19 64ApJ. . .139. .

909H

ON RELATIVISTIC ASTROPHYSICS

F. Hoyle
St. John’s College, Cambridge, England, and
California Institute of Technology, Pasadena, California
William A. Fowler
California Institute of Technology, Pasadena, California

G. R. Burbidge and E. Margaret Burbidge

University of California, San Diego, La Jolla, California
Received October 5, 1963
ABSTRACT
In this paper we have attempted to discuss the relation of massive highly condensed objects to astro-
physics in general, rather than only to the radio-source problem. Because situations in which general
relativistic effects play a dominant role have not received much attention in astrophysics, a brief review
of the relativistic properties of collapsed objects is given in Section I of the paper.
In Section II we have given especial attention to fwo problems: (1) Galaxies may contain considerable
quantities of inert or “hidden” mass, not simply in the form of white dwarfs or neutron stars. (2) Highly
collapsed objects may act as energy sources, not just explosively, but over extended time intervals.
The first of these problems has implications both for nucleosynthesis and for the mass-to-light prop-
erties of stellar systems. The second appears applicable to a wide range of phenomena, ranging from super-
novae in our Galaxy to whole radio galaxies. A continuous source of optical synchrotron electrons appears
necessary both in the Crab Nebula and in the jet of M87. We suggest their origin lies in processes that
are physically similar and in which considerations of general relativity play a dominant role.

I. THE FATE OF MASSIVE BODIES

a) Introduction
Gravitation tends to pull aggregates of matter together. A logical question evidently
arises as to the ultimate outcome of this process. In astronomy it has usually been sup-
posed that astrophysical evolution proceeds in such a way that, given sufficient time,
a cold body in hydrostatic equilibrium remains a body that can maintain its equilibrium
over all further time. Yet it is known that no such equilibrium state can be found for a
mass appreciably greater than M®. The Chandrasekhar limit gives 5.76 Mq/p? for the
maximum mass that can be supported by degenerate electron pressure. A similar result
applicable at higher densities for neutron degeneracy was obtained by Oppenheimer and
Volkoff (1939)—in fact, the maximum mass was about 0.7 M&. Even the doubtful
assumption of a hard-core neutron potential, allowing the nuclear fluid at high densities
to behave like an incompressible liquid, only extends the maximum mass to about 3 Moj
a result obtained immediately from the Schwarzschild interior solution, using nuclear
densities. Evidently, however, it is possible for gravitation to pull together much larger
masses than these. The conventional view is that in all such cases a process of mass
ejection takes place, conveniently reducing the final mass below one or the other of the
limits just mentioned. We wish to emphasize that this view is no more than a supersti-
tion. The existence of white dwarfs has sometimes been taken as supporting evidence,
but the white dwarfs can be explained in terms of the evolution of stars whose masses
have never appreciably exceeded the limits mentioned above. For massive stars (and
very massive objects) there is no evidence, either from theoretical studies of stellar
evolution or from observation, that mass loss plays any such role. Indeed, it would be a
curious situation if astrophysical processes occurring long before the onset of the final
909

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

910 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 139

implosion crisis were to operate always to prevent the crisis from arising. This would
imply an unlikely “foreknowledge” on the part of natural processes.
In this paper we propose to accept the situation that stars with masses greater than
the critical mass can reach a stage of catastrophic implosion in which general relativity
becomes dominant. In the following section we give a brief review of what might be
called the classical implosion problem (Datt 1938; Oppenheimer and Snyder 1939). In
later sections we shall consider modifications demanded by (i) non-zero internal tem-
perature, (ii) rotation, (iii) the non-classical discussion of Hoyle and Narlikar (1963).
This will conclude Section I of the paper.
b) The Classical Implosion Problem
The object is assumed uniform and spherically symmetric and the internal pressure
is zero. Outside is empty space, and the line element is Galilean at infinity. The object
is taken to implode from rest without rotation.
The interior solution, obtained from Einstein’s equation, is found to be identical to
the simple elliptic cosmology with line element

¿í2 = ¿í!-S2(0(y3^¿+rw), CD

¿O2 = ¿ 02 + sin2 Od<p2 C = 1 . (2)

The coordinates r, 0, are intrinsic—that is to say, they always have the same values
for any particular particle. It is possible to choose such coordinates so that r = 0 for
any assigned particle. These conditions arise out of the special conditions of homogeneity
and isotropy treated in cosmology. The same conditions arise in the case of a finite
object from uniformity and spherical symmetry and from considering implosion from
rest without rotation.
Einstein’s equations give
1 SttGp
(3)
Si S2“*" 3 ’

where the proper density p satisfies the condition p53 = constant. The right-hand side
of equation (3) always has a zero. In the cosmological case the zero gives the value of
the scale factor S at maximum expansion of the universe, while in the case of a finite
object the zero refers to the initial state of rest. If quantities at the zero are referred to
by the subscript zero then
5o = (3^)-1/2 (4)

It is evidently possible to redefine the r coordinates so that S' = 1 at maximum phase.

The line element then takes the form

¿ 52 = - 52 ( o (t + r w), (5)
where
SttGp o
(6)
Equation (3) becomes
Â2_ (1-5)
(7)
52 a
5s

It is usual to work with equations (1) and (3) in cosmology, but equations (5) and (7)
are more convenient in the implosion problem. It should be noted that, whereas in cos-

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 911

mology, r = 0 can be any assigned particle, in the implosion problem r = 0 refers

uniquely to the center of the body.
A further difference is that an exterior solution must also be found for the implosion
problem. At first sight it might be thought most convenient to use Schwarzschild coordi-
nates for the exterior,
ds2 = ev dt2 — ex dR2 — R2 dti2. (8)

(The radial coordinate is characterized by the property that the sphere through R, T
has area 47rR2, independent of 7\) The advantage of these coordinates is that for a non-
radiating object the T coordinate can be chosen such that
constant
e* = = 1 (9)
R

and the line element is static. An awkward problem arises, however, in matching the
interior and exterior solutions at the boundary of the object. The boundary has a con-
stant r coordinate, n say, but the R coordinate is not constant. For this reason it is
convenient to complete the exterior solution in terms of the r, t coordinates. With the
solution obtained a transformation to R, T can be found.
The exterior solution has the remarkable property that Einstein’s equations, instead
of being partial in r, become ordinary equations with r~3/21 as variable. The line ele-
ment is
(10)

where K = (\/S)d/dr(rS), and S satisfies equation (7) but with differentiation taken
with respect to trhzl2/r*12. The exterior solution is, therefore, closely similar to the interior
solution.
Turning now to the transformation to Schwarzschild coordinates, our aim is to deter-
mine the constant in equation (9). This is achieved most easily by noticing that Schwarz-
schild coordinates can also be used for the interior—subject to a proviso to be mentioned
later—and that e% ex are continuous at the boundary. In fact

T (n)
R=rS{t) and = * (f J J&)

transform the line element from equation (5) to equation (8) with
, 1 S2#(l-ar2) (12)
e
1-ar -r2^’
2 6
( 1 - ar2 - r2#)^2 '

Here 3> is any differentiable function and 4>i is the derivative with respect to the argu-
ment. For the purpose of determining the constant in equation (9) nothing more need
be known about 3>.
Inserting for S2 from equation (7) we have
ar2 _ . ar3
e~x = 1 1 (13)
~R'

Applying equation (13) at the boundary r = nwz see that the constant in equation (9)
is ar&3. Remembering the definition of a we can write ex in the conventional form

e~x (14)

