Spectral variability of quasars from multi-epoch photometric data

Aims. The study of the ensemble properties of the UV/optical broadband variability of quasars is hampered by the combined effects of the dependence of variability on timescale, rest-frame wavelength, and luminosity. Here, we present a new approach to analysing the dependence of quasar variability on rest-frame wavelengths. Methods. We exploited the spectral archive of the Sloan Digital Sky Survey (SDSS) to create a sample of over 9000 quasars in the Stripe 82. The quasar catalogue was matched with the Light Motion Curve Catalogue for SDSS Stripe 82 and first-order structure functions were computed from the lightcurves. The structure functions are used to create a variability indicator that is related to the same intrinsic timescales for all quasars (about 1 to 2 yr in the rest-frame). We study the variability ratios for adjacent SDSS filter bands as a function of redshift. A quantitative interpretation of these relations is provided by comparing with the results of simple Monte Carlo simulations of variable quasar spectra. Results. We confirm the well-known dependence of variability on time-lag; the best power-law fit of the sample-averaged structure function has a slope β = 0.31 ± 0.03. We also confirm that anti-correlations exist with luminosity, wavelength, and redshift, where the latter can be fully explained as a consequence of the former two dependencies. The variability ratios as a function of redshift resemble the corresponding colour index-redshift relations. While variability is almost always stronger in the bluer passband than in the redder, the variability ratio depends on whether strong emission lines contribute to either one band or the other. We find that th