1 Introduction

The primary goals of the "Submillimeter Wave Astronomical Satellite'' (SWAS; Melnick et al. 2000) were to observe the abundance and distribution of gas-phase oxygen and carbon in the interstellar medium (ISM). The chemical composition of interstellar clouds is important in studies of ionisation fraction and chemical balance. These observations set out to confirm long standing predictions of gas-phase chemical models that H₂O and O₂ are major reservoirs of atomic oxygen in cold, dense clouds and are important coolants of the gas.

Gas-phase chemical models of quiescent cloud cores typically predict steady-state molecular oxygen abundances of $\sim$10^-5 and water abundances of a few times 10^-7, relative to H₂ (Bergin et al. 1996; Lee et al. 1996a; Millar et al. 1997). Results from SWAS, however, indicate that these species are actually far less abundant. Goldsmith et al. (2000) searched for O₂ in 20 sources, but made no convincing detections. The limits on O₂ abundance they determined were < $3\times10^{-6}$ in TMC-1 and L134N, and <10^-7 towards most other sources. A recent tentative detection of O₂ at a fractional abundance of 10^-5has been made (Goldsmith et al. 2001), but the source is a molecular outflow towards $\rho$ Oph A rather than a dense core. Snell et al. (2000b) detected H₂O towards several star-forming regions, with abundances of only $6\times10^{-10}$-10^-8, and did not detect water in the dark clouds TMC-1, L134N, and B335, with the strongest upper limit being < $7\times10^{-8}$.

The above observations impose a new set of constraints on interstellar chemical models. Bergin et al. (2000) discussed some explanations for the low oxygen and water abundances in dense cores. The simplest explanation is chemical youth, but although this is reasonable for molecular oxygen, a species that takes a long time to reach steady state, it is less reasonable for water, which reaches too large an abundance too quickly. One can assume that the branching fraction for production of H₂O in the dissociative recombination of H₃O⁺ is very low (Williams et al. 1996), but this result is in disagreement with other recent experiments (Vejby-Christensen et al. 1997; Neau et al. 2000; Jensen et al. 2000). Bergin et al. (2000) also considered models with a high C/O elemental abundance ratio and more dynamic gas-grain models in which O and C atoms accrete onto surfaces and form water and methane, respectively. The methane is released back into the gas phase, enhancing the carbon chemistry but the water remains on the grains. Finally, Bergin et al. mentioned other possibilities such as shocks, turbulence, and cycling of material between low- and high-density phases, the latter two mechanisms serving to maintain an apparent chemical youthfulness. The role of shocks has been studied to a greater degree by Charnley et al. (2001). The role of accretion has been explored by Viti et al. (2001), who considered cyclic gas-phase models in which CO and N₂ are allowed to accrete onto grains and return to the gas-phase when other heavy molecules remain on the grains. They also suggested that the high ionisation phase of bistable models provides a natural explanation for the low abundance of oxygen.

Another explanation of the low water and oxygen abundances is based on the low spatial resolution of SWAS. The SWAS beam can be as large as 4 arcmin. Unless the morphology of the region is completely determined, the observed abundances cannot be deconvoluted and may represent averages over areas with differences in physical conditions and/or significant gradients in molecular abundances. Spaans & van Dishoeck (2001) have constructed models of clumpy molecular clouds, both with and without nearby stars. For quiescent sources as well as sources near young stars, they find a significant decrease in the average abundances of oxygen and water compared with homogeneous models. The effect is mainly due to photodissociation. Their agreement with the SWAS results is reasonable for S140 but less so for $\rho$ Ophiuchi. Photodissociation also plays a large role in the explanation provided by Casu et al. (2001) who emphasize that dust coagulation can reduce the far-UV extinction of dust grains sufficiently that photodissociation removes gas-phase water and molecular oxygen rapidly.

The constraints on chemical models of dense cores become more stringent when combined with recent observations of interstellar ices. The Infrared Space Observatory (ISO) found water ice to be abundant in dense clouds, with a fractional abundance of $\sim$10^-4 with respect to H₂. CO₂ is also an abundant ice component towards both active star forming regions and quiescent clouds. Gerakines et al. (1999) found its abundance to be typically 10-20% of water ice, while Nummelin et al. (2001) found CO₂/H₂O to be up to 0.37 in the ices towards some sources. On the other hand, Vandenbussche et al. (1999) searched for O₂ ice towards cold, dense clouds, obtaining only an upper limit of O₂/H₂O <0.2 towards NGC 7538 IRS9.

We have recently developed models of gas-grain chemistry in cold cloud cores which simultaneously follow the chemistry that occurs in the gas and on grain surfaces in some detail (Ruffle & Herbst 2000, 2001a, 2001b). The model has previously been used mainly to understand observations of interstellar ices, but its predictions for gas-phase species are also germane. Despite the fact that desorption, induced by cosmic ray bombardment, is included, results strongly depart from pure gas-phase models at times greater than 10⁵ yr. Indeed, there is a strong interaction between surface chemistry and desorption. It seems reasonable to apply this model to the problem posed by the SWAS observations, especially those objects for which the model was designed - cold, dark clouds. We also use the gas-grain model to study chemical abundances in some warmer regions, also observed by SWAS, to probe the expected water and oxygen abundances in such sources.