Stellar Radial-Velocity Programs of the Lick Observatory

STELLAR RADIAL-VELOCITY PROGRAMS OF THE LICK OBSERVATORY

George H. Herbig Lick Observatory, University of California

The first director of the Lick Observatory, Edward S. Holden, fully recognized the importance of extensive radial-velocity information for the determination of the motion of the solar system in space. He regarded it as "the main [spectroscopic] research to be undertaken with the great telescope." In 1890, he drew up an extensive program of observations to be carried out visually at the 36-inch refractor by J. E. Keeler. The contribution of Holden to- ward starting radial-velocity work at the Lick Observatory is not generally appreciated. It is true that he played an administrative rather than a participating role, but it was an essential contribution nonetheless. Holden's attitude is indicated by the letter of instruction from him to Keeler, dated September 15, 1890, which reads in part : My dear Mr. Keeler : It is now time to begin the principal spectroscopic work for which the great telescope was designed; namely the observations of the motion of stars in the line of sight for the determination of the motion of the solar system in space. Your observations on the nebulae have shown that you can do this work better than it is done elsewhere . . . Such a work as this done properly will require several, perhaps many years, and it ought to be thoroughly thought out before it is begun and an adequate programme adopted and adhered to. I have therefore drawn up the following prelimi- nary scheme for you to look over and criticise . . . You have what I think to be by far the most important work with the great telescope, and I promise you on my part every official and personal assistance which it is in my power to give. Sincerely yours, E. S. Holden It is obvious to us now that the very extensive program pro- posed by Holden was unrealistic, considering the limitations of the visual method, but the approach was a sound one in the light of modern knowledge. Work by Keeler in 1890-91 with a spectroscope containing a plane Rowland grating gave good results for the radial velocities of three first-magnitude stars as well as for fourteen gaseous nebu-

191

lae. Keeler's velocities were incomparably superior to those obtained by earlier visual observers, whose results seem to bear little relation to the true stellar velocities. Keeler's procedure, in observing the stars, consisted of measuring with a micrometer the velocity shift of the D lines of sodium in the star with respect to the same lines in a comparison spectrum. The visual observations were exceedingly difficult even under optimum conditions, but it was planned, nevertheless, to extend the work to all feasible objects. This extended program was to be carried out by Keeler with the assistance of W. W. Campbell (who had helped Keeler in the earlier work on a voluntary basis) and Henry Crew, both newly appointed to the staff in the summer of 1891. Keeler, how- ever, left the Lick Observatory in mid-1891 to assume the direc- torship of the Allegheny Observatory, and following the departure of Crew a year later, the work fell into Campbell's hands. In 1891 the work of Scheiner and Vogel of Potsdam on the determination of radial velocities by the photographic method appeared. The advantages of photography over the eye at the telescope were very apparent : the Potsdam observers, with an 11-inch refractor, had reached fainter stars than had Keeler, visually, with the 36-inch. The probable error of a single Potsdam observation was given as ±2.6 km/sec from the internal agreement of the observations, but comparison with modern velocities demon- strates that the Potsdam results were less dependable than this figure would indicate. Campbell realized that the day of visual work in astronomical spectroscopy was over, and he attempted to adapt Keeler's grating spectroscope to photography, but he found it to be so subject to mechanical flexure that the attempt was abandoned. A limited amount of work was done with it, with prismatic dispersion, until it was completely superseded in 1895 by the original Mills spectrograph. Not long thereafter, Campbell's interests, which had until then ranged over a broad field of astronomical subjects, con- verged on the problems of radial velocities. The Mills spectrograph was reminiscent, outwardly, of the Potsdam instrument, but in its details and optical arrangement it represented a considerable improvement. Campbell was generally responsible for the design, but the spectrograph also owed a