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

912 F. HOYLE, W FOWLER, G. BURBIDGE, AND E BURBIDGE Vol. 139

by defining
,, 4ir „
— pon3. (15)

For an observer in the exterior, M is the gravitational mass of the object. It remains
constant throughout the implosion for a non-radiating object.
An event horizon develops at R = 2GM. Once the object has contracted inside the
sphere with this I? coordinate no further signals can cross the sphere. Writing i? = rbS(t)
for the boundary, we see that the boundary coincides with the event horizon when
2GM
an2, (16)
Tb
The proper time required for the object to fall inside the event horizon is
s-S—aru2 rarb2 /í ^ 1 F'ir 1
jT = = -^5^ —sin-1 a1/2r¡>+VC 1 — afi,2) J , U7)

and implosion into a singularity occurs in a proper time

r1 dS T
(18)
o «$ 2 a1/2-
The square root of equation (7), choosing the negative sign, is used in the evaluation of
these integrals.
Initially, we have R = At the event horizon R = 2GM = an*. Hence, if implosion
occurs from a dispersed state, ar&2 must be <$Cl, since by a dispersed state we mean that
the initial area of the boundary is large compared with the area of the event horizon.
The proper time required for implosion from the event horizon into the singularity is
— r&(l — art,2)172 + a“172 sin-1 a172 rh fan3, an2 <3C 1 , d9)

i.e., for implosion from a dispersed state. In terms of M, 2ar&3/3 = 4GM/3. So far units
with c = 1 have been used. In conventional units the proper time for implosion from the
event horizon into the singularity is
tGM M
——~6.6
3 X 10“6 77— sec^ 660 sec for M = 108 Mq. (20)
3c Mq

It was mentioned above that the transformation (11) is subject to a proviso, namely,
that ex, ev must not become negative. In fact the transformation is singular for S = ar2,
and ex, ev are negative for 5 < ar2. Since S —» 0 it is clear that the transformation to
Schwarzschild coordinates fails sooner or later at all points of the interior. The transfor-
mation also fails at all exterior points within the event horizon.
The transformation can, of course, be made even when S < ar2, but R, T are not the
usual Schwarzschild coordinates. Because of the negative values of ev, ex the space- and
timelike roles of R, T are inverted—R becomes the time coordinate and T a space coor-
dinate. These transformation difficulties explain why Schwarzschild coordinates are
unsuitable for the solution of the whole implosion problem.

c) The Effects of Internal Pressure

Even a uniform pressure is capable of modifying equation (7) for S. The equivalent
equation (3) can be written as
>S2 1 8xGE
(21)
S2 2
5 ^ 3

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No 3, 1964 RELATIVISTIC ASTROPHYSICS 913

since the energy density E is just p in the absence of internal pressure. When there is a
non-zero pressure, E is not equal to p, however. Equation (21) holds for uniform pressure,
but not when there is a pressure gradient.
Now at very high densities, >^1016 gm cm-3, quantum statistics prevent E — p.
Thus even for zero temperature we have E cc p4/3 at very high densities. For p oc S~z,
S2 is determined by a term of order S~2 for small S> not by a term of order S-1. This evi-
dently only accentuates the onset of the singularity.
Pressure is also important in a discrete object because a pressure gradient is always
present, on account of the necessity for the pressure to be zero at the boundary. The
effect of a pressure gradient is to change the mathematical structure of Einstein’s equa-
tions from ordinary equations to partial equations in the coordinates r, t. The problem
is then harder to handle and less is known about its solution, except in the special case
of equilibrium when the t coordinate drops out.
One might be tempted to speculate that pressure gradients will ultimately halt an
imploding object, since this is the case in the Newtonian approximation whenever the
ratio of specific heats is greater than 4/3. But the same result is not true if relativity
effects are sufficiently important. While a formal solution, applicable in detail to the
most general case, has not yet been given, a disproof follows from a simple reductio ad
absurdum argument. For example, consider a cold object with degeneracy pressure—
that is, P ^ oc p4/3. If such a pressure were able to halt an implosion the resulting
motion would be oscillatory between radii R2, say—where R is once again the
Schwarzschild coordinate. An equilibrium solution for the object must then exist with
a radius between i?i, R2. But this is not possible if the total mass exceeds the limiting
values mentioned in the Introduction. Hence implosion cannot be halted for sufficiently
large mass, >~Mo.
A similar argument can be used for a non-zero temperature. Although an equilibrium
solution with a radius between Ri and R2 may be possible, the thermal energy necessary
to maintain the equilibrium state becomes impossibly large as M is increased, a result
recently demonstrated by Iben (1963). The mass in question depends on the value of R
at which the equilibrium solution is sought, failure occurring when 2GM/R becomes of
order unity.
A hot object evolves through the escape of energy. This leads to R being ultimately
reduced to the stage where 2 GM/R is of order unity. Hence a pressure gradient must
finally fail to prevent implosion; the only exception to this statement in the spherically
symmetric case being when the mass falls below one or the other of the limiting values,
~Mo.
d) Neutrino Losses^
Because of the development of the event horizon one might take the view that singu-
larities are of no consequence to an external observer, since the observer loses contact
with an imploding object after it has fallen inside the event horizon. But while this is
true for electromagnetic communication the external observer still has contact with the
object in the sense that he can detect a gravitational field—unless there is some way in
which the object can radiate its gravitational mass. This question, with particular refer-
ence to neutrinos as the radiating agent, has recently been discussed by Michel (1963).
In the present section we shall re-examine this question.
To understand the issue more clearly it is convenient to use the Schwarzschild form
(8) for the line element outside the object, and to consider a situation in which the
energy-momentum tensor is small at all points between the object and the observer.
Over a small time interval the line element can be considered static, with ev and ex given
by equation (9). However, over a long time the constant in equation (9) may change
secularly. Thus the 41-component of Einstein’s equation is

f,Ç--S*CTS.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

914 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 139

If we define M by
e-\ = 1 — 2GM (23)
R
equation (22) gives
dM
-4wR2TaK (24)
dt

The right-hand side of equation (24) is just minus the energy flow through the sphere
of coordinate R—for example, in the electromagnetic case T41 is the radial component
of the Poynting vector.
We continue by estimating the energy flow in the form of neutrinos, on the basis that
an object of initial mass M0 implodes from a comparatively dispersed state in which the
density p and the temperature T$ = T/IO90 K are related by
/a\3/4 T3
p = 2.0
\Gj Mo1/2
(25)
'MqV/2 3
= 2.8 X 106 To gm cm-3.
Mo)

This equation was obtained by Hoyle and Fowler (1963#, b) for polytropes of index 3.
Iben (1963) has investigated the matter density-temperature relation for the equilibrium
state using the adiabatic approximation and has shown that equation (25) holds quite
accurately up to If ^ 105 Mo and is only a factor of 2 too high at if ~ 109 Afo. Evi-
dently in an imploding situation the temperature corresponding to a given density is less
than it is in an equilibrium state. Hence the adoption of equation (25) must tend to
exaggerate the neutrino emission, as it also does for masses above 105 if o. Since we shall
show that neutrino emission does not reduce if significantly, pur conclusion applies a
fortiori,
A sample of material moves in the p, TVplane along a track determined by the usual
thermodynamic relation,
—,dp-dU = ^-dt, (26)
p2 dt

in which P is the pressure, U the internal energy per mass, and dUv/dt the neutrino loss
per unit mass per unit time. Equation (26) holds for an observer who is co-moving with
the material and who uses locally flat coordinates. According to Fowler and Hoyle
(1964) the left-hand side of equation (26) takes the simple form

a,t T9 appreciably in excess of unity—the case with which we shall be concerned. The
expression (27) includes contributions to pressure and internal energy both from radia-
tion and from electron pairs. The pairs contribute 7/4 of the contributions of the radia-
tion in both cases. We also have from Fowler and Hoyle (1964) for r9 > 2
dUj, = 4._3 )<_1_0^ erg gm_1 sec_1 (Ta>2) (28)

for p in gm cm~3 provided all neutrinos and antineutrinos generated by e~ + e+—>v+v

escape, and provided the neutrinos are not seriously redshifted by the gravitational field.
Below T9 ^2 neutrino losses are much smaller than given by equation (28).

