Table 5 (click here) summarizes the morphological information on the SULIRGs, based on our optical and infrared observations. Data from the literature have
been added. The projected angular and, in parentheses, the linear separation
between the nuclei of interacting galaxies is indicated in Col. 4. The
infrared images have
preferentially been used to measure them.
For objects with a single nucleus, we have estimated an upper limit calculated
from the FWHM of stars in the field. In loose pairs, we also give, within
brackets, the angular
and linear distance of the companion. When available the velocity
difference between the interacting partners is indicated in Col. 5. The velocities
have been determined using the mean redshift of [OI]
, H
, [NII]
and [SII]
\
high resolution lines.
In our sample of 24 systems, 13 (54%) are close interacting galaxies (with a linear separation less than 12 kpc) or merging objects, having two nuclei enshrouded in a common envelope. 5 (21%) have a single nucleus both in the optical and in the infrared but exhibit typical tidal tails of mergers. 3 (12.5%) do not show direct evidences of interaction but their slightly disturbed morphology - for instance, an asymmetry in their elliptical isophotes (IRAS 14378-3651), an eccentric (IRAS 16090-0139) or elongated (IRAS 23230-6926) nucleus - suggest that they are the remnants of complete mergers. Their tidal tails would have had time to dissipate since the collision or they are too far and faint to be visible. Finally, 3 (12.5%) objects have an ambiguous nature, discussed hereafter.
A clear proof of a merger origin for these objects could undoubtedly be obtained in disentangling their nucleus. Most of the galaxies in the sample of Sanders et al. (1988a), first thought to be single nucleated objects, appeared to have twin nuclei when observed in the infrared, a wavelength range less affected by obscuration, which is prominent in IRAS galaxies. However, apart from IRAS 22491-1808 (Carico et al. 1988), none of the SULIRGs, for which we have R and K images, show discrepant nuclear morphology between the two bands.
Compared to other ULIRGs, the ambiguous objects in our list
appear as systems still in an early
phase of an interaction process. Their companions are situated at a distance
of 40-70 kpc whereas most ultraluminous galaxies have disk-disk separation
less than 10 kpc. One galaxy, IRAS 15462-0450, exhibits clear signs of
tidal disruption; the two others, IRAS 09111-1007 and IRAS 20414-1651
seem to show an unperturbed morphology. More details on these systems are
discussed in Sect. 5 (click here). The enhancement of the infrared
luminosity in ULIRGs is generally explained by the heat of dust
following either starbursts episodes in colliding gaseous clouds or
the fueling of an active nucleus during the merging process. Such
mechanisms imply a small distance between the colliding galaxies.
The unexpected properties of our "ambiguous'' objects could question
this interpretation. However one cannot exclude the hypothesis that
they are already merger remnants and
that their companions do not take part in the ultraluminous phenomena.
The result that most, if not all, SULIRGs are facing a current interaction
or have been subject to one in the past, agrees with previous analysis
of ultraluminous galaxies
(e.g. Sanders et al. 1988a;
Clements et al. 1996a). The conclusion of
Leech et al. (1994)
who obtained, in their sample of 42 galaxies with 60 
m luminosity
greater than 
, a significant 17% fraction of apparently
isolate systems, was recently questioned by
Clements & Baker (1996) who reobserved some of their systems
and found signs of interactions in them.
In this section, we try to classify the SULIRGs among
the different classical spectral types:
starburst (SB), LINERS (LI), Seyfert 2 (SII) and Seyfert 1 (SI).
While Seyfert 1 galaxies are easily recognizable thanks to their extremely
large permitted emission lines, other objects are classified using
diagnostic flux line ratios. As recommended by
Veilleux & Osterbrock (1987), to get rid
as much as possible of extinction problems, we have used criteria involving
ratios between nearby lines. The related three diagnostic diagrams, [NII]/H - [OIII]
/H
,
[OI]/H - [OIII]/H and [SII]/H - [OIII]/H,
are shown in Fig. 6 (click here). To avoid line blending problems we have
used, when available, the high resolution data. Care has been taken not to
mix LR and HR
lines. The fluxes have been corrected for Galactic extinction, and,
if possible, for interstellar extinction, using a
reddening curve parameterized by Miller & Mathews (1972) as:
and an absorption 
derived from the Balmer decrement with the formula:
H/H has been determined
from the LR data. However the LR H line is strongly blended with
the [NII]
line. To take this effect into account the flux has been
multiplied by the mean ratio between the HR and LR H fluxes.
