We describe the Sloan

The fourth generation of the Sloan

We report on the diversity in quasar spectra from the Baryon Oscillation Spectroscopic Survey. After filtering the spectra to mitigate selection effects and Malmquist bias associated with a nearly flux-limited sample, we create high signal-to-noise ratio composite spectra from 58,656 quasars (   z 2.1 3.5), binned by luminosity, spectral index, and redshift. With these composite spectra, we confirm the traditional Baldwin effect (BE, i.e., the anticorrelation of C IV equivalent width (EW) and luminosity) that follows the relation µ , and −0.41±0.005 for z=2.25, 2.46, and 2.84, respectively. In addition to the redshift evolution in the slope of the BE, we find redshift evolution in average quasar spectral features at fixed luminosity. The spectroscopic signature of the redshift evolution is correlated at 98% with the signature of varying luminosity, indicating that they arise from the same physical mechanism. At a fixed luminosity, the average C IV FWHM decreases with increasing redshift and is anti-correlated with C IV EW. The spectroscopic signature associated with C IV FWHM suggests that the trends in luminosity and redshift are likely caused by a superposition of effects that are related to black hole mass and Eddington ratio. The redshift evolution is the consequence of a changing balance between these two quantities as quasars evolve toward a population with lower typical accretion rates at a given black hole mass.

The Baryon Oscillation Spectroscopic Survey (BOSS) has collected more than 150,000 2.1 ≤ z ≤ 3.5 quasar spectra since 2009. Using this unprecedented sample, we create a composite spectrum in the rest-frame of 102,150 quasar spectra from 800Å to 3300Å at a signal-to-noise ratio close to 1000 per pixel (∆v of 69 km s −1 ). Included in this analysis is a correction to account for flux calibration residuals in the BOSS spectrophotometry. We determine the spectral index as a function of redshift of the full sample, warp the composite spectrum to match the median spectral index, and compare the resulting spectrum to SDSS photometry used in target selection. The quasar composite matches the color of the quasar population to 0.02 magnitudes in g − r, 0.03 magnitudes in r − i, and 0.01 magnitudes in i − z over the redshift range 2.2 < z < 2.6. The composite spectrum deviates from the imaging photometry by 0.05 magnitudes around z = 2.7, likely due to differences in target selection as the quasar colors become similar to the stellar locus at this redshift. Finally, we characterize the line features in the high signal-to-noise composite and identify nine faint lines not found in the previous composite spectrum from SDSS.