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 915

We shall find that all significant neutrino loss occurs before nuclear densities are
reached. Implosion can be considered to take place in accordance with
Â^SttGpo
(29)
S2~ 3S3 ’

where we neglect the —S~2 term in equation (7) since we are concerned with 5 <3C 1, and
where we also neglect the slowing of the implosion due to the pressure gradient, and the
slight acceleration due to the contribution of pairs and of radiation to E. This last con-
tribution is llaT4/4 and becomes equal to pc2 when p = 23TV in our units. For stars
with M < 109 ilfo which follow equation (25) the condition p = 23 TV is never reached
before the event horizon.
From pS3 = constant, we have
dS .dp_
(30)
T+y-

Combining equations (29) and (30), and remembering that the negative sign must be
taken in the square root of equation (29),
^ / 3 V/2^_ 1 dp
d 1/2
\8tGp) S ( 2 4irGp ) p '

Inserting in equation (26) leads immediately to

dp F. 1 dUyJdt I _ dT
p L (247rGp)1/2 llaT14/3pJ T ’

and using equation (28) this can be expressed numerically in the form
dp dT*
-7.10-Æ)-3 (32')
P o n ‘
This can be integrated straightforwardly and the constant of integration determined
with the aid of equation (25) to yield
■3/5
p^2.8X 106 10 -ÆV (33)
pW

This result shows that the material continues to follow a track p oc TV so long as the
second term in the second parentheses is small compared with unity. However, with
increasing temperature the second term increases in importance. In the limit for high
temperature the term in the second parentheses yields
p ~ 10~8 TV0. 04)

The track (33) of the material in the p, TV-plane is a curve with equations (25) and (34)
as asymptotes. The asymptotes intersect at the point
/MoVA
P^I.ÓXIO11 gm cm' (35)
\M0J ¡ r—(tr

We note at this point that pressure gradients can be effectively treated as decreasing
the value of G used in equation (32). However, we see that a pressure gradient which
reduces G to 10 per cent of its full value will only increase the time of fall, (247rGp)“1/2,
by a factor of 101/2. Thus the calculations made below for neutrino losses using the free-

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

916 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 139

fall time cannot be seriously low in value for this reason, and in fact other approxima-
tions more than compensate for the free-fall assumption.
Subject to the assumption that all neutrinos escape without redshift effects arising,
we are now in a position to estimate the total neutrino loss along the path (33). This
proviso means that the integration must not be carried beyond the event horizon where
the neutrinos are redshifted to zero energy. According to the work of Section II the
event horizon has Schwarzschild coordinate R — anz, while Mq— 47rpo/'&3/3, S = ar&2.
The proper density at the event horizon is obtained from p = po5~3, which gives
p = poa~Vb“6 at the event horizon. Using the definition of a, a = SirGpo/S,

Pmax
(36)
KSttGJ po2rt6 32ttG3 M02'

It will be recalled that units with c = 1 were used in the former work. To express p in
conventional units, a factor c6 must be introduced on the right-hand side of equation
(36). Numerically, we obtain
Pmax = 1-85 X 1016 0^) gmcm-3. 07)

Equations (25) and (37) yield a maximum temperature

(r9)max = 4.1 x 103 (Mo/Mo)112. (37')

Thus stars with M > 107 Mo do not reach 2 and neutrino losses are much smaller
than calculated in what follows.
The expression for the energy loss per gram of material along the path (33) from
p = 0 to Pmax or from ¿ = 0 to ¿(pmax) is

dUy
E ie dt
-’f, Tr -l a
Pn a
> * dU,/dt dp
(38)
’I (2 41rGp)V2 p

’max J’g9
18
= 1.92 X 10 /' dp erg gm“1,
n

where dtr is the time interval and dUv/dtf is the energy loss measured by an external
observer. It will be clear that the redshift decrease in the energy loss is just canceled
by the time dilation for such an observer. It is simpler to use the calculations of the
local co-moving observer and to cut off the loss which he calculates at the time he
measures, t(pmax)y corresponding to infinite time for the external observer. Eliminating
the temperature by means of equation (33) yields

77 Z,nax
£,= 5.0X 1018 / MoY f Z^dZ (39)
\Mq) Jo d+Z)V5’
where
/ Mo \5/6
z=8-5xl014G¿) p7/6

and
'M©Y/2
8.1 X 10®
.Mo)

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 917

The integral in equation (39) is T(9/7) T( 18/35)/T(9/5) ^ 1.67 for Zmax = 00. For our
purposes a sufficiently accurate integration of (39) for Zmax > 1 or M0 < 104 Mo is
Mo_
8.3 X1018 io-3 erg gm“1, <104. (40)
Mo
This expression has a maximum value at (M0/Mo) ^ 2 X 103 at which value E^(max) ^
1.37 X 1020 erg gm_1 = 0.15 â. Thus only 15 per cent of the rest mass is lost in the
maximum case. As might be expected it can be shown that Ev(max) is independent of
the numerical coefficient, 4.3 X 1015, in equation (28). Furthermore, it can be shown
that E„(max) does not depend critically on the power of Tq in equation (28). For very
large exponent Ey(max) approaches ^22 per cent. The point is that any mechanism of
great energy loss is quenched by the large redshift which arises when collapse to the
event horizon occurs.
These arguments apply to nuclear processes which customarily depend on a high power
of the temperature. High temperatures are reached only near the termination of collapse
when redshifts are large. On the other hand the second expression in equations (38) indi-
cates for any dU/dt = constant, for example, that the energy release diverges on evalua-
tion at the lower limit of integration. Thus it is not possible to rule out large losses for
processes such as gravitational radiation which do not depend critically on temperatures.
For Zmax < 1 which corresponds to stars with M0 > 104Mo one finds

E,~1.5 X 1026 0^^ 7

= 1.7 X 105 0^) c2erggm-1, 104<-^<107. (4i)

The upper limit, 107 Mo, occurs because equation (28) grossly overestimates neutrino
losses along the evolutionary path of such stars. Thus for M0 = 106Mo, Ev = 1.5 X
1017 erg gm-1 = 1.7 X IO-4 â and the fractional energy loss is very small indeed. Michel
(1963) has suggested that the energy gained by the envelope if the imploding core loses
all of its rest-mass energy and thus no longer acts gravitationally on the envelope is 2.9 X
1049 (recoil (M/Mo)3/2 erg, where M is the total mass of the star and (r9)coii is the tem-
perature at which collapse of the core begins. This expression must be multiplied by
Ev/â on the basis of our analysis and the mass of the collapsing core (M0 in our notation)
must be estimated. Michel (1963) used M0^0.37 M. The explosion energy of the
envelope thus becomes
Eenv^ 2 x io55 (r9)coii erg ,
or
^e^iow-o, (42)

which is ~10~3 at M ^ 104Mo and only 10“6 at M ^ 107Mo when the reasonable
estimate (r9)c0ii ^ 1 is employed. Thus neutrino losses from collapsing cores cannot
lead to envelope explosions with appreciable fractions of the rest-mass energy of a star.
It is worth noticing, by way of concluding the present section, that the track (33)
applied to ordinary stellar masses determines the situation under which neutrino emis-
sion could produce a free-fall implosion. This track lies far below the track actually fol-
lowed by evolving stellar material (Hoyle and Fowler 1963a, b). This shows that free-
fall implosion is never produced in ordinary stars by neutrino emission.
e) Rotation
The implosion of a discrete body including rotational effects has not been solved in
general relativity. Rotation effects have been studied in cosmology, and we are encour-
aged to think that what is known about the cosmological case may be adequate for a

