A&A 399, 1063-1072 (2003)
DOI: 10.1051/0004-6361:20021868

Gas-phase CO2 toward massive protostars[*]

A. M. S. Boonman 1 - E. F. van Dishoeck 1 - F. Lahuis 1,2 - S. D. Doty 3

1 - Sterrewacht Leiden, PO Box 9513, 2300 RA Leiden, The Netherlands
2 - SRON National Institute for Space Research, PO Box 800, 9700 AV Groningen, The Netherlands
3 - Department of Physics and Astronomy, Denison University, Granville, Ohio 43023, USA

Received 4 November 2002 / Accepted 16 December 2002

We present infrared spectra of gas-phase CO2 around 15 $\mu $m toward 14 deeply embedded massive protostars obtained with the Short Wavelength Spectrometer on board the Infrared Space Observatory. Gas-phase CO2 has been detected toward 8 of the sources. The excitation temperature and the gas/solid ratio increase with the temperature of the warm gas. Detailed radiative transfer models show that a jump in the abundance of two orders of magnitude is present in the envelope of AFGL 2591 at T>300 K. No such jump is seen toward the colder source NGC 7538 IRS9. Together, these data indicate that gas-phase CO2 shows the same evolutionary trends as CO2 ice and other species, such as HCN, C2H2, H2O, and CH3OH. The gas-phase CO2 abundance toward cold sources can be explained by gas-phase chemistry and possible freeze-out in the outer envelope. Different chemical scenarios are proposed to explain the gas-phase CO2 abundance of 1- $2 \times 10 ^{-6}$ for T>300 K and of $\sim$10-8 for T<300 K toward AFGL 2591. The best explanation for the low abundance in the warm exterior is provided by destruction of CO2 caused by the passage of a shock in the past, combined with freeze-out in the coldest part at T<100 K. The high abundance in the interior at temperatures where all oxygen should be driven into H2O is unexpected, but may be explained either by production of OH through X-ray ionization leading to the formation of abundant gas-phase CO2, or by incomplete destruction of evaporated CO2 for T>300 K.

Key words: stars: formation - ISM: abundances - ISM: molecules - infrared: ISM - ISM: lines and bands - molecular processes

1 Introduction

Carbondioxide is predicted to be among the more abundant carbon- and oxygen-bearing gas-phase species in massive star-forming regions (e.g. Charnley 1997). However, the lack of a permanent dipole moment restricts observations of this molecule to infrared wavelengths. Due to its ubiquitous presence in the Earth's atmosphere, it was not until the launch of the Infrared Space Observatory (ISO) that a systematic search for CO2 toward star-forming regions could be performed. Van Dishoeck et al. (1996) made the first search for gas-phase CO2 in absorption toward a few deeply embedded massive protostars. Since then, ISO has detected gas-phase CO2 toward many astronomical objects, including other massive protostars (Dartois et al. 1998; van Dishoeck 1998), planetary atmospheres, and Asymptotic Giant Branch stars (e.g. Lellouch et al. 2002; Justtanont et al. 1998; Cami et al. 2000).

Table 1: Observed sources and their properties.
\noalign{\smallskip }
\end{array} \end{displaymath} \end{table}

For sources where multiple IDs are listed, the spectra presented in this paper are a combination of all of these.
Continuum flux at $\sim$15 $\mu $m in the ISO-SWS aperture, obtained by extrapolating the continuum flux in the 14.5-14.8 $\mu $m region.
The first reference refers to the luminosity, the second to the distance.
References: 1. van der Tak et al. (2000b); 2. Lahuis & van Dishoeck (2000); 3. Henning et al. (1990); 4. Henning et al. (1992); 5. Giannakopoulou et al. (1997); 6. Thompson et al. (1981); 7. Anthony-Twarog (1982); 8. Azcarate et al. (1986); 9. Persi et al. (1987); 10. Helmich et al. (1994).

