1. HISTORICAL INTRODUCTION

The study of active galactic nuclei is one of the most rapidly growing subjects in present-day astrophysics. As the sites of the release of energy on the most powerful sustained rates and compact scales we know, they clearly are of great intrinsic interest. As the most luminous objects we know, they are the best markers we have of the distant reaches of the universe. From both points of view, understanding their physical nature and structure, and thus how to use them to measure cosmological distances and times are two of the most important aims of current astrophysical research. The greatest difficulty in attempting to review work in this field is the huge number of papers, covering so many different methods of research, that are being published. An excellent recent book on the subject is Quasar Astronomy by Weedman (1986).

The observational study of active galactic nuclei (AGNs) began with Fath (1909), who noted in a spectroscopic survey of the brightest spiral `nebulae' that although most of them had absorption-line spectra, which he correctly interpreted as the integrated light of large numbers of stars, one, NGC 1068, also had six emission lines in its spectrum. He recognized them all as characteristic emission lines of planetary nebulae; today we know them as [O II] 3727, [Ne III] 3869, H, and [O III] 4363, 4959, 5007. Other astronomers, especially Slipher (1917), soon obtained much better spectra of NGC 1068 and of the somewhat similar NGC 4151. Hubble (1926), in his epoch-making paper on the `extragalactic nebulae', emphasized the planetary-nebula-like emission-line spectra of three AGNs (as we call them today), NGC 1068, 4051 and 4151. Then seventeen years later Seyfert (1943) studied these and other galaxies, and isolated the small fraction of them which show many relatively high-ionization emission lines in their nuclear spectra. These nuclei are invariably highly luminous, and their emission lines are wider than the lower-ionization emission lines that occur in the spectra of the nuclei of many `normal' galaxies, he reported. These properties, broad emission lines covering a wide range of ionization, arising in a small, bright (`semi-stellar') nucleus, became the defining characteristics of the class of objects we call Seyfert galaxies, the most numerous known type of AGNs.

They were very little studied, however, until after the optical identification of several of the strongest radio sources with galaxies. One of the first of these was Cyg A = 3C 405, identified by Baade and Minkowski (1954) with a cD galaxy at redshift z = 0.057. Its rich emission-line spectrum was observed to be very similar, in line widths and high levels of ionization, to the spectra of Seyfert galaxies. Soon these characteristic features in the spectrum became the recognized signature by which many (but not all) radio galaxies could be identified. These are the next most numerous type of AGNs we know, after the Seyfert galaxies.

In addition to the radio galaxies, a fraction of the early identified radio sources appeared to be stars, with no trace of a galaxy or nebula in their images. Their spectra were continuous, without absorption lines, but with broad emission lines which could not be identified. Many attempts were made to understand these `stellar radio sources' as peculiar stars, perhaps white dwarfs or subdwarfs with highly deviant abundances of the elements, but no physically consistent interpretation was found along these lines. Then Schmidt (1963) solved the puzzle, identifying several well-known nebular emission lines with the then unusually large redshift z = 0.158 in the `stellar' radio source 3C 273. Immediately after this, Greenstein and Matthews (1963) identified several similar emission lines in 3C 48, with redshift z = 0.367. This was larger than the redshift of any galaxy known at the time, but 3C 48 appeared to be a 14th magnitude star. It was immediately clear that both these objects are highly luminous, and could be observed to very great distances. They are not stars, but quasi-stellar radio sources, usually referred to as `quasars'. They are now understood to be AGNs, so luminous and so distant that the galaxy in which they are cannot (or could not) be detected on available photographic images. Now with CCDs and other high-quantum efficiency, linear two-dimensional imaging detectors, the quasi-stellar nucleus can be subtracted with good accuracy, revealing the galaxies around many of the nearer quasars.