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

918 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 139

general discussion. As emphasized in Section II, the implosion of a finite body without
rotation is closely similar to the cosmological case.
We have so far regarded S(t) as a scale factor applicable to all three space coordinates.
Now we regard S as applicable only to a coordinate taken parallel to the axis of rotation.
A different scale factor R(t) will be used for two coordinates taken perpendicular to the
rotation axis. In Newtonian mechanics the differential equations for R, S are of the form

R2^ a Q2 _ SttGpo
a
(43)
~ 3

¿i2,, 2a
(44)
S*—R*S'

where fí is the angular velocity at R = 1, and terms which become small as the implosion
proceeds have been omitted.
The right-hand side of equation (43) has a zero for sufficiently small R, implying that
implosion is halted and reversed to explosion so far as the coordinates perpendicular
to the axis of rotation are concerned. But no such effect occurs for the coordinate parallel
to the rotation axis. In the cosmological case implosion continues indefinitely for this
coordinate (Narlikar 1963). However, for a finite body a different situation can arise
due to a pressure gradient, as will now be seen.
Suppose that at the onset of implosion from a dispersed state the body is essentially
spherical in shape, with R = S = 1. Rotation eventually causes S/R to decrease below
unity. Now the pressure gradient necessary to maintain equilibrium in a direction paral-
lel to the rotation axis can be considerably less than would be necessary for the whole
mass in the absence of rotation. In fact the problem of support is reduced to that for a
mass of only M(S/R)2. For sufficiently small S/R this can be reduced below the limiting
value of order Mo, and equilibrium is possible either through electron degeneracy or
nucleon degeneracy. Hence it is in principle possible for a rotating object to maintain
itself whatever the total mass M, as was pointed out by Hoyle (1947). For M^>Mo
the shape must be that of a thin disk. Although in the imploding case oscillations must
occur, we expect they will eventually become damped away, perhaps by repeated bursts
of neutrino emission.
At first sight one might suppose that, because of this effect, implosion to a singularity
never occurs in any actual case, since presumably there is never a complete absence of
rotation. This view is incorrect, however. If p attains nuclear densities, the energy density
E cannot be taken as p and the Newtonian equations (43) and (44) require modification.
We must have E oc p4/3, and the terms in a are changed to the proportionality Rr* for
the È2 equation and to Rr*n S~4/z for the S2 equation. Hence both terms on the right-
hand side of the È2 equation behave in the same way as i? —» 0, and there is no root if
0 is small enough. The gravitational field can be strong enough even to crush rotation,
a result that never occurs in the Newtonian case.
The situation is further complicated, even in the case Ec^ p, by the circumstance
that a rotating disk can be locally unstable. Contraction parallel to the rotation axis
increases the density by R/S. A local aggregation is then unstable against contraction
perpendicular to the rotation axis. Indeed, a spherical aggregation of mass ~(S/R)2 M
can shrink as a whole by a factor R/S before rotation again prevents contraction per-
pendicular to the axis. Hence the density can rise by a total of (R/S)*. We may state
this in the following way: If a rotating disk of mass M and radius a breaks up into more
or less spherical fragments of mass %~2 M (x^> 1), the density in each fragment can
increase to M/az.
The important issue is whether nuclear densities are reached before or after the
fragment masses become of order Mo. If before, no equilibrium is possible, since rotation

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 919

becomes ineffective before the fragment masses are small enough for pressure gradients
to maintain equilibrium. If after, we expect the implosion to be halted, and a final state
reached in which the object has divided into stable fragments with masses o that
move in nearly coplanar orbits about a common center.
Set x~2 M Mo. The density then increases by (M/M®)2 above the value at which
rotation first impeded shrinkage perpendicular to the axis. Write pi for this density, so
that the fragments have density pi (M/Mo)2. Our two cases depend on whether this value
is > or < ^lO16 gm cm“3. Hence our cases are given by

Pi ^ ^ 1016 gm cm* (45)

This density is close to that which occurs at the event horizon. Hence we arrive at the
conclusion that, if rotation does not become dynamically important until after an im-
ploding object has retreated within the event horizon, singularities develop.
When the “less than” sign applies in equation (45) two cases appear to arise. If pi is
sufficiently small compared with the right-hand side of equation (45) the relevant pres-
sure can arise from electron degeneracy and the object fragments into white dwarfs.
At larger values of pi nucleon degeneracy occurs, however, and the object fragments into
neutron stars.
/) The Prevention of Singularities in C-Field Cosmology
Hoyle and Narlikar (1963) have shown that equation (21) is modified to

^ _!+MË_ü
52 3 S«

by the presence of the C-field of steady-state cosmology where ^4 is a small constant.

Since E is proportional only to S“4 at very high densities, the right-hand side of equation
(46) can have two zeros. In fact, for any physical motion there must be two roots, a
larger root with the —S~2 term essentially compensating the E term—this is just the
dispersed state—and a smaller root with the — ^42 5“6 term essentially compensating the
E term. This result arises from the negative-energy density of the C-field. Although the
concept of a negative-energy density is strange, several arguments can be advanced in
its support:
1. A negative-energy field is gravitationally repulsive. The expansion of the universe
can be taken as evidence of the existence of such an effect—otherwise we must take the
unsatisfactory step of imposing special initial boundary conditions.
2. Matter must be “created” either continuously or at the “origin” of the universe.
A negative-energy field appears necessary to obtain a mathematical description of the
creation process.
3. A negative-energy field now seems to be necessary if singularities are to be avoided.
The singularities are prevented through the gravitational repulsion of the field.
The two roots of the right-hand side of equations (46) imply that the object oscillates
between a maximum Si and a minimum Sz instead of imploding into a singularity. It is
to be expected (Hoyle and Narlikar 1963) that the oscillations are gradually damped
in the sense that S\ decreases. Neutrino emission is probably the main damping agent,
although charged particles may also be emitted. The object never attains a static state
—that is, Si can never be reduced to 8% because the damping ceases when the redshift
cuts off the emission. The maximum Si then takes its value, ar&2, at the event horizon,
whereas S2 an2, that is, at minimum the object lies far inside the event horizon. This
is because the coefficient A2 is very small.
According to Hoyle and Narlikar it seems unlikely that the maximum density can be
less than 1030 gm cm“3, corresponding to an interparticle separation of 10-18 cm, or less.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

