THE EMISSION-LINE PROPERTIES OF LOW-REDSHIFT QUASI-STELLAR OBJECTS

Spectra covering the region lambda-lambda-4300-5700 have been obtained of all 87 QSOs in the BQS catalog having redshifts less than 0.5. An empirical technique which allows the measurement and subtraction of the many Fe II lines in this region has been developed and applied to these spectra. Measurements of the strengths of H-beta, [0 III]lambda-5007, and He II-lambda-4686, and a four-dimensional parameterization of the H-beta profile have been combined with optical, radio, and X-ray continuum information from the literature to try to understand how these properties are related. An analysis including the complete correlation matrix of the measured and compiled properties and a principal component analysis reveal the following results: Most of the variance is connected to two sets of correlations, the first being a strong anticorrelation between measures of Fe II and [ O III]. The asymmetry and width of the H-beta line are also associated with this eigenvector. The second group of correlations involves the optical luminosity, the strength of He II-lambda-4686, and alpha(ox) The next three eigenvectors are each dominated by a single property: H-beta equivalent width, the shape of the H-beta line, and the shift of the peak of the H-beta line from the systemic velocity. We conclude (1) the dominant source of variation in the observed properties of low redshift QSOs is a physical parameter which balances Fe ii excitation against the illumination of the narrow line region. We argue that this property (and the observed properties of QSOs in general) is not driven by external orientation, i.e., our viewing angle, for three reasons. First, the dominant eigenvector is highly correlated with [0 III] luminosity, an isotropic property. Second, the steep-spectrum radio sources fall on logical extrapolations of relations defined by the radio-quiet objects, arguing against a distinction in the basic physical parameters driving the observables in the two types of objects. Third, the Fe ii lines have the same widths as the H-beta lines, suggesting that they must arise in the same region and have the same degree of anisotropy, a finding inconsistent with the currently popular AGN unification models. Our best guess is that this dominant parameter is related to the fraction of the Eddington luminosity at which the object is emitting. (2) The second parameter is explained as a correlation between luminosity and the slope of the ionizing continuum, in the sense that lower luminosity objects have harder spectra. ( 3) A comparison of subsamples defined by their radio properties suggests that radio-quiet and radio-loud objects cannot be parallel sequences because there is a large deficit of radio-quiet analogs to the steep-spectrum QSOs. Instead, radio-loud objects should be thought of as the extreme end of some (unknown) property which also determines Fe II and [ 0 III] strength and the asymmetry of the H-beta line. The one significant distinction is that in radio-loud objects the peak of H-beta is systematically shifted to the red, while for radio-quiet objects, there are equal numbers of red- and blueshifts. (4) In confirmation of Sulentic et al. we find no correlation between H-beta asymmetry and centroid shift, indicating that a relativistic accretion disk explanation for the line profiles is not justified in general.