© Astronomical Society of the Pacific · Provided by the NASA Astrophysics Data System LICK RADIAL-VELOCITY PROGRAMS 193 great deal to Keeler, who initiated Campbeirs spectroscopic train- ing. The dispersion, produced by three dense flint-glass prisms, was 12.5 A/mm at Hy. The results, after some initial difficulties had been located and removed, fully justified the design : the probable error of a single observation of a bright star with good lines was reduced to ±0.5 km/sec, and for bright stars with the best lines, the probable error was even smaller.1 With this instrument Campbell and his colleagues ushered in a new era in radial-velocity work. One would certainly expect new developments in any field in which the accuracy of observation is suddenly improved by an entire order of magnitude, and in which the number of objects observable with such precision is increased perhaps fiftyfold. Successful as the Mills spectrograph proved to be, after it had been in operation for some time Campbell, Wright, and others at Lick recognized that increased efficiency would result if several modifications could be made. For example, the mechanical con- struction was such that a certain amount of internal flexure was present ; furthermore, the spectrograph operated in a wave-length region (centered at λ 4340) at which the transmission of the 36- inch objective and the flint-glass prisms was not high. The New Mills spectrograph, a completely new instrument except for the prisms, was put into operation in 1903, and the original Mills was retired, except as will be mentioned later. The New Mills incor- porated a superior support system, based on a suggestion of Wright's, by which internal flexure was reduced to undetectability. · The central ray was now a-t λ 4500, a choice that involved a com- promise between the opposing requirements of the failing transmission of the optics to shorter wave lengths, the decreasing sensitivity of ordinary photographic emulsions to longward, and consideration of the spectral lines that fell in that part of the spectrum in stars of various spectral types. Other mechanical improvements were also made. The dispersion of the new spectrograph is 10.9 A/mm at λ 4500, somewhat higher than the original Mills at Η γ, but no large improvement in the quality of the velocities of bright stars resulted on this score by replacement of the original Mills. Instead, the increased mechanical and ther- mal stability and somewhat higher efficiency of the new instrument enabled essentially the same standards of accuracy to be extended

to fainter stars. At the present time, through the use of coated optics and faster plates, the precision obtainable fifty years ago with the original Mills only for first- and second-magnitude stars is now attained for sixth-magnitude stars with the New Mills. Campbell had realized in the mid-'nineties, when the radial- velocity work was getting under way at Mount Hamilton, that regardless of how well the velocities of the stars observable from Lick were determined, the investigation of the solar motion would be seriously handicapped by the lack of data from the inaccessible southern third of the sky. In 1900, the necessary funds were obtained from D. O. Mills of San Francisco and New York, donor of the Mills spectrographs, to set up, equip, and maintain for two years an observing station in the southern hemisphere for the express purpose of determining the radial velocities of the brighter southern stars. In 1903, the expedition, under the direction of Wright, established an observatory near Santiago, Chile. The only telescope was a 37-inch Cassegrain reflector that was equipped, at first, with a three-prism spectrograph similar to the new Mills ; later, one- and two-prism instruments were provided. It was soon apparent that an extension of the life of the expedition was desirable ; support was provided by Mr. Mills until his death in 1910, and by his son, Ogden Mills, until 1927. From that year until 1929, the station was supported by gifts from friends of the Lick Observatory. In 1929, the station and its equipment were sold to the Catholic University of Chile. During the twenty-six years of its operation, about 11,000 spectrograms of planets, stars, and gaseous nebulae were obtained. The major velocity program at the two observatories was the determination of the radial motions of the brighter stars. For a period in the early years of the program, the limiting magnitude was taken to be 5.01 in the northern sky (the authority for the magnitudes was the Revised Harvard Photometry), but this was later extended to magnitude 5.51 for the entire sky. From 1896, when the program was initiated, until 1926, when the last spectrograms were taken, some 15,000 plates for the determination of radial velocity were obtained at Mount Hamilton and about 10,000 in Chile. The velocities of the stars were discussed by Campbell and Moore in Volume 16 of the Lick Observatory Publications,