920 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol 139

Nuclear physics gives no real guidance as to what might happen under such conditions.
Boson fields are to be expected with energy densities proportional to S“4, individual
bosons have energies proportional to 5“1. At 1016 gm cm-3 the energies are of order
1 BeV, while at p ^ 10 gm cm energies of order 105 BeV would be expected. The pos-
30 -3

sibility exists that such bosons emerge from the surface of the object. At first sight it
might seem as if nothing of this kind could be detected by an outside observer, since the
object lies far inside the event horizon at stages where the density is very high. How-
ever, particles can be pushed outside the event horizon by the expansion of the object
itself. Any massless boson that emerges radially from the surface always stays ahead of
the surface. So long as the surface crosses the event horizon the particle is ultimately
disgorged into the outside world. Decay could provide electrons and protons or neutrinos
and neutrons if appropriate weak coupling interactions exist.
It is clear that at just this point a serious gap exists in our knowledge, a situation that
has been strongly emphasized by Wheeler (cf. Wheeler, Wakano, and Harrison 1958).
A filling of this gap would provide an interesting connection between high-energy physics
and astronomy.
II. ASTROPHYSICAL CONSEQUENCES
In the previous sections we have described the various situations which have been
envisaged as the final stages of evolution of a star which reaches the end point of ther-
monuclear evolution with a mass greater than the mass which can be supported either
by degenerate electron pressure or degenerate neutron pressure. While it is not clear
which of these theoretical possibilities is the correct one, in each case the star eventually
reaches a situation in which it ceases to communicate with the outside world except
through the action of its gravitational field. In the theory of Hoyle and Narlikar there
is the possibility that such an object will be able to radiate some fraction of its mass
energy as it pulsates, but such pulsations will eventually damp out and the remnant will
disappear.
Thus we conclude that the following facts should be taken into account.
1. Invisible mass is likely to be present wherever the condensation and evolution of
stars have occurred. How much mass is present will depend on the masses of the stars
which condensed, and on what fraction of this mass can be ejected in the stages of evolu-
tion prior to catastrophic implosion.
2. In these circumstances star formation is a one-way process which removes uncon-
densed material in the universe and transforms it into a form which makes it immune to
further evolution.
In modern theories concerning the evolution of stars and galaxies it has been supposed
that much of the material which is condensed into a first generation of stars will be ejected
and condensed into a second generation, etc. While such processes do occur, it is necessary
to take into account that fraction of the mass which goes into an invisible form and no
longer plays a direct role in the evolution.
3. In the theory of Hoyle and Narlikar mass energy can be radiated by the object after
it has reached the collapsed state. This energy is bound to be radiated in the form of
high-energy quanta. It thus provides a continuous energy input for non-thermal radio
sources.
We now discuss the implication of these results for astrophysics in more detail.
a) The Presence of Invisible Mass
We first discuss the stellar mass function and the mass-to-light ratios for aggregates
of stars and for galaxies.
In our Galaxy the number of stars born with masses between ikf and M + dM is pro-
portional to ^M~lA d log M at any rate for masses up to 5 If o (for a review of the work
on clusters cf. Burbidge and Burbidge 1958). This means that, whereas most stars lie
at the bottom end of the mass range, the total mass of a group of stars comes mainly

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 921

from the upper end of the range. Except for a few stars in very young clusters, stars
near the upper end have already evolved. Hence most of the mass originally condensed
into stars has evolved, in the sense that the stars into which it was originally condensed
are no longer visible. The question evidently arises as to what has happened to this
material. There seem to be only two possibilities: (i) The bulk of the material was
ejected from the stars in the course of their evolution and is now condensed into further
stars, (ii) The stars evolved into imploding objects and the mass is now invisible. The
possibility of the material being stored permanently in the interstellar medium as gas
appears to be excluded, since not more than 10 per cent of the mass of the Galaxy seems
to be in the form of gas.
The first of these possibilities can be tested by considering what would be expected
for the mass-to-light ratio in our own Galaxy. Consider first the simple case of main-
sequence stars with the mass distribution if“2-4 dM extending from if 0 ~ 0.1 if o up
to if of the order of, or greater than, ifo. Write if* for the upper limit of mass and L*
for the upper limit of luminosity. The mass-to-light ratio is

~ / LM~24dM. (47)

WithL = L*(M/M*y on the main sequence, equation (47) is approximately evaluated

to give 6.5(M*/L*) [(if*/if0)0 4 — 1]. Next, we require values for if*, L*. In the
Galaxy Mo for the old star distribution, while L*c^2Lo at the so-called break
point of the main sequence. Hence in solar units we have a mass-to-light ratio of
^3[(if */if o)0-4 — 1]. Similar data do not exist for other galaxies. However, the theory
of stellar evolution would not permit a significantly larger value to be taken for if */L*,
unless the galaxies in question were substantially older than our own.
The mass-to-light ratio calculated in the previous paragraph must be reduced to allow
for young clusters and for stars that have evolved off the main sequence. Giants, in
particular, contribute appreciably to the light but not to the mass. Hence the expected
mass-to-light ratio cannot be much greater than ^3.
The mass-to-light ratio for the solar neighborhood is about 4 (Schmidt 1963a), while
Burbidge, Burbidge, and Prendergast, and Page (cf. Burbidge 1961) have found values
of order 3 or less for a number of spiral galaxies, mainly of type Sc. These values are
consistent with a choice of 0.1 ilfo for M0. On the other hand, the mass-to-light ratios
for some spirals are ^40-15 (M31, NGC 253, M81, and others; cf. Burbidge 1961) while
the average value for elliptical galaxies (Page 1962) is about 30. Such values point to
the existence of hidden mass, unless the mass function for stars is grossly different in
other galaxies from what it is in our own. Obscuration of light might perhaps falsify the
observed values in dusty spirals, but obscuration cannot be important for the ellipticals.
Hidden mass in the above sense could come from (i) faint white dwarfs, (ii) very faint
main-sequence dwarfs, corresponding to a very small value for if 0, (hi) imploded evolved
stars that have not lost appreciable fractions of their original masses, (iv) imploded ob-
jects of very large mass. As far as our Galaxy is concerned the number of white dwarfs
in the solar neighborhood is not sufficient for (i) to make more than a modest contribu-
tion to the total mass. Nor is there any direct evidence from the solar neighborhood to
support (ii). However, it seems probable that considerable numbers of red dwarfs are
present in some galaxies (Spinrad 1962) and that the mass function is not the same as
it is in our own Galaxy. We have pointed out elsewhere the difficulties associated with
the assumption of a universal mass function (Burbidge, Burbidge, and Hoyle 1963). On
the other hand, either a very low value for JkTo or a very gross difference in the mass func-
tion for main-sequence stars has to be assumed to obtain the mass-to-light ratios for
ellipticals, and the observations do not indicate that such large differences as these, in
fact, exist.
An upper limit for the importance of (iii) can readily be calculated. Suppose the

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

922 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 139

M~2 4 dM distribution extends at birth to an upper limit if**, if* being the upper limit
to which it applies at present. Then the total mass born over the whole range from iffl
to if** exceeds the range from if0 to if*—the latter being the present-day observed
range—by the factor

