Non-Periodic Phenomena in Variable Stars
                                               IAU Colloquium, Budapest, 1968


                 NON-PERIODIC PHENOMENA IN BINARY SYSTEMS.
                          CONVENTIONAL BINARIES


                         Introductory Report by

                          FRANK BRADSHAW WOOD
         Department of Physics and Astronomy, University of Florida
                        Gainesville, Florida, USA


  This paper is primarily for those astronomers who are not specialists in 
the field of close double stars. It will review some of the evidence for 
non-periodic phenomena in eclipsing systems, especially those changes connected 
with the orbital period, and it will attempt to set a background for the more 
specialized papers on specific topics which follow.
  The detection and the interpretation of non-periodic changes in the radiation 
received from stars present many problems. This is especially true if the 
changes are small or if they occur only rarely. In the case of small changes 
in particular, we have the question of whether these have really occurred 
in the star or whether they have been introduced by the earth's atmosphere or 
the observational equipment.
  The earlier photometric observers of variable stars found various irregular 
changes chiefly in the form of "humps" or intermittent increases in brightness 
over normal light. In time, these showed secular changes, decreasing in number 
and size in direct proportion to the increase in the precision of observing 
techniques. In other words, they were not changes in the stars themselves. 
The realization of this, combined with the notorious temperamental behavior 
of the early photoelectric photometers, led to a tendency to believe that all 
such observed changes were a result of atmospheric or instrumental effects. 
In efforts to obtain light curves of the highest possible precision, observers 
frequently applied arbitrary "corrections" called night errors or seasonal 
errors. This produced beautiful light curves, but probably prevented the 
discovery of many non-periodic phenomena of the type with which we are 
concerned in these discussions.
  The study of period changes is made difficult because the observed effects 
can be detected only long after the changes have taken place. If a change of 
period of 0.1 seconds occurs, usually several thousand epochs must pass before 
the differences between observed and computed minima can be detected. For a 
star with a three day period, this can amount to well over a year. Thus we can 
never be certain whether we are observing one large period change or a series 
of smaller ones. Careful photoelectric monitoring of selected systems should 
help at least partially to answer this question.
  In principle, the study of period changes is extremely simple; in practice it 
is frequently more difficult. We simply plot the (O-C)'s -differences between 
observed times of minima and those computed from linear light elements- against 
the epoch of observation. If the period is constant, these scatter about a 
straight line; if the period is constantly changing, a curved line is needed 
to fit the observations; if an abrupt change (or closely spaced series of small
changes) has occurred, the (O-C)'s before the epoch of change will be fitted 
by one straight line and those following it by another of different slope.
  The difficulties arise in the usual scarcity of observed minima at the 
critical times which frequently leaves a wide choice in interpretation 
possible. The situation would be even more serious were it not for the 
long series of minima observed during the past fifty years at the Cracow 
Observatory. Working with relatively small instruments the Polish astronomers 
have observed a large number of times of minima to which modern observers using 
instruments of high precision turn as a first step in the interpretation of 
their own observations. When properly used, observations like wine grow more 
valuable with age, and while it is recognized that visual and photographic 
estimates -as distinct from measures- are of little use in obtaining accurate 
solutions, the times of minima from such observations are frequently the only 
data we have to combine with modern methods for studies of period changes.
  Instead of attempting the lengthy task of discussing all types of 
non-periodic phenomena in eclipsing stars, I shall concentrate on one system 
as an example of the sort of behavior we at times encounter, and for this 
purpose, select R Canis Majoris. In one sense, this is not a good choice, 
because R CMa certainly cannot be considered a "type" star-that is, as an 
example of some sub-class of eclipsing stars. Any attempt so to use it would 
have to rest on very superficial comparisons. On the other hand, it is a system 
which has been studied for more than seventy years and a great deal of 
observational data exists. The earliest reports of a non-periodic phenomenon 
came from observations made in 1898-9 when Pickering (1904) and Wendell (1909) 
reported an unusual increase in brightness immediately after the following 
shoulder of primary minimum. The plotted normal points of the two observers 
show a marked discrepancy but fortunately the individual observations were 
published and it is possible to show that this is a result of different 
combinations of the observations on the nights when the "hump" was present 
with those on the nights when it was not. When the individual observations are 
compared directly, any discrepancy is smaller than the observational scatter. 
Each observer used a polarizing photometer; while the accidental errors using 
such an instrument are considerably larger than those from photoelectric 
observations (probable errors of +-0.04 magnitudes or larger for an individual 
observation were not uncommon) the technique was remarkably free from 
systematic error, and the magnitude of the hump -when it was present- indicates 