and the velocities of the gaseous nebulae in Volume 13. Velocities were assigned to 2743 stars as a result of this great project, which occupied the efforts of a major part of the Observatory staff for thirty years. The "New" Mills spectrograph is still in active use. Since the completion of the large program, its time has been taken up mainly by the observation of stars of variable velocity. At the present time, some thirty spectroscopic binaries that were discovered at Lick are being followed. Most of these stars have long periods (up to 40 years) and small velocity amplitudes, so that their study requires the use of a high-dispersion spectrograph whose charac- teristics are well known and that possesses long-term stability, conditions that probably are met as well by the Mills as by any existing instrument. An example of such a binary is the long- period system of Polaris : the period is 29.6 years, while the velocity semiamplitude is only 4 km/sec. Polaris is the classic example of a Cepheid variable (the pulsation period is 3.97 days) whose normal velocity is itself variable in a long period. Both components of the velocity variation were discovered with the original Mills in 1899, and during the 1% revolutions in the long- period orbit that have been completed since discovery, over 1000 Lick three-prism spectrograms of the star have been collected. The definitive study of this huge accumulation of material has not yet begun. Polaris is not the only Cepheid known to possess a variation of the systemic velocity of the pulsation. A study has recently been completed at Lick of the Cepheid S Sagittae, on the basis of published Michigan and Lick observations as well as more recent Mills material. In this star, the 8-day Cepheid velocity variation is accompanied by another variation with a period of 1.85 years. Several other bright Cepheids, either known or suspected to possess such long-period variations, are under observation with the Mills. An interesting single-line binary, now being studied by K. L. Franklin, is the K-type subgiant 26 Aquilae. The period is rather long—267 days—and the eccentricity of the orbit is very large, which is unusual for a late-type star. The eccentricity, is 0.83, a value exceeded at the present time by only two other binaries.

From Mills observations made in 1931-33, G. F. Paddock showed that the A2 supergiant α Cygni was subject to small, non- periodic fluctuations of radial velocity that seemed to have a tend- ency to favor cycle lengths of 0.2 and about 12 days. At Mount Wilson, R. F. Sanford has found that the radial velocity of the B8 supergiant β Orionis also fluctuates irregularly with a small range. In order to gain more information on this phenomenon, eight supergiants of different spectral types and luminosities and with variable velocities are currently under regular observation with the Mills. A number of other stars of individual interest are also being followed. Among these are the bright metallic-line star τ Ursae Majoris, the irregular variable R Coronae Borealis, the Be star Pleione, and the peculiar binary 3 Puppis. When the New Mills spectrograph was put into commission in 1903, the frame, slit, and collimator section of the original Mills were preserved, and around this structure a number of interchangeable prism and camera units of low to moderate dispersion have since been built. One of these combinations, pro- ducing a dispersion of 58 A/mm at Η γ, was used extensively in the 5.51-magnitude program on those stars of early type whose lines were either too poor or too few in the λ 4500 region to be observed with the three^prism dispersion. A considerable number of other radial velocity projects have been and are being carried out with this general-purpose spectrograph, with disper- sions varying from 37 to 200 A/mm. Three of these programs were extensive enough to deserve individual mention here, 1. R. J. Trumpler began, in 1921, an extensive study of the stars in galactic clusters. The measurement of velocities was only a part of this large program, which included 670 stars in 80 clusters north of declination —35°. Special attention was given to those clusters which are of importance in the study of the galactic rotation. The observing for this program was completed in 1947, and the final results are now being prepared for publication. 2. F. J. Neubauer's program for the determination of the radial velocities of faint B-type stars began in 1933. Earlier work by Neubauer and others had given results for O- and B-type stars brighter than about magnitude 6.5, and this program for the fainter objects extended to magnitude 9, and even fainter in some

