JULV

1,

1935

PIIVSICAL

REVIEW

VOLUME

48

The Particle Problem in the General Theory of Relativity

A.

EINSTEIN AND

N.

Institutefor Advanced Study, Princeton

(Received May 8, 1935)

ROSEN,

The writers investigate the possibility of an atomistic

theory of matter and electricity which, while excluding
singularities of the field, makes use of no other variables
than the g~. of the general relativity theory and the <PI' of
the Maxwell theory. By the consideration of a simple
example they are led to modify slightly the gravitational
equations which then admit regular solutions for the static
spherically symmetric case. These solutions involve the
m~thematical representation of physical space by a space
of two identical sheets, a particle being represented by a
"bridge 11 connecting these sheets. One is able to understand why no neutral particles o negative mass are to be

found. The combined system of gravitational and electromagnetic equations are- treated similarly and lead to a
similar interpretation. The most natural elementary
charged particle is found to be one o zero mass. The rnanyparticle system is expected to be represented by a regular
solution of the field equations corresponding to a space of
two identical sheets joined by many bridges. In this case,
because of the absence of singularities, the field equations
determine both the field and the motion ofthe particles_
The many-particJe problem, "hich would decide the value
of the theory, has not yet been treated.

spite of its great success in various fields, the

I Npresent
theoretical physics is still far from

One would be inc1ined to answer this question

in the negative in view of the fact that the
Schwarzschild solution for the spherically symmetric static gravitational field and Reissner's
extension of this solution to the case when -an
electrostatic field is also present each have a
singularity. Furthermore the last of the Maxwell
equations, which expresses the vanishing of the
divergence of the (contravariant) electrical field
density, appears to exc1ude in general the existence of charge densities, hence also of electrical
partic1es.
For these reasons writers have occasional1y
noted the possibility that material particles might
be considered as singularities of the field. This
point of view, however, we cannot accept at all.
For a singularity brings so much arbitrariness
into the theory that it actually nullifies its laws.
A pretty confirmation of this was imparted in a
letter to one of the authors by L. Silberst@in. As
is well known, Levi-Civita and Weyl have given
a general method for finding axially symmetric
static solutions of the gravitational equations.
By this method one can readily obtain a solution
which, except for two point singularities Iying on
the axis of symmetry, is everywhere regular and
is Euc1idean at infinity. Hence if one admitted
singularities as representing particles one would
have here a case of two partic1es not accelerated
by their gravitational interaction, which would
certainly be exc1uded physically. Every field
theory, in our opinion, must therefore adhere to
the fundamental principie that singularities of
the field are to be exc1uded.

being able to provide a unified foundation on

which the theoretical treatment of all phenomena
could be based. We have a general relativistic
theory of macroscopic phenomena, which however has hitherto been unable to account for the
atomic structure of matter and for quantum
effects, and we have a quantum theory, which is
able to account satisfactorily for a large number
of atomic and quantum phenomena but which by
its very nature is unsuited to the principie of
relativity. Under these circumstances it does not
seem superfluous to raise the question as to what
extent the method of general relativity provides
the possibility of accounting for atomic phenomena. It is to such a possibility that we wish to
cal1 attention in the present paper in spite of the
fact that we are not yet able to decide whether
this theory can account for quantum phenomena.
The publication of this theoretical method is
nevertheless justified, in our opinion, because it
provides a c1ear procedure, characterized by a
minimum of assumptions, the carrying out of
which has no other difficulties to overcome than
those of a mathematical nature.
The question with which we are concerned can
be put as follows: Is an atomistic theory of matter
and electricity conceivable which, while exc1uding singularities in the field, makes use of no
other field variables than those of the gravitational field (g.) and those of the electromagnetic
field in the sense of Maxwel1 (vector potentials, 'PI')?

73

74

A.

EINSTEIN

AND

N.

ROSEN

In the following we shall show that it is possible

to do this in a natural way, that the question we
are raising can be answered in .the affirmative.

that in g2Rk! there is no longer any denominator.

If then we replace (3) by

1. A

this system of equations is satisfied by (1) at all

finite points. This amounts to introducing in place
of the g~' the cofactors [g~.J of the gIL' in g in
order to avoid the occurrence of denominators.
One is therefore operating with tensor densities
of a suitable weight instead of with tensors. In
this way one succeeds in avoiding singularities of
that special kind which is characterized by the
vanishing of g.
The solution (1) naturally has no deeper
physical significan ce insofar as it extends into
spatial infinity. It allows one to see however to
what extent the regularization of the hypersurfaces g=O leads to a theoretical representation of matter, regarded from the standpoint of
the original theory. Thus, in the framework of
the original theory one has the gravitational
equations
(4)

SPECIAL KIND OF SINGULARITY AND

I TS

REMOVAL

The first step to the general theory of relativity

was to be found in the so-called "PrincipIe of
Equivalence": If in a space free from gravitation
a reference system is uniformly accelerated, the
reference system can be treated as being "at
rest," provided one interprets the condition of
the space with respect to it as a homogeneous
gravitational field. As is well known the latter is
exactly described by the metric field 1
ds 2= -dXI2-dx22-dx3 2+a 2xI2dx42.
(1)
The g~. of this field satisfy in general the equations
(2)

and hence the equations

(3)

Rkl=Rmklm=O.

