Next Contents


Terminology: MK Diagram: The two-dimensional diagram relating MK spectral type and luminosity class.

The present Atlas was prepared to satisfy three requirements: (1) To furnish an improved version of the out-of-print Atlas of Stellar Spectra by Morgan, Keenan, and Kellman (University of Chicago Press, 1943) for stars earlier than the Sun; it thus complements the Keenan-McNeil Atlas for the stars of later type. (2) To decrease the classification "noise"; for the sake of determining spectral types and luminosity classes as precisely as possible from plates of relatively low dispersion. (3) By means of (2), to demonstrate how rich the prospects can be for classification in the future, by making use of conceptual improvements developed here, and incorporating them in what we describe as the MK-78 system. These processes, when taken together, will have the result of introducing a finer structure over that part of the MK Diagram occupied by stars earlier than the Sun.

The process described under (3) brings us to a major characteristic of the present Atlas: it is by no means a definitive work. In the early stages of its preparation there was a feeling that it might furnish all that needed to be said about its field for an indefinite period -- or even forever. As the work of preparing the Atlas plates progressed, however, it became clear that the field was being opened up increasingly to new fine structure in the MK Diagram, and to new localized third dimensions. These developments now indicate convincingly that there can be no such thing as a "definitive" spectral atlas - so long as a body of unrecognized, and uninvestigated, specimens exists in the observable region of the universe.

When we label the revised system defined by this Atlas as the MK-78 system, we are taking a step that requires further comment. The earlier MK "domain" has been the spectral region Hbeta-Ca II K; here, we extend the spectral range used to the neighborhood of lambda 3500 in the cases of certain categories of stars. The Atlas plates show the spectral range used, by the features marked. This increase in spectral range has been possible because of the remarkable quality of the low-dispersion spectrographs designed by a great astronomical instrumentalist, Dr. A. B. Meinel, the first director of the Kitt Peak National Observatory. These spectrographs were installed on several of the smaller reflectors at Kitt Peak and at the Cerro Tololo Inter-American Observatory, and were used by Dr. Abt to obtain all of the spectrograms of lower dispersion illustrated here. Such spectrograms, having dispersions in the range 63 A/mm-125 A/mm, when greatly widened, represent an almost ideal compromise between the factors of differing line-widths due to stellar rotation and the line visibility necessary for accurate spectral classification.


The basic framework of the MK Diagram has changed little between the 1943 MKK Atlas and the present one; in each case, we have a field defined by a certain number of standard stars - an autonomous, self-consistent field that requires no measures of any kind to establish its validity in use. In the earlier atlas, the procedure was generally followed of defining each box by the means of several stars that had been classified in it. Since these stars are not identical in spectral characteristics, such a procedure introduces a certain amount of noise in the definition of the box. In the present Atlas, we have adopted the procedure of defining each box by a single star; this procedure permits a higher order of discrimination than does the use of multiple standards for each box.

Let us consider the situation where we have obtained classification spectrograms for a number of unclassified specimens. With a single standard for each classification box, we find that we are able to classify more precisely than with multiple standards - both in the two-dimensional MK plane, and in the localized third dimensions discussed below. This is true because the single most important characteristic of the Revised MK Diagram is: Each box is defined by the patterns and intensities of spectral lines, bands and blends which exist in the spectrum of the standard star. The fundamental standard stars have been observed in a limited spectral range; but this principle holds for any spectral interval observed - as far as the MK system itself is concerned. It may develop later that a standard is found to show peculiarities in some spectral region, (x), other than the standard one. In such cases, the unsatisfactory standard in spectral region (x) can be classified in terms of other stars classified in the same spectral region which give self-consistent types in that region.

The convention of assumed constant spectral type for standards in all spectral wavelengths is the pivot on which the entire MK-78 system rests. It makes possible the precise localized definition of the MK plane in a manner that approaches the practicable limit in accuracy, and at the same time provides a method for re-classifying fundamental standards that show peculiarities in some wavelengths other than the standard spectral range. So, by this approach, we are able to consider the unobserved realm of specimens as candidates for precise classification by means of the MK boxes, through a simple, visual confrontation between the "unknown" specimen and the fundamental standards nearest to it in appearance - this confrontation to be made by use of spectrograms obtained with the same spectrograph, the same camera, the same photographic emulsion, and the same processing in development. And, in this confrontation, we use the language of lines, bands and blends - not the language of stellar atmospheres - this last being a separate process, to be carried out later.


Some time ago it appeared that a general third dimension might be applicable to the whole MK Diagram; this would have been a dimension depending on line intensity, such as the weak metallic-line stars in classes F - G. In the case of these weaker-lined stars, such a procedure is possible (but has not yet been carried out satisfactorily); however, it does not seem practicable for O and B stars - and probably not for the A stars. In place of a line-intensity parameter, we find categories of spectra which show various peculiarities and cannot be classified satisfactorily in the ordinary two dimensions. And so it is necessary to develop sharply localized third dimensions for such categories; examples are illustrated in the present Atlas. Some of these localized third dimensions are extremely small when projected on the MK plane (the MK - Hg stars), while others are found over a considerable range in spectral type (Sr II stars).


