7. Discussion and conclusions
As shown by the present CO study, the distribution and kinematics of molecular gas in the center of NGC 3593 is characterized by a gaseous ring (the CND), which hosts a massive dusty starburst, and a one-arm spiral which starts at the edges of the CND, opening in the disk as a slow (0) leading arm (the S-R feature). The small pitch angle of the spiral might change with radius; there might be also trailing arcs superposed in the disk (the N-R feature).
The observed characteristics of the gas instabilities can be now compared with the results of our numerical simulations, adapted to the particular case of NGC 3593, with a two-fold aim: 1) to select the mass model which best fits the observations and 2) to understand the evolutionary status of the gas accretion process in NGC 3593. Although an accurate fit between the observations and any of the two models would be unrealistic, an overall comparison is already illustrative.
According to the results of numerical simulations, both models develop marked perturbations during the entire run. These modes are preponderant during the so-called transitory regime (T=0-500Myr), especially in the halo-dominated model, where the inner nucleus (rrcrit) hosts a stationary (0) one-arm wave leading with respect to the most massive component in disk II (using the adopted NGC 3593's nomenclature). The wave propagates outwards through rcrit, being leading with respect to the gas at all radii. At the end of the transitory regime we have a mixture of trailing+leading modes which share a low (see Fig. 11). Only during the transitory phase of the halo-dominated run, the gas response for rrcrit agrees with the observations. As expected, in the halo-dominated model modes develop more efficiently for the gas and the stars and they also last longer than in the disk-dominated model. The latter is a less plausible scenario to account for the observed CO arm. No nuclear ring has formed in the gas disk at this first stage of evolution. However, as all modes of the transitory regime have 0, they have no Inner Lindblad Resonances and the absence of rings in the models is hardly surprising. More importantly, as the gas remains inside corotation for the leading arm (the main instability), it should be gaining angular momentum from the wave. Therefore, the leading instabilities at this stage of evolution cannot cooperate in the formation of the ring.
Simulations do show the formation of gas rings linked with Lindblad resonances of modes, but these appear only after the transitory regime (for T500Myr). As expected, the bar+two-arm spiral instability is more apparent in the disk-dominated model. The modes follow the rotation sense of the majority of the mass, with negative pattern speed in the center, and positive in the outer parts. For both mass models the waves are coupled to slow modes which are always trailing with respect to the counter-rotating gas, and leading with respect to the main stellar rotation for rrcrit. Therefore, neither of the simulations reproduces the observed winding sense of the CO one-arm spiral at this epoch. In summary, gas rings connected to modes appear too late (for T500Myr) to allow for the survival of leading modes in the gas, which are only present at T500Myr. Moreover, the latter are stable enough only in the halo-dominated model. Most noticeably, the absence of a flat profile in the NIR images of NGC 3593, typical of barred systems, argues also against the disk-dominated model (Moriondo et al. 1998b). To resolve the paradox we should envisage other mechanisms of ring formation which could act efficiently on smaller time scales.
Alternatively, formation of rings in counter-rotators has been suggested to be an end product of a process involving large-scale collisions between the infalling counterrotating gas and the directly rotating gas of the accretor galaxy (Thakkar et al. 1997; Corsini et al. 1998). Since the two components have opposite rotating directions, there can be angular momentum annihilation when they get in contact. Therefore large amounts of gas can fall towards the nucleus and form a molecular gas ringed disk, whose radius would depend on the mass ratio of the two components. Gas should be swept up from the disk at intermediate radii creating a hole. In the course of this process, a starburst might be triggered in the nuclear ring. The time-scale for gas infall is extremely short, being close to the free-fall time, i.e., Myr. Therefore, the formation of a nuclear ring, the onset of the starburst and the leading solution for the gas instabilities outside the nucleus could coexist simultaneously.
NGC 3626 is another example of a counterrotating galaxy where a leading spiral wave and a nuclear ring are present in the disk. In NGC 3626, 12CO(1-0) emission is concentrated in a compact nuclear disk of radius r1.2kpc. The distribution and kinematics of the molecular gas indicates a density wave response in the form of an asymmetrical two-arm spiral pattern, suggesting an interplay between and modes (see García-Burillo et al. 1998). The nuclear ring of radius 10" is seen in ionized (Haynes et al. 2000), atomic (Haynes et al. 2000) and molecular gas (García-Burillo et al. 1998). The position of the dust lane going across the nucleus of NGC 3626, allows to identify the west side as the near side. According to the CO velocity field, the asymmetric spiral pattern in NGC3626 would be also leading with respect to the gas.
High-resolution maps of counterrotating disks are scarce, so observational support for the different scenarios is still insufficient. However, the few studied galaxies share a characteristic radial distribution of neutral gas. The bulk of H2 resides in circumnuclear disks, and it is practically absent outside the nuclear regions. (NGC 3626: García-Burillo et al. 1998; NGC 3593: this work and Wiklind & Henkel 1992). Recently, Haynes et al showed that the bulk of the counterrotating HI gas in NGC 3626 is in an outer ringed disk, which extends well beyond the optical radius of the galaxy. The outer HI disk has a different orientation than the rest of the galaxy. Although a smaller HI disk is associated with the CO circumnuclear disk, there is a hole in the gas distribution between the two HI disks, which is not filled by H2. At least for NGC 3593 and NGC 3626, it seems the accretion process ended up forming circumnuclear H2 disks of radius500-1000pc. Outside these compact sources, the counterrotating gas displays leading and instabilities that would last only for T500 Myr. Beyond the optical disk of NGC 3626, the existence of a decoupled HI disk betrays the old accretion episode. In the case of NGC 3593, a starburst has been onset in the CND. In view of the SFR estimated in this work for the CND (SFR0.6-0.7yr-1), all the molecular gas would be converted into stars in 600 Myrs. It is likely that NGC 3593 has accreted a gas-rich dwarf satellite 1 Gyr ago, and thus recent stars had time to form a central counter-rotating disk in the settling gas.
© European Southern Observatory (ESO) 2000
Online publication: December 5, 2000