Astron. Astrophys. 350, 529-540 (1999)

5. Discussion

5.1. Excitation mechanisms

In the region we are investigating, which is at the base of the outflow driven by L1448-mm, at least three different excitation regimes are present. In particular CO J=10 and 21 maps show the presence of two different components: a low-velocity outflow ( 20 km s^-1) having a biconical morphology centered on the mm source, and a much faster moving (projected velocity 70 km s^-1) jet-like emission along the axis of the outflow, which consists of a chain of well defined clumps (molecular "bullets", Bachiller et al. 1990). The LWS beam encompasses only those bullets lying closest to the central source (R1 and B1, symmetrically displaced in the red and blue lobe respectively, see Fig. 1). Such bullets also emit strongly in SiO (Guilloteau et al. 1992), a molecule which is detected only in regions where strong shocks are able to release Si from dust grains into the gas phase, thus allowing the enhancement of the SiO abundance by orders of magnitude (Schilke et al. 1997; Caselli et al. 1997). In addition to millimeter line emission, H₂ 10 S(1) emission is also present, with an arc-shaped morphology (knots A and B in Davis & Smith 1995) having the appearance of a bow shock. This region lies at about 30" from the mm source in the blue lobe, and therefore just at the border of the 75" LWS beam and not in the smaller field of view of SWS.

The physical conditions estimated from the LWS spectrum of L1448-mm, can be used to investigate whether the warm gas emission can be associated to any of the above shock environments.We have inferred that the warm component traced by the ISO lines originates from a relatively compact ( 12") region with temperature less than 1400 K. These characteristics exclude the possibility that the bulk of the observed emission comes from the region outlined by the H₂ knots A and B (at the edge of the LWS beam), which are rather extended and excited at temperatures in excess of 2000 K (Davis & Smith 1995). These knots could however be associated with the second CO emission component at higher temperatures, which is traced by the rotational lines with J23. We can also exclude that the observed molecular excitation is associated with the low velocity CO, which appears to have a greater spatial extent and smaller column densities than those derived from our analysis. More likely is that the observed molecular emission is associated with the dense molecular bullets: indeed there are several pieces of evidence in favour of this hypothesis. Firstly, the spatial extent of the molecular bullets closest to L1448-mm is about 215 arcsec² each (Guilloteau et al. 1992), consistent with our derived emission area if we reasonably assume that both the redshifted and blueshifted bullets contribute to the emission. Moreover, the detection of abundant SiO clearly indicates the presence of a shock propagating with 25 km s^-1, i.e. a shock fast enough to remove the silicon from the dust grains (Schilke et al. 1997): these are the same conditions needed to produce water, since the endothermic reactions for its production become efficient at 300-400 K, corresponding to shocks with 15 km s^-1.This interpretation is corroborated by the fact that the high-J CO emission lines observed in other low mass young sources have been successfully explained in terms of high-density gas in the presence of compact shocks with velocities from 10 to 25 km s^-1 (Nisini et al. 1998), indicating that such regions are often associated with outflows, even if they are not necessarily observed at millimeter wavelengths in the form of molecular bullets.

We also consider the possibility that the large water abundance is formed in the original protostellar jet (Glassgold et al. 1991). In favour of this possibility there is the evidence that a large part of hydrogen can be in atomic form, as shown by the high observed value of the CO/H₂ abundance ratio: in addition, the strong water maser spots are aligned with the molecular jet and lie at distances of only 30 AU on each side of the central source, suggesting that they may originate directly from the protostellar jet (Chernin 1995). We note however that under such an hypothesis the emitting region, being located very close to the protostellar source, is fully sampled by both the SWS and the LWS beams, and is best fitted with 1400 K and 3 10⁴ cm^-3. Such a low density is comparable to that of the dense core within which L1448-mm is embedded, and it is unlikely to be found in the innermost regions in the envelope of the protostar, where densities in excess of 10⁶ cm^-3 are expected.

5.2. Comparison with shock models

The complex shock structure with multiple excitation regimes present in the surroundings of L1448-mm renders difficult any detailed comparison with simple shock models. Moreover, a further complication is that the presence of episodic phenomena with timescales as short as few hundred years, would require specific time-dependent modelling. It is however instructive to see to what extent the existing models are able to reproduce the spatial and time averaged physical quantities estimated from the ISO data.