[1 - (ifo/if **)° 4]/[l - (ifo/if *)0*4] « 1/[1 - (ifo/if *)0-4]

for sufficiently large if**. Taking if*^ifo, ifo —0.1 if©, this factor is 5/3. The
mass-to-light ratios, including hidden mass in the form (iii), then becomes ^5. Such a
value is clearly compatible with the mass-to-light ratios of galaxies such as our own, and
perhaps with M31, M81, NGC 253, etc. However, the much larger ratios which are found
for many elliptical galaxies might indicate that very massive objects have imploded and
are present in the form of invisible mass, that is, possibility (iv) may be required to ex-
plain the observational results.
It is self-evident that direct observational evidence for the presence of invisible mass
can never be forthcoming. The arguments given above concerning the mass-to-light ratios
are indirect but give some indication of phenomena which may be explained by these
developments in the theory. However, we do not believe that the discussion in the first
part of the paper is strongly dependent on the indirect observational arguments. We do
assert that masses which arrive at the end point of evolution and implode must follow
one of the paths described in that section. Another phenomenon which might suggest
that some mass is present in the form of invisible matter is the presence of invisible (often
called “infrared”) members of binary star systems.
Another problem of some significance is that concerning the stability of clusters of
galaxies (cf. the papers published in A.J., 66, 533-636, 1961). It is found by applying
the virial theorem to clusters and groups of galaxies that in nearly all cases it must be
supposed either that the systems are expanding, or else that a very large amount of
invisible matter must be present. Probably systems of both types occur. In the case of
clusters and groups which are stable by virtue of their containing a large fraction of
mass in invisible form, it is quite reasonable to suppose that much of this is made up
of large masses >105 JkTo which have imploded.

b) The Effect of the Concept of Imploded Mass on Stellar Evolution and Nucleosynthesis
The rate at which the material in a galaxy is enriched in the heavy elements is de-
pendent, first, on the efficiency of the processes of nucleosynthesis in the interiors of
stars and, second, on the processes which will redistribute this material into the inter-
stellar medium. The discussion given in Section I of this paper strongly suggests that
the latter process is much less efficient than has been thought up to the present, since
in some proportion of the stars with masses above the critical mass we may expect that
the implosion will occur and the mass will vanish. Thus some modification of the calcu-
lations of the rate of enrichment of heavy elements as a function of star formation as
they have been made by Schmidt (1963&) is required, since he made the assumption that
all of the mass in a star in excess of the white dwarf mass would be returned to the
interstellar medium. In the case of massive supernovae (Burbidge, Burbidge, Fowler,
and Hoyle 1957; Hoyle and Fowler 1960; Fowler and Hoyle 1964) the possibility that
mass vanishes when these stars have evolved is of particular interest. It is usually sup-
posed that there have not been more than ^108 supernovae in the lifetime of the
Galaxy, whereas the number of stars that have been born with M > 2M® is probably
of order 1010. It therefore seems necessary to argue that only a small fraction of implod-
ing stars, about 1 per cent, become supernovae. A similar result is obtained if we simply
take the known number of star deaths among A, B, and O stars in young clusters, together
with an assumed present-day supernova rate of ^1 per 102 years. Rotation may supply
the restraining factor. Imploding stars with too much rotation may attain stable disklike

© American Astronomical Society • Provided by the NASA Astrophysics Data System

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 923

structures in accordance with the considerations of Section 1(e) and these also have the
property of hiding mass. It may well be necessary for a star to satisfy the “greater than”
condition of equation (45) in order that it become a supernova. The physical reason could
be that superdense conditions are needed in order that relativistic particles be generated.
It has been stressed previously (Fowler and Hoyle 1960) that, if we suppose that the
r-process elements are made in supernovae in sufficient amount to explain the light-
curves on the californium hypothesis, then taking an average rate of supernovae as 1
per 300 years far too much r-process material would be produced. This difficulty can be
avoided if Type I supernovae are taken to have such small mass and such long evolution
times that none evolved in the Galaxy before the solar system formed. However, an
alternative explanation which involves the rejection of the californium hypothesis will
be explored in the next section. A similar difficulty is encountered when we consider the
production of iron peak elements in Type II supernovae, that is, too much iron is pro-
duced. However, it now appears possible that much of the elements synthesized in these
massive stars are never ejected but are contained in the imploded mass, so that this dif-
ficulty can be surmounted.
Finally, in this section we turn briefly to the evolution of a globular cluster. In a
recent analysis of the dynamics of M3, Oort and van Herk (1959) came to the conclusion
that more than half the mass was originally in the form of stars with average masses of
4.4 Mo. Since these stars must have evolved long ago they argued that the excess mass
above the white dwarf mass, i.e., 3.8 Mo per star, must have been ejected and lost from
the cluster. This assumed mass loss through stellar evolution is far greater than the
mass loss due to the evaporation of stars. However, according to our earlier considera-
tions, this mass may not have been lost to the cluster but may still be present in invisible
form. While the observable dynamical effects would be small (a change in mass by a
factor of 2 corresponds to a change in velocity dispersion by -\/2) the possible presence
of such mass rather than its assumed ejection should be considered; although it could
still be the case that the dynamical processes involved produced a sufficient recoil on
the remnants to cause escape from the cluster.
c) The Injection of Energy into the Interstellar and Intergalactic Medium from Collapsed Stars
The imploding bodies discussed in Section I have the property that gravitation is in
principle capable of yielding ^9 X 1020 erg gnr*1, much greater than the energy yield
from any nuclear reaction (Hoyle and Fowler 1963a, h; Burbidge 1962a). But classical
implosion into a singularity does not seem to provide an atomic or nuclear mechanism
whereby the full dynamical energy stored in a collapsing body can be returned in an
observable form to the outside world.
As pointed out by Hoyle and Fowler (1963a) it is probable that hydrogen burning
can retard the implosion at a central temperature near 8 X 107 ° K for a limited time
(^lO6 years) and supply the positive internal energy (^1059 ergs for M ~ 10Wo)
necessary for hydrostatic equilibrium under general relativistic conditions (Iben 1963)
as well as that required to match the radiated energy (^1060 ergs for M ~ 10Wo).
In fact, radio stars may well represent the early stage of the implosion of massive objects
where the luminosities predicted by Hoyle and Fowler, Z, = 2 X 1038 {M/Mo) —> 2 X
1046 ergs sec-1 for M = 10Wo, are of the order of magnitude observed for 3C48 for
example (Greenstein and Matthews 1963). The radius and effective surface temperature
depend critically on the structure of the star during hydrogen burning. Hoyle and
Fowler (1963a) give R ~ IQ11 {M/Mo)l/2 cm for polytropic index w = 3, but this could
be low by as much as a factor of 102 to 104, since a larger index is indicated if general
relativistic considerations are taken into account. Thus 1014 (M/Wo)1/2 cm with
considerable uncertainty. Similarly the surface temperature Te~ 7 X 104° K given by
Hoyle and Fowler (1963a) may be too high and a value Te ^ 104 ° K is probably to be
preferred with an uncertainty of a factor of 3 either way.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

