The interacting system Arp 105 gathers several noteworthy properties, among others, a complete spatial segregation between its different gas components, a perturbed dynamics of its tidal tails, the accretion of gas by an elliptical galaxy and the formation of dwarf galaxies out of tidal debris. Each of the above mentioned items will be discussed in turn. Also, an atemot will be made to sketch the 3D orientation of the objects involved.
4.1. Origin of the gas segregation in Arp 105
One of the most striking properties of the system Arp 105 is the complete segregation between the HI and CO gas distributions. Huge quantities of HI are found along and at the tip of two tidal tails ( , and ), while no atomic gas, up to a limit of , is detected towards the spiral galaxy. In contrast, massive quantities of molecular gas, , are found in the core of the spiral and nowhere else.
The close association of the HI clouds with the optical tidal tails, as suspected in Paper I on the basis of single dish observations and confirmed by our VLA data, clearly suggests that the HI depletion in the core of the spiral was tidally induced by interactions. In that respect, Arp 105 is similar to classical colliding and merging galaxies. Hibbard & van Gorkom (1996) have mapped in HI several systems of the Toomre Sequence of colliding galaxies ( Toomre (1977) ) and have shown that the atomic gas extends along tidal tails from 2 to 7.5 times their standard blue radius, and is deficient in the central regions.
Other characteristics which NGC 3561A has in common with merging galaxies are the high quantities of molecular gas in its core, and the compactness of its distribution (e.g. Young & Scoville (1991) ). In the central 1 kpc, we measured a mass surface density of M pc-2, in the range of that of ultraluminous infrared galaxies (see review by Sanders & Mirabel (1996) ). As noted in Paper I, NGC 3561A is luminous in the infrared, and has a far infrared to molecular mass also typical of starburst galaxies. A straightforward explanation for the lack of HI in the inner region and the unusual quantity of H2 there is that, as a result of shocks following the encounter, the atomic gas dissipates energy, loses angular momentum, and sinks to the central regions where it is transformed into a molecular phase. From numerical simulations, Hibbard (1995) estimates that in the merger NGC 7252, at least half of the initial gas may have been collected in the core of the galaxy.
Besides gravitational forces, other environmental effects, due to the influence of the intergalactic medium, could contribute to the peculiar distribution of the gas in Arp 105. In particular, ram pressure stripping by the intracluster gas ( Gunn & Gott (1972) ) has been considered to account for HI truncated disks or even anemic disks, observed in cluster spirals (e.g. Cayatte et al. (1994) ). Although Arp 105 belongs to a rich cluster, the density of the intracluster hot gas at its location, at the edge of the cluster X-ray emission mapped by the Einstein satellite, is too low ( Mahdavi et al. (1996) ), and therefore ram-pressure far too weak, to induce the removal of a cloud as massive as from a spiral galaxy. However, the clear S-N asymmetry of the column density profile, shown by Cloud A105N, could well be due to gas compression at the interface between the moving, tidally expelled, HI cloud and the intergalactic medium. Phookun & Mundy (1995) have argued that ram pressure was responsible for a similar asymmetry, observed in the stripped HI disk of the Virgo galaxy NGC 4654.
Another effect which might play a role is ionization by a background radiation field. This mechanism was originally proposed by Maloney1992) and Dove & Shull (1994) , to account for the sharp decrease of the HI column density in the outskirts of spiral galaxies. Here we would be dealing with a similar, but spatially reversed phenomenon. When moving from the northern tip of the atomic hydrogen cloud coinciding with A105N towards the spiral galaxy, NGC 3561A, the HI surface density declines along the northern tail, to a point where the gas becomes so dilute that it could get ionized by the interstellar radiation field from sources within the spiral galaxy.
Still, in Arp 105, cluster environmental effects have, in comparison with tidal forces, only played a marginal role to shape the overall gas distribution. In Sect. 4.3, we will describe in more detail the interaction taking place in Arp 105, and present our best guess for the relative positions of the objects and their orientations (see sketch, Fig. 9).
4.2. Kinematics of the tidal tails
From the smooth spatial distribution of HI along the tidal tails, it is tempting to argue that all the HI clouds detected in Arp 105 have been expelled from the spiral galaxy, and form a unique structure. In that case, the complex kinematics of the system has to be understood.