that it was a real although transient feature of the curve. Several astronomers 
have studied these observations independently and have reached this same 
conclusion.
  This remarkable discovery of "flaring" (although not in the modern use of 
the term) attracted little comment and was not detected by other observers; 
however, it appears that no systematic attempt was made to monitor this part 
of the light curve. Then, Jordan (1916) observed radial velocity curves in the 
years about 1912. These showed two features. The first was a large orbital 
eccentricity when interpreted by conventional methods. The location and 
duration of secondary minimum on the light curve had indicated a circular orbit 
in distinct contradiction to this spectrographic data. The second unusual 
feature of the velocity curves was a large variation in the eccentricity from 
year to year. At the time there was no known explanation for this, although 
a number of cases have been found later and work by Struve (1944) and by
Hardie (1950) has shown the cause to be distortion of the spectral lines 
in circumstellar gas streams.
  About this time, Dugan (1924) began a photometric study, possibly in an 
effort to solve this discrepancy between photometric and spectrographic data. 
Instead of doing so, his observations not only confirmed the discrepancy, but 
added another problem. The relative depths of his minima were shallower than 
those found by Wendell, and the smaller secondary in particular, where the 
addition of a small amount of extra light would be relatively more noticeable, 
could not be reconciled with Wendell's observations. Little attention was paid 
to this problem for twenty years, although Gadomski (1930) did suggest a 
shortening of the period which was confirmed by other observers. Then 
photoelectric observations made in 1938-9 (Wood, 1946) agreed with Wendell's 
rather than Dugan's observed depth of secondary. These also confirmed the 
changed period, which seemed to occur at about the time of Jordan's work. 
At that time, the only conclusion was that further work in the infrared might 
help in reaching some conclusion.
  In light of current ideas concerning mass ejection and the frequent 
existence of circumstellar material in the neighborhood of close double 
stars, a reasonable explanation of these data is possible. If the 
intermittent flaring observed in the 1890's be considered an indication 
of developing instability which culminated in the ejection of material 
shortly before Jordan's observations, this ejection could be responsible 
for the period changes and the circumstellar material responsible for 
the distortion of the spectral lines and for the added light which 
caused the discrepancy between the earlier and later light curves and 
the one observed at this epoch. Many problems are solved by this 
hypothesis; however, I do not want to imply that this system presents no 
further problems; later work, in particular that by Koch (1960) and by 
Kitamura and Takahashi (1962), discusses some of these in more detail.
  I turn now to a general non-periodic phenomenon of many eclipsing 
systems-that of sudden and unpredictable period changes, as distinct from 
those caused by the rotation of the line of apsides of an elliptical orbit or 
changing light time in a three-body system. The first general study of these 
was by Dugan and Wright (1939). From period studies of a number of systems, 
they showed that in nearly all there were period changes which could neither 
be predicted nor explained by a periodic change. This conclusion has been 
confirmed for many other systems by later observers. In particular, Plavec 
(1960) has made a thorough study and has concluded that no reasonable 
combination of periodic terms can explain the observed changes.
  In interpreting these changes, it is natural to turn first to Newtonian 
mechanics, since relativistic effects such as gravity waves have been shown 
to be too small to be detected.* If Newtonian mechanics suffice, then by 
Kepler's third law change of period indicates either change of mass or change 
of separation of the components or both. Of these, a sudden change of mass 
roughly analogous to major prominence activity on the sun seems most reasonable 
and this possibility was postulated in 1950 (Wood, 1950). Earlier workers 
(e.g., Kuiper 1941) had considered mass loss but, if Kepler's third law alone 
is considered period changes in one direction only can take place whereas in
reality changes in either direction are observed to occur. If, however, we 
consider the effect of the high velocity ejection of matter upon the motion of 
the star itself, it can be shown that loss of the order of 10^-7 of the mass 
of the star per ejection can cause the larger period changes observed, with 
much smaller values possible if we are really observing the cumulative effects 
of many small changes. It also was shown that systems showing erratic period 
changes had one component near the Roche stability limit, whereas those showing 
long intervals of constant period had both components well outside these 
limiting surfaces. This rather primitive early work has been carried further 
by a number of astronomers making detailed computations of particle 
trajectories; a recent summary of this work has been given by Piotrowski 
(1967). The figure of 10^-7 of the mass of the star, which looked uncomfortably 
large at the time, looks far more reasonable in terms of recent work on mass 
loss from stars. As early as 1956, Deutsch from spectrographic data found 
ejection of at least 3 X 10^-8 needed in his work on Alpha Herculis and Dadaev 
showed a series of small changes would not produce photometric effects large 
enough to be noticeable. Batten's (1964) discussion of pronounced disturbances 
in the light curve at the time of a sudden period change of U CrB and related 
spectrographic evidence is also of considerable importance.

