2.2. Doubts and Confirmation

After the original BE discovery paper (Baldwin 1977a), doubts were raised about its nature, reality, or utility for cosmological studies. Although we now regard the BE as well established, there are several effects that must be considered in any analysis of the BE:

1. Selection Effects. Jones & Jones (1980) pointed out that faint quasars with weak emission lines would be difficult to observe and therefore cause a selection effect that could artificially enhance the apparent BE. Osmer (1980) showed that the slitless spectrum technique would systematically favor the discovery of strong-lined quasars at faint magnitudes in a manner that was also in accord with the apparent BE. More recently, Yuan et al. (1998) have shown that different dispersions in observational parameters can artificially produce luminosity correlations in data sets.

2. Variability. Murdoch (1983) discussed how the variability of flat-spectrum quasars could account for much of the BE. For example, if the line emission stayed constant while the continuum decreased in brightness, the equivalent width would appear large when the object was fainter. In reality, both the line and continuum vary with time, which makes the situation more complicated. In any case, the effect of variability must be considered in analyses of the BE.

3. Population Effects. Baldwin himself (1977a) noted in the BE discovery paper that the flat-radio-spectrum quasars showed the tightest correlation in the log L(1450 Å) - log W(C IV) diagram. A long-standing question has been whether other types of quasars exhibit the same BE as do the flat-spectrum objects. This question bears on the physical nature of the different types of quasars and AGNs as well as their usefulness as cosmological indicators.

4. Evolution. If the physical nature of quasars varies with cosmic time, then the BE observed at one redshift might not be appropriate for quasars at a different redshift. The separation of evolutionary (redshift) effects from luminosity effects has been difficult in the flux-limited samples studied to date. The steep rise of quasar number counts with magnitude causes most of them to appear near the magnitude limit. This in turn will introduce a correlation of redshift and intrinsic luminosity into the data set in that the most luminous quasars will be those with the largest redshifts.

Now we will consider how all these effects have played out in subsequent studies of the BE.

Baldwin, Wampler & Gaskell (1989, BWG), building on earlier work, addressed the problem of selection effects for radio-loud quasars by obtaining spectra of all objects in a well-defined radio sample, so that objects with weak lines would not be omitted. For optically selected quasars, they used quasars selected by the ultraviolet excess (UVX) technique, which was based on quasars (primarily those with z < 2.2) being much brighter in the ultraviolet than most stars and was less dependent on emission-line properties than the slitless spectrum technique. ⁽¹⁾ BWG demonstrated that the BE was indeed present in radio-loud quasars. They also pointed out that their data could not establish differences between the BE in the radio- and UVX-selected quasars because of the differences in the luminosity limits of the samples.

A next important step in confirming the reality of the BE was made by Kinney, Rivolo, & Koratkar (1990, KRK), who used IUE archival data for quasars and Seyfert galaxies to obtain a range of 10⁷ in continuum luminosity. Their data showed conclusive evidence for the BE being a real, physical effect (Fig. 2, upper). Furthermore, they were able to address the variability problem and show that it caused much of the scatter in their overall set of data. When they averaged multiple observations of different objects into single points, the scatter was much reduced (Fig. 2, lower).

Figure 2. The Kinney et al. (1990) results for the C IV BE. The upper panel shows results for the full IUE data set, supplemented by ground-based measurements by Sargent et al. (1989) and by BWG. The lower panel shows the same plot for a subset of data selected for high signal-to-noise ratio, with repeated observations averaged into a single point per object.

Thus, by 1990, there was convincing evidence for the reality of the BE as well as for the importance of selection effects and variability, which is now known as the intrinsic BE. The intrinsic BE is an important subject in itself, and we discuss it in more detail below in Section 5.

The question of population effects on the BE has been more difficult to settle, with different investigators finding different results. One problem is that the intrinsic scatter in many of the properties of quasar and AGN spectra can mask effects like the BE, especially in small samples or ones that do not cover a sufficiently large range in luminosity, as the work of BWG and KRK made clear. A further complication when comparing radio-loud and radio-quiet systems is the necessity of working with samples that span the same range in luminosity, a point emphasized by BWG. Sargent et al. (1989) and Steidel & Sargent (1991) compared the behavior of radio-loud and radio-quiet sources matched in luminosity, and found indications supporting Baldwin's claim of a stronger BE correlation for radio-loud QSOs. ⁽²⁾ However, a visual scrutiny of their results suggests that this is a question that would benefit from the study of larger samples spanning a broader luminosity range; the significance of correlations in the published plots often rests on the location of a very small number of points.

Zamorani et al. (1992) combined all data sets available at that time for optically selected quasars to make an improved investigation of the possible differences between radio- and optically-selected quasars. From data on the C IV line in 316 quasars, they found the presence of a BE, but with a slope about half the value of BWG for radio-selected quasars. Their slope was in good agreement with the KRK results, but the normalization was about half as large. They considered that variability of the quasars in the BWG sample with the strongest CIV emission was an important factor causing the difference between the radio and optically selected quasars. In the end, however, they did not think it possible to conclude there was a physical difference between the two classes of objects. A difficulty with this type of analysis is the inherent inhomogeneity in the data when equivalent widths are compiled from the literature; a large scatter in observed equivalent widths can result artificially from the disparate measurement methods employed by different researchers. Regrettably, we still appear to be in the situation where a comparison based on even larger and more carefully selected samples is needed.

Finally, the question of evolutionary effects in quasar spectra continues to be important and is just now becoming feasible to address. Here the difficulty has been in obtaining samples and especially follow-up slit spectroscopy of faint quasars/AGNs at high redshift. For samples with limiting magnitudes of, say, magnitude 21, objects at z = 3 will have M < -25 (H₀ = 75, q₀ = 0.5), far more luminous than nearby AGNs. A key project for large ground-based telescopes will be the discovery and spectroscopy of quasars with z 3 and m 25 - with such data a systematic and significant investigation of the evolutionary properties of quasar spectra will become possible.

² The radio-loud objects studied by Steidel & Sargent (1991) were primarily flat-spectrum (core-dominated) systems, although they found no significant differences between the behavior of flat- and steep-spectrum quasars within their sample. Back.