5 Discussion

We overlap two age histograms in Fig. 8. The dotted line histogram shows objects from Table 2 with ages via CMDs. There are three peaks, at about 80, 220 and 450 Myr. The first two peaks appear to be present in Grebel et al.'s (2000) sample for 200 SMC clusters based on OGLE-II data, reported at t $\approx$ 100 and 200 Myr respectively. They suggested them as enhanced star formation epochs. Considering Pietrzynski & Udalski's (1999b) sample with CMD ages for 93 SMC clusters, no peak is seen at t = 80-100 Myr, having their histogram a maximum at the youngest bin. In their histogram there occurs a local maximum at t $\approx$ 250-300 Myr, but the statistics is low. The third peak in the present study has no counterpart in both previous analyses, and it is probably an artifact from the low isochrone age step resolution in that range.

The solid line histogram in Fig. 8 shows additionally the embedded objects in HII regions for which we assumed an age of 3 Myr, and the objects from Table 2 with lower or upper limits by assuming them as the ages themselves. We can observe again a three peak distribution but the first peak is now shifted to the youngest bin similarly to Pietrzynski & Udalski's (1999b) histogram. This suggests that only the 200 Myr peak is relevant, being related to the SMC/LMC last encounter (Gardiner et al. 1994; Grebel et al.'s 2000). The maximum seen in the youngest bin is possibly related to the cluster formation/destruction rates. Since the present sample deals with pairs and multiplets a fast destruction rate might be caused by the internal dynamical evolution in each cluster complex, caused by merger and/or other effects.

Approximately 55% of cluster pairs and multiplets in Table 2 present similar ages between their members indicating that they are coeval. This suggests that most of the pairs and multiplets had a common origin, possibly from the same molecular complex. Note that about 60% of the pairs and multiplets embedded in OB associations (H-A, Hodge 1985), as indicated in Col. 9 of Table 1, have comparable ages between their members. This could be an explanation for the origin of cluster systems.

Considering triplet and multiplet members, we found that about 70% of them are younger than 100 Myr. These results suggest a possible binary cluster formation scenario: clusters can be born in multiplet systems and coalesce by mergers and tidal disruptions forming binary clusters in a timescale of $\approx$ 100 Myr. This time is in agreement with dynamical times required for an interacting pair to merge into a single cluster (ODB98, de Oliveira et al. 2000).

The pairs with a bridge in the isophotal maps have comparable ages for their components (Table 2). As examples, the pairs NGC 241/NGC 242 and B39/BS30 in Fig. 2 show a bridge linking their members which could be interpreted as a sign of interaction (see the similarity with the N-body simulation model in Fig. 11 of ODB98). A typical timescale for the bridge phenomenon is $\approx$ 30 Myr, as deduced from a series of N-body simulations related to bridge formation and evolution (ODB98).

Another interesting isophotal feature is related to the cluster triplet NGC 220/NGC 222/B23 which shows distortions for the small cluster in a direction almost perpendicular to the line connecting itself to the large components NGC 220 and NGC 222. This configuration and morphologies are compatible with a fast hyperbolic encounter with small impact parameter (e.g. Fig. 12 of ODB98).

Figure 9 shows the angular distribution of pairs and multiplets together with SMC HI column density isophotes from Mathewson & Ford (1984). It can be seen that most of the objects are concentrated in the SMC main body, close to the higher concentration of HI, so it is not unexpected that in general they result young (Sect. 4). The objects appear to be distributed along an axis. Such distribution is present in the overall SMC cluster sample and there is growing evidence that it is related to a nearly edge-on disk containing the bulk of the young stellar population in the SMC (Bica et al. 1999; Westerlund 1990).

A nearly edge-on disk in a low internal reddening galaxy like the SMC would imply an increase of projection effects as compared to a simulation such as that carried out by Bhatia & Hatzidimitriou (1988) for the nearly face-on LMC disk where the physical pairs would be about 50%. Consequently the fraction of physical pairs in the SMC would be lower. The present approach including morphological evidence of interaction is an attempt to constrain this aspect. Indeed the fraction with isophotal distortions is only 25% (Sect. 3.1). Projection effects can be responsible for the age spread in some multiplets. For example the sextuplet (Table 2) has component ages 80 Myr (B72), 200 Myr (H86-143), 50 Myr (BS257), 3 Myr (SMC-N52A and SMC-N52B) and 400 Myr (H86-148). Possibly only the 3 or 4 younger components could be related to OB-Association H-A35, the remaining objects would be captures or projection effects. This age spread is also present in IC 1611's quadruplet and in some triplets. On the other hand the quintuplet in the star forming complex NGC 395 has all its members with the same age (3 Myr) thus forming a physical system.

In Fig. 10 we show the distribution of the diameter ratio between members for all pairs in the sample. The diameter ratio is mostly in the range 1 - 2, with a peak at 1 indicating that the majority of pair members have a comparable size. This effect was also observed in the LMC (Bhatia et al. 1991).

Figure 11 shows the distribution of the centre-to-centre angular separation between pair members. The separation range is $\approx$ 10 - 80 arcsec ($\approx$ 3 - 22 pc) with a pronounced peak at $\approx$ 45 arcsec (13 pc). A similar peak was also observed by Bhatia et al. (1991) and de Oliveira (1996) who found a bimodal distribution for the LMC pairs with peaks at $\approx$ 5 and 13 pc. The observed upper limit of the projected centre-to-centre linear separation $\approx$ 23 pc (80 arcsec) is comparable to Bhatia & Hatzidimitriou's (1988) separation criterion for pairs in the LMC (18.7 pc). The frequent separation value around 13 pc may reflect a preferred survival distance for the systems, combined to projection effects.