On stars, their evolution and their stability

S. Chandrasekhar
The University of Chicago, Chicago, Illinois 60637

INTRODUCTION
2. THE ROLE OF RADIATION PRESSURE

A central fact concerning normal stars is the role which

radiation pressure plays as a factor in their hydrostatic
When we think of atoms, we have a clear picture in our equilibrium. Precisely the equation governing the hydro-
minds: a central nucleus and a swarm of electrons sur- static equilibrium of a star is
rounding it. We conceive them as small objects of sizes
dI' GM (r)
measured in angstroms ( —10 cm); and we know that (4)
some hundred different species of them exist. This pic- dI p'

ture is, of course, quantified and made precise in modern where P denotes the total pressure, p the density, and
quantum theory. And the success of the entire theory M(r) is the mass interior to a sphere of radius r Th. ere
may be traced to two basic facts: first, the Bohr radius of are two contributions to the total pressure P: that due to
the ground state of the hydrogen atom, namely, the material and that due to the radiation. On the as-
sumption that the matter is in the state of a perfect gas in

me
-0.5 ~ 10 cm, the classical Maxwellian sense, the material or the gas
4m pressure is given by
where h is Planck's constant, m is the mass of the elec- k
tron and e is its charge, provides a correct measure of
pgas= ~pT
pH
~

atomic dimensions; and second, the reciprocal of where T is the absolute temperature, k is the Boltzmann
Sommerfeld's fine-structure constant, constant, and p is the mean molecular weight (which
under normal stellar conditions is —1.0). The pressure
hc —137, due to radiation is given by
(2)
2me 2
prsd = aT 4
1
3 (6)
gives the maximum positive charge of the central nucleus
where a denotes Stefan's radiation-constant. Consequent-
that mill allow a stable electron-orbit around it. This —
maximum charge for the central nucleus arises from the ly, if radiation contributes a fraction (1 p) to the total
effects of special relativity on the motions of the orbiting pressure, we may write
electrons. I'= —1 —1 4= —
1 k
'We now ask: can we understand the basic facts con- 1 P3 aT PpH pT .
cerning stars as simply as we understand atoms in terms
To bring out explicitly the role of the radiation pressure
of the two combinations of natural constants (1) and (2).
in the equilibrium of a star, we may eliminate the tem-
In this lecture, I shall attempt to show that in a limited
perature, T, from the foregoing equations and express I'
sense we can.
in terms of p and p instead of in terms of p and T. We
The most important fact concerning a star is its mass.
f)nd.
It is measured in units of the mass of the sun, O, which is '
1/3
2X 10 g: stars with masses very much less than, or very k 3 1 —P
much more than, the mass of the sun are relatively infre-
@Ha P
quent. The current theories of stellar structure and stellar
evolution derive their successes largely from the fact that
the following combination of the dimensions of a mass
provides a correct measure of ste11ar masses: —P 1/3
k 3 1
4n=C(p) 4

3/2 pH a p4
hc 1
—
~22 29. 2G (3) (say).
The importance of this ratio, (1 —p), for the theory of
where 6is the constant of gravitation and H is the mass stellar structure was first emphasized by Eddington.
of the hydrogen atom. In the first half of the lecture, I Indeed, he related it, in a famous passage in his book on
shall essentially be concerned with the question: how The Internal Constitution of the Stars, to the "happening
does this come about' ? of the stars" (Eddington, 1926, p. 16). A more rational

Reviews of Modern Physics, Vol. 56, No. 2, Part I, April 1984 COpyRlGHT THE NOBEL FOUNDATlON 1984
138 Chandrasekhar: On stars, their evolution and their stability

version of Eddington's argument which, at the same time, 4

' 1/2 3/2
isolates the combination (3) of the natural constants is the (135) ~ Qc
pM z~3 6 H'
following:
There is a general theorem (Chandrasekhar, 1936) 3/2
which states that the pressure, P„at the centre of a star hc
=0. 1873 Il' (14)
of a mass M in hydrostatic equilibrium in which the den- G
sity, p(r), at a point at a radial distance, r, from the centre
We observe that the inequality (14) has isolated the
does not exceed the mean density, p(r), interior to the
combination (3) of natural constants of the dimensions of
same point r, must satisfy the inequality,
a mass; by inserting its numerical value given in Eq. (3),
— m)' —
2 G( 3~ p M (Pc (—— m)' —
T 6( 3~ p
pc M 7
we obtain the inequality,
4 1/2
(10) c
) 5.480 . (15)
where p denotes the mean density of the star and p, its
density at the centre. The content of the theorem is no This inequality provides an upper limit to (1 —P, ) for a
more than the assertion that the actual pressure at the star of a given mass. Thus,
centre of a star must be intermediate between those at the
centres of the two configurations of uniform density, one 1 —P, &1 —P, , (16)
at a density equal to the mean density of the star, and the where (1 — determined by the mass M of
P, ) is uniquely
other at a density equal to the density p, at the centre (see the star and the mean molecular weight, p, by the quartic
Fig. 1). If the inequality (10) should be violated then equation,
there must, in general, be some regions in which adverse
density gradients must prevail; and this implies instabili-
ty. In other words, we may consider conformity with the
p M=5. 48 (17)
inequality (10) as equivalent to the condition for the stable
existence of stars. In Table I, we list the values of 1 —P, for several values
The right-hand side of the inequality (10) together with of p M. From this table it follows in particular, that for
P given by Eq. (9), yields, for the stable existence of stars, a star of solar mass with a mean molecular weight equal
the condition, to 1, the radiation pressure at the centre cannot exceed
3% of the total pressure.
4 1/3 1/3 What do we conclude from the foregoing calculation?
1 P GM'/', We conclude that to the extent Eq. (17) is at the base of
p4 the equilibrium of actual stars, to that extent the com-
bination of natural constants (3), providing a mass of
or, equivalently, proper magnitude for the measurement of stellar masses,
is at the base of a physical theory of stellar structure.
' 1/2 4 1/2
M&
6 31
7T pM a p4 63~2 3. DG STARS HAVE ENOUGH ENERGY TO COOL' ?