The g~. corresponding to (1) are regular for aII

finite points of space-time. Nevertheless one
cannot assert that Eqs. (3) are satisfied by (1)
for all finite values of Xl, , X4. This is due to
the fact that the determinant g of the g~. vanishes
for Xl = O. The contravariant g~' therefore become infinite and the tensors Rik1m and Rkl take
on the form O/O. From the standpoint of Eqs. (3)
the hyperplane Xl = O then represents a singularity of the field.
We now ask whether the field law of gravitation (and later on the field law of gravitation and
electricity) could not be modified in a natural
way without essential change so that the solution (1) would satisfy the field equations for all
finite points, i.e., also for Xl = O. W. Mayer has
caIIed our attention to the fact that onecan make
Ri k1m and Rk! into rational functions of the gp..
and their first two derivatives by muItiplying
them by suitable powers of g. It is easy to show
1 It is worth pointing out that this metric field does not
represent the whole Minkowski space but only part of it.
Thus, the transformation that converts
ds' = -d~I'-d~2'-d~3'+da
into (1) is
h=xI cosh aX4,
~3=X3,
~4 = XI sinh aX4.
~,=X2,

It follows that only those points for which

spond to points for which (1) is the metric.

tl'?:~42

corre-

(3a)

where T ik is the tensor of mass or energy density.

To interpret (1) in the framework of this theory
we must approximate the line element by a
slightly different one which avoids the singularity
g=O. Accordingly we introduce a small constant
u and let
ds 2= -dXI2-dx22-dx32+(a2xI2+u)dx42;

(la)

the smaIIer 0'( >0) is chosen, the nearer does this

gravitational field come to that of (1). If one
calculates from this the (fictitious) energy
tensor T ik one obtains as nonvanishing components
T 22 = T 23 =a 2/ u/O +a2XI2/ 0')2.
We see then that the smaller one takes u the
more is the tensor concentrated in the neighborhood of the hypersurface Xl = O. From the standpoint of the original theory the solution (1)
contains a singularity which corresponds to an
energy or mass concentrated in the surface Xl = O;
from the standpoint of the modified theory, however, (1) is a solution of (3a), free from singularities, which describes the "field-producing mass,"
without requiring for this the introduction of
any new field quantities.

PARTICLE PROBLEM IN RELATIVITY

It is clear that aH equations of the absolute
differential calculus can be written in a form free
from denominators, whereby the tensors are replaced by tensor densities of suitable weight.
It is to be noted that in the case of the solution
(1) the whole field consists of two equal halves,
separated by the surface of symmetry XI=O,
such that for the corresponding points (Xl, X2, Xa,
X4) and (-Xl, X2, xs, X4) the gik are equal. As a
result we find that, although we are permitting
the determinant g to take on the value O (for
Xl = O), no change of sign of g and in general no
change in the "inertial index" of the quadratic
form (1) occurs. These features are of fundamental importance from the point of view of the
physical interpretation, and will be encountered
again in the solutions to be considered latero

2.

THE SCHWARZSCHlLD SOLUTION

As is weH known, Schwarzschild found the

sphericalIy symmetric static solution of the gravitational equations
1

ds 2 =

dr 2 -r 2(d0 2 +sin 20d4>2)

1-2mjr
+ (1-2mjr)dt 2,

(5)

(r>2m, O from O to 'Ir, 4> from Oto h); the variables Xl, X2, Xa, X4 are here r, O, 4>, t. The vanishing
of the determinant of the g". for 0=0 is unimportant, since the corresponding (spatia!) direction is not preferred. On the other hand gIl for
r = 2m becomes infinite and hence we have there
a singularity.
If one introduces in place of r a new variable
according to the equation
u 2 =r-2m,

one obtains for

the expression
2
2
ds = -4(u +2m)du2
ds2

- (u 2+ 2m)2(d0 2+sin 20d4>2)

u2

dt2.
u 2 +2m

(Sa)