* However, see comments by L. Detre in the discussion following.

  We might end this part of the discussion by classifying our knowledge 
into three degrees of certainty. That irregular and -as far as predictability 
goes- erratic fluctuations of period occur in close double stars is a 
well-established fact of modern observational astronomy. Less certain, but 
still looking reasonably firm, is that as a general rule these changes are 
found to occur in systems in which one component is near the Roche lobe and 
will occur only rarely (e.g., AR Lac) when both systems are well separated 
from these lobes. The old suggestion (Wood, 1950) of a rough separation of 
close binaries into these two classes seems a reasonable working hypothesis and 
fits well with much later work on the evolution of close binaries, where the 
eventual growth of the more massive system to fill this limit seems an almost 
inescapable consequence of stellar evolution for all but the most widely 
separated systems. The later suggestion of a third classification of "contact" 
binaries is difficult to support on present observational evidence and it is 
not easy to rationalize the creation of such systems in any number from known 
evolutionary procedures. Finally, the details of the mechanism causing these 
changes are less certainly established but are a fruitful field for further 
investigation, and we may hope for an increase in our knowledge of these in 
the future. Mass loss from at least one component is almost certainly involved, 
but the precise physical mechanism or mechanisms involved and the ultimate 
disposition of the mass lost are still matters to be investigated.
  In conclusion, the importance of precisely observed light curves in known 
spectral regions and of times of minima continue to be of considerable 
importance. Theoretical advances are hindered by lack of enough observational 
evidence concerning the non-periodic phenomena which we know to occur. 
Interesting information can come from studies of the association of close 
double star systems with clusters of various ages, and this should be a 
profitable field of investigation which should aid in evolutionary studies. The 
absence of eclipsing systems in globular clusters has long been noted, and if 
they prove to be absent or very rare in the older galactic clusters, this will 
be significant information. Even in the younger clusters, the relative number of
close double stars compared to that of non-cluster systems will give help in 
studying their formation and evolution.
  The idea of eclipsing systems being important to such studies is not a new 
one although it is now receiving much emphasis. I should like to close with 
the statement: "The study of Algol should bring us to the very threshold of 
the question of stellar evolution..." The writer was A. W. Roberts in volume 
24 of the Proceedings of the Royal Society of Edinburgh. The year was 1902.


                              REFERENCES

Batten, A. H., 1964, Q. J. R. astr. Sec. 5, 145.
Deutsch, A. J., 1956, Astrophys. J. 123, 210.
Dugan, R. S., 1924, Princeton Contr. No. 6, 49, 1924. 
Dugan, R. S., Wright, F. W., 1939, Princeton Contr. No. 19. 
Gadomski, J., 1930, Astr. Nachr. 239, 96. 
Hardie, R. H., 1950, Astrophys. J. 112, 542. 
Jordan, F. C., 1916, Publ. Allegheny Obs. 3, 49.
Kitamura, M., Takahashi, C., 1962, Publ. astr. Sec. Japan 14, 44. 
Koch, R. H., 1960, Astr. J. 65, 326.
Kuiper, G. P., 1941, Astrophys. J. 93, 133.
Pickering, E. C., 1904, Ann. Harv. Coll. Obs. 46, 172.
Piotrowski, S. L., 1967, Commun. Obs. R. Belgique, Ser. B, No. 17, 133. 
Plavec, M., 1960, Bull. astr. Inst. Csl. 11, 197. 
Struve, O., 1944, Astrophys. J. 99, 89. 
Wendell, O. C., 1909, Ann. Harv. Coll. Obs. 69, 66. 
Wood, F. B., 1946, Princeton Contr. No. 21, 31. 
Wood, F. B., 1950, Astrophys. J. 112, 196.


                              DISCUSSION

Bakos: It appears that not only globular but also galactic clusters appear to 
      have fewer eclipsing binaries, about a factor of 3, than expected in a 
      general stellar field.
Detre: I have only the same comment I already said in my introductory paper: 
      if at least one component is of high temperature, a lot of ionized matter 
      is flowing between the components and the high temperature star is 
      losing ionized matter. Mass loss of this kind might be more effective 
      in causing period changes than simple "gravitating" matter. I refer to 
      Schatzman's papers on "magnetic braking".
Wood: I am in complete agreement with Dr. Detre's remarks and had intended 
      to say this in the body of the paper.