where in the foregoing inequalities, p, is a value of p at The same combination of natural constants (3) emerged
the centre of the star. Now Stefan's constant, a, by virtue soon afterward in a much more fundamental context of
of Planck's law, has the value resolving a paradox Eddington had formulated in the
form of an aphorism: "a star will need energy to cool. "
8~ k
(13) The paradox arose while considering the ultimate fate of a
15h c
gaseous star in the light of the then new knowledge that
Inserting this value a in the equality (12) we obtain white-dwarf stars, such as the companion of Sirius, exist,
which have mean densities in the range 10 — 10 g cm
As Eddington stated

TABLE I. The maximum radiation pressure, (1 —p, ), at t&e

centre of a star of a given mass, M.
Mp /0 Mp /0
0.01 0.56 0.50 15.49
(a) (c) 0.03 1.01 0.60 26.52
0. 10 2. 14 0.70 50.92
FICs. 1. A comparison of an inhomogeneous distribution of 0.20 3.83 0.80 122.5
density in a star (b) with the two homogeneous configurations 0.30 6.12 0.85 224. 4
with the constant density equal to the mean density (a) and 0.40 9.62 0.90 519.6
equal to the density at the centre (c).

Rev. Mod. Phys. , Vol. 56, No. 2, Part l, April 1984

 Chandrasekhar: On stars, their evolution and their stability 139

I do not see how a star which has once got into this
(p) = h3 p
8m
compressed state is ever going to get out of it. . . . It (p &po»
would seem that the star will be in an awkward predica-
ment when its supply of subatomic energy fails (Ed- =o (p &po) . (22)
dington, 1926, p. 172).
The paradox posed by Eddington was reformulated in The value of the threshold momentum, p0, is determined
clearer physical terms by R. H. Fowler (1926). His for- by the normalization condition
mulation was the following:
Pp
The stellar material, in the white-dwarf state, will have n = 0
n(p)dp=
Svr
po, (23)
radiated so much energy that it has less energy than the 3A
same matter in normal atoms expanded at the absolute where n denotes the total number of electrons per unit
zero of temperature. If part of it were removed from the volume.
star and the pressure taken off, what could it do? For the distribution given by (22), the pressure p and
Quantitatively, Fowler s question arises in this way. the kinetic energy Ek;„of the electrons (per unit volume),
An estimate of the electrostatic energy, Ez, per unit are given by
volume of an assembly of atoms, of atomic number Z,
ionized down to bare nuclei, is given by (24)
]0 11
Z 2p4/3 (18)
while the kinetic energy of thermal motions, Ek;„, per unit
volume of free particles in the form of a perfect gas of (25)
density, p, and temperature, T, is given by
where Uz and T~ are the velocity and the kinetic energy of
3 k 1.24' 10' an electron having a momentum p.
(19)
k1Q
2 ~ P
If we set
Now if such matter were released of the pressure to which
v~ =p/m and Tz
—p /2m, (26)
it is subject, it can resume a state of ordinary normal
atoms only if appropriate for non-relativistic mechanics, in Eqs. (24)
and (25), we find
Ekin + Ev ~ (20)
2/3
or, according to Eqs. (18) and (19), only if I'= 877
'm P0
—1
5 —
3 2
n (27)
3 15h 2o m m

