Proc. B . Soc.Lond . A.

338, 439-446 (1974)

Printed in Great Britain

Cosmological models and the Large Numbers hypothesis

B y P. A. M. D i r a c , O.M., F.R.S.
Florida State University, Tallahassee, Florida

{Received 3 December 1973)

The Large Numbers hypothesis asserts th a t all the large dimensionless

Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

numbers occurring in N ature are connected with the present epoch, ex­
pressed in atomic units, and thus vary with time. I t requires th a t the
gravitational constant G shall vary, and also th a t there shall be continuous
creation of m atter.
The consistent following out of the hypothesis leads to the possibility of
only two cosmological models. One of them, which occurs if one assumes
th a t the continuous creation is a multiplication of existing m atter, is
Einstein’s cylindrical closed Universe. The other, which occurs if one
assumes the continuous creation takes place uniformly through the whole
of space, involves an approximately flat Minkowski space with a point
of origin where the Big Bang occurred.

1. T h e L a r g e N u m b e r s hypothesis

The hypothesis has been discussed by the author (1973 a) as the revival of an
old idea. I f one expresses the age of the Universe in terms of a unit of time provided
by atomic constants, say e2/mc3, one gets a large dimensionless number t, which is
somewhere around 1039. I t characterizes the present epoch in a natural way,
independent of man-made standards.
The fundamental data provided by astronomy and atomic physics enable one
to obtain other large dimensionless numbers. The Large Numbers hypothesis
asserts th a t such numbers are connected with t, and therefore vary as t varies.
A number which is roughly (1039)w m ust vary in proportion to tn. The reason for
believing the hypothesis is th a t without it one does not see how these large numbers
could ever be explained.
The hypothesis is not easy to fit in with cosmological theories. I t puts a severe
restriction on the permissible models of the Universe. A popular model involves the
Universe expanding to a certain maximum size and then contracting again. The
time of maximum expansion, expressed in atomic units, will provide a large dimen­
sionless number which does not depend on the present epoch I t is just such numbers
th a t the hypothesis rules out. A model with a maximum size for the Universe is
thus not permitted.
Various models have been proposed by Friedman (1922, 1924) and others, in­
volving non-static spherically symmetric solutions of the Einstein equations. They
each refer to some characteristic epoch appearing in the equations, governing the
expansion. This epoch, expressed in atomic units, again gives a large number
[ 439 ]
 440 P. A. M. Dirac
independent of t,the present epoch, and thus a constant. Hence, the mo
allowed.
The only surviving models are those whose equations do not refer to a particular
constant epoch. They are essentially static universes.

2. P h y s i c a l c o n s e q u e n c e s of t h e h y p o t h e s i s

The electric force between the electron and the proton in a hydrogen atom is
e2jr2.The gravitational force between them is 2. Their ratio is the dimension­
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

less number e2IGmv>me.Its value is about 2 x 1039. According to the hypot

should be increasing proportional to t.Thus G, expre
decreasing proportional to t~x.
This is a physical effect which should show up with sufficiently accurate measure­
ments. The most hopeful chance of observing it is with Shapiro’s (1968) radar
measurements of the distances of the planets. If there is a secular variation of these
distances which cannot be explained in any other way, it will provide evidence for
a variation of G. Shapiro’s observations are sufficiently accurate for an effect of
the expected order of magnitude to show up in a few years’ time.
There is also a possibility of observing the variation of G directly from laboratory
experiments. Measurements of G are being carried out by Beams (1971) by an im ­
proved technique. They are not yet very accurate, b u t variations of G can be
observed with greater accuracy than G itself and it may be th a t the apparatus can
be improved sufficiently to show up the desired effect.
If one estimates the total number of nucleons in the Universe (or the total number
in the galaxies with a speed of recession of less than if one considers the Universe
to be infinite) one gets a number somewhere around 1078. According to the hypo­
thesis, this must be increasing proportional to l2. I t follows th a t new nucleons m ust
continually be created.
Some time ago the Steady State model of the Universe was very popular. This
required continuous creation of m atter to balance the m atter th a t was moving
away from us with the recession of the galaxies. The Steady State model of course
requires th at G shall be constant, and is thus not consistent with our basic hypothesis.
The theory being developed here works with the Big Bang model, b u t it also requires
continuous creation of matter.
The continuous creation th a t we are forced to adopt is a new physical process,
a kind of radioactivity, which is quite different from all the observed radioactivity.
We must face the question of where this new m atter is created. There are two
alternative assumptions th a t one might make.
One might assume th a t nucleons are created uniformly throughout space, and
thus mainly in intergalactic space. We may call this additive creation.
One might assume th a t new m atter is created where it already exists, in propor­
tion to the amount existing there. Presumably the new m atter consists of the same
kind of atoms as those already existing. We may call this midtiplicative creation.
 Cosmological models and the Large Numbers hypothesis 441

