The Contraction of Gravitating Spheres - Bondi
The Contraction of Gravitating Spheres - Bondi
The Contraction of Gravitating Spheres - Bondi
B y H. B ondi, F.R.S.
King's College, London
1. I n t r o d u c t io n
Renewed interest in the contraction of massive objects has been caused by the
discovery of star-like radio sources (cf. Fowler & Hoyle 1963). Since deviations
from spherical symmetry are likely to be incidental rather than basic features of
the processes involved, the discussion may be confined to spherically symmetrical
contraction. This is reasonably tractable in general relativity even when gravita
tional fields and pressures are too intense for the Newtonian approximation to be
sufficient. However, attention has usually been confined to consideration of a
succession of static models (cf. Iben 1963). In early work on this subject by
Eddington (1930) it was suggested that an invariant integral I of the density of
the energy invariant T through the configuration represents the number of
nucleons. If all the energy released during contraction travels away from the
configuration without emission of nucleons (as electromagnetic radiation,
neutrinos, etc.), the number of nucleons and hence I should accordingly remain
constant throughout the contraction. However, this suggestion was recently shown
to be wrong (Bondi 1964a).
In fact, the identification of / as a multiple of the nucleon number is wholly
fallacious. General relativity is a pure continuum theory and says nothing whatever
about particles, nucleons, etc. I t is true th at both I and the nucleon number are
invariants in the strictly technical sense that they are independent of the system
of co-ordinates chosen, but this in no way implies th at they are related to each other,
nor th at either or both of them should be constant in time. A purely physical
assumption about the system under investigation (non-emission of nucleons)
leads to the result th at the nucleon number is constant, and no statement about the
time variation of I can be made without full physical specification of the material
of the configuration. If we assume that the contraction is adiabatic, so that no energy
is radiated away, then the Schwarzschild m must remain constant. For the space
[ 39 ]
Downloaded from http://rspa.royalsocietypublishing.org/ on November 27, 2017
40 H. Bondi
outside the m atter is then permanently empty, and, by Birkhoffs theorem, a
spherically symmetric empty space-time is necessarily static. Since it is also fully
characterized by m, this quantity is constant. In the initial highly diffuse state
in which the Newtonian approximation must apply, m is simply the total rest mass
of the matter, and in this case therefore m also describes the nucleon number.
Note th a t in the Newtonian approximation adiabatic contraction is possible for
a material in which density pand pressure p are related by
V ~ p y, (!)
provided that, y ^ f , and is indeed catastrophic if y < f . In the relativistic case
the neutral adiabatic relation (corresponding to y = f) is no doubt not only more
complicated than (1), but may also differ between different pieces of material accord
ing to their position in the configuration and according to the structure of the con
figuration itself. I t seems impossible to imagine, however, th at there is in no sense
a relativistic counterpart to (1) with y = f, and in 3 of this paper it will be found
for a particular type of configuration.
For other materials the dependence of internal energy on pressure and density
may be such th at in the contraction there is a copious release of energy which must
be radiated away, and will therefore lend to a substantial reduction in (as discussed
further in 4). I t may be now imagined th at two static configurations have the same
dependence of p andp on a suitably defined radius r, but while one sphere was formed
by adiabatic contraction, the contraction of the other led to substantial radiation.
Statically the two spheres are indistinguishable by macroscopic means. (Note th a t
the set of spherically symmetric static models is no richer in relativity than in New
tonian theory, being completely specified in either case by p(r) and p(r) linked by a
single equation of hydrostatic support (Bondi 1964&.)) Both spheres have the same
final m, but the adiabatic one will have formed from diffuse m atter with this same m,
the other from m atter with an originally much larger m. Since in the diffuse state
m is proportional to the nucleon number which is supposed to stay constant during
contraction, the adiabatic sphere, though macroscopically indistinguishable from
the other in the static state, will in fact have far fewer nucleons. The nucleon number,
therefore, cannot be deduced from any integral over macroscopic variables in the
static case, such as / , which is the same for both spheres. Nothing can be said about
the nucleon number by macroscopic means, except by following through the con
traction and specifying the dependence of the internal energy of the material on the
state variables. I t should be noted th at it is quite unnecessary to suppose th at the
sphere is homogeneous in the sense th a t the m atter at every depth has the same
thermodynamic functions. Since only p, p, and the energy emitted during the con
traction process enter the relativistic equations, any assumption of homogeneity
is inappropriate.
