Annu. Rev. Astron. Astrophys. 1999. 37: 445-486 Copyright © 1999 by Annual Reviews. All rights reserved

3.2. The Chemical Composition Connection

A most significant development in the decade of the 1950s was the emergence of what is now perhaps the most important idea in stellar evolution and whose ramifications cover parts of cosmology as well. The mantra, summarized from its various versions, would read:

"The chemical elements heavier than H and He are made in stellar interiors and are spread throughout the interstellar medium by supernovae explosions. Because this is a continuous process, the metallicity of all stars subsequently made from the ISM will increase with time. Hence, on average, the oldest stars will be metal deficient compared with newly forming stars. The age-metallicity relation is expected to vary from place to place. Study of metallicity as a function of age, kinematics, and position can give clues to the problem of the formation of the Galaxy in particular, and the chemical evolution of galaxies in general, at all redshifts."

The opening campaign to reach this advanced form of the mantra was first set out by Hoyle (1946, 1954 in his Figure 1). The Hoyle program was, of course, brought to its classic maturity in the monumental paper by Burbidge et al (1957, B²FH), from which the concept took flight. A proper history of how the ideas of varying metallicity and its implications, as developed by the early 1960s, requires a separate article. Here, we simply give the barest of the historical highlights of how, within less than two decades, the subject of the chemical evolution of cosmic objects grew into the present maelstrom. For the purposes here, the theme is to show only why our early main-sequence fits were incorrect. The history in greater detail has been given elsewhere (Sandage 1986, Sections 4 and 6).

The early study that began the study of variations in the chemical abundances of stars was by Chamberlain and Aller (1951). They analyzed Mount Wilson 100-inch high dispersion Coude spectra of the extreme subdwarfs HD 140283 and HD 19445, finding metallicity deficiencies of factors between 6 and 10 relative to the sun.

The famous underground story, ⁽²⁾ often told, is that the referee of the paper was so astounded that he disbelieved the factor of approximately 100 originally derived by Chamberlain and Aller.

It is said that the referee suggested that their stellar temperatures should be adjusted upward so as to decrease the metallicity deficiencies. Chamberlain and Aller did this partially, and finally published metallicities of [Fe/H] = -0.8 and -1.0 for HD19445 and HD 140283 respectively (their Table 8). But even they were astounded at this result. They wrote:

"The one possible undesirable factor in our interpretation is the prediction of abnormally small amounts of Ca and Fe. As Greenstein suggests, the observed deficiency of some elements could possibly be caused by an excess second ionization - i.e., much higher ionization than that predicted by the Saha formula."

Hence, even as late as 1951 there was no clear idea of the extent (or even the existence) of gross deviations of the chemical abundances relative to the sun. However, the Chamberlain & Aller paper was the beginning of the current remarkable era of abundance analyses (see McWilliam 1997 for a review). Their paper was followed by Burbidge & Burbidge (1956) where the abundance variations, taken by them to be real, began to be put in the context of chemical evolution of the galaxy. The later detailed analysis of HD 140283 by Baschek (1959) using high dispersion plates taken in 1957 with the Mount Wilson 100-inch Coude spectrograph by Unsold, was definitive and gave [Fe/H] = -2.32. The value was closely confirmed by Aller & Greenstein (1960) using new, highly widened plates taken both with the Mount Wilson and the Palomar Coude spectrographs.

In the meantime, Roman (1954) had made the central discovery of the general ultraviolet excess in the spectral energy distribution of particular high-velocity stars. In a prescient paper she had isolated a homogeneous group of 17 "ultra" high-velocity stars that had ultraviolet excesses (relative to the UBV two-color locus defined by stars of low velocity with the same B-V colors) reaching 0.2 mag in delta U-B. The excess UV radiation was suggested by Stromgren (unpublished) to be due to lower Fraunhofer line absorption shortward of 4000 A in the spectrum, presumably due to lower metallicity. This was shown to be the case for less extreme metal deficiencies (Schwarzschild et al 1955). A model with the introduction of "blanketing lines" in the UBV two-color diagram (Sandage & Eggen 1959) generalized the case. Roman (1955) greatly added to the study of the UV excess for high velocity stars for which she had made a catalog that contained both photometric and kinematic data.

The first calibration of the UV excess in terms of the [Fe/H] metallicity deficiency was made by Wallerstein & Carlson (1960). They used the small number of abundance analyses that were just coming from the stellar abundance program of Greenstein and his team of postdoctoral fellows, using both the Mount Wilson and the Palomar Coude spectrographs. Wallerstein (1962) made a later calibration of the UV excess/metallicity relation that has been widely used because of the much larger number of calibrating stars it contains. Examples of the early literature are the papers by Helfer et al (1959), Wallerstein (1962), Wallerstein et al (1963), Wallerstein & Greenstein (1964), Wallerstein & Helfer (1966).