3. T h e two m e t r ic s

Einstein’s theory of gravitation demands th a t G shall be a constant. In fact, with

a natural choice of units G — 1. Now Einstein ’s theory is very successful in account­
ing for observations and we do not wish to abandon it. We therefore have to face
the problem of how to modify it to make it agree with a slowly varying G while not
spoiling its successes.
One can achieve this by supposing th a t the metric dsE occurring in the Einstein
field equations is not the same as the metric dsA measured by atomic apparatus.
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

W ith both metrics we take the velocity of light c = 1. All distances determined by
atoms, e.g. the wavelengths of spectral lines and lattice spacings in crystals, refer
to dsA, so all laboratory measurements of distances and times give dsA. One cannot
measure dsE directly. I t comes into play only in equations of motion. For example,
calculations of the motions of planets involve dsE. I f one measures the distances of
the planets with laboratory apparatus, as Shapiro is doing, one gets directly the
ratio of the two ds.
Let us determine the connexion between the two d Take as an example the
motion of the E arth around the Sun, in Newtonian approximation. The basic
equation is
GM = v2r,
where M is the mass of the Sun, ris the radius of the
E a rth ’s velocity. The formula applies both in Einstein units, when we may write it
^e =
and in atomic units, when we may write it
G*MA = v lr A.
Referred to Einstein units, all the quantities GE, ME, vE, rE are constants. Now v
is dimensionless, it is just a certain fraction of the velocity of light, so v K = con­
stant. From our previous discussion GA:: £-1.T h e
which assumption about creation we adopt. W ith additive creation, the number of
nucleons in the Sun is constant, so MA is constant. W ith multiplicative creation the
number of nucleons :: t 2 so MA :: ft. Thus with additive creation we get rA :: t~x and
with multiplicative creation rA::
rE. The general result is
dsA = 1dsE additive creation, (I)
dsA = tdsE multiplicative creation. (II)
W ith additive creation, the E arth is approaching the Sun (in atomic units) and
the whole Solar System is contracting. W ith multiplicative creation, the E arth is
receding from the Sun and the whole Solar System is expanding. These effects are
cosmological and are to be superposed on other effects arising from known physical
causes. Shapiro’s observations should show them up if they exist, and should
enable one to distinguish between the two kinds of creation.
 442 P. A. M. Dirac
Let us adopt the Einstein metric and proceed to examine the possible cosmological
models. We can then use the Einstein field equations.
If we take the field equations with the cosmological constant A, then the theory
involves a large distance E = A~i, of the order of the radiu
SsE be an atomic distance, such as the wavelength of a certain spectral line, ex­
pressed in the Einstein metric. Then R j 5sE is dimensionless, and it is a large number
of the order of 1039. Thus it increases proportional to t. Now is? is a constant, so
5 sEy .t-1. I f the atomic distance is expressed in atomic units, to give dsA, it is, of
course, constant. I t follows th a t we have case II, multiplicative creation.
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

We can infer th a t if we have the cosmological term in the Einstein equations, we

must have multiplicative creation. Additive creation can occur only if A = 0.