2. T h e c o n t r a c t io n i n Sen w a r z s c h i l d - t y p e c o -o r d in a t e s
This is an energy equation, showing clearly how the total (fluid and radiation)
pressure does work on a material sphere across its moving boundary. The radiative
loss term e is perhaps a little harder to interpret, but it also seems intelligible when
it is remembered th at e("~A) is the velocity of light in our co-ordinates. In New
tonian theory where mass itself is conserved, Dm/Dr = 0, i.e. the mass inside a
particular shell of particles is constant in time.
I t is interesting to note th at we get a simple connexion (17) with the metric and a
sensible energy equation (19), not by the use of an invariant mass and an in
variant radius, but with our quantities r and m. Note particularly th at when (19)
is written, with e =0, . . .
Am = - 477T2pAr, (2 0 )
3. S l o w c o n t r a c t io n o f a h ig h l y o p a q u e m a t e r ia l
If the material of the sphere is sufficiently opaque, 0, and hence every piece
of m atter contracts adiabatically. As the contraction proceeds, both density and
gravitational force increase, and therefore the pressure gradients, and in fact the
pressure, have to increase. I t is only if this increase in pressure required for equili
brium is just equal to th at due to the adiabatic compression of the material th at such
Downloaded from http://rspa.royalsocietypublishing.org/ on November 27, 2017
The contraction of gravitating spheres 43
slow contraction with e = 0 can take place, and so such contraction depends on the
material having the correct adiabatic pressure-density relation. In Newtonian
gravitational theory, and with homologous contraction, this comes out to be
p~ (21)
In relativity we would expect a relation th at for low pressures and weak gravita
tional field reduces to (21), but we shall not be surprised if the relation for a piece of
material varies according to the position of the piece in the configuration, on the
configuration and, as in Newtonian theory, on its manner of contraction.
The assumed high opacity implies e= 0, while th
means th at w and time derivatives are so small th at their products (as well as second
time derivatives) may be neglected.
Next we notice th a t all time derivatives in (6) are negligible, and therefore equa
tions (4) to (6) are identical with those of a static sphere as previously discussed
(Schwarzschild 1916; Buchdahl 1959; Bondi 19646). At any value of r the con
figuration is therefore a static one. This sequence of static models is linked by
(7) and (14), which determine w, and thus how the material moves. Of course
we knowfrom (19), as well as from Birkhoffs theorem, th at the external mis constant.
Subject to this one restriction we are, however, quite free. In other words, if we
take any continuous one-parameter family of static spheres, all of the same m,
then suitable internal motions of the material as described by (7) and (14) will
gradually deform the models into each other, provided the parameter of the family
is treated as a function of time r only.
Unfortunately only one kind of relativistic sphere is known to be representable
in terms of elementary functions: namely, the Schwarzschild models. Although
their constant density makes these models rather unrealistic, the working through
is yet quite instructive.
Thus we take
{ Pit) < r < a ( < ) } f r p J - m . - constant). (22)
1.0 a(t) < r J
(1 %npr2)%(1 #7Tyoa2)I
(23)
m= rr3p, (24)
e~A= 1 2 m\r= 1 %npr2, (25)
eU = 1(3(1 7 rpa2)i (1 (26)
The limits to the contraction allowed are set by the physical assumptions made.
The central pressure will be finite if and only if ms, but the central energy
invariant (p 3p)will be positive if and only if the more stringent inequalit
a > -58-ms is satisfied.
The motion of the particles may be found from (7) and (14), or alternatively from
(18), which gives
\rp = (p+p)dr/dT. (27)
Downloaded from http://rspa.royalsocietypublishing.org/ on November 27, 2017
44 H. Bondi
Substituting for p and noting th at
r 1 _ i (1 _ *2) (1 _ (1 _ 2m>j*} (
where k , the constant of integration equals the ratio r j a in the initial highly diffuse
state a-
> oo. As is clear from (22) and (24), radii of constant m keep a constant
ratio throughout the contraction, but the material itself moves outward relative
to the spheres of constant m.
However, this relative motion is never very large. The ratio v/tca reaches its
highest value for low k and low a, but even in the limit -> 0, -> t he ratio only
tends to 3/2.
