The form taken by interstellar silicon in molecular clouds is still a considerable puzzle. Ultraviolet observations (e.g. Sofia et al. 1994) have shown that a large fraction of the silicon is in solid form, even in diffuse clouds, in apparent contradiction with the timescale for destroying silicate grains in shocks in the diffuse ISM. This timescale is so short that the supply of such grains, through mass loss from late type stars, is inadequate to explain the fraction of silicon observed to be in the solid state (see McKee 1989, Jones et al. 1994, Draine 1995, Tielens 1998). A possible resolution of this discrepancy is that processes taking place in molecular clouds are responsible for refurnishing silicon-bearing grains. If this is the case, silicon that is taken up by a molecular cloud in gaseous form is ejected from the cloud in solid form: rather effective depletion of silicon must be taking place within the clouds. Gas phase silicon in molecular clouds is expected to be in the form of SiO (Herbst et al. 1989). However, the observed abundance of SiO is extremely low in some dense dark clouds, with typical upper limits to [SiO]/ of (e.g. Ziurys et al. 1989). If most of the gas phase silicon is in the form of SiO, then silicon is depleted on to grains to a surprising degree (to a much greater degree than C, N and O): the fraction of silicon remaining in the gas phase is less than , on the above assumptions. One concludes that either the silicon depletion processes are surprisingly efficient, and/or the gas phase chemistry models are surprisingly wrong. A related question is: which form does the silicon take in the solid state?
While SiO "vanishes" in some molecular clouds, it reappears under certain circumstances. In particular, SiO has proven to be a good tracer of shocked gas in outflows (see e.g. Schilke et al. 1997, Pineau des Forêts and Flower 1997, Martin-Pintado et al. 1992, Gueth et al. 1998). The SiO abundance in the high velocity gas associated with young outflows is typically of order ; hence, an appreciable fraction of the "lost" silicon has reappeared in the form of SiO. In a previous study (Schilke et al. 1997), we considered the possibility that silicon reappears owing to the sputtering of grains in C-type shocks of speeds between 10 and 40 . We found that 25 shocks with pre-shock densities of order were consistent with the observations. However, it was not clear from that study whether the silicon which returned to the gas phase originated in refractory material or in some form of "ice" in the grain mantle. In the latter case, the supposition was made that the Si-bearing material was present either as silane (, see MacKay 1995, 1996) or as "dirt" mixed into other ice material (perhaps in the form ; an interesting by-product of the Schilke et al. study was the realization that can be the major gas phase form of Si for a considerable length of time in the post-shock material).
In order to decide which form silicon takes within molecular clouds, one clearly needs to study quiescent regions, where shocks do not confuse the chemistry. However, in cold quiescent regions, silicon seems usually to be absent from the gas phase. An alternative approach, which we shall pursue in this article, is to study the silicon chemistry in a PDR (Photon Dominated Region). PDRs are interstellar regions where the gas is predominantly neutral but the gas and dust are heated by UV (1000 Å) photons from neighbouring O-B stars (see the recent review of Hollenbach and Tielens 1997). The dust temperature in PDRs can be as high as 100 K, and hence the evaporation of dust mantles becomes possible. On the other hand, the available evidence suggests that PDRs are quiescent regions: shocks, which might liberate Si from the solid form, are absent.
Detailed PDR models have already been constructed (Sternberg and Dalgarno 1995, Jansen et al. 1995) which include the silicon chemistry; these assumed steady state conditions, as well as a given abundance of gas phase silicon. The latter must be sufficiently high (of the order of 10 percent of the solar abundance of silicon) to account for the observations of the Si+ fine structure line at 35 wavelength which now has been detected in several PDRs (eg Haas et al. 1986, Meixner et al. 1992, Stacey et al. 1995, Steiman-Cameron et al. 1997). It appears to be fairly certain that this line is emitted from a partially ionized (neutral) layer and not from the adjacent ionized gas (see the discussion in Sect. 4 below). There is in particular good evidence for this in the case of the Orion Bar where the spatial structure of the Si+ emission is similar to that of other PDR tracers. The inferred Si+ column density is of order .
The results for the Orion Bar represent moreover an interesting test case. Jansen et al. (1995) put an upper limit on the SiO abundance of [SiO]/[H] less than (N(SiO) ), which corresponds to less than of the solar abundance of silicon. However, in recent observations of the Orion Bar (Schilke et al. 1998), SiO has been detected with the IRAM 30-m telescope at levels consistent with the previous upper limits. For example, at the origin of the strip perpendicular to the ionization front discussed by Jansen et al. (1995), Schilke et al. detect a line (2 width) of SiO(2-1) with intensity of order 0.1 K in the main-beam brightness temperature scale. The emission appears to be extended at roughly this level both parallel and perpendicular to the bar. We estimate that this corresponds roughly to N(SiO) of order or almost five orders of magnitude less than seen in Si+. These results will be presented in detail elsewhere but it is clear that it is difficult to reconcile the SiO measurements with the column density inferred from the fine structure line.
The aim of the present study is to consider possible solutions of the paradox outlined above. We assume that Si is in a form other than gas phase SiO far from the HII region ionization front, which bounds the PDR on one side. We examine the consequences for the silicon chemistry of advection of material toward the ionization front from the adjacent molecular cloud. In Sect. 2 of this paper, we describe the computational scheme which we have adopted and, in Sect. 3, we present our results. Finally, in Sect. 4, we consider the extent to which the "silicon paradox" has been resolved and the implications for future studies.
© European Southern Observatory (ESO) 1999
Online publication: February 22, 1999