4.2.1. The kinematics of the northern tidal tail, A105N
In Sect. 3.1.2, we already pointed out that Cloud A105N is made of several kinematically distinct components. Inspecting the HI data cube along the velocity axis, we can distinguish in the position-velocity (PV-)diagrams (see Fig. 8) at least two systems. One seems associated with the optical tidal tail. Its structure which appears as a partial ellipse in the PV-diagrams is perhaps best visible in the panel labeled 20 at . The highest (recession) velocities are found mid-way between the spiral and the tip of the tidal arm, with lower velocities on either side. Such a highly curved structure in position-velocity space is a consequence of the interaction, and resembles that observed by Hibbard et al. (1994) in NGC 7252. Hibbard & Mihos (1995) did extensive numerical modeling of the interaction in NGC 7252 and could explain in reasonable detail its structure. They found that energy and angular momentum have a monotonic relationship along the arm which naturally leads to a relationship between radial period and pericentric distance. A curved tail seen edge-on would have a projected velocity field similar to that observed in the first component of the HI tail of Arp 105.
The second component seems to be associated with the dwarf galaxy near the tip of the northern arm. We can identify this system in the position-velocity diagram labeled 24 (taken at ). This feature is oriented almost exactly North-South, which implies that we are seeing this subsystem almost edge-on. It seems to be extending from and a velocity of 8640 km s-1 to and a velocity of 8770 km s-1. The PV-map would therefore suggest the presence within the tidal tail of a kinematically decoupled cloud, having a gradient of about 130 km s-1 over a distance of 40 kpc. If we assume that we are dealing with (solid body) rotation, we derive a dynamical mass of of . Most of the HI would belong to this feature and, corrected for He and heavier elements, the gas mass associated would account for about 25% of the dynamical mass.
All the above calculations were made assuming that we could separate perfectly both HI components which, given the limited VLA resolution, both spatially and in velocity, is clearly not the case. The spatial extent, and velocity gradient of the independent feature, are therefore rather uncertain. The dynamical mass, derived above, should be considered as an upper limit only. It supersedes the one quoted in Paper I, estimated from single-dish observations.
4.2.2. The kinematics of the southern tidal tail, A105S
Fig. 7 gathers all the multiwavelength information available regarding the velocity distribution of the southern region of Arp 105, along a South-North axis, encompassing A105S, the optical filament, the elliptical galaxy, the spiral galaxy, and the beginning of the northern tidal tail. Plain dots represent the velocity measurements of the ionized gas, determined from optical longslit spectroscopy. Plain lines represent the HI velocity profile, along the S-N axis, for both emitting and absorbing components. Finally the CO velocity curve along the S-N axis is shown with a dashed line. Table 7 gives for each component the velocity measured on their optical nuclei and, in parentheses, the total velocity range, for the H , HI and CO observations.
Table 7. Velocity field
The velocity field of the southern region is strikingly complex. Cloud N3561B seen in absorption in front of the elliptical, and Cloud A105S seen in emission, coinciding with A105S, have mean velocities which differ by 170 km s-1. The optical velocity of NGC 3561B, in turn, is lower by 250 km s-1 than that of Cloud N3561B. The ionized gas in the blue compact galaxy, A105S, features a strong velocity gradient.
All these features raise the question whether the different clouds observed along the tidal tails of Arp 105, especially in the southern one, belong to the same overall HI distribution, are overlapping each other, or are kinematically and spatially distinct.
