Ann. Rev. Astran. Astrophvs. 1978.

16: 1-13
ANNUAL
REVIEWS Further
Copyrrght © 1978 by Annual Reviews Inc. All rights reserved Quick links to online content

THE DEVELOPMENT OF x2123

OUR KNOWLEDGE OF
VARIABLE STARS
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

Cecilia Payne-Gaposchkin
Center for Astrophysics, Cambridge, Massachusetts 02138

In 1977 we commemorated the centenary of the birth of Henry

Norris Russell. Almost sixty years ago, already at the height of his powers,
the young Princeton professor delivered the Vice-Presidential address to
the American Association for the Advancement of Science in December
1918. He chose as his subject "Variable Stars. " His address serves as a
milestone: his encyclopedic knowledge and prodigious memory provided
a summary of what was known about, and thought about, variable stars
in that day. My own interest in variable stars has spanned almost the
whole interval between 1918 and the present, and I propose to survey
the changes that the subject has undergone.
Russell's paper contains several generalizations, from which I select for
comment:
1. "It is now possible to classify almost all variable stars into fairly
definite . . . natural groups. " This is true today, but with a complexity
undreamed of in 1918.
2. "It appears certain that almost all variable stars are . . . giants. . . .

This would mean that variability, while not confined to any one stage of
a star's evolution, is a characteristic of its early life-not the flickering
of the dying flame of age, but the exuberance of extravagant youth. "
This conclusion is no longer tenable.
3. Cepheid variables may provide "evidence that the rate of stellar
evolution is exceedingly slow and that the time scale of cosmical processes
is of very great length. " The conclusion is correct, but not merely on
the basis of the Cepheids.
4. To account for the outbursts of novae, "after what we now know
and believe regarding the stores of energy that are locked up in the
nuclei of atoms, the hypothesis of an explosive release of some such form

1
. 0066.4146/78/0915-0001$01.00
 2 PAYNE-GAPOSCHKIN

of energy within a star cannot be neglected." This suggestion shows

remarkable prescience, though the details are not as Russell anticip ated.
5. "The humiliating admission must still be made that . .. we know
extremely little about the real causes of stellar variation." Happily this is
no longer true, though much still remains to be understood.
The change that has taken place in the intervening years is reflected
in the progress of raw discovery. About 1700 variable stats were known
in 1918; since then, more than 26,000 have been catalogued. Russell
remarked that "dozens of [eclipsing] stars are still to be investigated."
Today we know almost four thousand. Most of the discoveries are
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

products of optical techniques already known in 1918, such as the blink

microscope and the photoelectric photometer. But photometric studies of
variable stars are now extending the frequency base into the infrared
and radio regions and the ultraviolet and X-ray domains, and there is a
new emphasis on "optical counterparts."
The understanding of the outer and inner condition of the stars was
still incomplete. Astrophysics, the outcome of analysis of laboratory and
stellar spectra -a subject to which Russell himself was to make no small
contribution -was virtually unborn in 1918. But by 1925 the surface con­
dition and composition of the stars were well on the way to formulation.
Two advances finally placed the variable stars in the pattern of stellar
development: the application of nuclear physics to the study of the
sources of stellar energy, and the development of computer techniques
that have led to the calculation of "evolutionary tracks" and to the
arrangement of the color-magnitude arrays of coeval groups of stars in
a developmental context and a time frame.
Meanwhile the classification of variable stars has undergone a change
that recalls the supersession of the Linnaean system by the modern
system of botanical classification. The five classes of variable stars
described by Russell (novae, long-period variables, irregular variables,
Cepheids, and eclipsing stars) have been replaced by a complex hierarchy
of physically significant groups. Three great phyla (irregular variables,
pulsating variables, and binary systems) are subdivided into over forty
types, which comprise virtually all the known variable stars. Very few are
not included: the'}' Cassiopeiae stars, the R Coronae Borealis stars, the
supernovae, and the pulsars. But besides these, the accumulating lists
of variable stars include ob jects that are not stars at all: optically variable
nuclei of galaxies (such as X Comae) and peculiar extragalactic radio
and infrared sources (such as BL Lacertae and W Comae). These optically
detectable variable objects are on the fringe of the growing lists of
optically variable galactic nuclei and quasars, which are not covered by
or included in the catalogues of variable stars proper. Even the optically
 V ARIABLE STARS 3
variable pulsars do not find a place there: CM Tauri is listed as a
supernova.
It can no longer be maintained that variable stars are all "giants."
Russell himself noted that some eclipsing stars, and the "Orion variables, "
are of low luminosity.Today we know that the W Ursae Majoris stars
are the most numerous of eclipsing stars, and the irregular and UV Ceti
variables probably preponderate per unit volume of space. Indeed, the
star of lowest known luminosity, V 1298 Aquilae, the fainter component
of +4°4048 (van Biesbroeck's star), is a UV Ceti variable.
Accordingly we cannot conclude, as Russell did, that variability is
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

