Galaxy mergers are recognized as the triggers of nearby starbursts,
especially the ultraluminous far infrared-selected galaxies. These
systems are powered in large part
by star formation rather than by an embedded AGN,
as confirmed by far infrared spectroscopy, and have star formation
rates of 100 or even 1,000
M
per year. Near
infrared mapping
reveals de Vaucouleurs profiles and CO mapping reveals a central cold
disk or ring with ~ 1010 M of molecular gas within a few
hundred parsecs. Can one generalize from the rare nearby examples that
ellipticals, and more generally spheroids, formed via merger-induced
starbursts?
Evidence that gives support to this contention requires a component of star-forming galaxies that is sparse locally to account for three distinct observations of galaxies, or of their emission presumed to be at z > 1. Far IR counts by ISO at 175 µm 42 and submillimeter counts by SCUBA 43 at 850 µm require a population of IR-emitting objects that have starburst rather than normal disk infrared spectra. Moreover identification of SCUBA objects demonstrates that typical redshifts are one or larger, but mostly below 2. 44
A powerful indirect argument has emerged from
modelling of the
diffuse far infrared background
radiation. This amounts to
41
i ~ 20 nw/m2sr, and
exceeds the diffuse optical background light of about 10
nw/m2sr that
is inferred from deep HST counts. The local population of galaxies,
evolved backwards in time fails to account for the diffuse infrared
light, if one only considers disk galaxies, where the star formation
history is known from considerations of their dynamical evolution.
The starburst population invoked to account for the FIR counts can account
for the diffuse infrared background radiation.
45
If this is the case, one
expects a non-negligible contribution near 1 mm wavelength from
ultraluminous FIR galaxies to the diffuse background radiation. For
example, the predicted FIR flux peaks at ~ 400 µm if the
mergers occur at z 
3. The extrapolation to longer wavelengths
tracks the emissivity, or decreases roughly
as 
3. Hence there
should be a
contribution at 1 mm of order 1 nw/m2sr, which may be compared with
the CMB flux of ~ 2000 nw/m2sr. One can measure fluctuations of
T/T ~ 10-6,
and one could therefore be sensitive to a
population of ~ 106 ultraluminous FIR sources at high
z. The
inferred surface density (~ 20 per square degree) is comparable to
the level of current SCUBA detections. Hence CMB fluctuations on an
angular scale of ~ 10' near the CMB peak could be generated
by the sources responsible for the diffuse FIR background. Moreover
these are rare and massive galaxies, and hence are expected to have a
large correlation length that should give an imprint on degree scales.
One can evidently reconcile submillimeter counts, the cosmic star formation history and the far infrared background together with formation of disks and spheroids provided that a substantial part of spheroid formation is dust shrouded. A difficulty that arises is the following: where are the precursors of current epoch ellipticals? A few are seen at z < 5 but are too sparse in number to account for the younger counterparts of local ellipticals. 46 Dust shrouding until after the A stars have faded (~ 2 x 109 yr) would help. Other options are that the young ellipticals are indeed present but disguised via ongoing star formation activity, and mostly form at z > 5 or else possess an IMF deficient in massive stars.