The Atlas of Stellar Spectra and the accompanying outline have been prepared from the viewpoint of the practical stellar astronomer. Problems connected with the astrophysical interpretation of the spectral sequence are not touched on; as a consequence, emphasis is placed on ``ordinary'' stars. These are the stars most important statistically and the only ones suitable for large-scale investigations of galactic structure. The plan of the Atlas can be stated as follows:
a) To set up a classification system as precise as possible which can be extended to stars of the eighth to twelfth magnitude with good systematic accuracy. The system should be as closely correlated with color temperature (or color equivalent) as is possible. The criteria used for classification should be those which change most smoothly with color equivalent.
b) Such a system as described under (a) requires a classification according to stellar luminosity, that is, the system should be two-dimensional. We thus introduce a vertical spectral type, or luminosity class; then, for a normal star, the spectrum is uniquely located when a spectral type and a luminosity class are determined. The actual process of classification is carried out in the following manner: (1) an approximate spectral type is determined; (2) the luminosity class is determined; (3) by comparison with stars of similar luminosity an accurate spectral type is found.
As it may not be immediately apparent why an increase in accuracy in spectral classification is desirable, a short digression on some problems of stellar astronomy will be made.
The problem of stellar distribution in the most general sense does not require any spectroscopic data. Stars of all types and temperatures may be considered together, and some general features of the distribution of stars in the neighborhood of the sun can be found. For this purpose a certain frequency distribution of stellar luminosities must be assumed. This luminosity function has a large dispersion and must be varied with Galactic latitude. In addition, there are certain regional fluctuations in the frequency of stars of higher luminosity of classes B, A, and M.
As a result of these considerations (and because of difficulties with interstellar absorption) the general method has very definite limitations; the large dispersion of the luminosity function means we must have a large sample, and this in itself precludes detailed analyses of limited regions. In addition, there is evidence of clustering tendencies for stars of certain spectral types - a cluster or star cloud might be well marked for stars of type A, for example, and be not at all apparent from a general analysis of star counts.
There is, then, for certain kinds of problems a great advantage in the use of spectral types of the accuracy of the Henry Draper Catalogue. Consider, for example, the stars of classes B8-A0 as a group. The dispersion in luminosity is far less than in the case of the general luminosity function, and the space distribution of stars of this group can be determined with a correspondingly higher accuracy. In addition, we are able to correct for systematic errors due to interstellar absorption from observations of the color excesses of these stars. We have thus gained in two particulars: we have limited at one time the dispersion in luminosity and in normal color.
The further refinement of a two-dimensional classification makes possible an even greater reduction in the dispersion in absolute magnitude for groups of stars. The mean distance of a group of stars of the same spectral type and luminosity class can be determined with great precision, even when the group consists of a relatively small number of stars. Even for individual stars distances of good accuracy can be derived. A corresponding gain is made in problems concerned with intrinsic colors and interstellar absorption.
In the fifty-five prints which make up the accompanying atlas an attempt has been made to show most of the common kinds of stellar spectra observed in stars brighter than the eighth magnitude. The dispersion selected is intermediate between that used for very faint stars, where only a few spectral features are visible, and the larger-scale slit spectra which show a multitude of details. A sufficient number of lines and bands are visible to allow an accurate classification to be made, both by temperature and by luminosity equivalent, while the relatively low dispersion makes it possible to observe bright and faint stars in a uniform manner and avoids the possibility of appreciable systematic differences in their classification.
A small one-prism spectrograph attached to the 40-inch refractor was used to obtain the plates. The reduction of collimator to camera is about 7; this makes it possible to use a fairly wide slit and still have good definition in the resulting spectra. The spectrograph was designed by Dr. Van Biesbroeck and constructed in the observatory shop by Mr. Ridell. The camera lens was constructed by J.W. Fecker, according to the design of Dr. G.W. Moffitt. The usable spectral region on ordinary blue-sensitive plates is from the neighborhood of K to H ( 3920-4900).
The dispersion used (125 A per mm at H) is near the lower limit for the determination of spectral types and luminosities of high accuracy. The stars of types F5-M can be classified with fair accuracy on slit spectra of lower dispersion, but there is probably a definite decrease in precision if the dispersion is reduced much below 150 A per mm.
The lowest dispersion capable of giving high accuracy for objective-prism spectra is greater; the limit is probably near 100 A per mm. The minimum dispersion with which an entirely successful two-dimensional classification on objective-prism plates can be made is probably near 140 A per mm. This value was arrived at from a study of several plates of exquisite quality taken by Dr. J. Gallo, director of the Astronomical Observatory at Tacubaya, Mexico; for plates of ordinary good quality the limit is probably considerably higher.
The Atlas and the system it defines are to be taken as a sort of adaptation of work published at many observatories over the last fifty years. No claim is made for originality; the system and the criteria are those which have evolved from a great number of investigations. Specific references to individual investigations are, as a rule, not given.
