The paradigm of QSO physics is that the energy is freed by accretion of matter in a massive black hole. This paradigm allows some simple estimates:
The Eddington luminosity for which gravitational attraction compensates radiation pressure is:
Using the bolometric luminosity of 3C 273 deduced in Sect. 3.5 we thus estimate that provided that the bulk of the luminosity is emitted isotropically the mass of the central black hole is ~ 109 solar masses. The corresponding gravitational radius is ~ 3 . 1014cm.
The mass accretion rate can be estimated from the luminosity L by:
where is the efficiency of the conversion of rest mass to radiation.
Using the same luminosity as above and an efficiency ~ 10% typical of accretion onto black holes we deduce a mass accretion rate ~ 1.3 . 1027g s-1 or about 10 solar masses per year. Not surprisingly, these numbers are close to those deduced from accretion disc models.
Going beyond these estimates requires understanding of how the energy liberated by the accretion process is transformed in the radiation we observe across the electromagnetic spectrum. This understanding is still widely lacking, we can, however, list some of the elements that do play a role.
Highly relativistic electrons and magnetic fields must belong to any model of 3C 273 (and other radio loud AGN) as shown by the presence of synchrotron radiation. Their energy densities are very inhomogenous and partly organized in small packets some of which at least are accelerated to relativistic bulk velocities along complex paths to form the observed small scale jet. It is probable that the electrons are accelerated to highly relativistic energies in shocks and that they thus aquire the energy that they radiate from the kinetic energy of some underlying flows.
Another indication of fast flows is the presence of the broad lines which indicate that the material surrounding the black hole has velocities of the order of 104km/s. It has not been possible to find in the line variability pattern the signature for a dominantly ordered velocity field (expansion, accretion or rotation). One, therefore, concludes that a large fraction of the velocity field is of a turbulent nature. In these circumstances, the presence of shocks where streams of matter collide is difficult to avoid. It is interesting to note that thermalising Hydrogen gas with bulk velocities of the order of 104km/s will produce a gas of T ~ 5 . 109K. A temperature close to the one needed to Comptonise the UV photons to X-rays with a slope as observed [Walter & Courvoisier 1992].
A fraction of the UV and higher energy radiation is reprocessed by gas to form the broad lines and by dust to give the thermal infrared radiation. The organisation of the broad line emitting clouds is unclear, no cloud confining medium having been found. Whether the optical-UV emission forming the blue bump is itself due to reprocessed X-ray emission is also unclear, mostly because of the absence of the signature of Compton reflection in the X-rays.
There are many different timescales at play. Among those we know there is the few days delay between the UV and optical light curves which imply that the signals ruling the blue bump emission travel at the speed of light. There is also the presence of much longer timescales (of several years) in the visible light curves and correlations which delays of the order of a year or so, between emission components. These timescales are long compared with light crossing times or dynamical times in the vicinity of the black hole. They are, however, short compared to viscous timescales of standard accretion discs. [Courvoisier & Clavel 1991]. The presence of these timescales may either indicate that the size of the continuum emitting accretion is of the order of a parsec (similar to the size of the broad line region) or that there exist characteristic velocities of the order of few percent of the speed of light in a region of several gravitational radii. One may also note that the amplitude of the variations at short timescales ( one day ~ 10 gravitational radii over c) is small (few percent; [Paltani et al. 1998]). This may indicate that the variations on this timescale are not associated with the regions closest to the black hole, but rather to small regions in an extended object.
There have been many attempts to understand the geometry of the emission regions considering one or several emission components. Most have been based on the presence of accretion discs. Many of the arguments are revised in [Blandford 1990]. The addition of a corona being discussed by [Haardt et al. 1994].
[Camenzind & Courvoisier 1983] attempted to understand the continuum emission of 3C 273 in terms of a mildly relativistic wind originating in the core of the object and shocked at some distance. Most of the observed emission in this model was the by product of the shocked material. This model predicted that the variation time scales of the different components was such that the UV varied faster than the X-rays which in turn varied faster than the optical emission. The infrared and gamma ray variability timescales were expected to be the longest. These predictions were soon disproved by observations which led to a revision of the geometry [Courvoisier & Camenzind 1989]. In this revised geometry the wind is channeled in such a way that the shocked material covers only few percents of the UV source. The shocked material is heated to temperatures such that the UV photons crossing it are Comptonised to X-ray energies. The lag between X-ray and UV fluxes may be understood naturally in this geometry [Paltani et al. 1998].
[Courvoisier et al. 1996] have considered whether accretion of matter could be in the form of stars rather than gas. In their model the gravitational energy is radiated following collisions between stars in the vicinity of a black hole. First order considerations showed this to be a possible alternative to understand the variability of AGN and its dependence on luminosity. This also points to the little studied question of the interaction between the active nucleus and the surrounding stellar population.
It is thus clear that although the main elements of the AGN model have been in place for more than 30 years, often following pioneering observations of 3C 273, much remains to be understood. AGN are considerably more complex than many of us anticipated. This complexity together with the extreme properties they show make them fascinating object to study.
This work is based on a long term effort by a large set of colleagues who have participated in the gathering of data and in many discussions over the years. I owe a particular debt to M. Camenzind and M.-H. Ulrich for sharing their knowledge with me when this effort began. I would never have been able to write this review without the benefit from many interactions over the years and around the world. Several of my colleagues at the ISDC have given me some very direct help in preparing this review and in particular the figures. They are S. Paltani, M. Polletta and M. Türler. I thank them and also T. Krichbaum, R. Walter and L. Woltjer for reading and commenting the manuscript. A. Aubord and M. Logossou have been of much help in the typesetting.