4. M u l t i p l i c a t i v e creation

Let us consider further the case of multiplicative creation. We may introduce the
Einstein epoch r, *
r = d 5e , (1)

taken along the world-line of a galaxy. From (II), putting dsA = d£,

T £_1d£ = Ini. ( 2)
/■
The variable r is the dynamical time, as distinct from the atomic time t. The Big
Bang occurred a t t= 0, corresponding to r = —go, so referred to dynam
the Universe has always existed.
The idea of two time variables related in this way was first introduced by Milne
(1937), but his theory was based on quite different assumptions from the present
one, and it is just a coincidence th a t they both give formula (2).
The definition of energy in the Einstein theory has an ambiguity connected with
the coordinate system, but this does not come into play in cosmological models.
We must then have conservation of energy or of mass (if we ignore the possibility
of an appreciable pressure working on the expansion of the Universe). We shall then
have the Einstein equations applying with the metric d.sE and with a suitable unit
of mass. This unit must be such th a t the mass of a body such as the Sun is constant.
The mass of a nucleon must then be proportional to t~2. All atomic particles m ust
have their masses varying in this way. I t just compensates the multiplicative
creation to make the masses of classical bodies constant.
One can easily work out how all the atomic constants vary when referred to
Einstein units. We had in § 2 e2/Gm2•• t
In Einstein units G is constant and m :: t~%.Hence, e :: We then find h :: t~3.
We saw in § 1 th a t we must have a static model. The only static model with
positive mass is Einstein’s cylindrical model, which we are thus forced to adopt
with multiplicative creation.
 Cosmological models and the Large Numbers hypothesis 443
We now have a picture in which, referred to Einstein units, the galaxies are not
receding, but keep approximately a t a constant distance. To understand how the
red-shift arises with this picture we must take into account th a t an atomic clock,
marking out units A£ = 1, will mark out units Ar = f-1. W ith increasing t these
units get continually smaller, so the atomic clock is continually speeding up.
The light coming from a distant galaxy was emitted in the past when atomic
clocks were slower. The wavelength of the emitted light referred to these slow
atomic clocks. As the light travels to us the wavelength remains constant in
Einstein units. When it arrives here it is referred to the present atomic clocks and
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

the wavelength appears longer. The increase is in the proportion t/te, where te is
the time of emission. Thus, the red-shift is t/te—1.

5. A d d i t i v e creation

Let us now examine the alternative kind of creation. We now have m atter, pre­
sumably hydrogen atoms, created uniformly throughout space. This will give
violation of conservation of mass whatever units we use. The only way in which we
can preserve the Einstein equations, which demand conservation of mass, is to
suppose that, together with the H atoms, a uniform distribution of negative mass
is created, so as to make the total density of created m atter zero. In order not to
have violent disagreement with observation we m ust suppose th a t the negative
mass is not observable and so is not quantized, like the H atoms. I t m ust not in­
teract with other m atter, except gravitationally, and m ust have no physical effects
a t all, apart from producing a curvature of space. The H atoms condense into
nebulae and stars and form the m atter th a t we observe. The negative mass remains
uniform and unobservable.
The total density of m atter is zero apart from local irregularities arising from
condensations of the H atoms. If we smooth out these irregularities we get a model
in which space-time, referred to the Einstein metric, is flat. I t is thus just Minkowski
space and the Einstein metric becomes the Minkowski metric.
There is one special point, the origin O, where the Big Bang occurred. The
physical world lies within the future light cone from O. The world-lines of the galaxies
are the straight lines through 0 lying within the future light cone.
The Einstein epoch r at a physical point P is the Minkowski distance from P to O.
Formula (1) still applies and must now be used in conjuction with I, so

Jd^-J td t ¥*• (3)

The three-dimensional world a t a given epoch consists of all those points P with
a given r-value. I t is of infinite extent and has a negative curvature.
The total number of nucleons within th a t part of the world whose speed of re­
cession is < | is proportional to t2or to r. The volume o
is :: t3. The number per unit volume is thus :: t~2. W ith additive creation the mass of
a nucleon in Einstein units in constant, as is shown by the constancy of MF in § 3.
 444 P. A. M. Dirac
Thus the density of the nucleons :: r~2. The density of the continuous distribution
of negative mass m ust therefore also be proportional to 7~2.
The Einstein field equation is
Rf*- Ig^R = Siz(T^ -
where is the material energy tensor for the ordinary physical m atter and —U/lv
is the corresponding tensor for the uniform negative m atter. We have
= kr-Hh?,
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

where v? is the velocity vector corresponding to the recession speed for the point
concerned and kis a constant. Thus we may rewrite equation (4)
R /I V _
1 gp > R + J ' - ^
c t vV = (5)
with another constant k '.
Because of the uniformity of the negative m atter, the term k'r~2vtlvv m ay be
looked upon as an intrinsic property of space, like the cosmological term Agp" th a t
is introduced into other forms of the Einstein theory. Its effect on the motions of
the planets is just as small as th a t of the usual cosmological term, so it is much too
small to be observed. The only reason we have to p u t it into equation (5) is to
preserve consistency with the Bianci identities when involves continual creation.

