2003ApJ...586...52S

We have performed a spectral principal component analysis (SPCA) for an essentially complete sample of 22 low-redshift QSOs with spectral data from Lyα to Hα. SPCA yields a set of independent principal component spectra, each of which represents a set of relationships among QSO continuum and line properties. We find three significant principal components, which account for ∼78% of the total intrinsic variance. The first component, carrying ∼41% of the intrinsic variance, represents Baldwin relationships: anticorrelations between equivalent widths of broad emission lines and continuum luminosity. The narrow-line core (FWHM ∼2000 km/s) of the broad emission lines dominate this component. The second component, accounting for ∼23% of the intrinsic variance, represents the variations in UV continuum slope, which is probably the result of dust reddening, with possible contributions from starlight. The third principal component is directly related to the Boroson & Green ``eigenvector 1'' (their first principal component), clearly showing the anticorrelation between strengths of optical Fe II and [O III] λ5007, and other relationships previously found in the Hβ-[O III] region. This third component shows the expected strong correlation with soft X-ray spectral index. The widths of C III] λ1909, Mg II λ2798, and Balmer emission lines are also involved and clearly correlated, relating this component to black hole mass or Eddington accretion ratio. We find an inverse correlation between the strengths of the UV and optical Fe II blends, as suggested by some photoionization models. We also find correlations of the strengths of several low-ionization UV lines with Fe II(opt), and a strong positive correlation of C IV λ1549 with [O III] strength. The wide wavelength coverage of our data enables us to see clearly the relationships between the UV and optical spectra of QSOs. The Baldwin effect and Boroson & Green's eigenvector 1 relationship are clearly independent. We demonstrate how Baldwin relationships can be derived using our first principal component, virtually eliminating the scatter caused by the third principal component. This rekindles the hope that the Baldwin relationships can be used for cosmological study.