STARS, SPECTRAL CLASSIFICATION JANET ROUNTREE Spectral classification is a branch of astronomy that has the goal of arranging the spectra of the stars into meaningful and self-consistent groups. Several purposes are served by this procedure: (1) a detailed study of the prototype star in each group is, in principle, sufficient for an understanding of all of the members, thus avoiding the necessity of studying each star separately; (2) the astrophysical parameters (effective temperature and luminosity, and eventually the mass and radius) of the stars can be derived through a suitable calibration of the spectral types; (3) peculiar objects can be isolated from the groups of "normal" stars for further study; (4) the distribution of stars in various groups can be used for statistical studies of stellar populations and stellar evolution. HISTORY The continuous spectrum of the Sun, our nearest star, was discovered by Isaac Newton in 1666, in the course of his optics experiments. But the absorption lines, on which most spectral classification systems are based, were first seen in 1802 by William H. Wollaston, and first described in detail some years later by Josef von Fraunhofer, in whose honor they are still called Fraunhofer lines. The first extensive classification program using stellar absorption spectra was carried out in the 1860s by Father Angelo Secchi, the observatory director at the Jesuits' Roman College. After making visual observations of over 4000 stars, Secchi concluded that the great majority of them fell into four main spectral types, which he called I-IV and which we would now identify as spectra dominated by (I) hydrogen lines, (II) metallic lines, (III) titanium oxide bands, and (IV) molecular carbon bands. Secchi's classification remained in general use until it was supplanted by the Harvard, or Henry Draper, system. The system developed at Harvard College Observatory used letters of the alphabet rather than Roman numerals to denote the spectral types; originally, the stellar spectra were arranged from A to Z in order of "increasing complexity." The series was later rearranged in order of decreasing color temperature, from blue to red, and extraneous types were discarded. The types that were retained, with their principal characteristics in parentheses here, form the well-known spectral sequence: O (ionized helium lines); B (neutral helium lines); A (predominantly hydrogen lines); F (both hydrogen and metallic lines); G (strong CH band); K (similar to G, but with more prominent metallic lines and weaker hydrogen); M (titanium oxide bands); R and N (molecular carbon bands); and S (zirconium oxide bands). The position of a stellar spectrum along this sequence is indicated by a letter and a number-for example, O9,B5, G2. Annie Jump Cannon classified some 325,000 objective prism spectra on this system. Her work was published as the Henry Draper Catalogue and the Henry Draper Extension, beginning in 1918. The Harvard spectral-type sequence is well correlated with the stars' effective temperature. But it became increasingly apparent in the 1920s and 1930s that a second dimension, correlated with stellar luminosity, was needed for a complete classification system. Work along this line was inspired partly by the development of the Hertzsprung-Russell diagram, which showed that all the common stars are restricted to several small areas in the temperature-luminosity plane. Two schools of classifiers at the Mt. Wilson Observatory and the Dominion Astrophysical Observatory attempted to graft a second dimension onto the Harvard spectral sequence by deriving an absolute magnitude for each star from a set of calibrated line ratios. But the two-dimensional system that eventually received broad acceptance was the MK classification, developed by William W. Morgan and Philip C. Keenan. In the MK system, first published in 1941, each star is assigned a spectral type (sometimes called a temperature type) in a notation similar to the Harvard system, and a "luminosity class" on a scale of I-V, where I is the most luminous. Members of the different luminosity classes are commonly called dwarfs (class V, the main sequence stars), giants (class III), and supergiants (class I). (Ia and Ib are the more luminous and less luminous supergiants, respectively; II are luminous giants and IV are subgiants.] Stars hotter than the Sun are sometimes called early-type stars, and those cooler than the Sun are late-type stars. The assignment of spectral type and luminosity class is made by comparing the spectrum of the "unknown" star with a predetermined set of standard stars, which serve as paradigms. PRINCIPLES AND PROCEDURE The principal difference between the MK classification and the systems that preceded it is not the use of two dimensions rather than one. The MK system introduced two unique principles into spectral classification: (1) Only information obtained from a visual