Hypothesis 1: A105S and N3561B are two independent clouds
In the following, we assume that Clouds A105S and N3561B are independent, having different mean velocities (resp. 8890 km s-1 and 8720 km s-1 ). Both clouds might spatially overlap along the line of sight towards NGC 3561B. Cloud A105S is seen in emission along the southern tail up to the nucleus of the elliptical, where a faint emission feature can still be seen (see Fig. 4). Its weakness could result from a low intrinsic HI column density there and/or contamination by absorption. Extrapolating the HI column density profile of Cloud A105S, assuming that like in Cloud A105N, the flux declines linearly along the tail when approaching the spiral, and taking into account the fact that no HI is seen in emission north of the radiosource, we get at the position of the radiosource an upper limit of 2 mJy beam-1. The faint emission feature is only 1 mJy beam-1 which suggests that it might be attenuated by absorption against the bright nucleus of the elliptical. In that case one can in principle determine the spin temperature, , of the gas, using the method described by Dickey et al. (1992) . is given by:
where , the optical depth, can be determined from the formula:
In these equations, , is the continuum flux, , the line flux measured on the radiosource, affected by the absorption ( ), and , the line flux that would have been measured at the position of the radiosource if there were no absorption ( ). The factor 0.82 results from the gain of the telescope and depends on the array configuration (and thus beam size). With the known flux density of the continuum source of 42.4 mJy we derive an optical depth less than 0.02, and a spin temperature higher than 100 K.
Cloud N3561B is only seen in absorption. No HI in emission outside the radiosource was detected at the velocity of the absorption line, up to a limit of 0.5 mJy beam-1. This means that either the HI column density of cloud N3561B drops sharply outside the radiosource, or that the gas temperature is extremely low, so that the HI cannot be seen in emission. Given our sensitivity, the lowest column density measurable in emission at 2.5 is . is given by:
Assuming , we derive at the absorption peak. Since will be greater or equal than 10 km s-1, our VLA velocity resolution, one obtains, from Eq. 3, an upper limit for the spin temperature of 20 K.
In the above calculations, it was assumed that the two HI clouds were extended enough Cloud A105S in the direction of the elliptical, and Cloud N3561B outside of it so that the determination of spin temperatures is relevant. If that hypothesis is correct, it implies that both clouds have rather different physical properties and origins: they could have been expelled from several regions of the spiral, during different passages of the elliptical; the latter galaxy could also have accreted HI gas from other cluster galaxies.
Hypothesis 2: N3561B and A105S form a single cloud
We now make the hypothesis that N3561B and A105S form a single HI cloud, corresponding to the counter tidal tail of NGC 3561A. We can estimate the spin temperature, under the simple, naive assumption that it remains constant all along the tidal tail. Like in hypothesis 1, we take for the interpolated emission flux, an upper limit of . The absorption flux at its peak is . From Eq. 2, we derive , and from Eq. 1, .
The different velocities measured for Clouds N3561B and A105S imply that the HI tail is pulled strongly towards the elliptical. Taking into account the VLA beam size, and the fact that the HI in emission, at the velocity of A105S, is seen up to north of the radiosource, the velocity gradient should be of order 170 km s-1 over less than . Even if this drop along the tail is not directly visible because of the limited spatial resolution, it should appear in the HI spectrum as an asymmetry towards the higher velocities, in addition to a similar wing seen towards the blue. This is definitely not seen (see Fig. 4).
An alternative interpretation for the velocity field presented in Fig. 7 is that the southern HI tail might have a roughly constant mean radial velocity, equal to that measured for Cloud N3561B, . An argument in favor of that hypothesis is that the optical HII knots seen along the southern tail north of A105S as well as the very tip of the filament, have velocities that differ from by less than 50 km s-1, and therefore might be associated with a single HI tail. Only the gas near the elliptical, which seems to accrete part of it (see Sect. 4.4), and in the very core of A105S might have a discrepant behaviour. The velocity determined for Cloud A105S, 170 km s-1 higher than , is coincident with the centre of the rotation curve measured in H on the dwarf galaxy. Due to the large VLA beam size, direct evidence for rotation of the atomic hydrogen cannot be provided. However the HI spectra show "blue" wings that are indicative of a global motion of the gas (see Fig 2). Hence our velocity data might support the idea that HI along the southern tail has been gravitationally pulled away, to form the blue compact galaxy A105S. The latter object would then have already gained its kinematic independence.
Obviously the lack of spatial resolution for the HI observations make it difficult to give a definitive answer to the issue whether the HI clouds seen south of NGC 3561A belong to the same tidal tail, were ejected from different regions of the spiral, or have different origins, although the interaction has undoubtly played a crucial role in their formation. Unfortunately, VLA configurations which provide better spatial resolution require to get a high enough signal-to-noise ratio much higher fluxes than those seen in Arp 105.