By far the most important are those of the investigators at Harvard and Mount Wilson. The idea of a temperature classification is based on the work of Miss Maury and Miss Cannon at Harvard and of Sir Norman Lockyer. We owe to Adams the first complete investigation of luminosity effects in stellar spectra. If we add to this the work of Lindblad on cyanogen and the wings of the Balmer lines in early-type stars and the investigations of the late E.G. Williams, we have the great majority of the results on which the new classification is based. References to individual papers are given in the Handbuch der Astrophysik.
The present system depends, then, to a considerable extent on the work of these investigators, combined with data which were not available until recently. These data are of two kinds: accurate color equivalents for many of the brighter stars and accurate absolute magnitudes for a number of the same stars. These results have been used to define the system of classification more precisely, both in the temperature equivalents and in the luminosity class. The most important of the determinations of color equivalents for this purpose are the photoelectric colors of Bottlinger and of Stebbins and his collaborators and the spectrophotometric results of the Greenwich Observatory and those of Hall.
The absolute magnitudes used depend on a variety of investigations. There are the classical catalogue of trigonometric parallaxes of Schlesinger; the catalogue of dynamical parallaxes of Russell and Miss Moore; various cluster parallaxes, principally due to Trumpler; and, in the case of the stars of earlier class, parallaxes from interstellar line intensities and from the effects of galactic rotation.
Throughout the discussion emphasis will be laid on the ``normal'' stars. A number of peculiar, objects are noted; but the main aim of the investigation has been to make the classification of the more frequent, normal stars as precise as possible for the use of the general stellar astronomer. This investigation is not concerned with the astrophysical aspects of stellar spectra or with the spectra of the dwarfs of low luminosity. Relatively few of the latter are met with among stars brighter than the eighth magnitude, and their classification can be considered as a separate problem.
There appears to be, in a sense, a sort of indefiniteness connected with the determination of spectral type and luminosity from a simple inspection of a spectrogram. Nothing is measured; no quantitative value is put on any spectral feature. This indefiniteness is, however, only apparent. The observer makes his classification from a variety of considerations - the relative intensity of certain pairs of lines, the extension of the wings of the hydrogen lines, the intensity of a band - even a characteristic irregularity of a number of blended features in a certain spectral region. To make a quantitative measure of these diverse criteria is a difficult and unnecessary undertaking. In essence the process of classification is in recognizing similarities in the spectrogram being classified to certain standard spectra.
It is not necessary to make cephalic measures to identify a human face with certainty or to establish the race to which it belongs; a careful inspection integrates all features in a manner difficult to analyze by measures. The observer himself is not always conscious of all the bases for his conclusion. The operation of spectral classification is similar. The observer must use good judgment as to the definiteness with which the identification can be made from the features available; but good judgment is necessary in any case, whether the decision is made from the general appearance or from more objective measures.
The problem of a classification according to luminosity is a difficult one. In the first place, lines or blends which may be useful at one spectral type may be quite insensitive at another. In fact, some lines which show a positive absolute-magnitude effect for some spectral classes may show a negative one for others. This is true for certain lines of H, Sr II, and Ba II.
Besides the variation with spectral type, there is also a very marked change in appearance with the dispersion of the spectrograms used. Some of the most useful indicators of absolute magnitude are lines and blends which can be used only with low dispersion. The hydrogen lines, for example, show marked variations with absolute magnitude in spectra as early as B2 and B3 on plates of low dispersion; with higher dispersion the wings which contribute to the absolute-magnitude effect are not apparent to the eye, and the lines look about the same in giants and dwarfs. In stars of classes G2-K2 the intensity of the CN bands in the neighborhood of 4200 is one of the most important indicators of absolute magnitude. The band absorption has a different appearance on spectrograms of high and low dispersion, and it is doubtful whether high-dispersion plates show the luminosity effects of CN as well as those of low dispersion.
On the other hand, a considerable number of sensitive line ratios are available on high-dispersion spectra which cannot be used with lower dispersion. One of the most sensitive lines to absolute-magnitude differences for the F8-M stars is Ba II 4554; this line is too weak to be observed on low dispersion spectra. A number of the other ratios found by Adams to be sensitive indicators of absolute magnitude are also too weak to be used with low dispersion.
These considerations show that it is impossible to give definite numerical values for line ratios to define luminosity classes. It is not possible even to adopt certain criteria as standard, since different criteria may have to be used with different dispersion. In the Atlas some of the most useful features for luminosity classification have been indicated, but it should be emphasized that each dispersion has its own problems, and the investigator must find the features which suit his own dispersion best.
The luminosity classes are designated by Roman numerals; stars of class I are the supergiants, while those of class V are, in general, the main sequence. In the case of the B stars the main sequence is defined by stars of classes IV and V. For the stars of types F-K, class IV represents the subgiants and class III the normal giants. Stars of class II are intermediate in luminosity between the supergiants and ordinary giants.