Corresponding radio-quiet high-luminosity stellar-appearing objects were found soon afterwards. At first they were called quasi-stellar objects, or QSOs, but gradually the distinction faded out, and they are now referred to simply as quasars by most research workers in the field. In this review we shall often preserve the distinction. As a result of many systematic discovery programs, at present we know many quasars and QSOs with redshifts up to z 4, but with a fairly strong cutoff around z 3.5, so that at the present writing only ten are known with z > 4 (Schneider et al 1989). They are the most distant objects we know in the universe, but there appears to be a limit to the distance, or light-travel time, at which we can observe them. They have a wide range of luminosities. One of the chief aims of AGN research is to understand them physically so that the luminosity of an individual object can be estimated from its spectrum with sufficient accuracy to calculate its distance.

For rough orientation purposes table 1 lists the approximate space densities here and now of the various types of AGNs we have discussed, in comparison with representative densities of two classes of normal galaxies (Osterbrock 1982). The quasars and QSOs are the rarest but most luminous types of AGNs; radio and Seyfert galaxies are more numerous in space but intrinsically less luminous. Observation and theory alike have made it more and more apparent that QSOs and Seyfert galaxies are not different types of objects, but rather names we use for AGNs at the high- and low-luminosity ends of a continuous sequence of physically similar objects. In fact Schmidt and Green (1983) have introduced the terminology that AGNs in galaxies with total absolute magnitude more luminous than M_B = -23 are called quasars (or QSOs), while those less luminous than M_B = -23 are called Seyfert galaxies. Note, however, that to say QSOs and Seyfert galaxy nuclei belong to one physically continuous sequence does not imply that they are identical except in luminosity, size and other scales, any more than to say O stars and M dwarfs belong to one physically continuous main sequence implies that they are identical except in luminosity, size and other scales. The luminosities of AGNs are very high; a nucleus with M_B = -23 has, in order of magnitude, L 10¹² L 10^45.6erg s^-1 integrated over the ultraviolet, optical and infrared spectral regions 0.1 µm 100 µm (Edelson and Malkan 1986), or L 5 x 10¹² L 2 x 10⁴⁶ erg s^-1 integrated over the entire `observed' range (including interpolations and sketches below upper limits) 10¹⁰ Hz 10²⁵ Hz (Ramaty and Lingenfelter 1982, Urry 1990). This is released within a very small volume, typically with a dimension of less than a few tens of light days, as will be discussed quantitatively in the next section. The energy release is far larger than that which can be derived from stars we know, or any stellar-like objects operating on thermonuclear reactions. The only energy-release process that seems possible is the liberation of gravitational energy. The one that seems most plausible by far is the gravitational release of energy in a rotating accretion disk around a massive black hole, as has been well reviewed by Rees (1977, 1978, 1984).

No matter what the energy-liberation process is, radiation pressure will be important, and any spherically symmetric object which is stable must be gravitationally bound against being blown apart by it. This is the Eddington condition

Here _T is the electron-scattering cross section, and the condition has been written for a pure hydrogen, completely ionized object; any larger opacity would correspond to a smaller limit on the luminosity. For an AGN with L 10⁴⁶ erg s^-1, this corresponds to M 10⁸ M, the minimum mass for a spherically symmetric situation. In more complicated geometries, for instance a roughly cylindrically symmetric structure, it is still useful as an order of magnitude estimate. Furthermore, if enough fuel is available, the luminosity tends to build up too close to this self-limiting maximum value. The luminosity gives the rate at which mass is converted to energy

in solar masses per year, with the efficiency or fraction of the energy which escapes disappearing into the black hole and is radiated. For 0.1, probably a high estimate of the efficiency, 1.8 M yr^-1 for an AGN with L = 10⁴⁶ erg s^-1.

AGNs radiate their energy over a wide range of energies, from -ray and x-rays through the ultraviolet, optical and infrared spectral regions to the far-infrared and radio-frequency regions. This is the reason the luminosity quoted above is so much higher integrated over the entire energy range. Thus observations at all frequencies yield important information on the structure and nature of AGNs.

Because the AGNs are so small in angular scale as to be unresolved except for the very outermost parts of a few of the very nearest examples, we can observe only the integrated radiation of the entire object. Thus interpreting the measured data to deduce the structure of the AGN is far from simple. Nebulae, nova shells and supernova remnants are valuable `laboratories' from which spatially resolved information is available, to suggest and also to test physical ideas that may be important in understanding AGNs.