4.3. A 3D view of the interaction
From the shape and the kinematics of the tidal tails, it is possible to build a rough overall picture of the interaction.
One should note first that we are dealing here with a fundamentally different situation than discussed most often in the literature, namely the encounter between an elliptical and a spiral galaxy, rather than an interaction between two spirals. In the latter situation, two giant tails spread out, one from each disk. Shorter counter-tails form on the opposite side of each galaxy, and then rapidly collide and merge (e.g., Barnes (1988) ; Hibbard (1995) ; Mihos & Hernquist (1996) ). In a spiral-elliptical collision, only one long tail and a short counter-tail form from the spiral ( Elmegreen et al. (1993) ). Going into more detail, numerical models of such collisions are characterized by the sparseness of material between the spiral and the elliptical, and the accumulation of tidally expelled material in the vicinity of the elliptical. A good example for this is presented by Barnes & Hernquist (1992) who, in their Fig. 3, show the fate of one disk as it is perturbed by its partner, which is approximated by a point-mass. The similarities between this case and Arp 105 are remarkable. In the latter system, the northern arm is the main tidal feature; a very diffuse optical counter-tail is seen between NGC 3561A and NGC 3561B (Paper I), and HI clouds have gathered south of NGC 3561B.
From the comparison of the straight curve of the northern arm, with the highly curved tails produced by numerical models, one can conclude that the interaction taking place is seen almost edge-on (i.e., the observer is located close to the plane of the interaction). This is confirmed, as shown in the previous Sect. 4.2.1, by the kinematics of the northern tail.
We must be past perigalacticon, but still within the first passage. The tidal arm is still developing and most material (gas and stars) is moving outward and has not started to fall back towards the spiral. Relative to the observer, the elliptical is coming from behind the spiral and is now moving towards the observer; the spiral is moving away with respect to centre of gravity of the system. The tidal arm is still moving mostly away from the spiral, no gas having reached its maximum separation from the spiral yet. The elliptical is likely embedded within the counter-tail. A possible sketch for the interaction is presented in Fig. 9. In this figure, the kinematically decoupled components seen in the gas clouds have been represented, and their sense of rotation indicated. A105N is counterrotating with respect to NGC 3561A; A105S had the same sense of rotation as the spiral.
Without a more thorough analysis including numerical simulations, it will be difficult to define more precisely the true three dimensional shape and the motions of the gas in the tails. Arp 105 should be an excellent case for numerical modeling as it is in in an early stage of an interaction between a spiral and an elliptical, an encounter which results in only one major tidal arm, thus avoiding the chaos normally seen in disk-disk collisions.
4.4. Gas accretion towards the elliptical galaxy
The HI absorption spectrum towards NGC 3561B discloses the presence of atomic hydrogen in front of the elliptical. The shape of the line a blue wing towards the systemic velocity of the galaxy suggests inflow towards the nucleus. In addition, indirect signs of former accretion by the elliptical are seen. The optical image exhibits dust lanes (see Fig. 3 in Paper I) and ionized gas was found in the core of the galaxy, which is classified as a LINER from optical spectroscopy. Our interpretation is that part of the HI may feed a central engine.
The mass of the absorption cloud associated with NGC 3561B may be estimated from the following formula:
R is the cloud size; the column density is given by
Finally taking values characteristic of Cloud N3561B, one obtains:
For a typical age for the formation of the system of 1 Gyr and taking reasonable values of between 10 K and 100 K, the accretion rate would be of the order of 0.2-2 M yr-1. Clouds of atomic hydrogen associated with early type galaxies, ellipticals or lenticulars, are now routinely detected. From a sample 26 ellipticals, Huchtmeier (1994) has derived a mean HI mass to blue luminosity, equal to 0.01, to be compared to , for NGC 3561B.
It is commonly believed that the atomic gas detected in early type galaxies has an external origin: for instance, gas rich dwarf galaxies which are swallowed up or HI clouds torn away from another galaxy ( Knapp et al. (1985) ; Wardle & Knapp (1986) ), although most of the time all traces of the suppliers of this gas have disappeared. As discussed in the previous Sect. 4.2.4, it is likely that the HI detected towards NGC 3561B, was tidally pulled out from its companion, NGC 3561A. Arp 105 could therefore provide direct evidence of a transfer of gas between a spiral and an elliptical galaxy.