characteristic of extreme stellar youth. Present knowledge of stellar

development has made it possible to date related groups of stars such as
clusters and associations, at least roughly, and to arrange them in con­
tinuous sequence.It follows that variable stars that are members of such
groups can be similarly dated.
Without exception, all the types of variable star that are included in
the three great phyla have been found in clusters or associations. They
range from types always (with the rarest exceptions) found in such groups,
the T Tauri and UV Ceti stars, to types which, though they may be
frequent in clusters, are found in even greater numbers in the galactic
field-the RR Lyrae and Mira variables. Only the R Coronae Borealis
stars and the pulsars (supposedly relics of supernovae) have not been
found in clusters or associations.
If we make the distinction between chronological and developmental
age, we draw the broad conclusion that variables always found in clusters
are (developmentally) the youngest; those least confined to clusters, the
oldest. This conclusion is fortified by the completely independent evidence
of their situation in the color-magnitude arrays of the groups. The very
youngest lie above and to the right of the main sequence and bear the
recognized marks of the pre-main sequence stage. According to this view,
"variability, while not confined to any one stage of a star's evolution, "
passes through successive stages as the star develops from the "exuberance
of extravagant youth" to the "dying flame of old age."
Stars that are approaching the main sequence for the first time have a
wide variety of luminosity and surface properties. Imbedded in a nebula,
AE Aurigae has a spectrum of Class 0 and a fairly high luminosity,
while R Coronae Austrinae is lower in luminosity, later in spectrum,
and deep in nebulosity. Such stars are rare compared to the T Tauri
stars with basically G and K spectra, and the UV Ceti stars, still later
in spectrum, which literally swarm in the great Orion and Taurus
associations. Many are associated with the Pleiades, a few with the much
older Hyades-a reminder that developmental, not chronological age is
 4 PAYNE-GAPOSCHKIN

the determining factor. The upper limit of the zero age main sequence
of such a cluster, where stars have just begun to veer off, provides a
terminus ad quem; the lower limit, beneath which such stars have not
yet arrived, a terminus a quo. Between them they help to define the
interval during which the cluster came to birth.
The spectra of many pre-main-sequence stars (and not only of those
obviously variable) have anomalous energy distributions. Ultraviolet
excesses are common, accompanied by "veiling," and may be associated
with the characteristic bright-line chromospheric spectrum. A large
number of these stars have infrared excesses too, and many show evidence
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

of dust envelopes, which are probably also responsible for the fact that
this is the only class of distant stars detectable in the radio range.(The
exceptional occurrence of radio radiation, as well as a large infrared
excess, for the massive eclipsing binary RY Scuti must be related to the
fact that it is deeply embedded in nebulosity.)
When "nebular " variables such as these were first discovered it was
difficult to understand the cause of their variation. Fifty years ago I
recall asking Eddington whether he thought a nebula could cause a star
to vary. "Ah, " he said slowly (as was his wont), "Probably the star
causes the nebula to cause the star to vary."
A star spends the major part of its traceable lifetime on or near the
main sequence-an interval of quiescence that is shortest for greatest
mass.It is passing through the first stage of thermonuclear development,
converting hydrogen into helium in a growing core.But our study of the
Sun reminds us that the quiescence is internal and that the surface may be
seething with activity.
Just above the main sequence is a small group of variable stars that
were once placed among the irregular variables, and are now classed as
y Cassiopeiae stars.Pleione (BU Tauri) is in the Pleiades, X Ophiuchi
in the association Scorpius II. They tend to be slightly evolved; their
variations are marked by shell phenomena and they rotate very rapidly.
The situation of Pleione shows that it is leaving, not approaching, the
main sequence.Probably the rapid rotation of these stars provokes the
shell phenomenon and the attendant variations.
The interval of gross quiescence ceases when the star reaches the first
of the many twists in the evolutionary track that mark a transition from
one thermonuclear mode to another.Here we encounter the first repre­
sentatives of the second great phylum, the pulsating stars: the f3 Canis
Majoris variables. Their early stage of development is indicated by their
presence in young associations (such as Perseus I).Some, though probably
not all, are members of binary systems. They are confined to a rather
narrow range of early B spectrum and slightly evolved luminosity classes
 VARIABLE STARS 5