p ( 0.94& 10
pZ
and
'
2/3
This inequality will be clearly violated if the density is 87T 5 3 3 ~2 5/3 (28)
sufficiently high. This is the essence of Eddington s para- 10h m 4o
dox as formulated by Fowler. And Fowler resolved this Fowler's resolution of Eddington's paradox consists in
paradox in 1926 in a paper entitled "Dense Matter" one this: at the temperatures and densities that may be ex-
of the great landmark papers in the realm of stellar struc- pected to prevail in the interiors of the white-dwarf stars,
ture: in it the notions of Fermi statistics and of electron
the electrons will be highly degenerate and Ek;„must be
degeneracy are introduced for the first time. evaluated in accordance with Eq. (28) and not in accor-
dance with Eq. (19); and Eq. (28) gives,
1.39&&10' (p/p)
Ek;„—— i (29)
4. FOWLER'S RESOLUTION OF EDDINGTON'S Comparing now the two estimates (18) and (29), we see
PARADOX; THE DEGENERACY OF THE
that, for matter of the density occurring in the white
ELECTRONS IN WHITE-DWARF STARS
dwarfs, namely p-10 gem,
the total kinetic energy is
about two to four times the negative potential-energy; and
Eddington's paradox does not arise. Fowler concluded his
In a completely degenerate electron gas all the available
paper with the following highly perceptive statement:
parts of the phase space, with momenta less than a certain
—
"threshold" value p0 the Fermi threshold are occupied — The black-dwarf material is best likened to a single
gigantic molecule in its lowest quantum state. Qn the
consistently with the Pauli exclusion-principle, i.e., with Fermi-Dirac statistics, its high density can be achieved in
two electrons per "cell" of volume h of the six- one and only one way, in virtue of a correspondingly
dimensional phase space. Therefore, if n(p)dp denotes great energy content. But this energy can no more be ex-
the number of electrons, per unit volume, between p and pended in radiation than the energy of a normal atom or
p+ dp, then the assumption of complete degeneracy is molecule. The only difference between black-dwarf
equivalent to the assertion, matter and a normal molecule is that the molecule can

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, April 1984

140 Chandrasekhar: On stars, their evolution and their stability

exist in a free state while the black-dwarf matter can p

V
only so exist under very high external pressure. m(1+p /m c )in
and (34)
5. THE THEORY OF THE WHITE-DWARF STARS; T~=mc [(1+p /m c )' —1],
THE LIMITING MASS
in place of the non-relativistic relations (26). We find that
The internal energy (=3P/2) of a degenerate electron the resulting equation of state can be expressed, parame-
gas that is associated with a pressure P is zero-point ener- trically, in the form
gy; and the essential content of Fowler's paper is that this
zero-point energy is so great that we may expect a star to P=Af(x) and p=Bx (35)
eventually settle down to a state in which all of its energy where
is of this kind. Fowler's argument can be more explicitly
c p, H
formulated
193 la).
in the following manner (Chandrasekhar, ~m4c&
3h3
B= 8am 3h3 (36)

According to the expression for the pressure given by

Eq. (27), we have the relation,

+=X,p 5/3 where K, =

1 —
3

where p, is the mean molecular weight per electron. An

2/3

equilibrium configuration in which the pressure, I', and

—,
m (p, H)
(30)
and

f (x) =x (x
And similarly
—Ag (x),
Ek;„—
+ 1)'~ (2x —3) + 3 sinh 'x . (37)

(38)
where
the density, p, are related in the manner,
g(x) =8x [(x +1)' —1]—f (x) . (39)
P = Kp'+'/" (31)
According to Eqs. (35) and (36), the pressure approxi-
is an Emden polytrope of index n. The degenerate config- mates the relation (30) for low enough electron concentra-
urations built on the equation of state (30) are therefore tions (x ~&1); but for increasing electron concentrations
polytropes of index 3/2; and the theory of polytropes im-
mediately provides the relation,
»
(x 1), the pressure tends to (Chandrasekhar, 1931b)
1/3
Z, =0.4242(Gm'"R) (32) P ——
1 3 acn4/3 (40)
8
or, numerically, for Ki given by Eq. (30), L

logio(R/Ro) = 3~
logio(M/C)) ——logiope 1 397 This limiting form of relation can be obtained very simply
by setting v~ =c in Eq. (24); then
(33)
For a mass equal to the solar mass and p, =2, the relation P= 8mc 0 2mc 4
(33) predicts R = l. 26 X 10 R ~ and a mean density of o
pdp=
3
poi (41)
3h 3h
7.0X 10 g cm . These values are precisely of the order of
the radii and mean densities encountered in white-dwarf and the elimination of po with the aid of Eq. (23) directly
stars. Moreover, according to Eqs. (32) and (33), the ra- leads to Eq. (40).
dius of the white-dwarf configuration is inversely propor- While the modification of the equation of state re-
tional to the cube-root of the mass. On this account, fin- quired by the special theory of relativity appears harmless
ite equilibrium configurations are predicted for all enough, it has, as we shall presently show, a dramatic ef-
masses. And it came to be accepted that the white dwarfs fect on the predicted mass-radius relation for degenerate
represent the last stages in the evolution of all stars. configurations.
But it soon became clear that the foregoing simple The relation between P and p corresponding to the lim-
theory based on Fowler's premises required modifications. iting form (41) is
For, the electrons at their threshold energies, at the cen-
tres of the degenerate stars, begin to have velocities com- 1/3
parable to that of light as the mass increases. Thus, al- P=K2p 4/3 where E2 =1
———3 (42)
ready for a degenerate star of solar mass (with p, =2) the 8 m (p, FI) i
central density (which is about six times the mean density)
is 4. 19)&10 gcm; and this density corresponds to a In this limit, the configuration is an Emden polytrope of
threshold momentum po —1.29mc and a velocity which is index 3. And it is well known that when the polytropic
0.63c. Consequently, the equation of state must be modi- index is 3, the mass of the resulting equilibrium configu-
fied to take into account the effects of special ration is uniquely determined by the constant of propor-
relativity. And this is easily done by inserting in Eqs. (24) tionality, K2, in the pressure-density relation. We have
and (25) the relations, accordingly,

