Average ultraviolet quasar spectra in the context of eigenvector 1: A Baldwin effect governed by the Eddington ratio?

We present composite UV spectra for low-redshift type 1 active galactic nuclei binned to exploit the information content of the eigenvector 1 (E1) parameter space. Composite spectra show high enough S/N and spectral resolution to permit a decomposition of the C IV lambda1549 line profile: one of the strongest high-ionization lines (HILs), and fundamental in defining E1 space. The simplest C IV lambda1549 decomposition into narrow-line region (NLR), broad-line region (BLR), and very broad line region (VBLR) components suggests that different components have an analog in Hbeta with two major exceptions. VBLR emission is seen only in population B [FWHM(Hbeta(BC))>4000 km s(-1)] sources. A blueshifted/asymmetric BLR component is seen only in population A [FWHM(Hbeta(BC))less than or equal to4000 km s(-1)] HILs such as C IV lambda1549. The blueshifted component is thought to arise in a high-ionization wind or outflow. Our analysis suggests that such a wind can only be produced in population A (almost all radio-quiet, RQ) sources, where the accretion rate is relatively high. We propose a model to account for several differences between low- and high-ionization line profiles. Part of the broad-line emission is attributed to a self-gravitating/fragmented region in an accretion disk. An inner, optically thick, geometrically thin region of the flow may give rise to a wind/outflow and produce the blueshifted HIL spectrum in population A sources. The fragmented region may produce all or most of the broad-line emission in population B, which contains RQ and the majority of radio-loud (RL) quasars. Comparison between broad UV lines in RL and RQ sources in a single, well-populated E1 parameter space bin (B1) shows few significant differences. Clear evidence is found for a significant NLR C IV component in most RL sources. The BLR/VBLR similarity in bin B1 provides circumstantial evidence in favor of black hole (BH) spin, rather than BH mass or accretion rate, as a key trigger in determining whether an object will be RL or RQ. We find a 10-fold decrease in EW C IV lambda1549 with Eddington ratio (decreasing from approximate to1 to similar to0.01), while N V lambda1240 shows no change. These trends suggest a luminosity-independent Baldwin effect'' in which the physical driver may be the Eddington ratio.