(IV and III), and their vanatlOns of light and velocity are evidently
governed by physical condition, perhaps stimulated by duplicity.
Lower in luminosity, and bearing a similar relation to the main
sequence, are the 'Y. Canum Venaticorum stars, marked by strong magnetic
fields and concurrent periodic variations of brightness and spectrum.
Many of them are binaries. It seems well established that the observed
periodicities are of rotation, not pulsation. A few are members of stellar
groups: LT Persei is in the Perseus moving cluster, GS Tauri is in the
"Pleiades group," and AI and UU Comae are in Coma Berenices. The
parallel with the f3 Canis Majoris stars is extremely suggestive; are these
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

stars pulsating as well as rotating?

As the star swings across the Hertzsprung gap it enters the instability
strip, and here we find the classical Cepheids. Less than a dozen are
known in clusters; many are known in the field. However, very few cluster
stars fall within the instability strip, and all those that have been studied
seem in fact to be Cepheids. The instability strip is defined by the surface
conditions necessary to the "heat engine" mechanism that maintains the
pulsations. Whether a particular star will enter the instability strip depends
on the course of the bends and loops undergone by the developmental
tracks. It appears that, for the range of chemical composition displayed
by galactic open clusters, such intersections take place for stars with
masses between about three and nine times the mass of the Sun..
Accordingly the ages of galactic Cepheids lie between 107 and 108 years.
Depending on mass and composition, a given star will theoretically
cross the instability strip three or five times, alternately expanding and
contracting at different rates in different parts of its course. Its period
(dependent on mean density) should therefore alternately increase and
decrease at different rates as development continues. Thus it is very
difficult to use the changes of period of Cepheids, as Russell suggested,
to measure the time scale of cosmical processes. However, his conclusion
remains correct: the rate must be "exceedingly slow." The situation is
complicated by the observed fact that the changes of period are not
secular, may be discontinuous, and may run in either direction.
Cepheid variables, especially the youngest ones with long periods, have
their uses in attempts to map the "structure of the Galaxy." All well­
observed classical Cepheids are within about 250 pc of the galactic plane,
and clearly belong to the flattened central system. There are, however,
a number of loose ends to be tucked in. The list of stars classified as
Cepheids contains a number that must be at a considerable distance
from the galactic plane if they conform to the period-luminosity relation.
Some, perhaps, belong in the W V irginis class. Others, I suspect, are
RR Lyrae stars, and their deduced periods rest on too few observations.
 6 PAYNE-GAPOSCHKIN

MostCepheids repeat their variations with fidelity. Those whose published

light curves show considerable scatter merit further study.
For stars with masses less than about three times the mass of the Sun,
bends and loops in theoretical tracks no longer intersect the instability
strip.Tracks for less massive and less luminous stars rise toward the red
giant stage and cross the extended instability strip just above the main
sequence. This position coincides with that of the (j Scuti stars, which
resemble the RRs (dwarf Cepheid) variables in period and spectrum, but
are formally distinguished by smaller light amplitudes. I am inclined to
think that there is no physical difference, and refer to all as (j Scuti stars.
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

