Introduction

Oxygen is the most abundant element in the Universe after hydrogen and helium and it is therefore important to determine its main reservoirs in the ISM. Steady state chemical models have suggested that the major oxygen-bearing species in molecular clouds should be CO, O, O$_{2}$ and H$_{2}$O (e.g. Lee et al., 1996). However, measurements with the Submillimeter Wave Astronomy Satellite (SWAS) have demonstrated that in most regions the gas phase O$_{2}$ and H$_{2}$O account for less than 1% of the available oxygen (e.g. Bergin et al., 2000, and references therein). In contrast, measurements of the atomic oxygen fine structure line at 63 $\mu$m have indicated that the atomic phase may be where most of the oxygen lies.

The fine structure lines of atomic oxygen are due to magnetic dipole transitions between electron spin states within its ground electronic level. This has an electron configuration of $1s^{2}2s^{2}2p^{4}$ where the letters $s$, $p$, $d$...indicate electrons with orbital angular momentum quantum numbers, $l=0$, 1, 2...The superscript denotes the number of electrons in each shell. In the unfilled $2p$ shell the resultant electron spin ($S$) and total electron orbital angular momentum ($L$) quantum numbers can both have non-zero values. In this case the interaction between spin and orbital angular momentum causes a splitting of each state into $2S+1$ levels. The lowest energy state where this occurs has $S=1$ and $L=1$, with spin-split levels denoted by $^{3}$P$_{0}$, $^{3}$P$_{1}$ and $^{3}$P$_{2}$ where the subscript gives the total angular momentum quantum number of the electrons (orbital $+$ spin) and the superscript gives the number of spin-split components ($2S+1$). The energy level diagram for these states is shown in Figure 5.1.

Figure 5.1: Diagram showing the fine structure lines in the $^{3}$P state of atomic oxygen and the $^{2}$P state of singly ionised carbon. The separation of the levels is not shown to scale.

The atomic oxygen 63 $\mu$m line has been detected in absorption against several bright background sources showing that large column densities of OI exist in many lines of sight. Poglitsch et al. (1996) observed this line towards the DR 21 star forming region using the KAO. They presented high-resolution data showing the 63 $\mu$m line in absorption in a foreground molecular cloud. The relative abundance of atomic oxygen in this cloud indicates that most of the gas phase oxygen must be in atomic form. The same effect has also been observed in several other lines of sight. Baluteau et al. (1997) observed the [OI] line in absorption towards Sgr B2 with the ISO LWS grating mode and derived a lower limit for the atomic oxygen column density of 10$^{19}$ cm$^{-2}$. This implies that more than 40% of interstellar oxygen is in atomic form in this line of sight. Kraemer et al. (1998) have also observed the [OI] 63 $\mu$m line in absorption towards a FIR continuum source in the massive star forming region NGC6334. In addition they observed anomalously low [OI] fluxes in emission from several other positions within NGC6334 implying that there is self-absorption taking place. This has implications for models of PDR regions where the atomic fine structure lines of OI and CII are used as diagnostics of the physical conditions. Until the models account for this possible self-absorption by treating the full radiative transfer through the cloud, the [OI] 63 $\mu$m line may not be a reliable PDR diagnostic (Kraemer et al., 1998).

Even higher abundances of atomic oxygen have been implied by comparing the OI in the molecular parts of clouds with CO measurements. Standard models predict a steady state ratio OI/CO$\sim1$ (e.g. Lee et al., 1996). However, Caux et al. (2002) compared the observed [OI] 63 $\mu$m emission towards the L1689N molecular cloud with measurements of CO and found a very high OI/CO ratio $\sim50$. This implies that up to 98% of the oxygen is in atomic form in the gas phase. High OI/CO ratios have also been found in molecular clouds along the line of sight towards the star-forming region W49N (Vastel et al., 2000). Only a small fraction of the [OI] 63 $\mu$m absorption was observed to be associated with atomic hydrogen clouds, implying a high atomic oxygen abundance in the molecular clouds with OI/CO$>15$. In this case gaseous oxygen would be almost entirely in atomic form and carbon would be deficient with respect to the local ISM. This may be via CO/CO$_{2}$ depletion onto dust grains.

It is therefore important to separate the various layers within PDR regions to determine where the atomic oxygen lies. This chapter describes high resolution L03 observations of the [OI] 63 $\mu$m line following up the initial results towards Sgr B2 with the LWS grating mode (Baluteau et al., 1997). This line has also been analysed using LWS L04 data by Lis et al. (2001). The study presented here improves on their results, using high signal to noise L03 non-prime data. Lis et al. (2001) used a Maximum Entropy Method (MEM) deconvolution algorithm to increase the resolution of the L04 observation by a factor 2-3. This method works by finding a model that reproduces the observed spectrum when convolved with the instrumental response function and consequently an accurate knowledge of this instrumental profile is required. Lis et al. (2001) used a Lorentzian profile with a width of 35 km s$^{-1}$. Measurements made on the ground and in orbit indicate that the resolving power defining the FP Airy function should be 6900 (see Figure 1.8). This gives a resolution element closer to 45 km s$^{-1}$. This will increase the uncertainty on the line depth in the deconvolved spectrum (they estimated that the 1$\sigma$ errors in level were only 3-5%). They divided the resulting spectrum into three main velocity ranges and calculated OI column densities for each range directly from the MEM spectrum. These were compared with HI and CO measurements to separate the atomic and molecular contribution to the absorption. No evidence was found for an intermediate cloud layer, predicted by PDR models to contain OI but not HI or CO (see later). Their results showed that OI/CO$\sim9$ in the line of sight clouds, lower than the values obtained towards L1689N and W49N.

The analysis presented here involves improved data reduction (including the use of L03 non-prime data and the improvements in calibration described earlier) and takes account of the [OI] 145 $\mu$m and [CII] 158 $\mu$m lines. This allows a much better determination of the continuum level and line shape. The lines are modelled using high resolution information from HI and CO convolved with accurate instrument response profiles. This leads to results that are much more consistent with the standard PDR geometry and are consistent with observations of CI along the line of sight. The observations and data reduction techniques are described with a discussion of systematic errors in Sections 5.2-5.4. The standard PDR model as applied to the clouds along the line of sight is then described and the contribution to the OI absorption from different cloud layers determined in Sections 5.5-5.7. The interpretation and analysis of results were carried out in collaboration with C. Vastel and will be published in Vastel et al. (2002).