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, AprII 1984

 Chandrasekhar: On stars, their evolution and their stability

3/2
K2 5.5—
M)jmjt 4' m. G
(2.018)
5.0-

=0. 197 "' 3/2

= 5.76@~ 0 (43) 4.5-
6 (p, M)
4.0—
[In Eq. (43), 2.018 is a numerical constant derived from
the explicit solution of the Lane-Emden equation for
3.5
n= 3.]
It is clear from general considerations (Chandrasekhar, 3.0-
1931c) that the exact mass rad-ius relation for the degen
crate configurations must prouide an upper limit to the 2.5—
mass of such configurations giuen by Eq. (43); and further,
that the mean density of the configuration must tend to in
ginity, while the radius tends to zero, and M~M&;;, .
1.5—
These conditions, straightforward as they are, can be es-
tablished directly by considering the equilibrium of con-
1.0—
figurations built on the exact equation of state given by
Eqs. (35)— (37). It is found that the equation governing 0.5—
the equilibrium of such configurations can be reduced to
the form (Chandrasekhar, 1934b, 1935) 0 ——
0 0.1 0.2 0.6
I
0.7
A
0.3 0.4 0.5 0.8 O.g 1 0
Al/AX)-
3/2
1 d 2dp p 1
'9 (44) FIG. 2. The full-line curve represents the exact (mass-radius)-
g d'rl d Y/ ye relation [1& is defined in Eq. (46) and M3 denotes the limiting
mass]. This curve tends asymptotically to the ———curve ap-
where
propriate to the low-mass degenerate configurations, approxi-
2 2
3'o=xo+& (45) mated by polytropes of index 3/2. The regions of the configura-
~
tions which may be considered as relativistic [p & (X~ /Ki) ] are
and mcxc denotes the threshold momentum of the elec- shown shaded. [From Chandrasekhar (1935).]
trons at the centre of the configuration and il measures
the radial distance in the unit
'
1/2 that required to be resolved was how to relate its existence
2A 1 to the evolution of stars from their gaseous state. If a star
=~iso (46)
mG &so has a mass less than M~;;„ the assumption that it will
eventually evolve towards the completely degenerate state
(say). appears reasonable. But what if its mass is greater than
By integrating Eq. (44), with suitable boundary condi- M~;;, '7 Clues as to what might ensue were sought in
tions and for various initially prescribed values of yc, we terms of the equations and inequalities of Secs. 2 and 3
can derive the exact mass-radius relation, as well as the (Chandrasekhar, 1932, 1934a).
other equilibrium properties, of the degenerate configura- The first question that had to be resolved concerns the
tions. The principal results of such calculations are illus- circumstances under which a star, initially gaseous, will
trated in Figs. 2 and 3.
The important conclusions which follow from the fore-
going considerations are: first, there is an upper limit,
Mi;m;„ to the mass of stars which can become degenerate
configurations, as the last stage in their evolution; and
second, that stars with M &M~;;, must have end states
which cannot be predicted from the considerations we
have presented so far. And finally, we observe that the
combination of the natural constant (3) now emerges in
the fundamental context of M~;;, given by Eq. (43): its
significance for the theory of stellar structure and stellar 1.0
evolution can no longer be doubted.

6.UNDER WHAT CONDITIONS CAN NORMAL STARS

DEVELOP DEGENERATE CORES?
FIG. 3. The full-line curve represents the exact {mass-density)-
relation for the highly collapsed configurations. This curve
Once the upper limit to the mass of completely degen- tends asymptotically to the dotted curve as M~0. [From
erate configurations had been established, the question Chandrasekhar (1935).]

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, April 1984

 142 Chandrasekhar: On stars, their evolution and their stability

develop degenerate cores. From the physical side, the the pressure p, given by the classical perfect-gas equation
question, when departures from the perfect-gas equation of state will be greater than that given by the equation if
of state (5) will set in and the effects of electron degenera- degeneracy were to prevail, not only for the prescribed p
cy will be manifested, can be readily answered. and T, but for all p and T having the same P, .
Suppose, for example, that we continually and steadily Inserting for a its value given in Eq. (13), we find that
increase the density, at constant temperature, of an assem- the inequality (49) reduces to
bly of free electrons and atomic nuclei, in a highly ionized
state and initially in the form of a perfect gas governed by
the equation of state (5). At first the electron pressure
960 1 Pe—
) 1 (50)

