ρ.
∂x ∂y ∂z

The remaining equations can all be found with a little algebra!

Towards Unification

Interestingly nothing in our tensor notation requires us to use four

dimensions. We can easily extend the notation to arbitrary dimensions. From
now on we will work in five dimensions; four of space and one of time. We can
easily extend our equations of electromagnetism to five dimensions, but we will
add a simple constraint. We will assume that all fields are independent of the fifth
dimension. This is exactly what was done in early Kaluza-Klein theory.
The Faraday tensor is an antisymmetric tensor which means that
ab ba
F = − F . As a result when we extend it to five dimensions we have a total of
ten independent components, instead of the six in 3+1 spacetime. We already
know what six of these components are; they are simply the E and B field
components. The remaining four naturally decompose into a three vector and a
ab
scalar. I will thus write the components of F as

 0 Ex Ey Ez φ 
 
− Ex 0 Bz − B y gx
 
F =  − E y − B z 0 B x
ab
g y

 − E z B y − Bx 0 gz
 − φ − 0 

gx − gy − gz 
 A crucial aspect to the recovery of Maxwell’s equations is the assumption
that the fields are independent of the fourth space dimension. In equation (1) we
see that when the index a is not four we simply regain four of Maxwell’s
equations with any change. Similarly in equation (2) when no index is four we
again have Maxwell’s equations. So we will begin by looking at equation (1) with
the index a equal to four. This leads to:

40 41 42 43 44 4
F +F
,0 ,1
+ F , 2 + F ,3 + F , 4 = J
−φ − g −g −g
4
,t x, x y, y z,z
= J
4
If we interpret J as mass density then we can rewrite this as

∂φ
4)∇ ⋅ g + = −ρ
∂t

Continuing in this fashion we can then derive:

5)∇ × g = 0
∂g
6)∇φ =
∂t

These are similar to, but not exactly equal to, Newton’s equations for gravity. This
is not too surprising since Newton’s equations seem to apply to static fields. If we
then apply these equations to the static case then the first two are Newton’s
equations exactly and the third says that phi has zero gradient. For the non-static
case we can derive a wave equation by taking the gradient of (4) and simplifying:

∇ g + ∂∂
2
2 g
2
= −∇ρ
t
One will notice that Newton’s gravitational constant is nowhere to be found.
However this is just an illusion created by our choice of units. The proper
coefficients can be restored by carefully selecting one’s units.
We still need to take a look at the generalization of equation (3). However,
we must make a slight modification of this:

2 a

7) m d x
ab

dτ
2
= F I b

where I b is the five-current of the particle in the field. In four dimensions this
reduces to equation (3) since current is just charge times velocity. This is no
longer true in five dimensions because the fifth component of five-current is
mass, just as the first component is charge.
We can expand (7) into a more familiar equation: ma = qE + jxB + mg . The
first two terms are simply the electromagnetic force and the third is the
gravitational force. Of interest is that the scalar field phi makes no contribution to
the motion of the particle (even if the gravitational fields are non-static). It is
however a factor in the work done on a particle as can be seen by expanding the
zero component of (7):

dW
= qE ⋅ v + mφ .
dτ

Conclusions

We have shown that with a straightforward generalization of Maxwell’s

equations to five dimensions and the assumption that all fields are independent
of the fifth dimension it was possible to derive a set of equations that would
formally reduce to Newton’s equations of gravity in the static case. This method
of unification leads to the prediction of a scalar field which has no effect on the
motion of a particle in the field but which contributes to the work done on a
particle. This scalar field has no source so it invites the interpretation that it
somehow represents the vacuum. It is possible to derive wave equations for the
gravitational field which suggests that the theory can be tested against
observations of binary pulsars.
This unification scheme opens up many doors for further research. Since
Maxwell’s equations can be quantized by well-known procedures it should be
possible to quantize gravity, something not current possible within General
Relativity. Furthermore, this process can be repeated with the currently accepted
Electro-Weak theory. The result would then be a unification of electromagnetism,
gravity, and the weak nuclear force.
Caution must be exercised. Newtonian Gravity, despite it’s stunning
successes, has many failings. It doesn’t correctly predict the precession of
perihelion of Mercury (and other bodies), it has nothing to say about the
interaction of light with gravity, disagrees with the rotation curves of numerous
galaxies, etc. This article makes no attempt to explain these effects and no
pretension is made that the present theory can explain them. These too are
issues left for future research. I welcome any comments on this paper. I can be
reached at [email protected].

Postscript

When I first conducted this research I did so simply because I had read
about Kaluza-Klein theory and was interested in what kind of results I might get
applying this idea to Maxwell’s equations. I was pleasantly surprised to see
Newton’s equations pop out. I later discovered that Gunnar Nordstrom had taken
a similar approach over 100 years ago (before there was Kaluza-Klein theory).