Rapid optical variability in active galactic nuclei and quasars

We monitored 23 quasars and Seyfert 1 galaxies on timescales of minutes, hours, days, weeks, 3 weeks, and 3 months. Observations were made in broad continuum bands of blue, yellow, and far-red light. We attained typical 1 a relative flux uncertainties of 2% through differential photometry of field stars in the same CCD frames as the active galactic nuclei (AGNs). In 77 intranight comparisons (23 in blue light, 2 in yellow, and 52 in red) on 20 different AGNs, we found no evidence for significant intranight optical variations. An zipper limit for the amplitude of variability for this sample of AGNs on timescales of 1 hr or less was derived to be 0.03 mag. The shortest timescales for which variations were detected, with confidence greater than 99%, were 25 hr in Mrk 79 and 27 hr in NGC 4151. Variations on even shorter timescales might have been missed, as we observed few AGNs over time intervals of less than 1 day. We found evidence for variability on timescales of 1-10 days in 11% of the AGNs in both the blue and yellow filters, and 26% of the AGNs in the red continuum. In fact, five of the six lowest luminosity AGNs were significantly variable in the red on a timescale of days. We found optical variability to be more common on a month-to-month timescale; detectable in 60% of the AGNs in the blue filter and 40% in the red filter. Averaging over all the variable objects in the three wave bands, the rms variation was 0.12 mag over 25 days and 0.21 mag over 75 days. We confirmed the overall trend for average variability amplitude to increase with timescale (from 1 to 100 days) by computing autocorrelation and structure functions. The average power density spectra of AGN fluctuations had a logarithmic slope similar to -1. +/- 0.5. The exceptions are the five low-luminosity Seyferts which varied rapidly in the red, displaying relatively flat power-density spectra of variability in that wave band. The fastest significant variations detected, on a timescale of days, are consistent with the dynamical (orbital) timescale of a black-hole accretion disk. Larger amplitude variations are observed on a timescale of 1 month, by which time the amplitude of the luminosity autocorrelation function has typically dropped to 0.5. This variability timescale is consistent with the predicted thermal timescale of accretion disks. The largest peak-to-peak changes observed were 0.3 mag in the red (NGC 4151) and 0.5 mag in the blue (MCG 8-11-11)-15 and 30 times our measurement uncertainties. When large variations were seen in one filter, they were usually detected, over the same interval and in the same sense, in other filters (12 out of 14 times). On timescales of months, the amplitude of blue variations is generally larger than that of the red. In many objects the amplitude difference is so large that the intrinsic spectrum of the active nucleus must have been bluer when brighter. The most rapid variations tend to occur in the least luminous AGNs. Luminosity was inversely correlated with the variability chi(2) in the red filter (r = -0.47, significant at the 98% probability level). Nonetheless, the luminosity dependence of variability is not strong. In fact, the amplitudes of variability that we measure on timescales of months extrapolate very well from those observed in qunsar variability monitoring in the blue over timescales of years (although they are somewhat smaller than the variations measured in the ultraviolet). Those AGNs with the strongest Fe rr emission lines were also marginally less likely to show strong variability (r = -0.39). There was little correlation between the incidence of variability and any other AGN property. These results suggest that the time variability of the nuclear continuum may be essentially similar in all Seyfert 1 galaxies, and that many apparent differences in their light curves are due to differences in sampling. We have examined the histograms of flux increases and decreases. This reveals, for example, whether or not the light curves would be statistically equivalent with their flux scales inverted. averaged over all variable objects, the light curves have a nearly symmetric distribution of points above and below the mean flux level. Physically this indicates that the variations are not predominantly due to flares and outbursts, on the one hand, nor to eclipses and dropouts, on the other. The light curves are nor a sum of impulsive shots superposed on a quiescent brightness level. We also plotted the flux-change histograms as a function of time interval. In general, the symmetry of flux increases and decreases does not depend on the observation interval, so the light curves would be statistically equivalent with their time axes reversed. A possible exception is made by the low-luminosity Seyferts with fast variations in the red. There is a marginal (3 sigma) indication that their red luminosity rises more rapidly than it decays.