Van Dishoeck et al. (1996) derive tentative gas-phase CO2 abundances of $\sim$10-7 averaged over the line of sight. Somewhat higher abundances of a few times 10-7 are found in the direction of Orion-IRc2/BN, whereas more than an order of magnitude lower abundances are found toward the shocked regions Peak 1 and 2 (Boonman et al. 2003). Envelope models by Doty et al. (2002) predict abundances of a few times 10-8 for temperatures $\la$100 K and a few times 10-7-10-6 for $T\sim 100$-300 K. Hot core models by Charnley (1997) also predict abundances of $\sim$10-7-10-6 for $T\sim 200$-300 K. On the other hand, Charnley & Kaufman (2000) show that shocks containing a high H/H2 ratio can destroy CO2, giving a possible explanation for the lower abundances found for the Orion shocked regions. Similarly, low gas-phase CO2 abundances of $\sim$10-8 are predicted by gas-grain chemistry in the post-shock gas for dark-cloud type environments (Charnley et al. 2001).

In addition to gas-phase CO2, abundant CO2 ice has been seen toward intermediate- to high-mass star-forming regions (de Graauw et al. 1996; D'Hendecourt et al. 1996; Gerakines et al. 1999; Nummelin et al. 2001). Abundances up to $\sim$10-5 have been found, much higher than the gas-phase CO2 abundances. The ice abundances are highest for the coldest sources, implying that grain-mantle evaporation is important for the chemistry in warmer sources (van Dishoeck 1998). One of the main questions is whether abundances of gas-phase CO2 as high as $\sim$10-5 are observed in any star-forming region.

In this paper, observations of the $\nu _2$ ro-vibrational band of gas-phase CO2 around 15 $\mu $m toward 14 embedded massive young stars are presented. All sources in our sample have luminosities between $\sim$ $ 10^4{-}10^5~L_{{\rm\odot}}$and do have complementary ISO data on ices and other gas-phase molecules (e.g. Lahuis & van Dishoeck 2000; Gerakines et al. 1999; Keane et al. 2001b). The observations and reduction of the data are summarized in Sect. 2. Section 3 describes the analysis, using pure absorption models to derive abundances and gas/solid ratios. Radiative transfer effects are taken into account in Sect. 4 and the results are discussed in Sect. 5. Finally, the conclusions are presented in Sect. 6.

\par\includegraphics[angle=-90,width=8.8cm,clip]{MS3258f1.eps} \end{figure} Figure 1: ISO-SWS spectrum of AFGL 2136 between 14.5 and 16 $\mu $m. The dashed line shows the adopted continuum, the solid line a good fitting ice mixture of polar + annealed ice (H2O:CO:CO2=100:3: $\rm 20 + H_2$O:CH3OH:CO2=1:1:1 respectively) to the observed CO2 ice band (Gerakines et al. 1999).
Open with DEXTER

\par\includegraphics[width=12cm,clip]{MS3258f2.eps} \end{figure} Figure 2: Normalised ISO-SWS spectra in the region of the gas-phase CO2 $\nu _2$ bending mode for all sources. The solid CO2 feature has been removed (see text).
Open with DEXTER

2 Observations and reduction

The $\nu _2$ ro-vibrational band of gas-phase CO2 around 15 $\mu $m has been observed with the Short Wavelength Spectrometer (SWS) in the AOT6 grating mode toward all sources. The spectra have been reduced using the ISO-SWS Interactive Analysis System SIA using the ISO Off-line Processing (OLP version 10) software modules and calibration files. In addition, the instrumental fringes have been removed by a combination of an optimized spectral response calibration and robust sine wave fitting (Lahuis & van Dishoeck 2000). The spectra have been rebinned to an effective spectral resolution of $\Delta \lambda = 0.0085~\mu$m. The S/N ratio on the continuum is typically 50-100 in the final spectra.

