4 Discussion

4.1 Comparison with V95

Because our VLIRGs sample are selected from 2 Jy redshift survey catalogue which covers the BGS ($\rm {\it f}_{60\,{\mu}m}\ge5~Jy$) sample, there are a large number of overlap galaxies between our sample and the V95 sample. The total overlap number is 37 if we count companions. We compare these overlap galaxies in the diagnostic emission line ratios, such as $\rm [OIII] \lambda 5007 / H\beta$,$\rm [OI]\lambda6300/H\alpha$, $\rm [NII] \lambda 6584 / H\alpha$ and $\rm [SII](\lambda 6716+\lambda 6731)/H\alpha$.The ratios of $\rm [NII] \lambda 6584 / H\alpha$ and $\rm [SII]\lambda6716+\lambda6731/H\alpha$ for these two samples agree very well, with mean difference less than 0.02 and scatter about 0.1; the difference is within our error bar (see Sect. 2.2). However, the mean difference and scatter for the ratios of $\rm [OI]\lambda6300/H\alpha$are much larger, about 0.06 and 0.21, respectively. The reason is that the $\rm [OI]\lambda6300$ emission lines for most sample galaxies are weak and blended, and are therefore more uncertain. For $\rm [OIII] \lambda 5007 / H\beta$, the difference and scatter are also large, about 0.05 and 0.23, respectively. This can not be solely attributed to the low S/N because the $\rm [OIII]\lambda5007$ line for most sample galaxies is strong. As we have discussed before, the reason could be due to several factors, such as aperture effects, underlying stellar absorption of the host galaxy, etc. There is a difference in the color excess E(B-V) between two samples. Our mean value of E(B-V) is 0.99, which is about 0.13 lower than that of 1.12 of V95 in the overlay objects. The Galactic extinction can account for one-fourth (0.03). The remainder could come from different estimates of underlying stellar absorption, aperture effect (e.g., IR23488+1949A) and possible low S/N of $\rm H\beta$ of spectra. Since spectral classification is based on emission line ratios, aperture effect and other uncertainties can also render different spectral classifications. A comparison shows that the our and V95's classifications agree for about 70% (26/37) of the overlap objects.

4.2 LINER-like VLIRGs

The original definition of LINERs by Heckman (1980) was based on the following conditions on oxygen forbidden line ratios, namely, $\rm [OII]\lambda3727 / [OIII]\lambda5007 \gt 1$ and $\rm [OI]\lambda6300 / [OII]\lambda5007 \gt 0.33$. However, the spectra for most of our galaxies do not cover the wavelength of $\rm [OII]\lambda3727$ and the large wavelength separation between $\rm [OII]\lambda3727$ and $\rm [OIII]\lambda5007$ could bring serious errors due to uncertain reddening correction. Therefore we use the criteria suggested by VO87 instead. In fact, for the galaxies with $\rm [OIII]\lambda5007$ line, the two definitions are consistent. As discussed at Sect. 2.3, some of our sources present properties of both HII galaxy and LINER, and could not be classified accurately as either type. We classify them into a mixture type. From the spectral analysis, we find that they are similar with LINERs for the properties of reddening and $\rm H\alpha$ emission, so we regard the mixture types as LINER-like galaxies. Considering that the mixture type VLIRGs show spectral features of both LINERs and HII galaxies, we can also assume that LINER-like galaxies could be the composite systems consisting of both AGN and HII region components. As discussed in Sect. 2.3, a few HII VLIRGs present high ionized level, the combination of such low ionized HII spectra and AGN spectra is similar to the spectra of LINER-like galaxies. Nearby galaxy NGC 7679 (V95) is a good example for this assumption. As the aperture size increases, the spectrum of NGC 7679 changes from that of Seyfert 2 to LINER-like type. This supports the idea that NGC 7679 has an active nucleus and circumnuclear HII region. Recent high resolution HST images (Surace et al. 1998) show that there indeed exist many blue "knots'' of star formation region in the inner few kpc of ULIRGs as well as possible active nuclear. In summary, it is very probably from our spectral analysis that the LINER-like VLIRGs are mixture type galaxies with both active nucleus and starburst regions. Also this type of galaxies may at the evolution stage from nuclear starbursts to AGNs. As AGN features become more dominant, they evolve toward the optical AGNs.

4.3 Relationship between spectral and interaction classifications

Spectral classifications provide information about the main energy output mechanism for galaxies whereas morphological classifications give the knowledge of evolution phase of interaction galaxies. The correlation between these two types of classifications may give important clue for understanding the possible evolution connection between starbursts to AGNs. In Sect. 3.2 we analyzed the statistical properties of optical and infrared properties as a function of the nuclear separation in the double or multiple-nuclear system. As the separation decreases, both infrared luminosity and $\rm H\alpha$ equivalent width increase. Since both infrared luminosity and $\rm H\alpha$ equivalent width are indicators of star formation, it is very clear that as the separation decreases, the star formation rate increases. Table 2 gives the results of spectral and morphological classifications. As we discussed in Sect. 3.1 we combine seven morphological types into three types, namely, galaxies with near or far companions (class 1 to 4), interacting or merging galaxies (class 5, 6) and isolated galaxies (class 0), and assuming that the isolate galaxies are at the last phase of merging, although there still exists the possibility that some of them are truly isolated galaxies and not at the end-state of merging. These three new types construct a simple merger sequence. From Table 2, we can see that the fraction of HII galaxies in VLIRGs decreases rapidly along the merger sequence, from 53% (10/19) of class 1-4 to 39% (16/41) of class 5-6 and then to 23% (3/13) of class 0. On the contrary, the fraction of AGN (LINER-like galaxies, Seyferts) increases dramatically. The corresponding fractions of LINER-like galaxies are 37%, 37% and 54% and those of Seyfert 1s are 5%, 7% and 15%, respectively. Therefore, we can conclude that along the evolutionary sequence of interacting galaxy, the spectral properties of VLIRGs change from HII-like to AGN-like. In the last merger stage, they could evolve either to Seyfert 1's or LINERs. We have discussed in Sect. 4.2 that quite a few VLIRGs are in loose or compact groups of galaxies. For the ten confirmed VLIRGs in group of galaxies in our sample, 8 are HII galaxies, one is a LINER-like galaxy and one is an unclassified AGN. There are no Seyfert in these groups. The high proportion of HII galaxies indicates that VLIRGs in group of galaxies are in the early evolution phase and the Seyferts or LINERs could also be formed in the more advanced stage of multiple-merger process. Therefore, the existence of relatively large fraction $\ge$14% (10/73) of VLIRGs in groups hints that multiple-merging is an important evolutionary process during galaxy formation.