ON THE ORIGIN OF PLANETARY NEBULAE

G. O. ABELL* AND P. GOLDREICH

Department of Astronomy
University1 of California, Los Angeles

Received March 5, 1966

There has been much interest of late in the evolution of planetary

nebulae; recent papers on the subject include those of O'Dell ( 1962,
1963 ), Harmon and Seaton ( 1964 ), and Abell ( 1966 ). The detailed
tracks of evolution are not yet clear, but the evidence strongly
suggests that the central stars evolve directly into white dwarfs.
There has been less quantitative discussion of the origin of
planetary nebulae. Shklovskiï (1956) suggested that planetary
nebulae are formed from the outer layers of some ( but not all ) red
giants, and he questioned whether the Τ Tauri stars, or such combi-
nation variables (or "symbiotic" stars) as Ζ Andromedae or V
Sagittae might not be intermediate stages in the planetary nebulae
formation. His arguments that red giants are the immediate pre-
cursors of planetary nebulae are based largely on following the
evolution of a typical nebular shell backward in time and recog-
nizing that at an early epoch its physical structure must resemble
the very tenuous envelope of a greatly distended red giant. More
recently, Deeming ( 1965 ) has similarly tried to trace the evolution
of a planetary nebula backward, but he does not seem to specifically
conclude that the nebulae are actually ejected from red giants.
(Moreover, Deeming's model for the early appearance of a plane-
tary nebula cannot be correct, for the forbidden radiation does not
disappear at a critical gas density as he has assumed. ) A possible
relation between planetary nebulae and red giants, and in particular
the "symbiotic" stars, has also been hinted at by Menzel (1946),
Aller (1954), and Merrill (1958), and possibly others.
In our opinion there are some rather compelling arguments that
most, or more probably all, population II red giants ultimately be-
come planetary nebulae and further that the nebulae are formed
directly from giant stars. To our knowledge, all of these arguments

#
At the Institut für Astrophysik, Max-Planck-Institut für Physik und Astro-
physik, Munich, Germany, 1965-66.

232

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 PLANETARY NEBULAE 233

have not been stated in the literature or at least have not been given
in one place. In any case, it is our experience that many investi-
gators working in the field are not familiar with the arguments, and
it is the purpose of this communication to summarize them.
a) Most, if not all, planetary nebulae represent stages of evolu-
tion of stars of about 1.2 9¾ after they have left the main sequence.
The galactic distribution of planetary nebulae matches that of old
stars (population II) such as those found in the nucleus, disk, and
corona of the Galaxy and not that of the young stars found in or
near the spiral arms (Minkowski and Abell 1963). It cannot be
ruled out that some nebulae evolve from population I stars, but at
least the vast majority must originate from older stars whose masses
must be near 1.2 9}?©. This point has been well discussed by O'Dell
(1963).
b) Most, if not all, stars of 1.2 ^c©, tchich are now evolving or
have recently evolved off the main sequence, must ultimately pass
through the planetary nebula stage. To verify this point, we need
only compare the current rate at which stars in the requisite mass
range are now leaving the main sequence to the rate of formation of
planetary nebulae.
The present rate at which stars are leaving the main sequence is
easily found from data given by Schmidt ( 1963 ), with main se-
quence lifetimes taken from Limber ( 1960 ). According to Schmidt,
the present rate of formation of stars of mass near 1.3 W© is about
10~10 stars per magnitude interval per year in a column of one
square parsec cross-section perpendicular to the galactic plane at
the position of the sun. If stellar formation and evolution in the
Galaxy has reached a steady state, this should also be the rate at
which stars are leaving the main sequence. From Schmidt's data,
we find that stars in the mass range 1.2 to 1.4 W© (comprising a
range of about 0.7 magnitude ) should be leaving the main sequence
at a rate of something under 1 χ 10-10 stars parsec-2 yr-1. On the
other hand, the rate of star formation was probably higher when
stars of this mass were first formed (about 5 X 109 years ago). As
an extreme alternative, suppose that all stars were formed at once,
and that just now those of mass 1.2 W© (potential parents of plane-
tary nebulae) are leaving the main sequence. Then the rate of
evolution from the main sequence should be dN/dt = (£(Mr)dMr/

