2.1. The power source for dusty galaxies
About 99% of the energy released by galaxies in the submm and far-IR wavebands is produced by thermal emission from dust grains; the remainder comes from fine-structure atomic and molecular rotational line emission. However, the source of the energy to power this emission by heating dust is often unclear. Any intense source of optical/ultraviolet (UV) radiation, either young high-mass stars or an accretion disk surrounding an AGN, would heat dust grains. Because dust emits a featureless modified blackbody spectrum, submm continuum observations can reveal little information about the physical conditions within the source. Regions of intense dust emission are very optically thick, and so little information can be obtained by observing optical or UV radiation.
In typical spiral galaxies, with relatively low far-IR
luminosities of several 1010
L
(for example
Alton et al., 2000,
2001),
the dust emission is known to be significantly
extended, on the same scale as the 10-kpc stellar
disk. 6 The emission
is certainly associated with molecular gas rich star-forming regions
distributed throughout the galaxy
(Regan et al., 2001),
in which dust is heated by the hot, young OB stars.
In intermediate luminosity galaxies, such as the interacting pair of spiral galaxies NGC 4038 / 4039 `the Antennae' (Mirabel et al., 1998; Wilson et al., 2000), the most intense knots of star-formation activity, from which most of the luminosity of the system emerges, are not coincident with either nucleus of the merging galaxies, but occur in a deeply dust-enshrouded overlap region of the ISM of the galaxies. This provides a strong argument that almost all of the energy in this system is being generated by star formation rather than an AGN.
In more luminous ULIRGs that are at sufficiently low redshift for their internal structure to be resolved, the great majority of the dust emission arises in a much smaller, sub-kpc region (Downes and Solomon, 1998; Sakamoto et al., 1999) within a merging system of galaxies. It is plausible that a significant fraction of the energy could be derived from an AGN surrounded by a very great column density of gas and dust that imposes many tens of magnitudes of extinction on the emission from the AGN in the optical and UV wavebands, and which remains optically thick even at near-IR wavelengths. Alternatively, an ongoing centrally condensed burst of star-formation activity, fueled by gas funneled into the center of the potential well of a pair of interacting galaxies by a bar instability (Mihos, 2000) is an equally plausible power source.
If the geometry of absorbing and scattering material is known or assumed, then radiative transfer models can be used to predict the SED of a galaxy, which should differ depending on whether the source of heating is a very small AGN with a very hard UV SED, or a more extended, softer-spectrum nuclear star-forming region (for example Granato et al., 1996). Note that the results are expected to be very sensitive to the assumed geometry (Witt et al., 1992). In merging galaxies this geometry is highly unlikely to be spherical or cylindrical, and is uncertain for the high-redshift galaxies of interest here. In the case of AGN heating, the SED would be is expected to peak at shorter wavelengths and the mid-IR SED would be expected to be flatter as compared with a more extended star-formation power source. Both these features would correspond to a greater fraction of hot dust expected in AGNs (see Fig. 2), and is seen clearly in the SEDs of low-redshift IRAS-detected QSOs (Sanders and Mirabel, 1996).
An alternative route to probing energy sources in these galaxies is provided by near- and mid-IR spectroscopy. At these longer wavelengths, the optical depth to the nucleus is less than in the optical/UV, and so the effects of the more intense, harder UV radiation field expected in the environs of an AGN can be observed directly. These include the excitation of characteristic highly-ionized lines, and the destruction of relatively fragile polycyclic aromatic hydrocarbon (PAH) molecules (Rigopoulou et al., 1999; Laurent et al., 2000; Tran et al., 2001), leading to the suppression of their distinctive emission and absorption features. Mid-IR spectroscopic observations with the successor to IRAS, the Infrared Space Observatory (ISO) in the mid 1990's indicated that most of the energy from low-redshift ULIRGs is likely generated by star-formation activity rather than AGN accretion. However, the fraction of ULIRGs containing AGN appears to increase at the highest luminosities (Sanders, 1999). This could be important at high redshifts, where the typical luminosity of dust-enshrouded galaxies is greater than in the local Universe. In addition, there may be duty-cycle effects present to make an AGN accrete, and perhaps to be visible, for only a fraction of the duration of a ULIRG phase in the evolution of the galaxy (Kormendy and Sanders, 1988; Sanders et al., 1988; Archibald et al., 2002). X-ray observations also offer a way to investigate the power source, as all but the densest, most gas-rich galaxies, with particle column densities greater than 1024 cm-2 are transparent to hard (> 2 keV) X rays.
Ultra-high-resolution radio observations provide a route to probing the innermost regions of ULIRGs (Smith et al., 1998; Carilli and Taylor, 2000). By detected the diffuse emission and multiple point-like radio sources, expected from multiple supernova remnants, rather than a single point-like core and accompanying jet structures expected from an AGN, these observations suggest that high-mass star formation contributes at least a significant part of the luminosity of the ULIRGs Arp 220 and Mrk 273.
It is interesting to note that the observed correlation between the
inferred mass of the black holes in the centers of galaxies and the stellar
velocity dispersion of the surrounding galactic bulges, in which most
of the stars in the Universe reside
(Fukugita et al., 1999),
might inform this discussion
(Magorrian et al., 1998;
Ferrarese and Merritt,
2000;
Gebhardt et al., 2000).
The mass of the bulge appears to exceed that of the
black hole by a factor of about 200. When hydrogen is processed
in stellar nucleosynthesis, the mass-energy conversion efficiency is about
0.007
*,
where
*
(
0.4) is the fraction of
hydrogen burned in high-mass stars. When mass is accreted onto a black
hole, the mass-energy conversion efficiency is expected to be about 0.1
BH, with
BH ~ 1
with the definition above.
If accretion and nucleosynthesis were to generate the same amount of
energy during the formation of a galaxy, then
the ratio of mass contained in both processed stars and stellar remnants
to that of a supermassive black hole is expected to be about 0.1
BH / 0.007
*.
For
*
= 0.4 and
BH = 1, this
ratio is about 36. As a mass ratio of about 200 is observed,
this implies that a greater amount of energy, by a factor of about 6, is
generated by high-mass star-formation activity than by gravitational
accretion.
If the bulge-to-black-hole mass ratio is in fact greater than 200, then
either the factor by which star formation dominates will exceed 6, or
the accretion must have been more than 10% efficient; that is
BH >
1. If low-efficiency accretion dominates the process of the build up of
mass in the central black hole, then less than 1 part in 7 of the
luminosity generated during galaxy formation
will be attributed to accretion as compared with high-mass star formation.
A greater amount of energy generated by star formation as compared with accretion processes appears to be favored by these circumstantial arguments.