In the previous section it was shown that the observed molecular emission is likely to be associated with the dense molecular jet traced by means of EHV CO and SiO emission. The velocity of such a jet, corrected for the outflow inclination, is about 200 km s^-1 (Bachiller et al. 1995); the leading shock due to such a high velocity jet would be strongly dissociative. Molecules can be copiously formed behind fast dissociative shocks, once the gas temperature has dropped sufficently to allow H₂ to reform (Hollenbach & McKee 1989; Neufeld & Dalgarno 1989). Although hydrogen becomes completely molecular at temperatures of only few hundred degrees, H₂ starts already to be formed for T 4000 K, rapidly followed by the reactions which bring to the formation of OH, H₂O and SiO:

Our derived temperature range, together with the evidence that the hydrogen may be still partially in atomic form, as testified by the derived high CO/H₂ abundance ratio, are compatible with the above scenario. There are however a number of difficulties to reconcile the derived gas coolings with the fast dissociative shock models. First of all the emission from atomic components like [OI ] and [SiII ] is too weak. In particular, in fast shock models with pre-shock densities implied by our observations, i.e. 10⁵ cm^-3, cooling due to O⁰ is never predicted to be less than that of water, as we in fact see in our case where L(H₂O)/L(O) is 4.5. In addition the [SiII ] 35µm line should be only a factor two or three weaker than the [OI ] 63µm emission, while our observed upper limit implies a ratio 35µm/63µm which is 0.07. More important, we note that neither the model by Hollenbach & McKee (1989) nor that by Neufeld & Dalgarno (1989) predict that the production of water behind the dissociative shock can bring the water cooling to exceed the contribution not only of [OI ] but also of CO, unless we consider a very dense shock, with pre-shock density 10⁶ cm^-3. Moreover, the presence of highly abundant atomic hydrogen favours a very rapid destruction of gas-phase H₂O through the reaction .

The observed high abundance of water molecules is predicted only by low velocity shocks influenced by the presence of a magnetic field; in this circumstance, if the shock velocity is lower than the Alfvén velocity of the ions, a continuos shock, or C-shock, is created, where the temperature across the shock structure changes continuously and never rises above 3000 K, thus preventing the molecular dissociation (Draine 1980). In this stuation, and if the shock velocity exceeds 15 km s^-1, water is rapidly formed from the reactions indicated above reaching abundances greater than 10^-4 (Kaufman & Neufeld 1996). To reconcile the possibility of the presence of a slow shock with the observation of gas velocities travelling at hundreds of km s^-1, we could suppose that the gas is travelling into a medium which has already set into motion by the passage of a previous shock; in this case the actual shock velocity (-) may be significantly lower. Alternatively, the slower shock could be produced by jet components which impact obliquely onto the ambient medium.

If we compare the parameters listed in Table 3 with the predictions of a simple planar and static C-shock model like that of Kaufman & Neufeld (1996), we firstly notice that a temperature range 700-1400 K is consistent with a shock velocity of 15-25 km s^-1. For pre-shock densities 10⁴ cm^-3 (implied by our estimated post-shock density of 10⁵ cm^-3), the model predicts the following cooling ratios: L(H₂O)/L(CO) 2-3, L(OH)/L(CO) 0.02-0.05 and L(H₂)/L(CO)2-3, which can be compared with the observed values of about 2, 0.2, and 0.1 respectively. Cooling by water is therefore substantially in agreement with that predicted. The fact that the H₂ luminosity is significatively lower than the predicted values is indicative that with the SWS we are observing much less H₂ than expected, which strenghtens the hypothesis that the small SWS beam has missed most of the emission region observed by the LWS, as already proposed in Sect. 3.3 to justify the anomalous CO/H₂ ratio. This could happen if, for example, the putative shock lies just at the edge of the molecular bullets rather than all along their lengths, as can be also visualised in Fig. 1. More significant is the greater amount of cooling deduced for the OH with respect to the model. However, the fact that the OH abundance has been estimated from only one line may introduce large uncertainties in the derived value which should therefore be taken with caution.