No attempt has been done to
correct our data from intrinsic stellar absorption line. This causes
an uncertainty in the [OIII]/H ratio, which however is not too dramatic
for the classification between SB and AGN (LINER or SII): the latter is
more governed by the [NII]/H ratio. For clearness concerns, only
the main galaxy or brightest nucleus, in systems with multiple components,
is displayed
in Fig. 6 (click here). Arrows correspond to galaxies
with one undetermined line ratio.
Table 5 (click here) presents, in Col. 6, the resulting adopted spectral classification for each system: the main galaxy or brightest nucleus in mergers, and below, either the faintest nucleus, or a possible interacting companion (in brackets). Since some galaxies do not have the same position in the different diagrams and others cannot be put into them, because some necessary lines were not detected, some rules had to be used, based on the appearance number of an object in a given spectral type area. The related convention for naming the spectral type is the following: "a>b'' means that two diagrams favor class "a'' and one class "b''. "a=b'' indicates that the galaxy is on the border line. "a?b'' means that no choice between "a'' and "b'' can be done because of an undetermined flux ratio.
To do some statistics, we have given a weight to a spectral
type t, proportional to
the appearance number of a given galaxy in the locus of t, for the
three diagnostic diagrams. An attempt to estimate
an uncertainty associated with t was made, based on
the number of discrepancies of a galaxy position between the different
diagrams.
For cases with undetermined line ratios,
an "undetermined type'' column was charged. Applying these rules,
we found, in our full sample, including two objects with data
taken from the literature (see Table 4 (click here)), that 
of SULIRGs are starbursts, 
, LINERs and 
,
Seyferts II.
Besides, there is one Seyfert 1 galaxy in our sample
(
), IRAS 15462-0450. The properties of that peculiar
object will be discussed in a future paper. For 20% of the galaxies,
no spectral type could be given.
Taking into account only
galaxies having sufficient spectroscopic data for a reliable spectral
determination, the distribution is the following:
for starbursts, 
for LINERs, 
for Seyferts II, 
for Seyferts I, and 6% for undetermined types.
The high fraction of AGNs among SULIRGs (
) agrees with the
trend found by Veilleux et al. (1995) of an increase in the proportion of AGNs
with the
infrared luminosity. Their sub-sample of ULIRGs extracted from the Extended
BGS and IRAS Warm Galaxy Survey has a similar proportion of
AGNs (62%, excluding the ambiguous objects), but a higher fraction of
Seyfert 1 (15%). Using low resolution spectra, Clements et al. (1996b)
found that at least 35% of the ULIRGs in their sample are likely
to contain an AGN-type central engine. However, when comparing
statistics on spectral type coming from different studies, one has to take into
account systematic biases, and among them, the way that individual nuclear
spectra were extracted. Contamination by off-nuclear regions might
significantly affect the determination of the spectral type. However,
up to now, only very few studies have been done to quantify this bias
(Kennicutt 1992; Lehnert & Heckman 1994;
Liu & Kennicutt 1995), and none of them were
actually devoted to AGN-like objects.
The question of the mechanism responsible for the infrared luminosity enhancement in ULIRGs has been widely addressed in the literature but no clear answer has yet emerged. Violent star forming episodes, nuclear activity, or a combination of both, have been proposed. In that context, one may wonder whether the high proportion of AGN nuclear spectral type apparently found among ultraluminous galaxies is an hint in favor of the second hypothesis. Some authors (Sanders et al. 1988a; Taniguchi et al. 1994) claim that the ultimate episode of the ultraluminous phase is the formation of a quasar buried in a merger remnant. However our candidates for complete mergers (IRAS 14378-3651, IRAS 20414-1651, IRAS 23230-6926), do not show particularly active nuclear activity. The only Seyfert 1 in our sample may even not be a merger. Finally, one point should be emphasized. The ultraluminous galaxies are heavily obscured objects, as indicated by their high visual absorption and by radio continuum observations (Condon et al. 1991). As a result, the optical spectra may not reflect the properties of the actual nucleus. In that context, the determination of the fraction of AGNs found in ULIRGs from optical observations might be biased.