924 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 13

During the hydrogen-burning period the central density is ^lO-1 to IQr2 gm cm“3 an<
the ratio of gas pressure to total pressure is ^10“2 to 10“3, these values holding fo
M ~ 106ilfo to 108ilfo using equations given by Hoyle and Fowler (1963a). Settinj
7a = 37 — 4 ^ ß/2 1 in equation (130.3), p. 192, of Eddington (1930), these value
lead to periods of pulsation H ^ (S0T/3Gßp)1/2 ~ 1.2 (Gß(p))1/2 ~ 106 to 107 sec. Smitl
and Hoffleit (1963) have found evidence in 3C273 for light variations with periods h
this range. In view of this relatively satisfactory model for radio-star behavior we mus
emphasize once again that somewhat less than 1 per cent of the rest-mass energies o
these objects is required to account for the energy requirements discussed above. Hydro
gen burning can supply this energy. However, if the light output of such an object i
to be attributed to the comparatively steady luminosity of a massive star during it
“main-sequence” evolution, at least one previous outburst involving the evolution am
collapse of another massive star must be invoked to account for the radio flux from th<
source.
It is also possible that rotation prevents the development of a singularity, at any rat<
in some cases, and that collapse to disklike aggregations of white dwarfs or neutron stan
can give rise to large energy sources. The mechanism whereby such aggregations act a¡
sources is far from clear, however.
Another possibility is the release of energy in the form of gravitational radiation. Be
cause of the critical dependence of this release on rotation and on the model of collapse
as discussed in Hoyle and Fowler (19636) it is difficult to make realistic estimates ol
the energy release and we have not attempted to do so. Our remarks in Section 1(d)
concerning the difficulty in abstracting energy from collapsing systems are subject to
the qualification that gravitational radiation rates have not yet been determined.
The non-classical oscillations discussed in Section 1(f) seem to provide a more con-
venient energy source. Energy can be derived through a damping of the oscillations until
the redshift cutoff is reached. The yield is then about 5 X 1020 erg gm-1 (Hoyle and
Narlikar 1963). If the fraction of all material involved in this process is/, the energy
density for the whole universe is ^5 X 1020 /pc, where pc is the cosmological mass
density, usually taken as ^3 X 10“29 gm cm“3. This gives ^10“8/ erg cm“3. As stated
in Section I, the main damping agent is probably neutrinos, and the present estimate
can be taken as referring to the energy density of v, v.
The speculations of Hoyle and Narlikar on the oscillatory problem lead to extraordi-
narily great densities at maximum compression, which has the advantage that very high-
energy particles would surely be generated. These speculations also have the advantage
of relating high-energy physics with astrophysics and cosmology. If k is the fraction of
the energy released as relativistic electrons and protons, the cosmic-ray energy density
for the whole universe would be ~10r8fk erg cm-3.
Burbidge and Hoyle (1964) have suggested that massive objects may be the source
of cosmic rays and that the energy density may be ^4 eV cm-3 everywhere in space.
This would require fk ~ 10~4. It seems unlikely that k could be more than a few per
cent, in which case Burbidge and Hoyle require / to be not much less than 1 per cent.
Since matter condensed into galaxies has mean density ^3 X KI-31 gm cm“3—i.e.,only
per cent of the value taken above for the total cosmological density, we evidently
require an appreciable fraction of the mass of the galaxies to have evolved into imploded
objects. This question has been discussed in the previous sections.
To account for the energy contained in non-thermal radio sources, if we take the
point of view that this is energy released in collapse to very high-density configurations,
it is therefore necessary to suppose that the model of Hoyle and Narlikar, and not the
classical implosion solution, is correct. If we make this assumption, then the supernova
remnants and the objects which give rise to the powerful extragalactic radio sources
continuously inject high-energy particles and quanta into the surrounding medium, and
the bulk of the energy injected must be in the form of neutrinos. This hypothesis leads

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 925

to quite a different picture of the evolution of a supernova remnant than that previously
developed.
First of all, it is possible to suppose that the interacting particles which are ejected at
the earliest phase are the source of visible radiation of the supernova, either through
direct synchrotron radiation or through the penetration of the particles into an expanding
envelope, ejected perhaps as a consequence of a nuclear explosion in the outer parts of
the star.
We now turn to the situation in the Crab Nebula, the remnant of the supernova of
AD 1054 which shows evidence of its continuing activity through the moving wisps
observed by Baade and the presence of high-energy electrons with synchrotron lifetimes
much less than the age of the remnant.
It is remarkable that no star brighter than about 18m appears at the present day to
be associated with the Crab Nebula. This, taken together with the estimates of Oster-
brock (1957) and O’Dell (1962) of no more than 0.3 Mo for the mass of the nebula, leads
to the challenging question of what has happened to the original star. Since it has not
been disintegrated, the major part of the mass must still exist as a compact object. It
cannot be a white dwarf, since cooling could not have reduced its luminosity sufficiently
in only 103 years. It can scarcely be a single neutron star since it is unlikely that the
original mass was less than Mo. The interesting possibility is that the star is now in the
oscillatory state discussed in Section I(/) and that the continuing activity of the Crab
Nebula is due to a steady output of high-energy particles.
The present picture allows a different point of view to be taken about the origin of
the filaments of the nebula. The volume of the nebula is ^1056 cm3. The mass of inter-
stellar gas normally present within such a volume, situated as it is^quite near the galactic
plane, would be ^3 X 1032 gm. If the filaments only have mass <0.3 Mo it follows that
they must be largely interstellar gas, unless the region of the Crab is extraordinarily void
of gas. We take the view that the whole filamentary structure is interstellar in origin,
that no significant expanding envelope was emitted, and the light of the supernova was
produced by particles of high energy, perhaps by the synchrotron process. A similar
supposition for all Type I supernovae explains the absence of detectable lines in the
spectra of these stars.
On this point of view, the approximately exponential decay of the light-curves of Type
I supernovae can then no longer be attributed to Cf254, or to other very heavy nuclei
subject to spontaneous fission. Damping of the oscillations of the central object may be
expected to proceed quite rapidly to the stage where redshift effects become important,
and exponential decay of the light-curve could arise from this damping. We urge that
observational astronomers turn their attention to the question of which of the two alter-
natives we have suggested for the light-curves of Type I supernovae is the correct one.
Is the source of the energy radioactivity or gravitation?
The present outward speed of the filaments is ^1000 km sec“1, the total kinetic energy
being ^TO48 erg. It has been argued in the past that the energy of the relativistic particles
associated with the Crab cannot be greater than the kinetic energy of the filaments;
otherwise the nebula would expand faster than it is observed to do. This argument is
only correct, however, if the relativistic particles are trapped within the filaments. This
has always been assumed because in the past it was felt to be difficult to explain the
origin of even 1047 erg in the form of high-energy electrons. The same difficulty does not
arise in the present picture. The rest energy of a star of mass 1-10 Mo is 1054-1055 erg.
Even though most of this energy emerges as neutrinos, the energy available for charged
particles could readily exceed 1050 erg. This leads us to the quite different point of view
that the filaments have derived their outward motion from the momentum of the rela-
tivistic particles, not from their energy. We suppose that the particles are emitted by
the central object and that they experience deflection by the filaments—the interstellar
magnetic field within the filaments forces the particles to fan out in much the shape of a

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

926 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E. BURBIDGE Vol. 139

comet’s tail in order to get around a filament on which they are incident.The filamentary
structure itself may arise from the ability of the high-energy particles to punch holes
in the gas. Indeed, the ravaged appearance of the nebula could be a consequence of such
violent fluting.
If the filaments take up a fraction 6 of the outward momentum of the relativistic par-
ticles the filamentary kinetic energy E/ will be related to the particle energy Ep by

