3.3.1 The variability of the blue bump
There exists a very long history of observations of 3C 273 beginning in 1887. The object is indeed bright enough to be measurable on a large number of photographic plates. The data up to 1980 have been collected, homogenised and analysed by (Angione & Smith 1985). This light curve shows variations by more than one magnitude and no strictly periodical signal.
The optical-ultraviolet emission of 3C 273 varies on many timescales. One form of variation, that due to the synchrotron flares, has already been mentioned when the infrared variations were discussed. The synchrotron flares are indeed observed at higher frequencies than the near infrared into the optical domain (Courvoisier et al. 1988). Outside of the periods of intense flaring the contribution of the synchrotron emission to the blue bump is negligible. This can be deduced from the fact observed by (Robson et al. 1986) that the near infrared emission is not affected when the synchrotron flux decreases. The synchrotron power law beeing steeper than the blue dump spectrum will contribute less to the blue bump than to the near infrared emission. Since its contribution is not measured in the near infrared it will therefore indeed be negligible compared to the other components making the blue bump.
Ultraviolet variability of 3C 273 was first discussed by (Courvoisier & Ulrich 1985). This discussion was expanded using 9 years of IUE data in (Ulrich et al. 1988). This work showed that it is not possible to account for the changes in the continuum spectral energy distribution by a variable uniform absorbing medium. Indeed such a medium would have to alter its reddening law (hence its composition) at the different epochs at which the flux varied. Difference spectra showed that the variations are more pronounced at short wavelengths and could be accounted for by a black body of 3-6 104 K that changes its emitting area. The recent data described below suggests more complex interpretations for the optical-ultraviolet variations.
The blue bump variability can now be well described using the 10 years of intense monitoring at optical and UV bands obtained since 1985 and shown in Figs. 1 and 3. Analysis of these data (Paltani 1995) and (Paltani et al. 1998) in terms of a structure function shows that the longest timescale on which the source varies is slightly shorter than a year at the shortest wavelengths available with IUE (1 200 Å) and longer than 3 years (i.e. longer than a third of the available timespan of controlled photometric observations) in the V band.
A cross correlation analysis of the light curves shows, furthermore, that all the light curves are very well correlated at short lags (less than a month, see below) but that a secondary correlation peak monotonically increases when longer wavelength light curves are correlated with the 1200 Å light curve.
Both of the above observations indicate that the blue bump variations are of a complex nature and cannot be due to a single physical component. Indeed, a single component like a black body of variable emitting area, is expected to show the same variation timescales at different wavelengths. (Paltani & Walter 1996) have proposed a decomposition into two components for a set of AGN including 3C 273 based on the suggestion that one component is stable or at least varies on timescales much longer than the other and that the spectral energy distributions of both remain stable, the variability being due to the changes of the relative normalisation. This decomposition has the interesting side benefit of giving a very high signal to noise spectrum that allows the measurement of the reddening, in the case of 3C 273 EB-V = 0.038, compatible with galactic reddening.
At very short lags, the lag of the peak of the cross correlation between the 1200 Å light curve with those at longer wavelengths increases with wavelength. The lag is of 2 days at 2 000 Å and 10 days for the V band (Paltani et al. 1998). These lags, although probably significantly different from zero (note that it is difficult to give a formal uncertainty on the lag at which a cross correlation peaks) is many orders of magnitudes less than that expected from viscously heated accretion disc models (Courvoisier & Clavel 1991). This result is insensitive to the details of the models and is also valid for other temperature distributions than those of standard accretion discs. In particular the lag is much shorter than the sound travel time in the accretion disc between the hot regions emitting the UV flux and the cooler regions emitting the V band flux. The lag between the UV and optical light curves of a few days implies that if an accretion disc is present it must be heated by an external source rather than by the internal dissipation of gravitational energy in an optically thick medium. More generally, this result states that the causal connection between the hot and cool regions that form the blue bump (i.e. those emitting in the ultraviolet and those emitting in the visible) must be based on information transported at speeds close to that of light. This observation is similar to that obtained for those Seyfert galaxies for which adequate data have been obtained.
The fact that the energy source should be located outside the disc has several implications. First the standard disc structure and spectra as deduced using the local gravitational energy dissipation Shakura & Sunyaev 1973) is not applicable, secondly the origin of the heating source must be sought. In other words, one should find a way of radiating the energy freed by the accretion process outside the disc.
One possibility that has been studied is the presence of hot coronae surrounding the discs (Haardt et al. 1994). In this paper Haardt et al. consider a structured corona in which a fraction of the accretion power is released through magnetic interactions. The hot blobs in the corona reprocess a fraction of the disc soft photons to X-rays.