inspection of the stellar spectrum is used in the classification. (2) The system is completely defined by its set of standard stars. The first principle means that knowledge of a star's color, distance, variability, cluster membership, apparent brightness, or any astrophysical parameter is not allowed to influence the classification-the spectral type is confronted with such information only after the fact. The second principle implies that the spectral type is independent of the specific criteria used to compare the "unknown" stars with the standards-all the information available in the observed spectrum is taken into account. The result of a strict adherence to these two principles is that spectral types are independent of any astrophysical calibration in terms of temperature, luminosity, mass, age, etc. - calibrations that tend to change as the state of astrophysical theory advances. Spectral types are also largely independent of the specific instrumentation used to obtain the spectra. Thus the MK system is stable and essentially invariant with time, except for continuing small refinements, and this fact has enabled it to retain its usefulness for the half-century since its inception. In performing a spectral classification program, the first step is to obtain spectra of as many as possible of the standard stars that occupy cells in that portion of the two-dimensional array (spectral type and luminosity class) that is likely to be of interest. The program stars are then observed with exactly the same instrumentation and processing techniques as the standards, where "processing" includes development procedures for photographic plates, or mathematical manipulation for digital data. Each "unknown" is then closely compared with the three or four standards most similar to it, and is assigned to the cell containing the standard that it most closely resembles. The two spectra will rarely be identical, but if the program star differs from the standard in some important way, it may be labeled "peculiar." If many stars exhibit the same peculiarity, but to different degrees, they may define a local third dimension in the classification system. Only after the classification is complete are the program stars assigned effective temperatures, absolute magnitudes, masses, radii, and so forth, which may be based on a prior calibration of the cells in the classification array. Figure 1 shows a typical set of photographic spectra taken at classification dispersion. The wavelength range covered is approximately 3800-4900 *. RESULTS Over 100,000 stars have been classified on the MK system, which is the only purely spectral classification system in wide use today (for photometric classification systems, see the entry on Magnitude Scales and Photometric Systems). In addition, the entire Henry Draper Catalogue is currently being reclassified on the MK system. The classification system has been calibrated in terms of effective temperature and luminosity, from which stellar distances, masses, radii, and ages can be derived with the adoption of suitable models. Thus the global astrophysical parameters of a large number of stars are known through their spectral classification, always assuming the accuracy of the calibration and the models. One of the first applications of the MK system was the study of the spatial distribution of the O and B stars near the Sun, leading to the discovery of at least two galactic spiral arms in the solar neighborhood. This was the first definite evidence of the spiral structure of the Galaxy. Combined with radial velocities and proper motions, distances derived from spectral types (spectroscopic parallaxes) have also been used to study the pattern of space motions of the early-type stars around the Sun, a pattern produced by a combination of differential rotation around the galactic center and local expansion. Many types of spectrally peculiar stars have been isolated by means of spectral classification techniques. In-depth study of these groups of objects often leads to new insights in the areas of stellar atmospheres, interiors, and evolution. Among the types of stars identified in this way are the metallic-line stars (Am, mercury-manganese, and helium-weak B stars); the peculiar or "magnetic" A and B stars (having enhanced absorption lines due to silicon or rare-earth elements, and later found to exhibit strong magnetic fields); and "shell" stars, whose extended atmospheres produce absorption lines with very narrow profiles. Other groups of objects that do not fit conveniently into a two-dimensional system include emission-line stars, Wolf-Rayet stars, subdwarfs, and white dwarfs. These objects have themselves become the subject of extensive spectral classification work. CURRENT PROBLEMS IN SPECTRAL CLASSIFICATION As better and better photographic plate material has become available, the Mk system has frequently been refined, without being basically