areas. The results for 433 stars in the declination zone 0° to —23° have appeared. The necessary spectrograms for 127 stars in the zone 0° to +20° have been obtained and measured, but the velocities have not been published. Approximately half of the necessary plates for the remaining zone from +20° to +90°, in which 319 stars were to be observed, have been taken, but because of Neu- bauer's retirement from the staff in 1950, that portion of the program has not been completed. 3. J. H. Moore initiated in 1928 a program for a study of the solar motion and galactic rotation from the radial velocities of faint stars. The spectrograms of 820 stars of spectral types F through M were obtained, mainly by Paddock and Moore, in 1928-37, and were measured by Paddock. The publication of the results was delayed by the death of Moore in 1949, but the results have now been published, with a brief discussion by N. U. Mayall. It may be appropriate here to speculate regarding the possi- bilities of further improvement in the determination of stellar radial velocities. Any reduction of the probable errors of the best modern velocities by an order of magnitude will probably require an instrumental revolution of the sort brought about largely by Campbell, Wright, and their co-workers over fifty years ago. There can be little doubt that one of the serious obstacles standing in the road to higher accuracy is the general use of elec- tric arcs and sparks for the production of comparison spectra. The use of such a reference source in which small local wave-length displacements occur can hardly be recommended if substantial improvement in the present results is desired. In the conventional forms of spectrograph, the identity of the optical paths followed by star and comparison light cannot always be guaranteed, since the comparison light is made to pass through all parts of the optical system while the star light may not do so because of poor guiding or maladjustment of the optical elements. These and other difficulties encountered in the use of an emission spectrum for the comparison would be eliminated by the superposition of a system of fine absorption lines upon the stellar spectrum. One might, as sug- gested by Langley many years ago, work in the red and infrared and use atmospheric absorption lines as standards. The stability of the wave lengths of the atmospheric lines with varying zenith

distance has been demonstrated by St. John and Babcock, and the technique has actually been employed by Schwarzschild and Rich- ardson in their study of the radial velocities of solar granules. Ab- sorption cells of other gases might be used either in front of or in the spectrograph.2 Such a change in the method of introducing the comparison spectrum is only one of the possible ways of improving radial- velocity accuracy. The entire problem requires a thorough recon- sideration of all the processes involved in the determination of a velocity. In particular, the employment of a powerful high-dispersion spectrograph alone will not be a sufficient guaranty of vast improvement. The situation was well stated by Campbell, many years ago, in another connection : Higher dispersion should of course make for greater accuracy in the determination of radial velocities, but it is only one of many factors to be con- sidered. Quality of optical parts, stability of apparatus, perfection of adjustments, methods of observing and measurement, and other considera- tions too numerous to mention are each of importance; but efficiency and accuracy of results depend more on the correct balancing of the several factors than upon the excessive elaboration of any one of them.

When instruments and techniques of high velocity-resolution become available, it may turn out that the relationship between the apparent velocities from individual lines and the true radial velocity of the center of mass of the object under observation is, in some cases, an evasive thing. We are familiar with the discrep- ancies between the velocities derived from lines of different levels of excitation and ionization in stars of high and moderate lumi- nosity. Such effects are to be expected in deep atmospheres in which differential motions exist ; they will depend upon the lines and elements in question, as well as the wave length (through the variation of the continuous absorption coefficient). Collisional effects would be expected to enter into the line displacements in a manner that will vary from line to line in a complicated way. Moreover, the observed lines are built up from the contribu- tions of all the absorbing gases on one hemisphere of the star. Consequently, surface activity on the star will affect the shapes and positions of the lines. In particular, if the activity produces appreciable velocity components of a systematic sort in the line

of sight (for example, if the motion is predominantly radial), it will influence the line positions. Other effects can readily be imagined. In conclusion, it seems that even though some resolute pio- neer is able to force routine radial-velocity precision out into the next significant figure by instrumental development, it may be that fundamental difficulties at the other end of the line of sight will cause uncertainty in the interpretation of his results. I am indebted to Dr. W, H. Wright, director emeritus of the Lick Observatory, for valuable comments on the early spectroscopic work at Mount Hamilton.

1 These probable errors are those obtained from the agreement of the velocities from different plates, and hence do not include the effect of any systematic error of constant amount. 2 R. W. Wood has used solutions of rare earth salts to produce absorption lines in objective prism spectra, and has experimented to some extent with gases for the same purpose {Ap. /., 31, 376, 460, 1910). The liquids produce lines that probably are too wide for high-dispersion work of the type being discussed here.