According to the ideas of the previous sections, an AGN requires a black hole plus fuel delivered at it. Some galaxies, such as M 31, M 32 and our Galaxy, have small black holes but are not active. There may be inactive galaxies with more massive black holes; they would be difficult to detect at larger distances. It seems likely that the black hole initially formed in an AGN episode, but this is not necessary. Once a black hole exhausts its fuel, it becomes quiet. However the n-body calculations of Byrd et al (1986, 1987) suggest that this may occur in an episodic manner, with `latency times' of a few times 108 y before the fuelling begins, and between episodes of fuelling. This behaviour is related to the free-fall time and gravitational instability in the disk. An existing black hole can be reactivated by further interactions which refuel it. The evolution of the mass of an individual black hole must be to increase (if it is fuelled) from Seyfert galaxy to QSO, according to the derived masses of section 2. This complex type of luminosity evolution of an individual AGN is quite different from the smoothly decreasing (with cosmic time) luminosity evolution generally assumed in fitting QSO magnitude and redshift counts, as for instance by Schmidt and Green (1983). From the scanty observational evidence, it is certainly conceivable that a large fraction of `normal' galaxies contain black holes ready and available for refuelling.
From the interaction picture, the rate of formation of AGNs increases rapidly with the number density of galaxies. Thus the observed increase in number density of luminous QSOs with redshift z can be qualitatively understood. Simplified calculations on this basis by De Robertis (1985) show that the density-evolution rates required by the counts can be approximately matched, with a maximum number density (in comoving coordinates) around z = 2-3. Further, as the expansion continues, if the more recently formed black holes have smaller mean masses, the mean luminosity of the QSOs will also decrease. This can mimic the luminosity evolution required by the counts, by a proper choice of assumed conditions. However, there is such a wide range of possible parameters, and the details of the interaction process are so simplified, that it is difficult to evaluate the significance of this result.
In the IRAS survey, a significant number of galaxies were discovered
with the bulk of
their radiation in the infrared spectral region. A significant number of
them, 10 of the
324 galaxies in the IRAS bright galaxy survey, have luminosity L >
1012 L
,
that is in
the QSO range. Their mean redshift is z = 0.055 and the maximum
is z = 0.08. Two
of them are previously known Seyfert galaxies, Mrk 231, a Seyfert 1, and
Mrk 273, a
Seyfert 2. Optical spectra show that one other is a Seyfert 1.5, another
a Seyfert 1.8,
and the rest, except for one with an H II region spectrum, are Seyfert
2s. All but one of
them have extended optical images, and seven of the 10 appear to be
`peculiar', being
either mergers (as Mrk 231 was previously described) or distorted
systems. Seven of
them have been observed in the mm-wavelength region for the CO (1 -> 0)
emission
line, and all of them rank among the most luminous CO sources observed.
Sanders et al
(1988b)
identify these ultraluminous infrared galaxies as QSOs in the process of
formation. They are still highly distorted, and shrouded with dust,
which is apparent
not only from the CO but from the strong reddening of their
emission-line spectra. As
the dust is blown away by the AGN radiation, supernovae, and stellar winds,
Sanders et al
(1988b)
hypothesize, and the nucleus becomes more directly apparent, there
will be a marked shift in the spectral energy distribution to shorter
wavelengths and
the characteristic AGN features will appear. The number of bright QSOs
in the same
redshift range with L > 1012
L is five,
comparable with the number of ultraluminous
infrared galaxies. Thus it seems quite likely that these are indeed the
first stage the QSO process.