The $\nu _2$ ro-vibrational band of gas-phase CO2 has been detected toward the sources AFGL 2136, AFGL 2591, AFGL 4176, MonR2 IRS3, NGC 7538 IRS1, NGC 7538 IRS9, W 33 A, and W 3 IRS5. The spectra of AFGL 2136, AFGL 2591, AFGL 4176, and NGC 7538 IRS9 have been analysed previously by van Dishoeck et al. (1996). The reduced spectra presented here are however of much higher quality. Both the instrument calibration and the reduction routines within the ISO-SWS pipeline as well as the SWS Interactive Analysis have improved significantly since 1996. In addition for AFGL 2136, AFGL 2591, and AFGL 4176, the spectra of multiple independent observations of the same source have been combined, leading to an additional increase in the final S/N ratio.

An example of the resulting spectra is shown in Fig. 1 together with a fit to the CO2 ice band, using an ice mixture similar to that found by Gerakines et al. (1999). Although the laboratory ice fit follows the solid-state feature quite well, it shows some small deficiencies, which could be due to the presence of other solid-state features that are not accounted for in the ice fit. Therefore the final spectra have been divided by a manual fit to the CO2 ice band, resulting in the normalised spectra presented in Fig. 2. Different fits to the continuum have been made to investigate their effect on the shape and depth of the gas-phase CO2 band and these have been taken into account in the analysis of the spectra.

3 Analysis

3.1 Homogeneous pure absorption models

The modeling of the spectra has been performed by computing synthetic spectra using the method described in Lahuis & van Dishoeck (2000). In this method, the source is assumed to be a homogeneous sphere with a single temperature $T_{{\rm ex}}$ and column density N. Here it is assumed that only absorption takes place and that emission can be neglected. The effects of adopting a non-homogeneous source and including emission are discussed in Sect. 4. The models are not sensitive to the CO2 line width as long as the lines are not saturated, which becomes important only for $b\la 1$ km s-1. A mean Doppler b-value of 3 km s-1 is adopted (see also Boonman et al. 2003), but values up to b=10 km s-1 have been explored. Similar values are used in the modeling of other observed ro-vibrational absorption lines in the same wavelength region toward these sources (e.g. Lahuis & van Dishoeck 2000; Keane et al. 2001a). For comparison, Mitchell et al. (1990) derive line widths of $b\sim 4$-7 km s-1 from high-resolution observations of the CO ro-vibrational band around 4.7 $\mu $m. The resulting synthetic spectra have been convolved to the nominal spectral resolution of the ISO-SWS spectra for comparison with the data.

\par\includegraphics[angle=-90,width=8.8cm,clip]{MS3258f3.eps} \end{figure} Figure 3: The strength of the CO2 (02 $^20)\leftarrow (01^1$0) hotband (thick line) compared to the fundamental (01 $^10)\leftarrow (00^0$0) band (thin line) for different excitation temperatures, using b=3 km s-1 and $N=1\times 10^{16}$ cm-2.
Open with DEXTER

\par\includegraphics[angle=-90,width=8.8cm,clip]{MS3258f4.eps} \end{figure} Figure 4: Example of a good fitting model for the observed CO2 spectrum toward AFGL 2136, using the best fit parameters from Table 2 (dashed line). The dotted line shows the contribution of the (02 $^20)\leftarrow (01^1$0) hotband. The solid line is the observed ISO-SWS spectrum divided by the continuum. The P, Q, and R-branches of the CO2 $\nu _2$ ro-vibrational band are indicated.
Open with DEXTER

\par\includegraphics[width=7cm,clip]{MS3258f5.eps} \end{figure} Figure 5: Example of the $\chi ^2$ distribution for AFGL 2136. The star indicates the minimum $\chi ^2$ and the contours are 2.5, 5, 10, and 22.5% above the minimum $\chi ^2$. The latter contour roughly corresponds to a 3$\sigma $ deviation of the model from the observed CO2 band for this source.
Open with DEXTER

Table 2: Model parameters for the $\nu _2$ band of CO2.
\item[$^{i}$ ] From Boonman et~al. (\cite{boonman02}).
\end{list} \end{table}