These new g". are regular functions for aH values

of the variables. For u=O, however, g44 vanishes,
hence also the determinant g. This does not prevent the field equations (3a) , which have no
denominators, from being satisfied for alI values
of the independent variables. We are therefore
dealing with a solution of the (new) field equations, which is free from singularities for all finite

7S

points. The hypersurface u=O (or in the original

variables, r = 2m) plays he re the same role as the
hypersurface Xl = O in the previous example.
As u varies from - 00 to + 00, r varies from
+ 00 to 2m and then again from 2m to + oo.
If one tries to interpret the regular solution (Sa)
in the space of r, O, 4>, t, one arrives at the following conclusion. The four-dimensional space is
described mathematicalIy by two congruent
parts or "sheets," corresponding to u> O and
u<O, which are joined by a hyperplane r=2m or
u=O in which g vanishes. 2 We call such a connection between the two sheets a "bridge."
We see now in the given solution, free from
singularities, the mathematical representation of
an elementary particle (neutron or neutrino).
Characteristic of the theory we are presenting is
the description of space by means of two sheets.
A bridge, spatialIy finite, which connects these
sheets characterizes the presence of an electricalIy
neutral elementary particle. With this conception one not only obtains the representation of an
elementary particle by using only the field
equations, that is, without introducing new field
quantities to describe the density of matter; one
is also able to understand the atomistic character
of matter as well as the fact that there can be no
particles of negative mass. The latter is made clear
by the following considerations. If we had started
from a Schwarzschild solution with negative m,
we should not have been able to make the solution regular by introducing a new variable u
instead of r; that is to say, no "bridge" is possible
that corresponds to a particle of negative mass.
If we consider once more the solution (1) from
the standpoint of the information we have acquired from the Schwarzschild solution, we see
that there also the two congruent halves of the
space for Xl> O and Xl <O can be interpreted as
two sheets each corresponding to the same physical space. In this sense the example represents
a gravitational field, independent of X2 and Xa,
which ends in aplane covered with mass and
forming a boundary of the space. In this example,
as well as in the Schwarzschild case, a solution
free from singularities at all finite points is made
possible by the introduction of the modified
gravitational Eqs. (3a).
2

Because oC the syrnrnetry about the hypersurface

g = 0, the sign oC g does not change at this hypersurface.

76

A.

EINSTEIN AND

The main value o the considerations we are

presenting consists in that they point the way to a
satisactory treatment o gravitational mechanics. One o the imperfections of theoriginal relativistic theory of gravitation was that as a fieId
theory it was not complete; it introduced the
independent postulate that the law of motion of a
particIe is given by the equation of the geodesic. 3
A complete field theory knows only fields and not
the concepts of particIe and motion. For these
must not exist independentIy of the field but are
to be treated as part of it. On the basis of the
description of a particIe without singularity one
has the possibility of a logicalIy more satisfactory
treatment of the combined problem: The problem
of the field and that o motion coincide.
If several particIes are present, this case corresponds to finding a solution without singularities
o the modified Eqs. (3a), the solution representing a space with two congruent sheets connected
by several discrete "bridges." Every su eh sol ution is at the same time a solution of the field
problem and o the motion problem.
In this case it will not be possible to describe
the whole field by means of a single coordinate
system without introducing singularities. The
simplest procedure appears to be to choose coordinate systems in the following way:
(1) One coordinate system to describe one o
the congruent sheets. With respect to this system
the field will appear to be singular at every bridge.
(2) One coordinate system for every bridge,
to provide a description of the fieId at the bridge
and in the neighborhood of the latter, which is
free from singularities.
Between the coordinates of the sheet system
and those of each bridge system there must exist
outside of the hypersurfaces g=O, a regular coordinate transformation with nonvanishing determinant.
3.

COMBINED FIELD. ELECTRICITY

The simplest method of fitting electricity into

the conceptual framework of the general theory
of relativity is based on the folIowing train of
3 To be sure, this weakness was formally avoided in the
original theory of relativity by the introduction of the
energy tensor into the field equations. It was clear, however, from the very beginning that this was' only a provisory completion of the theory in the sense of a phenomenological interpretation.

N.