altered. Thus new "fractional" spectral types (like B1.5) have been introduced, and luminosity classes have sometimes been split in two (e.g., B2 IV-V; G8 IIIa, IIIab, and IIIb; A0 Va and A0 Vb). But the question often arises as to whether two dimensions are sufficient to classify the spectra of all the normal stars. Because the location of a star in the theoretical Hertzsprung-Russell diagram is uniquely determined by mass, age, and chemical composition, there is some reason to expect that a third parameter (besides temperature and luminosity) may be involved in the empirical scheme as well. A local third dimension has been added to the MK system in some restricted areas of the HR diagram; most often, the third parameter is related to chemical composition. For stars of spectral type F8 to M, Keenan introduced a set of composition indices that describe the strengths of certain metallic absorption lines relative to those in the standard stars, which are assumed to be of solar composition. On the other hand, when classifying stars drawn from an entirely different population from that of the solar neighborhood-for example, stars in the galactic halo or in other galaxies-a new set of standards needs to be defined, and there may not be a one-to-one correspondence with the usual MK spectral types. New observing techniques present new problems and new opportunities for the spectral classifier. Many observers now routinely use electronic detectors instead of photographic plates, even in the visual-wavelength region, because electronic detectors have greater sensitivity and linearity of response. Although it is usually possible to use the digital output of these detectors to simulate a photographic spectrum, such a procedure is cumbersome and generally unnecessary. Spectral classification can be performed directly on plots of normalized intensity versus wavelength, provided that the usual precautions are taken: The standard spectra must be obtained and processed in the same way as the program stars, and all spectra should have the same signal-to-noise ratio. The same spectral lines that have been found to be temperature or luminosity sensitive on photographic plates can be used as classification criteria on tracings as well, but the classifier must become accustomed to their appearance in the new medium. In particular, the eye tends to respond to a line's equivalent width on the photographic plate, whereas on a tracing the central depth is the most striking feature. The development of efficient detectors for near-and far-infrared radiation, and of satellite-borne instruments, has made it possible to consider the use of spectral classification techniques in wavelength regions outside the original visual range, thus observing the stars at the wavelengths where they emit most of their energy. The vacuum ultraviolet, accessible only by spacecraft, may prove to be the most appropriate region for the classification of hot, blue stars, whereas the infrared may be useful not only for late-type stars but also for distant, reddened O and B stars. In principle, a set of standard stars can be defined for any desired two- or n-dimensional matrix of spectral types. If the MK standards are selected, the resulting system will approximate the MK system. If not, an entirely new system may be created. In the ultraviolet, the strongest lines in the spectra of early-type stars tend to be those of highly ionized carbon, silicon, and nitrogen, which are mostly formed in the stellar wind and which therefore are likely to be governed by parameters other than effective temperature and surface gravity. However, there are enough photospheric lines even in this region to perform a normal, two-dimensional classification. Figure 2 shows a set of ultraviolet spectra observed by the International Ultraviolet Explorer satellite, and displayed in a form resembling photographic data; although these spectra could be used for classification, it is more practical to display the data in the form of tracings. Similarly, in the infrared, the presence of extended dust shells around some stars requires care in the choice of classification criteria. But the systematic study of stellar spectra at these wavelengths, which is still in its infancy, offers the possibility of gaining information about new portions of the atmospheres of normal stars. Additional Reading Garrison, R.F., ed.(1974). The MK Process and Stellar Classification. David Dunlap Observatory, Toronto. Jaschek, C. and Jaschek, M.(1987). The Classification of Stars. Cambridge University Press, Cambridge. Kaler, J.B.(1989). Stars and Their Spectra: An Introduction to the spectral Sequence. Cambridge University Press, Cambridge. McCarthy, M.F., Philip, A.G.D., and Coyne, G.V., eds.(1979). Spectral Classification of the Future. Vatican Observatory, Vatican City. Morgan, W.W. and Keenan, P.C.(1973). Spectral classification. Ann. Rev. Astron. Ap. 11 29.