Specific C-shock models which take into account the Si chemical network and which are able to reproduce the SiO profiles observed in L1448 (Schilke et al. 1997) also predict that the water abundance should be as high as measured by us, although the SiO emission is better explained by a shock with densities of about 10⁶ cm^-3, i.e. slightly larger than those estimated from the present analysis.

More realistic models for the investigated regions are those which take into account a curvature and therefore a gradient in the excitation conditions along the shocked region ("bow" shock models). In particular, Smith (1991, 1994) considers both non dissociative J and C-type bow shocks; generally speaking our derived temperature and density, and the relative cooling of the various molecules, can be accounted for by these models but any specific constraints on the shock characteristics and geometry would require a more detailed comparison than we are able to make at the present time.

5.3. Comparison with other outflow driving sources

ISO has allowed us to observe, for the first time, the FIR molecular cooling of several shocked regions associated with outflows in young stellar objects (Nisini et al. 1998; Saraceno et al. 1998). In particular, water emission has only been observed in a few sources and always with abundances of the order of 1-5 10^-5 (e.g. HH54, Liseau et al. 1996; IRAS16293-2422, Ceccarelli et al. 1998; IC1396-N, Molinari et al. 1998; HH25MM, Benedettini et al. 1998). The notable exception to this is the Orion core, where a water abundance of 5 10^-4 has been found by Harwit et al. (1998) (although Cernicharo et al. (1997) derive for the same region a value X(H₂O)10^-5). The interesting aspect is that, although the abundances are fairly constant, the shock conditions inferred for the various sources are rather different; in particular in HH54 and HH25MM gas temperatures of 300-400 K are estimated, indicating shocks propagating with velocities not exceeding 10 km s^-1, while IRAS16293 and IC1396-N seem to have significantly higher temperatures, of about 1500 K. For these two sources, therefore, one would expect abundances as high as we observe in L1448-mm, since the shock conditions are very similar. Explanations for a lower water abundance than expected (with respect to shock models) in these sources have been proposed in Ceccarelli et al. (1998) and Saraceno et al. (1998) and include the hypothesis that most of the oxygen is locked into grains, thus rendering inefficient those endothermic reactions which bring all the available gas-phase oxygen into water, or that the time scale required for this process exceeds the dynamical time scale of the outflow. This last hypothesis is not supported by the L1448-mm observations: the dynamical timescale of the L1448-mm outflow has been originally estimated to be as short as 3500 years (Bachiller et al. 1990). Recently, Barsony et al. (1998) revised this value upwords to 32 000 years, which is still very similar to the age of other Class 0 sources, thus not explaining the apparent overproduction of water in terms of a longer available time for its formation. Beside that, we also remark that, for gas temperatures greater than 400 K, the timescale for the H₂O production from gas-phase chemistry is less than 10³ years (Bergin et al. 1998).

On the other hand the time scales of the individual clumps along the L1448 flow are of the order of 100 years, i.e. much shorter than the total dynamical time. It could be therefore argued that an opposite mechanism would lead to a decrease of the water abundance in older outflows; if the H₂O lifetime is short, the H₂O initially formed could be rapidly lost from the gas-phase either through depletion onto grains and through its conversion into OH. Time-dependent chemical models (Bergin et al. 1998) have shown that, if we consider only the gas-phase chemistry, the shock cooling time is much shorter than the time needed for the gas to return to its pre-shock chemical composition, this latter being in excess of 10⁵ years. If however gas-grain interactions are also included, such a time would be shorter for densities 10⁴ cm^-3 due to a rapid depletion of water onto dust grains. In L1448-mm, the occurrence of multiple shocks over a relatively short interval of time could have conspired to mantain a high water abundance. Indeed, the most recent shock episode is currently propagating into a medium which, having been pre-processed by an earlier shock, has already developed an enhanced water abundance. Specific comparisons with time dependent models of low-velocity shocks would need to be made to understand the influence on the relative cooling ability of the various molecular species over the lifetime of the outflow.

© European Southern Observatory (ESO) 1999

Online publication: October 4, 1999
helpdesk.link@springer.de