To find the pressure-density relation during the contraction, we substitute
(29) into (23), and change the independent variable from a to p. We obtain
4 V _1
(1 a:2 ) ap% |( 1 + ac2 ) (1 ocpi)i ( 1 ac2 ) (2 3
1
!^ 2) aP^ + ...,
where a is defined in (28). So long as p is very small, as in Newtonian theory.
For larger surface potentials (api = 2ms/a), p varies with a higher power of p. Thus
at m ja = Yg the power has risen to -Xg]- = 1-83 at the centre though only to = 1-54
a t the surface. In the extreme case msla f the power has risen to 4(1 + ac2)/3 ac2,
so th at even at the surface it is f , rising to oo at the centre.
By numerical work the gravitationally neutral variation of p with p could be
ascertained for other models, but it seems likely that, just as in the case here fully
investigated, the power law rises from the Newtonian f in the diffuse state to higher
values for configurations with higher potentials.
Incidentally, during the contraction neither of the invariant integrals
r ai
47 tp( 3p)e?xr2dr,
Jo
47rJ* (p 3 eI(A+lV2dr,
remains constant, although on any basis the particle number stays constant,
showing clearly the incorrectness of associating either invariant integral with the
nucleonic mass.
4. R adiation co-ordinates
When systems are contemplated in which 4= 0, and this radiation plays a
major role, it is advantageous to use radiation co-ordinates as introduced by
Bondi, van der Burg & Metzner (1962). In our spherically symmetrical case they
take the form ^ = e^{V
jr)dw2+ 2d77dr}-r2(dtf2+ sin20 d ^ 2), (31)
Downloaded from http://rspa.royalsocietypublishing.org/ on November 27, 2017
The contraction of gravitating spheres 45
where /? -> Oasr -*oo. Both /?and V are functions of and ( r, 6, $) = (0,1, 2, 3).
The components of the energy tensor in these co-ordinates are distinguished by a
tilde (~) , and differentiation with respect to r is denoted by suffixes 0, 1,
respectively.
- 8 nTm= - V'~ ^- ~(
e*> - F , + 2/?,
*11 (39)
*i *3 = ~V- (40)
Note also from (36) that for the motion of m atter
dr V
(41)
du r \ w'
We first apply these equations to the space outside the matter so th at 0. f
Equations (32) to (40) immediately show that (since /?->0 as r->oo)
VY)
ft = 0, F = r 2m(w), e= r(f_ 2w ) ' (42)
1w F r 2m
e"2^ : (46)
1+w (/o+P) 27jt2 A
A\ , 3A(l-2w1)-m11 (47)
* - - r! - ( 2A . + ^ A ) + 87TV
This is a rather more convenient form of the equations. We see from (45) th a t
't o * (48)
- i; 1+W
while from (11) and (16)
m (49)
f 0 4nr2 ( r ^ +j dr-
I t is easily established that, if the time rate of change of {f+g) is not too large,
both pressure and density are nowhere negative. The radiation must also be carrying
positive energy so th at e > 0 and hence
d(a3f)/du < 0. (56)
Since (54) and (55) are only two conditions on the three functions a(u), f(u),
g{u), even our highly restrictive model admits a variety of solutions. A large range
of phenomena awaits investigation. The following solution is perhaps indicative of
the situation:
Let |7ra2/ = msja = U (surface potential), (57)
and suppose th at
KU
so th at = (/c 2)f7 (by (54)), (58)
R eferen ces
Bondi, H ., van der Burg, M. G. J . & M etzner, A. W. K . 1962 Proc. Roy. A, 269, 21.
Bondi, H . 1964 a N ature Lond. 202, 275*.
Bondi, H . 19646 Proc. Roy. Soc. A (In th e press).
B uchdahl, H . A. 1959 Phys. Rev. 116, 1027.
E ddington, A. S. 1930 The mathematical theory of prelativy, . 121. Cambridge U ni
Press. E d d in g to n s rem arks h ad a g reat influence on other w riters. See for exam ple
Curtis, A. R . 1950 Proc. Roy. Soc. A, 200, 248, where th e identification of T and the
particle density is tak en for granted.
Fowler, W . A. & H oyle, F . 1963 M on. Not. Roy. Astron. Soc. 125, 169.
Iben, I. 1963 Astrophys J . 138, 1090.
Israel, W . 1958 Proc. Roy. Soc. A, 248, 404.
Schwarzschild, K . 1916 S .B . preuss. A kad. W iss. p. 424.
Synge, J . L. 1957 Proc. R . Irish Acad. 59 A, 1.