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 234 ABELL AND GOLDREICH

di, where φ (My) is the present luminosity function. We estimate

φ (My) and dMy/di from Schmidt's Table 1, and find άΝ/dtœ
2 χ 10-10 stars parsec-2 yr-1. Thus the present rate of supply of
stars that can become planetary nebulae ( in our part of the Galaxy )
probably lies in the range 1 X 10~10 and 2 χ 10-10 stars parsec-2 yi^1.
Adopting 75 ϊ)ίΘ per square parsec as the total surface density of
the Galaxy at the suns position (Schmidt 1963) and 1.1 X 1011 ϊ)!©
for the mass of the Galaxy ( Allen 1963 ), we find that in the entire
Galaxy the rate of such stars leaving the main sequence should be
less than 0.3 stars per year. Even if we assume that the formation
and evolution of all stars in a steady state in all parts of the
Galaxy — an assumption which must certainly greatly exaggerate
the rate of production of red giants — we find from Schmidt's data
that the total rate at which stars of all mass greater than 1.2 9Wg
leave the main sequence in the entire Galaxy is not over about 0.7
stars per year.
The total number of planetary nebulae in the Galaxy is more
difficult to determine. O'Dell (1963) estimates that there are
4
5 X 10 nebulae with radii R ^ 0.7 parsec. Shklovskiï ( 1956 )
gives a similar estimate. Typical nebulae expand about their parent
stars at velocities of about 30 km/sec (Wilson 1950). The lifetime
of a nebula (to reach a radius of 0.7 parsec) is therefore typically
about 2 X 104 years. If O'Dell's estimate is correct, planetary
nebulae are now forming in the Galaxy at the rate of 2 or 3 per
year. Since nebulae cannot form faster than the stars from which
they are ejected become available, there are very likely fewer
nebulae than 5 X 104. There can hardly be less than about 5 Χ 103,
however, since nearly a thousand planetaries are now known. Thus
it seems inescapable that most or all stars of mass near 1.2 or
slightly greater — those stars that are now leaving or have recently
left the main sequence throughout the disk population of the
Galaxy — will become planetary nebulae.
c) The observed typical expansion velocity cannot be a signifi-
cant underestimate due to observational selection. If some plane-
tary nebula shells were ejected very rapidly, so that their expansion
speeds were in excess of 100 km/sec, they would have short life-
times. The question might arise, therefore, whether we preferen-
tially discover only those nebulae of relatively low expansion

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 PLANETARY NEBULAE 235

velocity. However, there cannot be an appreciable fraction of

nebulae expanding at speeds in excess of 100 km/sec, for at least
two reasons: (1) The mean lifetimes for the nebulae would then
be substantially less than 2 χ 104 years, and their formation rate
would have to be even greater than the estimate given above,
which already seems to exceed the rate at which precursors of
planetary nebulae become available in the Galaxy. (2) If, say, all
expansion speeds between 10 and 1000 km/sec were equally likely,
since the probability of observing a nebula is proportional to its
lifetime, the probability that a randomly observed nebula would
have a velocity greater than 100 km/sec would be log 10/log 100 =
/2 — that is, approximately half of the measured expansion speeds
should exceed 100 km/sec. Actually, of 25 velocities listed by
Wilson (1950), only one is over 100 km/sec.
d) The mechanism of ejection of a nebular shell must arise from
internal conditions in the star producing a more or less ballistic
push, rather than a continual acceleration after the nebula has
left the star. Radiation pressure and a continuous ejection of a
"stellar wind" of corpuscular radiation from the central star prob-
ably influence the internal dynamics in a nebular shell, at least at
some stages of its evolution (Mathews 1966), but neither of these
can be the primary ejection mechanism. Even if the gas remained
optically thick to such electromagnetic and/or corpuscular radia-
tion, the amount of momentum required to be carried by that
radiation would be enormous. Typical nebular shells have masses
of 0.2 S)?© (O'Dell 1962, Osterbrock 1964). Just to support the
weight of such* a shell at the surface of a star of one solar mass
and radius R with the momentum carried by radiation, the lumin-
osity of the star would have to be

L = 8.3X 1013/ñ2 (1)

where L and R are in solar units. Luminosities and radii compatible

with equation (1) are outside the range of those possessed by
known stars. A shell might be produced by successive ejections of
smaller mass, but in any case the minimum momentum transfer
required to remove a unit mass from a distance r from a gravitating
body to infinity is easily shown to be the velocity of escape from