A number are known members of clusters: V 480 and V 483 Tauri are
in the Hyades, V 534 Tauri in the Pleiades, FM Comae in Coma Berenices.
It seems likely that their pulsations are maintained in the same way as
those of Cepheids. They conform to the period-luminosity relation and
the period-spectrum relation of the latter.
After crossing the instability strip the theoretical tracks finally rise to
the red giant stage. For the most massive and luminous stars, the first
bend already passes so far to the red that it briefly enters the red giant
domain, and after doubling back and forth, reaches the red supergiant.
Cluster stars at this stage are understandably rare, since it is very quickly
traversed.There may be an example in the very red VY Canis Majoris, a
member of the very young NGC 2362. This star (another rarity) seems to
be undergoing a secular decrease in brightness. Star and cluster are
embedded in a variable, dusty nebula. Another sample of red supergiants
in this stage is furnished by the considerable group in the great Perseus
association that surrounds h and X Persei. That these clusters and the
surrounding association are very rich in stars may account for the
presence of an unusually large group of red supergiants, most of them
semiregularly periodic and probably quite massive.
Less massive stars approach the red giant stage on converging tracks,
and the oldest open clusters exhibit color-magnitude diagrams that
terminate in red giants.About a third of all known variable stars are red
giants, and a large proportion of red giant stars are variable, so we might
expect to find red giant variables well represented in open clusters. But
this does not prove to be the case. Only one irregular red variable is
known in an open cluster, FG Vulpeculae in NGC 6940. Moreover, it is a
bright giant of luminosity class II, not the commoner red giant of
luminosity class III. And open clusters contain no known representatives
at all of the red variables commonest in the galactic field, the semiregular
red and Mira stars.
When interpreting the occurrence of variable stars in open clusters I
have spoken as though all are of the same chemical composition and can
 VARIABLE STARS 7

accordingly be related to the same set of theoretical tracks. This is

admittedly an approximation. But the available evidence suggests that
the hydrogen/metal ratio does not differ by a factor of more than two
or three from cluster to cluster-certainly not by an order of magnitude.
These clusters, and variable stars of the associated types, belong to the
flat component of the Galaxy. That component now appears to have an
appreciable composition gradient, such that the outer regions of the
Galaxy are relatively poorer in metals. Some hint of the effects of this
gradient may be seen among the types of variable star already discussed.
The median period of field Cepheids is greater toward the galactic center
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

than in the solar neighborhood, and less toward the galactic anticenter.
A similar trend is traceable in Messier 31. This small effect may well be a
consequence of the composition gradient, just as the much larger difference
of period frequency between the Galaxy and the Large and (even more)
the Small Magellanic Cloud are almost certainly due to difference of
composition, and its effects in modifying the bends and loops of the
evolutionary tracks.
Most of the remaining classes of pulsating variables either verge towards
or belong to the spherical component of the Galaxy, and here it is well
known that the hydrogen/metal ratio may have a range of at least two
orders of magnitude. It is no longer possible to picture the variable stars
in a two-dimensional scheme. The course of the tracks will depend not
only on mass and age but on composition.
The globular clusters furnish the background on which the most
numerous of the variable stars-Mira and RR Lyrae stars-are displayed,
as well as the less frequent W Virginis and RV Tauri variables. These
very populous stellar groups, representative of the spherical component
of the Galaxy, have a range of about two orders of magnitude in the
hydrogen/metal ratio. That composition is closely related to their gross
properties is well known, as is its relation to location within the galactic
system. Globular clusters are at least as old as the oldest known open
clusters, but composition is at least as important a factor as age in the
differences between them.
The red giant variables whose absence from open clusters is so con­
spicuous are found, though in small numbers, in globular clusters. Red
irregular variables and Mira variables are found in those richest in metals.
Both are well represented in the metal-rich NGC 104 47 Tucanae.
=

Mira stars in globular clusters characteristically have periods of about

200 days, whereas the very numerous Mira stars of the galactic field
have a median period of about 350 days. As there are slight indications
that the higher the metallicity of the cluster, the longer the periods of
its Mira stars, this observation suggests that a difference of composition
 8 PAYNE-GAPOSCHKIN

may lie behind the difference of period frequency-a surmise already

made for classical Cepheids. Mira stars are evidently pulsating variables
and occupy an instability strip comparable to that responsible for the
variations of the Cepheids.
The extension of variable star photometry into the infrared has greatly
enriched our knowledge of the irregular, semiregular, and Mira variables.
Virtually all the objects brigher than the third magnitude at 2 microns
are red stars with molecular spectra, and a large proportion are variable.
Mira stars are especially prominent, and characteristically vary by about
a magnitude at this wavelength, whereas optically their light may have a
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