The molecular line data have been taken from the HITRAN 2000 database (http://www.hitran.com). The model includes the fundamental (01 $^10)\leftarrow (00^0$0) band at 14.983 $\mu $m, the hotbands (02 $^20)\leftarrow (01^1$0) and (02 $^00) \leftarrow (01^1$0) at 14.976 and 16.18 $\mu $m respectively, and the (10 $^00) \leftarrow (01^1$0) band at 13.87 $\mu $m. The (02 $^00) \leftarrow (01^1$0) and (10 $^00) \leftarrow (01^1$0) bands are not detected. The (02 $^20)\leftarrow (01^1$0) hotband coincides with the fundamental Q-branch, making it difficult to detect. Figure 3 shows the shape and strength of the CO2 fundamental and the (02 $^20)\leftarrow (01^1$0) hotband for different excitation temperatures. This shows that for $T_{{\rm ex}}\ga 300$ K the (02 $^20)\leftarrow (01^1$0) hotband starts to play a role. This hotband was not included in the previous modeling of the CO2 spectra by van Dishoeck et al. (1996).

The best fit to the data has been determined using the reduced $\chi_{\nu}^2$-method. Since the P-branches of the C2H2 $\nu_5$ and HCN $\nu _2$ ro-vibrational bands extend into the region of the CO2 $\nu _2$ band, these bands are included in the modeling, using the best fit parameters from Lahuis & van Dishoeck (2000). For all sources, their contribution is less than the noise level, but they are included for consistency. The best fitting model parameters are listed in Table 2 for all sources. The uncertainty in the excitation temperature includes errors due to different continuum fits.

The results show that warm CO2 gas at $T \ge 200$ K is detected toward half of the sources and suggest that for the hottest sources, AFGL 2591, AFGL 2136, and AFGL 4176, also the (02 $^20)\leftarrow (01^1$0) hotband contributes. The gas-phase 13CO2 band near 15.4 $\mu $m is not detected in our sources.

A good fitting CO2 model is presented in Fig. 4. Figure 5 shows an example of $\chi_{\nu}^2$ contours for the source AFGL 2136. It illustrates that the column density of the gas-phase CO2 is well constrained, but that the excitation temperature shows a larger spread.

Comparison of our results for AFGL 2136 to those by Sandford et al. (2001) shows a lower excitation temperature than their $T_{{\rm ex}}$ of 580 K. Also, their column density in the 580 K gas is more than an order of magnitude higher than that found here. This discrepancy is likely caused by the low signal-to-noise in the spectrum presented by Sandford et al. (2001), which is also hampered by the presence of instrumental fringes. These fringes are carefully removed in our spectra. In addition, our AFGL 2136 spectrum shows the detection of a few P- and R-branch lines (Fig. 4), which poses an extra constraint on the excitation temperature and column density. The fact that the CO2 ice fit shown in Fig. 1 does not match the observed continuum very well between 14.6 and 14.8 $\mu $m introduces an uncertainty of $\sim$100 K in the excitation temperature, which is accounted for in the results in Table 2. Therefore, the newly reduced spectra presented here allow a more reliable estimate of the excitation temperature and column density of the CO2 $\nu _2$ ro-vibrational band.

Figure 6 presents a comparison of the CO2 excitation temperature $T_{{\rm ex}}$(CO2) and that of C2H2, a good tracer of the warm gas (Lahuis & van Dishoeck 2000). It is seen that $T_{{\rm ex}}$(CO2) increases with $T_{{\rm ex}}$(C2H2), indicating that CO2 also traces warm gas. However, the increase is not as strong as for C2H2 (Fig. 6). Note that for NGC 7538 IRS1 a C2H2 excitation temperature of $\sim$500 K is found from a new, updated reduction of the spectra, much lower than the $T_{{\rm ex}}$(C2H2)=800 K listed in Lahuis & van Dishoeck (2000).

