1. INTRODUCTION

It is 15 years since quasars were discovered, but there has been disappointingly slow progress towards achieving a consensus about what they are. There is now, however, wider agreement that quasars are not qualitatively different from other forms of activity in galactic nuclei, and this has deflated the more bizarre ideas that were aired in the early days. Radio and optical astronomers have accumulated a larger (and more systematic) body of data, from which we can infer details of the structure and radiation mechanism. Things become more conjectural as we attempt to extend the chain of inference back towards the central "power-house". But this, of course, is the most interesting and fundamental aspect of the quasar phenomenon.

Most interpretations of quasars invoke a mass of (say) 10⁶-10⁹ M in a region 1 pc in size. The three well-known categories of model are

: dense star clusters [1 - 6]

: massive stars, "spinars", or "magnetoids" [1, 7]

: accretion onto massive black holes [8 - 13].

One can assess the maximum likely efficiency of the three types of model. In case , this is essentially given by the maximum realizable energy of a supernova-style explosion, or by the maximum amount of binding energy that can be released via stellar collisions before the star cluster "dissolves" into a single massive gas cloud. For , the limit is set by the onset of dynamical instability: rotation can stabilize against the post-Newtonian instability, but a system that becomes too flattened is vulnerable to bar-mode instabilities. One cannot give firm estimates of the efficiencies attainable by models and , but there would, I think, be general agreement that they cannot be as high as the 10 percent conversion of rest mass into electromagnetic radiation that can be achieved via accretion into a black hole.

A prerequisite for any model is the accumulation of a large mass concentration, probably in the potential well as the center of a large galaxy. Once this entity has reached the stage when its power output (derived primarily from gravitational binding energy) becomes conspicuous, one cannot envisage any evolutionary end-point other than complete collapse of at least part of the material involved: in other words, even if an accumulation of gas does not collapse directly to a black hole, the progression (-> ) -> , or -> , seems inexorable. Given that any of the options lead to a black hole, and that, once formed, a black hole is a potentially more efficient power source than any conceivable progenitor, it seems plausible to attribute quasars the most powerful known cosmic phenomena - to black hole accretion processes: and then, as a secondary issue, to consider precisely how they form, and whether precursor stages resembling or can yield an explanation of some less spectacular type of activity in galactic nuclei.

Several new lines of evidence (some discussed by other contributors to this conference) have tilted the balance of evidence further towards models involving a single coherent and compact object (i.e., or above). For instance: (1) the VLBI data seem inconsistent with a simple "Christmas tree" model. (2) The compact radio components are (in, e.g. 3C 111 and Cygnus A) aligned along the same axis as the extended double structure, indicating that the central "engine" has maintained a preferred axis for millions of years. (3) There is evidence from line ratios (and from possible variations) that the broad emission lines come from regions of higher density and smaller dimensions than envisaged in most previous models; the central continuum source must be localized in a region of equally small dimensions. (4) The variability cannot be readily analyzed into successive standard supernova-type outbursts.

In this paper, I shall focus on some aspects of the class of models () in which the power supply results from accretion onto a massive black hole, omitting those aspects of the topic that are discussed elsewhere in these proceedings. The required mass M_h is typically in the range 10⁷-10⁹ M, and the minimum relevant length scale is then the Schwarzschild radius

(1)

If the efficience of energy conversion is , then the inflow rate needed to provide a luminosity L is

(2)

It is useful to define _crit as the value of corresponding to a luminosity L_edd 1.3 x 10³⁸ (M/M) erg s^-1.