4.5. Properties of the tidal dwarf galaxies
Recently an exciting result has emerged from various studies of the environment of interacting systems: the discovery of a particular class of dwarf galaxies: the tidal dwarfs, formed out of material pulled out from colliding galaxies. These recycled objects, often found at more than 100 kpc from their parent galaxies, are currently undergoing vigorous star forming episodes ( Schweizer (1978) , Mirabel et al. (1992) , Yoshida et al. (1994) , Duc (1995) ). They are also characterized by a rather high metallicity for their mass ( Duc (1995) ).
With arguments based upon the optical morphology of Arp 105, we pointed out in Paper I that the two small galaxies A105N and A105S were likely to have a tidal origin. Our VLA observations presented here stress this even more. First, the overall structure of the HI tails is remarkably similar to the distribution of the tidal material observed in classical interacting objects, in agreement with numerical models of spiral-galaxy encounters. If, on the contrary, A105N and S existed before the interaction between NGC 3561A/B, both hosting high quantities of HI, how could such low-mass objects have been able to hang on to it, whereas a massive spiral, like NGC 3561A, has almost completely lost its gas?
The HI and CO data reported here allow for a better characterization of the gaseous properties of tidal dwarfs. With an HI mass to blue luminosity ratio of 1.3 and 0.5 M /L , A105N and A105S may be classified as gas rich. The young tidal dwarf galaxies have not yet exhausted their gas supplies, and are currently forming their first generation of stars (excluding their old star population, pulled out from the parent galaxies). Since active star formation is taking place, one would expect molecular gas to be abundant as well. We did not detect with the IRAM 30-m single dish antenna (see Paper I) any CO in A105N. Other studies (e.g. Smith & Higdon (1994) ) also failed at finding molecular gas in tidal tails. Obviously due to the poorly known conversion coefficient between the column density and the CO flux in galaxies other than the Milky Way, any conclusion regarding the real absence of molecular gas in tidal dwarfs is premature. Classical dIrr galaxies seem also to be deficient in CO gas (e.g. Gallagher & Hunter (1984) , Israel et al. (1995) ). One explanation for this deficiency is their low metallicity, and therefore low content of CO. In the case of tidal galaxies, this argument is less relevant. Due to their very nature, "recycled objects" cannot be metal poor ( Duc (1995) ).
The HI gas clouds of Arp 105 have high velocity dispersions; the mean is 26 km s-1 for A105N, and 20 km s-1 for A105S. In A105N the velocity dispersion reaches much higher values in some places, but as emphasized before, this is due to line blending of a multiple system. For comparison a typical value of in isolated spiral galaxies is less than 10 km s-1. High velocity dispersions in tidal tails are actually predicted in a numerical model of tidal dwarf formation proposed by Elmegreen et al. (1993) . Following the Jeans criterion, high mass clouds, of , can only be formed if they have of the order of 20-30 km s-1. The asymmetry of the HI column density profile of Cloud A105N (Sect. 4.1) suggests that pressure exerted by the intra-cluster medium (ICM) may play a role in the collapse of these tidally expelled clouds. It might be useful to carry out X-ray observations in order to confirm the presence of shocks at the interface between the HI and the ICM.
Our VLA and H observations are consistent with the idea that the two tidal dwarfs A105N and A105S are already kinematically independent objects. They provide for the first time evidence that some tidal objects might be self-gravitating, and therefore be "true" galaxies. A105N/S show possible signs of rotation, that have to be confirmed. In Sect. 4.2.1, we have derived an upper limit for the dynamical mass of A105N of . According to Paper I, the blue luminosity is . Therefore if A105N is indeed self-gravitating, its ratio would be less than 5. In A105S, the lack of spatial resolution makes any mass determination based on HI data unreliable. However using the rotation curve determined in that object from H observations, we estimated in Paper I a dynamical mass of , and an ratio of the order of 1. These mass to luminosity ratio are indicative of a low dark matter content, in agreement with the predictions of numerical models by Barnes & Hernquist (1992) .
© European Southern Observatory (ESO) 1997
Online publication: October 15, 1997