range of from six to ten magnitudes. An exploratory survey of the stars

observed to be variable at 2 /lm has revealed that a large proportion
are optically variable too.
At 10 pm the relative representation of Mira variables is much greater
than for other red variables, either because they are very cool or as a
result of extended envelopes. The latter case is represented by the Mira
star WX Serpentis (P 425 days), which has an extended envelope, and
=

an apparent magnitude - 2.44 at 20 /lm. The intricate physical problems

presented by the Mira stars should be attacked from this angle.
When we pass to globular clusters of intermediate metallicity, we find
the W Virginis and RV Tauri stars occupying an instability strip that is
continuous with that for classical Cepheids. They display a period­
luminosity relation almost parallel to that for Cepheids, but about two
magnitudes fainter. The comparatively low metal content for these stars
is duplicated in that of field examples such as AC Herculis.
The RR Lyrae stars again are associated with an instability gap within
the horizontal branch that is an extension of that for the pulsating
variables of longer period. They are found throughout the range of
cluster composition. Rare in clusters of high metal content (47 Tucanae,
one of the most populous of globular clusters, has but three), their
numbers and the differences in their bimodal period frequency are evident
functions of metal content. But the vast majority of the RR Lyrae stars
known are in the galactic field. Most globular clusters contain few or
none, and only a handful (such as NGC 5139 and Messier 3) very many.
The preceding paragraphs have placed all the recognized types of
pUlsating variables, by reference to their occurrence in clusters, in a
sequence of developmental stages, from f3 Canis Majoris stars to RR
Lyrae variables. Russell's first generalization is fortified, his second has
been replaced, and his third is amply justified, not alone on the basis
of the Cepheids.
We now pass to the third great phylum, the binary systems. A large
part of Russell's paper dealt with eclipsing stars, and despite that early
 VARIABLE STARS 9
date his summary has a modern ring. Observation and theory have been
extended, as he predicted they would be (largely under his own stimulus),
but the study of eclipsing variables was already, as he said, in a satis­
factory state.
However, in 1918 little thought had been given to the proble m of
intrinsic variability in binary systems, save for an abortive attempt to
account for Cepheids in terms of spectroscopic binaries. In view of the
known commonness of duplicity it is of interest to enquire to what extent
the known types of variable star are members of binary systems, and if
so, how far binary nature affects their behavior.
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

There are a number of known binaries among pre-main-sequence

stars, such as S Monocerotis (also a member of the open cluster NGC
2264) and XY Persei, both irregular variables of early spectral type.
Later in spectrum are the irregular pa ir DR and DI Tauri; sixteen of
the late type UV Ceti stars are known binaries includin g the pair EQ
,

Pegasi a and b, V 645 Centauri ("Proxima"), WX Ursae Majoris (Lal

21258B), and the fainter component of van Biesbroeck's star. The small
but growing class of BY Draconis stars, which combine irregular and
periodic variations with spectra that recall the UV Ceti stars, may be
binaries whose periodicities are presentation effects, but none is observed
to eclipse. However, there seems no reason to suppose that duplicity is
the cause of the variations of any of the pre-main sequence stars.
About a quarter of the f3 Canis Majoris stars are known spectroscopic
binaries, including the type star. But others, such as f3 Cephei and y Pegasi
(one of the most thoroughly studied of stars) are not. It therefore seems
that duplicity, though it may well modify the pulsation, is not the primary
cause of it. Many IX Canum Venaticorum stars are periodic binaries,
but the observed photometric and spectroscopic periods have been shown
to be the much shorter rotational periods of the components, recalling
the BY Draconis stars. Duplicity might possibly stimulate and modify
the variation, but does not actually cause it.
Perhaps 15% ofCepheids ar e members of binary systems, whose periods
are very long compared to the periods of pulsation. Save for inevitable
effects on the color and amplitude of unresolved pairs, the duplicity does
not seem to affect the pulsation. Among the {) Scuti stars and dwarf
Cepheids, about a quarter are known members of binary systems.
Especially interesting is the brighter component of the Algol system AB
Cassiopeiae. In the light of the large number of single {) Scuti stars it
seems unlikely that the duplicity of a few is the cause of their pulsation.
Neither has it suppressed the pulsation of the brighter component of
AB Cassiopeiae, although the comparative closeness of the components
might lead to this expectation.
 10 PAYNE-GAPOSCHKIN