will increase linearly with p,' but soon departures will set
or, equivalently
in and eventually the density will increase in accordance
with the equation of state that describes the fully degen- 1 —P, ) 0.0921=1 —P„ (51)
erate electron-gas (see Fig. 4). The remarkable fact is that
(say). (See Fig. 5.)
this limiting form of the equation of state is independent
For our present purposes, the principal content of the
of temperature.
inequality (51) is the criterion that for a star to develop
However, to examine the circumstances when, during
degeneracy, it is necessary that the radiation pressure be
the course of evolution, a star will develop degenerate
cores, it is more convenient to express the electron pres-
less than 9. 2'of (p, +p„d). This last inference is so cen-
tral to all current schemes of stellar evolution that the
sure (as given by the classical perfect-gas equation of
directness and the simplicity of the early arguments are
state) in terms of p and p, defined in the manner [cf. Eq.
worth repeating.
(7)),
The two principal elements of the early arguments were
p, =
k Pe
pT= 1 — —
1
aT 4 (47)
these: first, that radiation pressure becomes increasingly
dominant as the mass of the star increases; and second,
p, H P, 3
that the degeneracy of electrons is possible only so long as
where p, now denotes the electron pressure. Then, analo- the radiation pressure is not a significant fraction of the
gous to Eq. (9), we can write total pressure —
indeed, as we have seen, it must not
4
31 —P, I /3
4~,
(48)
exceed 9.2%%uo of (p, +p„d). The second of these elements
in the arguments is a direct and an elementary conse-
p~H 0 Pe
quence of the physics of degeneracy; but the first requires
Comparing this with Eq. (42), we conclude that if some arnplification.
I' k 31—P, 1 3 hc
That radiation pressure must play an increasingly dom-
inant role as the mass of the star increases is one of the
&E
p, ,H a P, 8 H)4/3 earliest results in the study of stellar structure that was es-

(49)
1 027

027

1 025

1 025

10" =
0.5
1-P
0.06
0.05
0.01
1 021 I I ~ t I IIII I I I I IIIII I I I 1%IIII s s yssill s

10 '10 10 10
021 iii»l &»&i&l
p (gem, ~)
i
1 &

10 10 10' 10' 10'

p(gem ~) FIG. 5. Illustrating the onset of degeneracy for increasing den-
sity at constant p. Notice that there are no intersections for
FIG. 4. Illustrating how by increasing the density at constant 1— p~0. 09212. In the figure, 1 —p is converted into the mass
temperature degeneracy always sets in. of a star built on the standard model.

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, April 1984

 Chandrasekhar: Qn stars, their evolution and their stability

tablished by Eddington. A quantitative expression for or, equivalently

this fact is given by Eddington's standard model which 3/2
lay at the base of his early studies summarized in his The
Internal Constitution of the Stars.
3
16.
hc
G (p, H )'=1.
74p, 0. (55)

On the standard model, the fraction P ( = gas

We conclude that there can be no surprises in the evolu-
pressure/total pressure) is a constant through a star. On tion of stars of mass less than 0.43$ (if p, =2). The end
this assumption, the star is a polytrope of index 3 as is ap-
stage in the evolution of such stars can only be that of the
parent from Eq. (9); and, in consequence, we have the re- white dwarfs. [Parenthetically, we may note here that the
lation [cf. Eq. (43)]
inequality (55) implies that the so-called "mini" black-
3/2
holes of mass —10' g cannot naturally be formed in the
M=4 (2.018) (52) present astronomical universe. ]
mG

where C(P) is defined in Eq. (9). Equation (52) provides

a quartic equation for P analogous to Eq. (17) for P, .
Equation (52} for P = P„gives
' 3/2 7. SOME BRIEF REMARKS ON RECENT PROGRESS
THE EVOLUTION OF MASSIVE STARS
M =0. 197P„
1
= 6. 65p -20=% (53)
IN
AND THE ONSET OF GRAVITATIONAL COLLAPSE
G (pH)
(say). On the standard model, then, stars with masses It became clear, already from the early considerations,
exceeding 9R will have radiation pressures which exceed that the inability of the massive stars to become white
9.2% of the total pressure. Consequently stars with dwarfs must result in the development of much more ex-
M &K cannot, at any stage during the course of their treme conditions in their interiors and, eventually, in the
evolution, develop degeneracy in their interiors. There- onset of gravitational collapse attended by the super-nova
fore, for such stars an eventual white-dwarf state is not phenomenon. But the precise manner in which all this
possible unless they are able to eject a substantial fraction will happen has been difficult to ascertain in spite of great
of their mass. effort by several competent groups of investigators. The
The standard model is, of course, only a model. facts which must be taken into account appear to be the
Nevertheless, except under special circumstances, briefly following. '
noted below, experience has confirmed the essential quali- In the first instance, the density and the temperature
tative correctness of the conclusions drawn from the stan- will steadily increase without the inhibiting effect of de-
dard model, namely that the evolution of stars of masses generacy since for the massive stars considered
exceeding 7 — 80 must proceed along lines very different 1— P, & 1 — P„. On this account, "nuclear ignition" of
from those of less massive stars. These conclusions, carbon, say, will take place which will be attended by the
which were arrived at some fifty years ago, appeared then emission of neutrinos. This emission of neutrinos will ef-
so convincing that assertions such as these were made fect a cooling and a lowering of (1 — P, ); but it will still
with confidence: be in excess of 1 — P . The important point here is that
Given an enclosure containing electrons and atomic the emission of neutrinos acts selectively in the central re-
nuclei (total charge zero) what happens if we go on gions and is the cause of the lowering of (1 — P, ) in these
compressing the material indefinitely? (Chandrasekhar, regions. The density and the temperature will continue to
1932) increase till the next ignition of neon takes place followed
The life history of a star of small mass must be essential- by further emission of neutrinos and a further lowering of
ly different from the life history of a star of large mass. (1 — P, ). This succession of nuclear ignitions and lower-
For a star of small mass the natural white-dwarf stage is ing of (1 — P, ) will continue till 1 —P, & 1 — P„and a rela-
an initial step towards complete extinction. A star of tivistically degenerate core with a mass approximately
large mass cannot pass into the white-dwarf stage and that of the limiting mass (=1.4$ for p, =2) forms at the
one is left speculating on other possibilities. (Chan- centre. By this stage, or soon afterwards, instability of
drasekhar, 1934b) some sort is expected to set in (see following section, 8)
And these statements have retained their validity. followed by gravitational collapse and the phenomenon of
While the evolution of the massive stars was thos left the super-nova (of type II). In some instances, what was
uncertain, there was no such uncertainty regarding the fi- originally the highly relativistic degenerate core of ap-
nal states of stars of sufficiently low mass (Chan- proximately 1.4O, will be left behind as a neutron star.
drasekhar, 1934a). The reason is that by virtue, again, of That this happens sometimes is confirmed by the fact
the inequality (10), the maximum central pressure attain-
able in a star must be less than that provided by the de-
generate equation of state, so long as
1/3 ~I am grateful to Professor D. Arnett for guiding me through
'M 2'3 1 3 hc
—, G(-,4 ~) $ 3
~re, = (54) the recent literature and giving me advice in the writing of this
8 m (pH}~ section.