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 236 ABELL AND GOLDREICH

that body at the distance r. Thus the total momentum needed to

blow out a shell of mass 0.2 W© is at least 0.2 9¾ (2G 9Jto/K)V2,
which, for R in solar units, is 2.5 X 1040/fí1/2 gm cm sec-1. To supply
this momentum in the form of radiation pressure would require a
luminous output equivalent to 6 X 105 L© spread over 104 years if
the mass were ejected from a star of 1 fí©, and 2 X 104 L© over the
same time even if all of the mass were ejected from a giant star of
103 ñ©; such stars fall far outside the range of properties of observed
planetary nebulae nuclei. Indeed, just to accelerate a mass of 4 X
1032 gm from rest to a speed of 30 km/sec in 104 years, without any
gravitational drag whatsoever, would require a luminous output
over that time of 3 X 104 L©.
Moreover, even if radiation pressure were to be effective in ac-
celerating a nebular shell, that shell would have to be optically
thick to the radiation. The repulsive force on the nebula would thus
remain approximately constant, while the force of gravity on the
expanding shell would drop off as 1/r2. Ultimately, the nebular
shells would be accelerated to very high velocities unless they either
became transparent to the driving radiation or that radiation dimin-
ished in intensity at just the right moment to leave a nebula with
essentially the velocity of escape from its star. It is difficult to
imagine that such a coincidence could occur in every planetary
nebula.
Whereas there is little observational basis for estimates of the
momentum that can be carried by stellar winds from planetary
nuclei, it appears almost out of the question that corpuscular radia-
tion could blow 0.2 W© away from anything but a giant star by
gradual acceleration after the star and shell have separated. Even
if all of the energy of the ejected mass (in the corpuscular wind)
could be converted into mechanical thrust on the planetary shell,
a simple analysis of the energy requirements shows that at the
very minimum, a corpuscular flux equivalent to 103 L© would have
to work on the shell to eject it in 104 years from a star of 1 R©, and
then only if the stellar wind were ejected with a speed of at least
several times the escape velocity. These arguments do not rule out
that corpuscular radiation could play an important role in the ejec-
tion of a substantial mass from a star of very large radius, but this

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 PLANETARY NEBULAE 237

possibility, of course, is not inconsistent with our general conclu-

sions.
e) Stellar rotation is not important in the ejection of planetary
nebulae. Deeming ( 1965 ) has proposed a model in which stellar
rotation plays a role in the ejection of the nebula. He points out that
the rotational energy in a rapidly rotating star (say, 100 km/sec at
the equator) is of the order of 1047 ergs, and that this is near the
energy required to remove a nebular shell to infinity. If a major part
of the stellar mass contracts to a small dense core while the outer
stellar envelope grows (as in a star evolving to a red giant), the
core acquires a high rotation rate compared to the envelope. Now
if rotational energy can be transferred from the core to the envelope
(Deeming suggests a deep subsurface convection zone and mag-
netic fields), part of the envelope, according to Deeming's model,
gains enough rotational energy to be ejected. Deeming feels that
if just enough rotational energy is given to the shell to get the
ejection started, the nebular matter can be further accelerated to the
escape velocity by Lyman-a absorption.
We have seen, however, that no form of radiation pressure can
be important in ejecting a planetary nebula shell. There are, more-
over, several reasons why rotation cannot be important either. First,
the transfer of energy from the core to the envelope would have to
be sudden, for a gradual increase in rotation of the envelope would
at best merely expand and flatten it, not eject it. Moreover, we
would expect the material to come off in a ring from the equatorial
regions of the star, not in a shell, as is observed. Nor could material
in a ring diffuse into a shell distribution after ejection, for the
expansion velocities of the planetary nebulae are typically at least
2 or 3 times the thermal velocities in a gas at 104 0K.
The main argument against the rotation model for nebular
ejection is that the star can never have enough angular momentum.
The minimum momentum, p, needed to remove the nebular
material to infinity is, as we have seen, 0.2 (2G TIq/R)^,
where R is a representative radius in the matter to be ejected.
For circular motion, the angular momentum required is thus
pR = 0.2 Sí© (2G Now for an initially rigidly-rotating
star, the angular momentum is Iv/R#, where I is the moment of
inertia, ν is the equatorial speed at the surface of the star, and

