Among the various phenomena observed in interacting galaxies is the ejectio
n due to tidal forces of stellar and gaseous material into the intergalacti
c medium and its subsequent rearranging which can lead to the formation of
self-gravitating tidal dwarf galaxies (TDGs). We investigate this process w
ith a detailed multiwavelength study of the interacting system Arp 245 and
a numerical model of the collision computed with a Tree-SPH code. Our obser
vations consist of optical/near-infrared broadband imaging, Ha! imaging, op
tical spectroscopy, H I VLA cartography and CO line mapping. The system, co
mposed of the two spiral galaxies NGC 2992 and NGC 2993, is observed at an
early stage of the interaction, about 100 Myr after perigalacticon, though
at a time when tidal tails have already developed. The VLA observations dis
close a third partner to the interaction: an edge-on, flat galaxy, FGC 0938
, which looks strikingly undisturbed and might just be falling toward the N
GC 2992/93 system. Our H I map shows prominent counterparts to the optical
tails. Whereas the stellar and gaseous components of the plume that origina
tes from NGC 2992 match, the stellar and H I tails emanating from NGC 2993
have a different morphology. In particular, the H I forms a ring, a feature
that has been successfully reproduced by our numerical simulations. The H
I emission in the system as a whole peaks at the tip of the NGC 2992 tail w
here a gas reservoir of about 10(9) M-circle dot, about 60% of the H I towa
rd NGC 2992, coincides with a star-forming optical condensation, A245N. The
latter tidal object exhibits properties ranging between those of dwarf irr
egular galaxies (structural parameters, gas content, star formation rate) a
nd those of spiral disks (metallicity, star formation efficiency, stellar p
opulation). Although it is likely, based on our analysis of the H I and mod
el data cube, that A245N might become an independent dwarf galaxy, the dyna
mical evidence is still open to debate. Prompted by the questions raised fo
r this particular object, we discuss some issues related to the definition
and identification of TDGs and highlight some specific conditions which see
m required to form them. Finally, we outline what is needed in terms of fut
ure numerical simulations in order to further our understanding of these ob
jects.