Among the red giant variables we again find a small proportion of

binaries. Three naked-eye Mira stars are components of visual double
stars: X Ophiuchi, R Hydrae, and 0 Ceti. It seems unlikely that the
visual companions can cause the variations, but now for the first time
there is a suggestion that they can affect them. The period of R Hydrae
is recorded to have declined from 500 to 400 days in about two hundred
years-a rarity for a Mira star: though notoriously unpunctual they do
not display secular changes of period. The companion of 0 Ceti is the
faint blue variable star VZ Ceti, whose erratic eruptive behavior is
certainly affected and probably caused by material flowing from its red
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

giant companion, even at the great distance that separates them in their
100-year orbit. Still more compelling is the evidence from R Aquarii,
whose unresolved blue companion brightens at irregular intervals, and
while it is bright the otherwise normal variations of the red giant are
almost suppressed. A surrounding nebula testifies to a major eruption in
the far past.
During the past few decades the study of eclipsing stars has shown
that physical interaction between the components of a close binary is a
common, if not a universal phenomenon, and has led to sophisticated
theoretical treatments of light curves, velocity curves, and spectra that
were undreamed of in 1918. In the "early" types of intrinsic pulsational
variability, from f3 Canis Majoris stars to Cepheids, the variables may be
members of binary systems, but their companions do not cause, and
hardly affect, their variations. For the Mira stars that are binaries the
duplicity, though not the source of the variation, leads to observable
interaction. None of the pulsating variables, however, are invariably
binaries.
There is, however, a group of variable stars that are all binaries, and
their variations are the direct outcome of their binary nature. It comprises
the so-called cataclysmic variables, novae of all types (except the super­
novae), and the so-called symbiotic or Z Andromedae stars. The novae
(fast, slow, and recurrent) and the SS Cygni and Z Camelopardalis stars
have proved to be physically similar when not undergoing an outburst­
a red main sequence or giant star and a degenerate blue companion
forming a close binary in active interaction. Material flowing from the red
star surrounds the blue one with a ring, often with a "hot spot" on
which the inflowing material impinges. The isolated outbursts of novae,
the intermittent brightenings of SS Cygni and Z Camelopardalis stars,
and the rapid fluctuations of other stars, barely visible on a gross scale,
grade into one another. Each binary system preserves its individuality:
SS Cygni never erupts like a nova, and recurrent novae repeat their
outbursts with fair regularity.
 VARIABLE STARS 11
The Z Andromedae stars differ from the preceding in that the red
partner is not a main sequence star but a giant. They include at least
some of the recurrent novae, such as T Coronae Borealis. It is hard to
draw the line: perhaps R Aquarii falls in this category. And XX Ophiuchi,
long considered a prototype for an irregular variable of early spectrum,
has a red giant companion and is without doubt a symbiotic variable.
A third group, recently recognized, contains the ZZ Ceti stars, variable
white dwarfs without known red companions. They display complex
periods of the order of hundreds of seconds. Should they in fact find a
place among the pulsating variables? Assigned to this class is GP Comae,
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