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, April 1984

144 Chandrasekhar: On stars, their evolution and their stability

that in those cases for which reliable estimates of the that, again, the stability depends on an average value of
masses of pulsars exist, they are consistently close to 1.40. I &, but contrary to the Newtonian result, the stability
However in other instances —
perhaps, in the majority of now depends on the radius of the star as well. Thus, one
the instances
"
—
what is left behind, after all "the dust has finds that no matter how high I may be, instability will
&

settled, will have masses in excess of that allowed for set in provided the radius is less than a certain deter-
stable neutron stars; and in these instances black holes minate multiple of the Sehmarzschild radius,
will form.
In the case of less massive stars (M-6 — 8$) the degen-
erate cores, which are initially formed, are not highly
Rs —2GM/c (58)
relativistic. But the mass of the core increases with the
further burning of the nuclear fuel at the interface of the Thus, if for the sake of simplicity, we assume that I is a &

core and the mantle; and when the core reaches the limit- constant through the star and equal to 5/3, then the star
ing mass, an explosion occurs following instability; and it will become dynamically unstable for radial perturbations,
is believed that this is the cause underlying super-nova if R & 2.4Rs. And further, if I ~~ oo, instability will set
&

phenomenon of type I. in for all R &(9/8)Rs. The radius (9/8)Rs defines, in

From the foregoing brief description of what may hap- fact, the minimum radius which any gravitating mass, in
pen during the late stages in the evolution of massive hydrostatic equilibrium, can have in the framework of gen
stars, it is clear that the problems one encounters are of eral relatiuity. This important result is implicit in a fun-
exceptional complexity, in which a great variety of physi- damental paper by Karl Schwarzschild published in 1916.
cal factors compete. This is clearly not the occasion for [Schwarzschild actually proved that for a star in which
me to enter into a detailed discussion of these various the energy density is a uniform R & (9/8) Rs.]
questions. In one sense, the most important consequence of this
instability of relativistic origin is that if I (again as- ~

sumed to be a constant for the sake of simplicity) differs

8. INSTABILITIES OF RELATIYISTIC ORIGIN:
from and is greater than 4/3 only by a small positive con-
(1}THE VIBRATIONAL INSTABILITY
stant, then the instability will set in for a radius R which
OF SPHERICAL STARS
is a large multiple of Rs, and, therefore, under cir-
cumstances when the effects of general relativity, on the
I now turn to the consideration of certain types of stel-
structure of the equilibrium configuration itself, are hard-
lar instabilities which are derived from the effects of gen-
ly relevant. Indeed, it follows (Chandrasekhar, 1965)
eral relativity and which have no counterparts in the
from the equations governing radial oscillations of a star,
Newtonian framework. It will appear that these new
in a first post-Newtonian approximation to the general
types of instabilities of relativistic origin may have essen-
theory of relativity, that instability for radial perturba-
tial roles to play in discussions pertaining to gravitational
tions will set in for all
collapse and the late stages in the evolution of massive
stars.
We shall consider first the stability of spherical stars E 2GM
(59)
for purely radial perturbations. The criterion for such r, —4/3
stability follows directly from the linearized equations
governing the spherically symmetric radial oscillations of
stars. In the framework of the Newtonian theory of grav- where K is a constant which
depends on the entire march
itation, the stability for radial perturbations depends only of density and pressure in the equilibrium configuration
on an average value of the adiabatic exponent, I &, which in the Newtonian framework. Thus, for a polytrope of
is the ratio of the fractional Lagrangian changes in the index n, the value of the constant is given by
pressure and in the density experienced by a fluid element
following the motion; thus, 5 —n 2(11 n) .
1—
&~ d8 z
(56) 18 (n+1)g, ~8', ~3 "o dg
And the Newtonian criterion for stability is (60}
I &
——
I M
I &(r)P(r)dM(r) —:
f
M
P(r)dM(r) & —,