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 238 ABELL AND GOLDREICH

is the radius of the star before evolutionary effects produce any

differential rotation. Thus a necessary (but not suflBcient) condition
for ejection of the shell is

Iv/R# — 0.2 mo (2G Tlo^R*. (2)

It is instructive to express ν in terms of that circular velocity at

the stellar surface at which the star is barely rotationally stable. If
the star before nebular ejection has a mass of 1.2 S)?©,

vcir={eGmQ/5R^ (3)

so that (from equation (2))

υ0.2 me
—^ ,
i^cir *

Inspection of published models of stars, both homogenous and

inhomogeneous, in the mass range 0.6 to 2.5 SW© (Schwarzschild
1958 ), shows that the central concentration of density in actual stars
is very much underestimated by a density law of the form

P(r) (l-r/R0)2. (5)

Thus the moment of inertia computed from equation (5), which is

8 3K© K|/35? is an overestimate of 1. The inequality, equation (4),
thus becomes

— 1.3 (ñ/ñj^ (6)

vc{

Since the envelope would grow in a star whose core contracted,

giving rise to the needed differential rotation R/R# > 1, we find
that the initial star would have to be rotationally unstable. This
argument does not, of course, rule out the possibility of a large
star contracting and shedding mass (in the form of an equatorial
disk). The equatorial velocity of such a star, however, never
exceeds the circular velocity; thus it cannot, according to equation
(6), send 0.2 Sí© into a parabolic orbit.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 PLANETARY NEBULAE 239

f) The ejection of a planetary nebula shell almost certainly

occurs from a star of extremehj large radius. The velocity of escape
on
at a distance of 0.1 parsec from a star of 1 9¾ is ly about 0.3 km/
sec, which is negligible compared to the observed expansion speeds
of planetary nebulae. Thus the nebular shells now have virtually
the same kinetic energy, Ex? that they will have when they have
expanded to infinity. (We ignore the possibility of drag effects of
interstellar matter because the galactic distribution of planetary
nebulae places them in locations where such matter is not expected
to perturb appreciably their expansion. ) If E{) is the kinetic energy
of the shell at ejection,

E() = Eos -|- Εχ (7)

where E(>s is the escape energy from the star at the time of ejection.
Assuming 1.0 and 0.2 W© for the star and shell, respectively,
we find that Eos = 7.7 X 1047/R ergs, where R, the radius of the
star at ejection, is expressed in units of the solar radius, and the
ratio E^/E* is about 400/jR. If the central stars with their present
observed sizes ( R < 0.01 to 1 ) had ejected the nebular shells, Eos
would have to exceed Ex by factors of the order of nearly 103 to
105. The large disparities between E(>s and Ex would appear to rule
out sudden ejection from small or solar-sized stars as the mechan-
ism giving rise to the planetary nebulae, for it is diflScult to under-
stand how a star could throw off a nebula with the escape energy
to within one part in several hundred or, in extreme cases, to one
part in one hundred thousand. Shklovskiï (1956) made the point
that the ejection velocity of the nebula would have to be the
escape velocity of the star; actually this is not true if the star is so
very large that Ees and Ex are comparable. We would, in fact,
expect a spread in the energies of the ejecting mechanisms. For the
dispersion in ejection energies to be comparable to a typical ejection
energy itself, the stars would have to be as large as 400 RQ.
We are therefore virtually forced to the conclusion that plane-
tary nebulae are ejected more or less suddenly, although apparently
not catastrophically, from stars of extremely large radii. Shklovskiï
hypothesizes that red giants simply expand until their envelopes
"sluff off" at essentially zero velocity, becoming the nebular shells,
while the stellar cores quickly evolve, readjusting themselves into

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 240 ABELL AND GOLDREICH

the stars observed as the planetary nuclei. Shklovskii calls attention

to the fact that observed expansion speeds (which he quotes to be
of the order of 10 km/sec) do not greatly exceed the thermal
4 0
velocities in a gas at 10 K. On the other hand, to obtain the ob-
served hyperbolic velocities, which range up to 100 km/sec, a
somewhat less leisurely event than that envisaged by Shklovskii
might be required.
g) The parent stars of planetary nebulae are most probably red
giants. If a star of 200 to 1000 R© had a spectral class earlier than
that of the sun, its absolute magnitude would have to be brighter
than —6 to —10. It is unlikely that such luminous stars with a popu-
lation II or "disk" distribution in the Galaxy would escape detection
unless their lifetimes are very short indeed; in fact novae (which do
not give rise to planetaries) of lesser luminosity are commonly dis-
covered. We must conclude that the most probable precursors of
planetary nebulae are red giants.