-E/— 0 -~Ep> (48)

where V is the filamentary velocity. With E/O^ 1048 erg, V ^ 108 cm sec-1, 3 X 1050
-1
0 erg. Individual relativistic particles spend perhaps 10 years within the nebulosity—
the radius is about 3 light years while allowance for deflections and spiraling motions
reasonably increases the time to 10 years. Hence in the 103 years of existence of the Crab
we expect there to have been about 102 generations of relativistic particles. The energy
content at any moment is then ^10“2 Epc^3 X 1048 0”1 erg. Oort and Walraven (1956)
and Woltjer (1957) estimate the energy content, excluding protons and excluding elec-
trons that do not emit synchrotron radiation either in the radio band or the visible spec-
trum, as 1047 to 3 X 1048 erg depending on the intensity of the magnetic field. The energy
content could be as high as 1049 erg.
Deflection around a knot of gas arises from compression of the magnetic field. Deflec-
tion becomes appreciable when the magnetic pressure IP/Stt increases to a value com-
parable with the energy density of the incident beam of particles. Taking the total resi-
dent energy as ^1049 erg and the volume as 1056 cm3, the latter is ^10~7 erg cm-3. Hence
we expect compression to raise jff to a value of order 10“3 gauss, a value that has been
used in past investigations. Partly because of the compression of the field, and partly
because the electrons spend a comparatively long time in the region of compression, we
expect synchrotron emission to occur mainly at the inside faces of knots and filaments
of gas—i.e., the faces turned toward the central star.
If the central star emits particles at a steady rate, one might at first sight expect the
outward motion of the gas to be decreasing, because of the increasing quantity of inter-
stellar material swept up by the system. However, this need not be the case since B also
depends on the amount of gas in the system. In fact, so long as 0 < 1 we expect 6 to be
proportional to the total mass. Then it can easily be shown that the system accelerates.
Deceleration sets in when 0 increases toward unity.
It is also easy to see that interesting polarization effects can arise. For the simple case
of a coplanar fluting around a knot of gas, an observer with a face-on view finds the
polarization structure shown in Figure 1. This detail is mentioned because the polariza-
tion structure of the Crab contains an example of this special case (Woltjer 1957). A
general discussion of polarization lies far outside the present work. The problem is com-
plicated by the unknown three-dimensional structure of the system, by the lack of regu-
larity in the arrangement of the knots and filaments, and by uncertainty in our knowl-
edge of the initial run of the magnetic field lines.
In this picture of a supernova remnant tremendous fluxes of particles will escape into
the disk of the Galaxy, and this picture would be inconsistent with the estimates of the
relativistic electron energy density in the disk <5 X 10-14 erg/cm3 if the particles were
contained in the Galaxy. Taking 1069 erg as the total production in the lifetime T of the
Galaxy and taking 3 X 1066 cm3 as the volume of the disk the requirement is that

1069 ^ < 5 X 10-14 X 3 X 1066,

where t is the escape time. Thus t < ^10“6 T ~ 104 years. Thus the Galaxy must be
entirely open as is indicated on other grounds (cf. Ginsburg and Syrovatsky 1961).

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

No. 3, 1964 RELATIVISTIC ASTROPHYSICS 927

The extragalactic radio sources probably have short lifetimes ^lO6 years (Burbidge
1962Ô; Burbidge, Burbidge, and Sandage 1963), but in many cases much shorter time
scales within these are indicated. Examples are M87 which contains optical synchrotron
electrons with half-lives ^102-103 years, the nuclei of Seyfert galaxies, and the starlike
extragalactic objects (Matthews and Sandage 1963; Schmidt 1963¿;; Hazard, Mackey,
and Shimmins 1963; Oke 1963; Greenstein and Matthews 1963) which show light varia-
tions over time scales of months and years (Matthews and Sandage 1963; Sandage 1963;
Smith and Hoffleit 1963; Shklovsky 1963). These characteristics all indicate that con-
tinuous activity takes place in such objects. Since the tremendous energies required to
explain these objects most likely come from gravitational sources, the Hoyle-Narlikar
hypothesis of continuous energy injection appears to be indicated.

Fig. 1.—Magnetic field lines bent around a knot of gas by a beam of incident relativistic particles
produce the polarization vectors shown above for an observer with line of sight perpendicular to the
field lines.

We are indebted to Dr. John Gaustad for pointing out an error in the original draft
of Section 11(a) of this paper.
This work was supported in part by the National Science Foundation, the Office of
Naval Research, and the National Aeronautics and Space Administration.
One of us (W. A. F.) wishes to acknowledge support by the John Simon Guggenheim
Foundation during the period 1961-62, when this work was started.

REFERENCES
Burbidge, E. M., Burbidge, G. R., Fowler, W. A., and Hoyle, F. 1957, Rev. Mod. Phys., 29, 547.
Burbidge, E. M., Burbidge, G. R., and Hoyle, F. 1963, Ap. /., 138, 873.
Burbidge, G. R. 1961, A.J., 66, 619.
. 1962a, Ann. Rev. Nuclear Sei., 12, 507.
. 1962fr, Prog. Theoret. Phys. Japan, 27, 999.
Burbidge, G. R., and Burbidge, E. M. 1958, Handbuch der Physik, 51,134.
Burbidge, G. R., Burbidge, E. M., and Sandage, A. R. 1963, Rev. Mod. Phys., 35, 947.
Burbidge, G. R., and Hoyle, F. 1964, in preparation.
Datt, B. 1938, Zs.f. Ap., 108, 314.
Eddington, A. 1930, Internai Constitution of the Stars (Cambridge: Cambridge University Press).
Fowler, W. A., and Hoyle, F. 1960, Ann. Phys., 10, 280.
. 1964, Ap. J. Suppl., in preparation. This paper gives an extensive discussion of the topics covered
by W. A. Fowler in the Henry Norris Russell lecture given at the Alaska meeting of the American
Astronomical Society, July, 1963.

© American Astronomical Society • Provided by the NASA Astrophysics Data System

19 64ApJ. . .139. .909H

928 F. HOYLE, W. FOWLER, G. BURBIDGE, AND E BURBIDGE

Ginsburg, V. L. and Syrovatsky, S. I. 1961, Suppl, to Prog. Theoret. Phys. Japan, No. 20.
Greenstein, J. L., and Matthews, T. A. 1963, Nature, 197,1041.
Hazard; C., Mackey, M. B., and Shimmins, A. J. 1963, Nature, 197, 1037.
Hoyle, í\ 1947, M.N., 107, 231.
Hoyle, F., and Fowler, W. A. 1960, Ap. J., 132, 565.
. 1963a, M.N., 125,169.
. 19636, Nature, 197, 533.
Hoyle, F., and Narlikar, J. V. 1963, Proc. Roy. Soc., in press
Iben, I. 1963, Ap. J., 138, 1090.
Matthews, T. A., and Sandage, A. R. 1963, Ap. J., 138, 30.
Michel, C. 1963, Ap. J., 138, 1090.
Narlikar, J. V. 1963, M.N., 126, 203.
O’Dell, C. R. 1962, Ap. J., 136, 809.
Oke, J. B. 1963, Nature, 197, 1042.
Oort, J. H., and Herk, G. van. 1959, B.A.N., 14, 299.
Oort, J. H., and Walraven, T. 1956, B.A.N., 12, 285.
Oppenheimer, J. R., and Snyder, H. 1939, Phys. Rev., 56, 455.
Oppenheimer, J. R., and Volkoff, G. M. 1939, Phys. Rev., 55, 374
Osterbrock, D. E. 1957, Pub. A.S.P., 69, 227.
Page, T. L. 1962, Ap. J., 136, 685.
Sandage, A. R. 1963, private communication.
Schmidt, M. 1963a, private communication.
. 19636, Ap. J., 137, 758.
. 1963c, Nature, 197, 1040.
Shklovsky, I. S. 1963, private communication.
Smith, H. J., and Hoffleit, D. 1963, Nature, 198, 650.
Spinrad, H. 1962, Ap. J., 135, 715.
Wheeler, J. A , Wakano, M., and Harrison, K. 1958, La Structure et révolution de Vunivers, ed. R. Stoops
(Solvay Conf Rept., Brussels).
Woltjer, L. 1957, B.A.N., 14, 39

© American Astronomical Society • Provided by the NASA Astrophysics Data System