5.1. Line variability
Some carefully designed observation campaigns using the IUE satellite and additional ground based data have shown that the emission line variations in Seyfert galaxies follow those of the continuum. The lag is short and increases as the level of ionization decreases. The amplitude of the variations in the high ionization lines and in particular of Ly are similar to those of the continuum. See [Peterson 1993] and references therein for a review of these results. This is one of the strongest arguments demonstrating that the emission lines are indeed due to photoionization of gas close to the continuum source.
In 3C 273, the picture is quite different. Whereas the UV continuum varies by a factor of about 2 (see above), the amplitude of the Ly variations is only 15% or less [Ulrich, Courvoisier & Wamsteker 1993], [Ulrich et al. 1988], [O'Brien et al. 1989]. [O'Brien et al. 1989] studied the timescale of the Ly variations and the possible Ly continuum correlations. They claimed that the observed timescales are less than one year and that there is some correlation between line and continuum variations. The existence of such correlations and the measurement of any lag between continuum and line light curves using variations as small as those observed in 3C 273 are, however, barely possible based on the IUE data base [Ulrich, Courvoisier & Wamsteker 1993]. The small amplitude of the line variations compared with the continuum variations is confirmed by a study of the IUE data on 3C 273 up to 1991 by [Türler & Courvoisier 1997] which shows that when performing a principal component analysis of the spectra of 3C 273, the principal component does not show a line, but only the continuum. This analysis, contrary to previous ones considered all the spectra of a single object as the matrix in which the principal component is to be sought. The principal component then gives for the given object the most variable "spectrum". The result obtained for 3C 273 is in contrast with other well studied objects for which the principal component has the same shape as the average broad line. In the case of 3C 273 the principal component is essentially flat, indicating that the continuum varies, not the lines. This is possibly due to the fact that in 3C 273, an intrinsically bright object, the broad line region is further from the central source than the characteristic time of the continuum variations (of the order of a year) times the velocity of light.