Astron. Astrophys. 342, 542-550 (1999)

4. Discussion

In the previous section, we have attempted to throw light on what we have called the "silicon paradox", that is the weakness or absence of SiO in the molecular layers of PDRs, despite strong Si⁺ emission from the neighbouring atomic layer. One might ask how certain it is that the observed Si⁺ emission comes from the neutral gas, rather than from the HII region. We consider this question next and then summarize our conclusions on the silicon chemistry within a PDR.

4.1. Does the observed Si II emission originate in neutral gas?

An assumption which underlies all of the above discussion is that the observed [Si II] 35 emission is produced in neutral gas (the PDR). However, one cannot entirely exclude the possibility that the 35 emission originates within either the HII region or the ionization front separating the ionized and neutral material.

It is instructive, in this context, to compare with observations of iron. Unlike silicon, an iron-containing molecule has yet to be found in molecular clouds (e.g. Turner 1991). Iron is more depleted than silicon in diffuse gas, with a logarithmic depletion of 2.6 (relative to cosmic) in one of the Oph components, as compared to 1.3 for Si (see Sofia et al. 1994). Therefore, it seems reasonable to suppose that silicon is deposited in a less refractory form than iron. On the other hand, [Fe II] lines are observed which peak in intensity close to the ionization front in the Orion Bar (see Marconi et al. 1998 and references therein). Iron is estimated to have an abundance of 20 percent of solar in the Orion HII region (Osterbrock et al. 1992). It seems plausible therefore that an iron-containing component of the dust is destroyed at the edge of the HII region. Whilst the precise composition of such a mantle is not known, it may reasonably be expected to contain Si as well as Fe and thus will be of importance for the Si abundance within the HII region.

However, we do not believe that the main contribution to the observed [Si II] lines comes from the ionized gas. The Stacey et al. 35 [Si II] map and the Wyrowski et al. C91 map are quite similar, and the C I recombination line emission undoubtedly arises in neutral gas (based upon the line width). Thus, the coincidence in spatial distribution suggests strongly that an important fraction of the [Si II] emission is produced in the PDR. Both the [Si II] and the C I emission appear to come from a layer to the SE of the ionized bar (their spatial coincidence is confirmed by a more recent [Si II] map: Stacey, priv. comm.). Thus, we are convinced that Si in some form is released from the mantle within the neutral layer.

4.2. Si chemistry within the PDR

The models which we have presented are not exhaustive but they do contribute to resolving the "paradox" discussed above. Based on our understanding of the chemistry, we find that, in order that SiO should be currently unobservable in PDRs with large Si⁺ column densities, silicon has to be ejected from the solid into the gas phase close to the ionization front. The models suggest that such a transformation occurs at depths corresponding to a visual extinction of around 2-3 magnitudes. In this case, any SiO produced is photo-dissociated rapidly. The most likely mechanism responsible for the ejection of Si into the gas phase seems to be photo-desorption, but there remains a large uncertainty in the yield, and laboratory studies would be very useful.

A sensitive issue in our analysis (as one sees comparing models 7 and 8) is the adopted temperature profile. This is critical for OH production and hence, in our model, for SiO. There is a well-known discrepancy between theoretical and observed temperature distributions in PDRs (Lis et al. 1997), in that temperatures inferred from molecular lines tend to be higher than model predictions. In view of the theoretical uncertainties, we have adopted the observed values, although these are not without their own internal contradictions. Our adopted distribution (1000/(1 + 2A_v)) is based on the recent spectroscopic results of Luhman et al.(1998). It is consistent, close to the ionization front, with the v=0 results of Parmar et al. (1991), who estimated a value of 1000 K at the front. At large depths (A_v of order 7 mag. for an edge-on geometry), where the gas becomes molecular, Hogerheijde et al. (1995) estimate a temperature of 85 K on the basis of formaldehyde observations, which is again roughly consistent with our adopted profile. On the other hand, at depths corresponding (for a density of ) to A_v=5, Parmar et al. find much higher temperatures (500 K) than predicted for this depth. Also, the measured ¹²CO(6-5) brightness temperature of 175 K (Lis et al.) shows that some molecular gas is much hotter than estimated on the basis of formaldehyde measurements. The latter result and some of the data may be influenced by emission from hotter gas in dense clumps, but convincing evidence in favour of this view is lacking, in our opinion. We conclude that our adopted temperature distribution is reasonably consistent with current data and, if anything, tends to underestimate the kinetic temperature.