(57)
where 0 is the Lane-Emden function in its standard nor-
malization (8=1 at /=0), g is the dimensionless radial
coordinate, g~ defines the boundary of the polytrope
If I ~ &4/3, dynamical instability of a global character (where 8=0) and 8~ is the derivative of 8 at g~.
will ensue with an e-folding time measured by the time
taken by a sound wave to travel from the centre to the
surface.
When one examines the same problem in the frame-
work of the general theory of relativity, one finds (Chan-
drasekhar, 1964a; see also Chandrasekhar, 1964b, 1964c) It is for this reason that we describe the instability as global.

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, April 1984

 Chandrasekhar: On stars, their evolution and their stability 145

TABLE II. Values of the constant K in the inequality (59) for

various polytropic indices, n.

0 0.452 381 3.25 1.285 03

1.0 0.565 382 3.5 1.499 53
1.5 0.645 063 4.0 2.253 38
2.0 0.751 296 4.5 4.530 3
2.5 0.900 302 4.9 22.906
3.0 1.12447 4.95 45.94

I
In Table II, we list the values of X for different po- lPIQ (pl4
Ip
lytropic indices. It should be particularly noted that K
increases without limit for n~5 and the configuration
pc {kg m'j
becomes increasingly centrally condensed. Thus, already
for n=4. 95 (for which polytropic index p, =8.09&& 10 p), FIG. 6. The variation of the period of radial oscillation along
IC-46. In other words, for the highly centrally con- the completely degenerate configurations. Notice that the
densed massive stars (for which I may differ from 4/3 period tends to infinity for a mass close to the limiting mass.
&
There is consequently a minimum period of oscillation along
by as little as 0.01), the instability of relativistic origin these configurations; and the minimum period is approximately
will set in, already, when its radius falls below 5X 10 Rs. two seconds. [From Skilling (1968), p. 59.]
Clearly this relativistic instability must be considered in
the contexts of these problems.
A further application of the result described in the
have resulted from super-nova explosions have masses
preceding paragraph is to degenerate configurations near
the limiting mass (Chandrasekhar and Tooper, 1964). close to 1.4Q!)
Since the electrons in these highly relativistic configura- Finally, we may note that the radial instability of rela-
tions have velocities close to the velocity of light, the ef- tivistic origin is the underlying cause for the existence of a
fective value of I will be very close to 4/3 and the post- maximum mass for stability: it is a direct consequence of
&

Newtonian relativistic instability will set in for a mass the equations governing hydrostatic equilibrium in gen-
slightly less than that of the limiting mass. On account
eral relativity. [For a complete investigation on the
of the instability for radial oscillations setting in for a periods of radial oscillation of neutron stars for various
mass less than M~;;„ the period of oscillation, along the admissible equations of state, see Detweiler and Lindblom
sequence of the degenerate configurations, must have a (1983).]
minimum. This minimum can be estimated to be about
two seconds (see Fig. 6). Since pulsars, when they were 9. INSTABILITIES QF RELATIVISTIC ORIGIN:
discovered, were known to have periods much less than (2) THE SECULAR INSTABILITY OF ROTATING STARS
this minimum value, the possibility of their being degen- DERIVED FROM THE EMISSION OF GRAVITATIONAL
erate configurations near the limiting mass was ruled out; RADIATION BY NON-AXISYMMETRIC MODES
and this was one of the deciding factors in favour of the OF OSCILLATION
pulsars being neutron stars. (But by a strange irony, for
reasons we have briefly explained in Sec. 7, pulsars which I now turn to a different type of instability which the
general theory of relativity predicts for rotating configu-
rations. This new type of instability (Chandrasekhar,
1970a; see also Chandrasekhar, 1970b, 1970c) has its ori-
Since this was written, it has been possible to show (Chan- gin in the fact that the general theory of relativity builds
drasekhar and Lebovitz, 1984} that for n~5, the asymptotic into rotating masses a dissipative mechanism derived
behavior of EC is given by from the possibility of the emission of gravitational radia-
tion by nonaxisymmetric modes of oscillation. It appears
E —+2. 3056/(5 —n); that this instability limits the periods of rotation of pul-
and, further, that along the polytropic sequence, the criterion sars. But first, I shall explain the nature and the origin of
for instability (59) can be expressed alternatively in the form
this type of instability.
It is well known that a possible sequence of equilibrium
1/3 figures of rotating homogeneous masses is the Maclaurin
R (0.2264 26M 1
(p, /p&10 . sequence of oblate spheroids [for an account of these
.P
c' I ) —4/3 )
matters pertaining to the classical ellipsoids see Chan-
drasekhar (1968)]. When one examines the second har-
48y reason of the dominance of the radiation pressure in these monic oscillations of the Maclaurin spheroid, in a frame
massive stars and of P being very close to zero. of reference rotating with its angular velocity, one finds