Conclusions

Planetary nebulae must be ejected from extreme giant stars,

probably red giants. Although Shklovskii reached this same con-
clusion a decade ago, it seems to us that his arguments are less
binding. Moreover, we conclude that all, or nearly all, stars of
1.2 9¾ must eventually become planetary nebulae. The central
stars of some planetary nebulae have spectra which according to
Greenstein and Minkowski (1964), are not incompatible with the
idea that they are relics of the interiors of red giants. The possi-
bility that combination variables like Ζ Andromedae, in which
bright lines appear to arise from a nebulosity in the vicinity of an
M star, are early stages of the planetary nebula phenomenon has
already been mentioned.
The ultimate fate of planetary nebulae nuclei appears to be white
dwarfs. We are not aware that it has yet been shown that the very
small nuclei observed for some planetaries are electron degenerate,
but in any case it seems unlikely that these stars will re-grow en-
velopes and expand again. If planetary nebulae come from red
giants and their nuclei evolve to white dwarfs, we are left with
the problem of explaining the horizontal branch stars. Our estimate
of the formation rate of planetary nebulae compared to that of red

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System
 PLANETARY NEBULAE 241

giants is not so precise that we can rule out the possibility that two
paths of evolution are open to red giants, one leading to planetary
nebulae and the other to a sequence of phases involving horizontal
branch and RR Lyrae stars, but we can rule out the planetary
phenomenon as a rare occurrence, accounting for only a small frac-
tion of the red giants. Moreover, it would be strange that stars
which, as far as is determined, have identical masses, chemical
compositions, and other characteristics should not follow similar
evolutionary patterns. We suggest that the possibility should be
taken seriously that stars evolve through a giant phase at least
twice — once immediately after leaving the main sequence, and
again in a post-horizontal-branch phase, prior to becoming plane-
tary nebulae.

We wish to thank Drs. L. H. Aller, D. M. Popper, Ε. K. Upton,

J. Hazelhurst, and H. Schmidt for their helpful discussions with us
concerning this interesting problem. One of us (P. G.) acknowl-
edges support from NASA 216-62.

REFERENCES

Abell, G. O. 1966, Ap. J. 143, 259.

Allen, C. W. 1963, Astrophysical Quantities (London; Athlone Press), p. 267.
Aller, L. H. 1954, Pub. Dominion Astrophysical Obs. 9, 321.
Deeming, T. J. 1965, Pub. A. S. P. 77, 443.
Greenstein, J. L., and Minkowski, R. 1964, Ap. J. 140, 1601.
Harmon, R. J., and Seaton, M. J. 1964, Ap. J. 140, 824.
Limber, D. N. 1960, Ap. J. 131, 168.
Mathews, W. G. 1966, Ap. J. 143, 173.
Menzel, D. 1946, Physica, 12, 168.
Merrill, P. W. 1958, in "Etoiles à raies d'émission" {Mém. Soc. Roy. Sei. Liège,
4th Ser. 20), p. 436.
Minkowski, R., and Abell, G. O. 1963, Pub. A. S. P. 75, 488.
O'Dell, C. R. 1962, Ap. J. 135, 371.
1963, Ap.J. 138, 67.
Osterbrock, D. E. 1964, Annual Rev. Astronomy and Astrophysics 2, 95.
Schmidt, M. 1963, Ap. J. 137, 758.
Swarzschild, M. 1958, Structure and Evolution of the Stars (Princeton: Prince-
ton University Press), pp. 254-259.
Shklovskiï, I. S. 1956, Astr. Zhurnal, U.S.S.R. 33, 315.
Wilson, O. C. 1950, Ap.J. Ill, 279.

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System