The nature of the fundamental standards that define the MK-78 system requires further comment; for this purpose we introduce the concept of the sui generis (literally: of its own kind) object. This can be described by the example of P Cygni (Plate 9). Here we have a complex emission and absorption line spectrum (the stellar radial velocity represented by the emission lines, approximately) and the absorption lines originating in a dense expanding shell. Now the members of the P Cygni class of stellar spectra resemble the prototype in some respects; but it seems unlikely that any member of this class will be found to have identical spectral characteristics with P Cygni itself. In addition, the spectrum of the latter varies with time; so that in this case we have the complication of intrinsic variations in the prototype spectrum. If we wished to go to the possible limit of discrimination in the use of such a standard, we would have to specify the date of the standard spectrograms being used. Here we have an obvious case of unsuitability for use as a standard - so long as we wish to make use of all features for classification; but, in spite of this, we can say that P Cygni is a sui generis object.

The situation is much more favorable in the case of the normal standards illustrated in this Atlas; there can be variations with time in the appearance of Halpha in the Ia supergiants of early type, but the other lines in these objects are fairly constant in appearance; in the case of the other fundamental standards of normal stars there is not likely to be difficulty from variable features. And so, when we apply the label sui generis to these stars we are emphasizing the fixed nature of the standards defining the boxes in the MK Diagram.


There are three areas of incompleteness in the present Atlas: (1) All stars illustrated in the Atlas are of Population I with the exception of HD 22879 (Plate 31), which is of Population II; two other stars, 31 Com (Plate 32) and v Peg (Plate 30) are intermediate between Population I and extreme Population II; it can be seen, therefore, that the Atlas deals effectively only with the classification of Population I stars. (2) Many of the boxes of the MK Diagram do not contain defining standard stars, and the local third dimensions are incompletely identified - or not yet even suggested. (3) The problem of illustrating the standard stars in other wavelength regions remains a major task for the future. In addition to these omissions, some of the standard stars adopted can only be observed in one hemisphere. We regret this, and hope to be able to furnish a more nearly complete list of standards at a later time.

In the light of this incompleteness, it should be emphasized that the present Atlas is in effect a report on work in progress; it develops a methodology of power and stability, and gives numerous examples of its use in spectral classification. We can look forward to a time in the future when three necessary steps have been taken. (1) To develop a unitary spectral classification for Population II stars. (2) To populate as many as possible of the boxes of the MK Diagram with a defining star. (3) To carry as far as justified the splitting of boxes (as the A0 V box has been split in Plate 18), for the sake of determining more precise luminosities and distances - and for adding the greatest amount of justified detail to the MK Diagram.


It can be seen from the preceding section that the present Atlas is not adequate to deal with the classification of stellar populations like globular clusters and the Galactic halo population. It is also not suited to precise classification of stars in extragalactic systems which are at differing evolutionary stages from our own Galaxy. In the general case of classification of stars in other galaxies, the most crucial requirement is to retain as much particular information as possible. To guarantee this, we should not take for granted any similarities with stars in the solar neighborhood, and should: (1) Create a new, self-consistent classification from the stellar spectra available in each object, using a notation completely different from that of the MK system; (2) when this is finished, we should confront the resulting classifications with the MK classifications as illustrated in the present Atlas; (3) from this confrontation, we should then assess the differences and similarities between the stars in the other galaxy and those in our own. The above procedure can be used only if a considerable number of stars have been observed in the particular galaxy; when only small numbers of spectrograms are available, they must be referred to the MK standards, and their similarities and differences noted.


A classical example of the approach to spectral classification in any new spectral region is furnished by Bidelman's recent paper on spectral classification from Copernicus ultraviolet data (Highlights of Astronomy, Vol. 4, Part II, 355-359, 1977). The approach described by Bidelman is a model for a procedure that guarantees preservation of the maximum justified information - and for a careful, descriptive comparison with spectral types from the normal MK spectral region.


We present herewith a revised structure for the classification of stars earlier than the Sun. The structure has greater precision than that of the MKK Atlas of 1943, partly because of the use of single standard-stars to define each classification box. There are still important problems awaiting attention: (1) The classification of very broad-lined early-type stars is intrinsically less exact than that of the narrower-lined objects; what is needed here is a frame of reference consisting entirely of very broad-lined stars, with this frame fitted to the standard MK-78 boxes by the use of low-dispersion (~ 150 A/mm), fine-grain, high-contrast spectrograms, which would minimize differences in appearance between n and s stars. (2) The MAT Atlas is applicable only to Population I stars; still needed is a general classification that would include Population II stars. (3) The calibration of the MK Diagram for the determination of physical parameters will have to await classification of large numbers of stars on the MK-78 system.

One of us (W.W.M.) wishes to make the following acknowledgments: for the support of a series of grants from the National Science Foundation, which made possible both the preparatory work and the publication of the Atlas: for discussions with Dr. Lewis M. Hobbs and with Dr. Dimitri Mihalas concerning the new Atlas; to Dr. Martin F. McCarthy, S.J. for a remarkable formulation of the conceptual structure underlying the present Atlas; to Constance Siebert for continuing help.

We wish to acknowledge here the outstanding quality of the printing of the thirty-two Atlas plates carried out by Tru Line Lithographing, Inc. of Sturtevant, Wisconsin. A careful comparison of the printed plates with the original photographic prints establishes the remarkable faithfulness with which the latter have been reproduced. The portfolios were constructed by Troy Plastics Corp. of Burlington, Wisconsin.

Next Contents