Another question that one might pose considering our results is the validity of our assumptions concerning the mode in which silicon comes into the gas phase. The comparison for example of models 6 and 8 shows that the ejection of Si in atomic form rather than as SiO causes the expected SiO column density to go down by more than a factor of five. On the other hand, we have considered variations of model 5 in which the silicon mixed with water ice is in a variety of forms (, , , ) and derive SiO column densities which differ by a factor of 2 or less from that given in Table 1. We conclude that the form in which Si comes back into the gas phase is not of great importance, although the most favorable case, from the viewpoint of reducing the expected SiO column density, is ejection as atomic Si.

Although our analysis suggests that silicon is probably not a significant constituent of ice mantles, our models 2-4 show that this possibility cannot be completely ruled out. We note that, in this case, the corresponding solid state spectral features might be observable in the infrared. However, the comprehensive ISO spectra of NGC 7538-IRS9 (Whittet et al. 1996) show no sign of any features other than those attributable to silicates. It has been argued that the so-called `XCN' feature at 4.6 might be due to SiH (Moore et al. 1991), but this proposal now seems unlikely to be correct (see Sandford 1996). On the available evidence, the column density of SiH ice towards sources such as NGC 7538-IRS9 is less than (for an equivalent width less than 10 cm^-1 and an absorption strength of cm molecule ^-1; Nuth and Moore 1988). The corresponding limit on the "Si-H" abundance is , which is a factor of 3 below that required to explain the fine structure line emission. Given the uncertainties, we cannot exclude the possibility that the evaporation of silane, associated with e.g. CO₂ ice, is responsible for the observed [Si II] line emission in PDRs, but we consider it to be unlikely. Alternatively, Si could be in the form SiO or SiO₂ in the ice mantles. In this case, one might hope to see a narrow feature at around 8 due to the SiO stretch, although the nearby silicate feature would make detection difficult. Nevertheless, a narrow (10 cm^-1) feature might be detectable in ISO spectra awaiting analysis (Tielens, priv. comm.).

In our opinion, a more likely explanation of the "silicon paradox" is that Si is present in a somewhat more refractory form (e.g. Tielens 1998) in grain mantles which are partially destroyed by the UV photons incident upon the PDR. Our model 8, in which Si enters the gas phase in atomic form, is consistent with current observations; a detailed comparison with the results of Schilke et al. (1998) is however needed. We find that the SiO abundance as a function of depth is closely linked to the temperature profile, and it will be useful in the future to compare with the temperature dependence derived from tracers such as . It is clear also that laboratory determinations of the photodesorption yield for Si-containing species are needed. The present calculations show that a rather small yield, of order , can have important astrophysical consequences.

4.3. Ice mantles in PDRs

A by-product of the present investigation has been the clarification of the relative importance of various processes which destroy ice mantles within PDRs. Our results show, for example, that apolar ice mantles with binding energies of order 0.1 eV will evaporate far from the ionization fronts of regions such as Orion and M17. Thus, one can expect that, in such regions of high mass star formation, CO, and will be in the gas phase, even if water ice is still present. Furthermore, using the Westley et al. (1995) measurements of the photodesorption yield of water, we find that direct photodesorption can remove even polar ice mantles. For the polar component, photodesorption is more important than evaporation; this has implications for the structure of PDRs, as the abundances of water and other coolants then increase close to the PDR surface (at A_v of order 6 in the Orion Bar model).

© European Southern Observatory (ESO) 1999

Online publication: February 22, 1999
helpdesk.link@springer.de