Let us discuss the red-shift with this model. Let OA represent the world-line
of our galaxy and let OB represent the world-line of another galaxy, and suppose
the points A and B are such th a t light em itted from B can arrive a t A. Let P be
the point on OA such th a t PB is orthogonal to OA. Then PB = PA. I f 6 is the
hyperbolic angle between OB and OA, we have
OP = OB cosh ,
PA = PB = OB sinh 0 ,
and hence OA = OB e°.
Let ta and r B denote the epochs of A and B in Einstein units, so th a t they are equal
to the distances OA, OB respectively. Then

t J tb = e0-

If tA
, tB are the epochs in atomic units, we get from (3)

W hi = e
 Cosmological models and the Large Numbers hypothesis 445
Suppose light is emitted from B with a wavelength A = I t is received a t A
with the wavelength 5 tA—ei 6 &B =
There would still be a red-shift with this model even if there were only one metric,
the Minkowski metric. The magnitude of the red-shift would then be

5t a / 5t b - 1 =

For small 6 it is just double the red-shift with the two-metric theory. We need the
two-metric theory, of course, in order to have G varying with the epoch.
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024

6. C o m p a r i s o n of t h e two m o d e l s

Two cosmological models have been obtained, corresponding to the two alter­
native assumptions th a t one might make for the continuous creation of m atter.
They are both very simple models. The question arises, which should one prefer?
A decisive answer may be obtained in a few years time from radar observations of
the planets. In the meantime we may discuss them and try to assess which is the
more likely.
Multiplicative creation requires th a t all forms of m atter shall be multiplying,
with the number of atoms increasing proportional to 2. I t is a little difficult to
understand how this can take place in the case of a crystal. Presumably the new
atoms must appear on the outside. The rate of multiplication is extremely small,
so there is plenty of time for the new atoms to appear in the places most suitable
for them. B ut during the course of geological ages the increase must be quite
appreciable, and should be taken into account in any discussion of the formation of
crystals in very old rocks. I t might lead to insuperable difficulties.
Multiplicative creation requires also th a t the number of photons in a given
beam of light shall increase, in order th a t the energy in Einstein units shall be
conserved. This will cause the apparent brightness of a distant galaxy or quasar to
be increased. The effect would not be directly observable because we do not know
the absolute brightness a t the time in the remote past when the light was emitted.
One may make statistical assumptions and try to get some evidence for this effect.
The question has been discussed in the author’s paper (19736).
Additive creation does not lead to such drastic departures from generally accepted
ideas and does not face us with such difficult problems. I t involves continual crea­
tion of intergalactic gas and presumably much of it remains in the gaseous form,
but its density is much too small to be observable. This theory is not so likely to
lead quickly to a clash with observations, as could happen with multiplicative
creation.
The foregoing- work is all founded on the Large Numbers hypothesis, in which
I have great confidence. I t also requires the assumption of two metrics, which is
not so certain. The only reason for believing in the two metrics is th a t up to the
present no alternative way of bringing in the Einstein theory has been thought of.
B ut this situation could change.

28
Vol. 338- A.
 446 P. A. M. Dirac

R eferences

Beams, J . W. 1971 Physics Today, May, p. 35.

Dirac, P. A. M. 1973 a Proc. R. Soc. Lond. A 333 , 403.
Dirac, P. A. M. 19736 Pont. Acad. Commentari1 1 -N .4 6 , 1.
Friedm an, A. 1922 Z. Phys. 10 , 377.
Friedm an, A. 1924 Z. Phys. 21 , 326.
Milne, E. A. 1937 Kinematic relativity, p. 36. Oxford U niversity Press.
Shapiro, 1 . 1 . 1968 Scient. A m . 219 , 32.
Downloaded from https://royalsocietypublishing.org/ on 26 October 2024