Rev. Mod. Phys. , Vol. 56, No. 2, Part I, April 1984

146 Chandrasekhar: Gn stars, their evolution and their stability

that for two of these modes, whose dependence on the az- tral at e=0.813, unstable beyond this point with an e-
imuthal angle is given by e '+, the characteristic frequen- folding time which depends inversely on the magnitude of
cies of oscillation, o., depend on the eccentricity e in the the kinematic viscosity and which further decreases
manner illustrated in Fig. 7. It will be observed that one monotonically to zero at the point, e=0.953 where the
of these modes becomes neutral (i.e., o =0) when e =0.813 dynamical instability sets in.
and that the two modes coalesce when e=0.953 and be- Since the emission of gravitational radiation dissipates
come complex conjugates of one another beyond this both energy and angular momentum, it does not induce
point. Accordingly, the Maclaurin spheriod becomes instability in the Jacobi mode; instead it induces instabili-
dynamically unstable at the latter point (first isolated by ty in the alternative mode at the same eccentricity. In the
Riemann). On the other hand, the origin of the neutral first instance this may appear surprising; but the situation
mode at e=0.813 is that at this point a new equilibrium we encounter here clarifies some important issues.
sequence of triaxial ellipsoids —
the ellipsoids of Jacobi- If instead of analyzing the normal modes in the rotat-
bifurcate. On this latter account, Lord Kelvin conjec- ing frame, we had analyzed them in the inertial frame, we
tured in 1883 that should have found that the mode which becomes unstable
if there be any viscosity, however slight. . . the equi- by radiation reaction at e=0.813, is in fact neutral at this
librium beyond e=0.81 cannot be secularly stable. point. And the neutrality of this mode in the inertial
Kelvin's reasoning was this: viscosity dissipates energy frame corresponds to the fact that the neutral deforma-
tion at this point is associated with the bifurcation (at this
but not angular momentum. And since for equal angular
momenta, the Jacobi ellipsoid has a lower energy content point) of a new triaxial sequence — the sequence of the
than the Maclaurin spheroid, one may expect that the ac- Dedekind ellipsoids. These Dedekind ellipsoids, while
tion of viscosity will be to dissipate the excess energy of they are congruent to the Jacobi ellipsoids, they differ
the Maclaurin spheroid and transform it into the Jacobi from them in that they are at rest in the inertial frame
ellipsoid with the lower energy. A detailed calculation and owe their triaxial figures to internal vortical motions.
(Chandrasekhar, 1968) of the effect of viscous dissipation An important conclusion that would appear to follow
on the two modes of oscillation, illustrated in Fig. 7, does from these facts is that in the framework of general rela-
confirm Lord Kelvin's conjecture. It is found that tivity we can expect secular instability, derived from
viscous dissipation makes the mode, which becomes neu- radiation-reaction to arise from a Dedekind mode of de-
formation (which is quasi-stationary in the inertial frame)
rather than the Jacobi mode (which is quasi-stationary in
1.4 the rotating frame).
A further fact concerning the secular instability in-
duced by radiation-reaction, discovered subsequently by
1.2 Friedman [(1978); see also Friedman and Schutz (1977)]
and by Comins (1979a, 1979b), is that the modes belonging
1.0 to higher values of m (=3,4, . . . ) become unstable at
smaller eccentricities though the e-folding times for the
instability become rapidly longer. Nevertheless it appears
0.8 from some preliminary calculations of Friedman (1983)
that it is the secular instability derived from modes be-
0.6 longing to m=3 (or 4) that limit the periods of rotation
of the pulsars.
It is clear from the foregoing discussions that the two
0.4
types of instabilities of relativistic origin we have con-
sidered are destined to play significant roles in the con-
0.2 texts we have considered.

10. THE MATHEMATICAL THEORY OF BLACK HOLES

0.2 04 0.6 0.8 1.0
e So far, I have considered only the restrictions on the
FIG. 7. The characteristic frequencies [in the unit (m.Gp)'~~] of last stages of stellar evolution that follow from the ex-
the two even modes of second-harmonic oscillation of the istence of an upper limit to the mass of completely degen-
Maclaurin spheroid. The Jacobi sequence bifurcates from the
erate configurations and from the instabilities of relativis-
Maclaurin sequence by the mode that is neutral (o. =0) at
e=0.813; and the Dedekind sequence bifurcates by the alterna- tic origin. From these and related considerations, the
tive mode at D. At 02 (e=0.9529) the Maclaurin spheroid be- conclusion is inescapable that black holes will form as one
comes dynamically unstable. The real and the imaginary parts of the natural end products of stellar evolution of massive
of the frequency, beyond 02, are shown by the full line and the stars; and further that they must exist in large numbers in
dashed curves, respectively. Viscous dissipation induces insta- the present astronomical universe. In this last section I
bility in the branch of the Jacobi mode; and radiation-reaction want to consider very briefly what the general theory of
induces instability in the branch DO2 of the Dedekind mode. relativity has to say about them. But first, I must define

Rev. Mod. Phys. , Vol. 56, No. 2, Part l, April 1984