reported to be a spectroscopic binary. Too few such stars are known at

present for placement in the pattern of development.
I have been guided in this enquiry by relating variable stars of
different types to the stellar groups in which they occur. Three novae
(one a very peculiar nova) and three dwarf novae have been observed in
or near globular clusters, and some of these, at least, are surely members.
But the dwarf nova BX Puppis is possibly a member of NGC 2482,
an open cluster not unlike the Hyades. So it cannot be concluded that
the novae necessarily belong to the spherical population, and in fact their
galactic distribution does not suggest it.
It seems that the one essential condition for a nova or other cataclysmic
variable is membership in a close binary system of a very special type.
We may note, perhaps, that close binaries of the W Ursae Majoris type,
though very common indeed in the galactic field, have been sought in vain
on or near the main sequences of globular clusters. Moreover, member­
ship in a globular cluster connotes low metal content, while current
theories of the nova outburst involve thermonuclear runaways in an
environment exceptionally rich in carbon, nitrogen and oxygen.
I have passed in review the major types of variable star, and have
attempted to relate them to the stages through which a star passes in
the course of its development. The summary has covered the representa­
tives of the three great phyla. Two classes of variable star, and those
not the least interesting, remain to be mentioned: the R Coronae Borealis
stars and the supernovae. We have called one subclass of the binary stars
the cataclysmic variables; the two remaining classes deserve the name
catastrophic. It seems that the composition and structure of the R Coronae
Borealis stars can only be the results of the loss of the outer portions
of an evolved star, leaving an almost hydrogenless nucleus rich in heavy
elements. The transition from supernova to pulsar may involve an equally
abrupt event. Abrupt or not, both processes seem likely to be irreversible.
The pulsar, detected and defined by its radio behavior (though, as with
CM Tauri, it may also be optically variable) owes its behavior to rotation.
 12 PAYNE-GAPOSCHKIN

As we have seen, the optical variations of the IX Canum Venaticorum

stars and the BY Serpentis stars are rotational phenomena too. We are
reminded of n5 Orionis, whose variations have long been ascribed to a
presentation effect.
Of Russell's fifth generalization it can be said that after almost sixty
years we know a great deal about the causes of stellar variation.
Emphatically this is the case for eclipsing stars. For Cepheids, as he
suggested, the great breakthrough was to be made by Eddington. Today
our understanding of Cepheids, RR Lyrae stars, and W Virginis stars is
gratifying. The cataclysmic variables, too, have reached a plateau, though
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.

much still remains mysterious. The group that intrigues me the most
comprises the Mira stars. Many years ago I used to tease Paul Merrill
by telling him that he took the obscurantist view that they were too
complex to be understood. At least they seem to have defied physical
understanding up to the present. Among known variables they are
equalled in numbers only by the RR Lyrae stars. They bridge the gap
between the flat and spherical systems, between the open and globular
clusters, between "Populations I and II. " Their spectra, combining high­
excitation bright lines and overlying molecular clouds, still present an
enigma.
In 1918 the study of variable stars was emerging, in Russell's words,
from being "once neglected and even rather despised " to become a major
tool of astrophysics. In 1977 there is a tendency to relegate it to the
status of "old-fashioned astronomy." A final word, therefore, on two
aspects of timely interest.
In his discussion of eclipsing stars, Russell pointed out their possibilities
in determining stellar densities. It seems strange that he made no mention
of stellar masses, perhaps the most important of their properties. In those
days the existence of high stellar masses had not been demonstrated,
and it was usual to assume that all were comparable to that of the sun,
for they were derived from visual binaries of low luminosity. Eclipsing
stars like Y Cygni changed all that. But there are still only a few dozen
directly determined stellar masses. A rich harvest is still ungarnered. Not
only are there many double-lined spectroscopic binaries that may well be
eclipsing stars, but today it would be possible to make a direct determina­
tion of the mass-luminosity relation for other stellar systems, such as the
Magellanic Clouds, which probably differ in chemical composition from
our G alaxy.
Not only eclipsing stars can give physical information about stellar
systems other than our own. The distribution of the periods of Cepheids
and RR Lyrae stars can throw light on the course of the development
 VARIABLE STARS 13

tracks and thus on the physical condition of systems that differ from
our own, particularly the dwarf eUipticals.
The Magellanic Clouds must still remain a happy hunting ground
for the student of variable stars. In 1918 the number of variable stars
known in the Galaxy was about half the number known in the Magellanic
Clouds, even though at that time only "giant" stars were accessible in the
latter. Few Mira stars are known there today, and very few RR Lyrae
stars. But if the variable star population of the Clouds is anything like
that of our own Galaxy, they should be found in myriads when the
Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org

survey is pushed beyond the twentieth magnitude. If I were beginning

Access provided by University of Houston - Main on 06/08/17. For personal use only.

again, it is there that I should search.

Annu. Rev. Astron. Astrophys. 1978.16:1-13. Downloaded from www.annualreviews.org
Access provided by University of Houston - Main on 06/08/17. For personal use only.