ROSEN

thoughts. If besides the pure gravitational field

other field variables are al so present, the field
equations of gravitation are
(4)

where T ik is the "material" energy tensor, i.e.,

that part of the mathematical expression of the
energy which does not depend exclusively on the
gil'. In the case of the phenomenological representation of matter-if it is to be considered as
"dust-like," that is, without pressure-one takes

where pis the density-scalar, dxi/ds the velocityvector of the matter. It is to be noted that T 44 is
accordingly a positive quantity.
In general the additional field-variables satisfy
such differential equations that, in consequence
of them, the divergence T ik ; mg km vanishes. As
the divergence of the left side of (4) vanishes
identicalIy, this means that among all the field
equations those our identities exist which are
needed for ,their compatibility. Through this condition, in certain cases, the structure of T ik, not
however its sign, is determined. It appears
natural to choose this sign in su eh a way that
the component 1'44 (in the limit of ihe special
relativity theory) is always positive.
The Maxwell electromagnetic field, as is welI
known, is represented by the antisymmetric
field tensor 'P!"( = 8'P!,/8x' - 8'P,/8x!') , which satisfies the field equations
'P!"; ug'U =

o.

(6)

These equations have the welI-known consequence that the divergence of the tensor
Tik=tgik'PaP'P afl - 'Pia'Pka

(7)

vanishes. The sign has been so chosen that T 44 is

positive or the case o the special relativity
theory. If one puts this T ik into the gravitational
Eqs. (4), then the latter together with (6) and
(7) form a theory .of gravitation and electricity.
It so happens that we are orced to put the
negative of the aboye into the gravitational equations if it is to be possible to obtain static sphericalIy symmetric solutions of the equations, free
from singularities, which could represent electrical particles. Making this change of sign one
finds as the required solution

PARTICLE PROBLEM rN RELATIVITY

1
ds 2 = -~------dr2

1-2m/r-t2 /2r 2

(8)

Bere m has obviously the significance of .a

gravitattng mass, E that of an electrical charge.
It turns out that also in this case there is no
difficulty in forming a solution without singularity corresponding to the solution just given. 4
Curiously enough, one finds that the mass m is
not determined by the electrical charge E, but
that E and m are independent constants of integration. It also turns out that for the removal of
dhe singularity it is not necessary to take the
ponderable mass m positive. In fact, as we.shall
show immediately, there exists a solution free
from singularities for which the mass constant m
vanishes. Because we believe that these massless
solutions are the physicaIly important ones we
will consider he re the case m = O.
The fieId equations without denominators can
be written

where in the last equation the term in R has been

omitted because it vanishes in consequence of
(7), by which Taa is zero.
If in Eq. (8) (with m=O) one replaces r by the
variable u according to the equation

u 2 =r2 -

E2

/2

one obtains
<Pl= <P2= <pa=O,

ds = -du
2

2-

<P4=e/(u 2+2/2):,

(u +e /2) (d8 2 +sin 2 8d<t>2)

2

+ [2u

2/

-(8a)

(2u 2 +e 2) Jdt 2 .J

This solution is free from singularities for all

finite points in the space of two sheets and the
charge is again represented by a bridge between
the sheets. It is the representation of an elementaryelectrical particle without mass.
4 If we had taken the usual sign for Tik, the solution
would involve +.2 instead of -.'. It would then not be possible, by making a coordinate transformation, to obtain a solution free from singularities.

77

4. SUMMARY AND GENERAL REMARKS

If one solves the equations of the general
theory of relativity for the static sphericaJly
symmetric case, with or without an electrostatic
field, one finds that singularities occur in the
solutions. If one modifies the equations in an
unessential manner so as to make them free fram
denominators, regular solutions can be obtained,
provided one treats the physical space as consisting of two congruent sheets. The neutral, as well
as the electrical, particle is a portion of space
connecting the two sheets (bridge). In the hypersurfaces of contact of the two sheets the determinant of the gllP vanishes.
One mightexpect that processes in which
several elementar:y particles take part correspond
to regular solutions of the field equations with
several bridges between the two equivalent
sheets corresponding to the physical space. Only
by investigations of these solutions will one be
able to determine the extent to which the theory
accounts for the facts. For the present one cannot
even know whether regular solutions with more
than one briqge exist at al!.
It appears that the most natural electrical
particle in the theory is one without gravitating
mass. One is therefore led, according to this
theory, to consider the electron or pro ton as a
two-bridge problem.
In favor of the theory one can say that it explains the atomistic character of matter as weJl
as the circumstance that there exist no negative
neutral masses, that it introduces no new variables other than the g~P and <P1l' and that in principIe it can c1aim to be complete (or c1osed). On
the other hand one does not see a priori whether
the theory contains the quantum phenomena.
Nevertheless one should not exclude a priori the
possibility that the theory may contain them.
Thus it might turn out that only such regular
many-bridge solutions can exist for which the
"charges" of the electrical bridges are numericalIyequal to one another and only twO different
"masses" occurfor the mass bridges, and for which
the stationary "motions" are subject to restrictions like those which we encounter in the quantum theory.
In any case here is a possibility for a general
relativistic theory of matter which is logicaIly
completely satisfying and which contains no ncw
hypothetical elements.