The slope of the blanketing lines in the UBV two-color diagram was calibrated by Wildey et al (1962) based on high dispersion spectrograms obtained by the Burbidges with the Mount Wilson 100-inch Coude. A table of blanketing-line slopes and the resulting corrections to colors was made using the Wallerstein (1962) calibration of the excess itself. Later, account was taken (Sandage 1969) of the effect of the guillotine on the observed excess due to the convergence of the slope values of the blanketing and intrinsic lines in the U-B, B-V diagram for subdwarfs redder than B-V = 0.7. A normalized UV excess was defined, as reduced to B-V = 0.6, to account for the guillotine. Subsequent calibrations of the normalized UV excess include those of Carney (1979), Cameron (1985).

After Roman's discovery of the UV excess in the 17 field high-velocity subdwarfs, the same type of excess was sought in globular cluster stars. It was soon found in the giant stars of the cluster NGC 4147 (Sandage & Walker 1955), and soon thereafter in M3 (Johnson & Sandage 1956), M13 (Baum et al 1959), M5 (Arp 1962), and M15 and M92 (Sandage 1970a). This closed the last test for the identification of the field high-velocity F and G stars with stars in globular clusters that had been suggested by Baade (1944a), based on Oort's (1926) PhD thesis on high-velocity stars.

As a consequence, the work also established the great age of the field subdwarfs because, from the age-dating of the globular clusters (Sandage & Schwarzschild 1952) it was known that these clusters, as a whole, were as old as the first stars formed in the galaxy. By extension, this identification led naturally to the model of the formation of the galaxy through collapse from a larger volume (Eggen et al 1962, Sandage 1990a).

With the knowledge gained that there is a large difference in the metallicity between globular cluster stars and the common field F and G stars of the population I, we became aware, but only gradually, that our initial fit of the globular cluster main sequence to the nearby trigonometric field stars discussed earlier was not correct. The suspicion was that the MSs of metal-poor subdwarfs were fainter than the population I MS by a progressive amount that depends on [Fe/H].

In hindsight, it turns out that the first evidence of that fact was in fact the position of HD 19445 and HD 140283 in the HR diagram of Adams et al (1935) where these two stars were among the six "intermediate white dwarfs" found in the Mount Wilson spectroscopic catalog of 4367 spectroscopic parallaxes (look carefully at their diagram: the six outriders at spectral type near A5 and absolute magnitude near +5 are easily missed).

The second reason to suspect an offset between the population I and II MSs was theoretical. Stromgren (1952) had set out homology relations from stellar interior theory that showed that MS stars of low metal abundance are expected to have fainter MS positions than those of higher metallicity. Although now a well-known proven proposition, the premise had not been proved observationally or indeed by calculated stellar models (rather than homology transformations) even in the mid 1950s. The first calculated model with zero metals was by Reiz (1954), where his model star was indeed a star lying fainter than the high metallicity MS (i.e. it was a subdwarf).

Recall that all we had to work with in the 1950s were stellar parallaxes that were sufficiently inaccurate for the few very high velocity field stars that the observational test for the existence of a faint subdwarf sequence was uncertain. However, analyzing the available data on less extreme stars that had only mild subdwarf characteristics (intermediate velocities and small but definite UV excesses), we could provide an observational proof of sufficient weight that was almost convincing that a relation does exist between UV excess and depression of the MS from that of the population I high metal stars (Sandage & Eggen 1959, Figs. 1 and 2).

However, the extreme case for the very low metallicity stars, similar to those in globular cluster stars, was only given by the moving group parallax of the Groombridge 1830 group (Eggen & Sandage 1959), a result not widely accepted at the time.

The problem of the MS position was discussed again three years later using the totality of known trigonometric parallaxes and the rapidly growing data on UBV photometry of the trigonometric parallax stars. By combining these data we could show that the departure from the main sequence of F and G stars was a progressive function of [Fe/H] (Eggen & Sandage 1962), reaching the order of 1 mag departure faintward for subdwarfs with [Fe/H] = -2. From this calibration, a new MS-fitting procedure could be used, as illustrated for the fit to M13 in Figure 10 of the cited paper (Eggen & Sandage 1962). Later fits were made (Sandage 1970a), the most complete being of 19 cluster diagrams to the continuum of MS positions as a function of [Fe/H] (Sandage & Cacciari 1990).

With this development, the RR Lyrae stars could now be calibrated by the MS fits, bringing to a close, in principle, that phase of Baade's Palomar program originally proposed in 1948. The most recent use of the method is reviewed by Reid (1999), based on post-Hipparcos analyses of the relevant new trigonometric parallaxes.

² The nearly universal opinion of the time was that there was a fixed chemical composition of all stars throughout the universe, a kind of cosmic palimpsest containing some hidden initial, but universal, mystery relating to origins and evolution. Back.