\par\includegraphics[width=6.5cm,clip]{MS3258f6.eps} \end{figure} Figure 6: Correlation between the CO2 excitation temperature $T_{{\rm ex}}$(CO2) and that of C2H2, a good tracer of the warm gas (Lahuis & van Dishoeck 2000). Only those sources are shown for which both excitation temperatures are determined. The cross denotes typical error bars for $T_{{\rm ex}}\protect\ga 200$ K.
Open with DEXTER

\par\includegraphics[angle=-90,width=8.8cm,clip]{MS3258f7.eps} \end{figure} Figure 7: The CO2 gas-phase abundances from Table 2 (left panel) and the CO2 ice abundances from Gerakines et al. (1999) (right panel) versus $T_{{\rm ex}}$(C2H2) from Lahuis & van Dishoeck (2000). The crosses denote typical error bars for $T_{{\rm ex}}\protect\ga 200$ K. For MonR2 IRS3 and S 140 IRS1 the 13CO temperatures from Giannakopoulou et al. (1997) and Mitchell et al. (1990) respectively, are plotted.
Open with DEXTER

3.2 Abundances

The CO2 column densities have been converted into abundances, using the total H2 column densities (cold + warm components) derived from infrared observations of 13CO (e.g. Mitchell et al. 1990). A 12CO/13CO ratio of 60 and a 12CO/H2 ratio of $2\times10^{-4}$ have been assumed (Lahuis & van Dishoeck 2000; Lacy et al. 1994). Typical derived CO2 abundances are a few $\times$10-7 (Table 2). The inferred abundances do not show a clear trend with temperature (Fig. 7). Using the H2 column densities in the warm gas only increases the gas-phase CO2 abundance by a factor of $\sim$2 for most sources, but does not change the lack of a trend with temperature. The abundances differ by less than a factor of $\sim$4 between the sources, except for G 333.3-0.4 (Fig. 7). However, for G 333.3-0.4 the H2 column density is determined from the C17O J=2-1 submillimeter transition, resulting in an upper limit on the H2 column density, and consequently a rather uncertain CO2 abundance.

The derived CO2 abundances for AFGL 2591, AFGL 2136, and AFGL 4176 are a factor of $\sim$2.5-4 higher than the tentative values from van Dishoeck et al. (1996). For NGC 7538 IRS9 the tentative value falls within the error bars of the CO2 abundance derived here. The inferred CO2 abundances agree well with the abundance range found toward Orion IRc2/BN (Table 2).

\par\includegraphics[width=6.5cm,clip]{MS3258f8.eps} \end{figure} Figure 8: Gas/solid ratio for CO2 versus $T_{{\rm ex}}$(C2H2) (filled circles). For MonR2 IRS3 and S 140 IRS1, $T_{{\rm ex}}$(13CO) is used (Giannakopoulou et al. 1997; Mitchell et al. 1990). The dot-dashed line shows the least-squares fit to the CO2 gas/solid ratios and the dashed line the least-squares fit to the H2O gas/solid ratios from Boonman & van Dishoeck (2003).
Open with DEXTER

3.3 Gas/solid ratios

Gas/solid ratios can be determined by combining the derived gas-phase CO2 column densities with those for CO2 ice for the same sources from Gerakines et al. (1999) (see Fig. 7). This ratio increases with temperature, consistent with the location of CO2 in the warm inner part of the envelope, above the evaporation temperature (Fig. 8). However, the increase is less strong and the ratios are lower than for H2O, although pure CO2 ice is more volatile than H2O (Fig. 8; Boonman & van Dishoeck 2003). This is probably due to the fact that toward our sources CO2 ice is mostly embedded in a H2O ice matrix and that its column density is only 10-23% of that of H2O ice (Gerakines et al. 1999). In addition, gas-phase H2O abundances of up to $\sim$10-4 are easily formed above T>230-300 K, thus rapidly increasing the H2O gas/solid ratios (Charnley 1997).

4 Radiative transfer effects

4.1 Filling in by emission

In this section, the effect of possible emission along the line of sight on the derived CO2 column densities and abundances is investigated. To this purpose a similar excitation model as that described in Boonman et al. (2003) has been set-up. This excitation model includes energy levels for CO2 up to J=40 in both the $\nu_2=0$ and 1 vibrational states. The level populations are calculated adopting a Boltzmann distribution using $T_{{\rm ex}}$ from Table 2. As a central radiation source, a blackbody with T=300 to 600 K has been explored.

Adopting a homogeneous source as before, but including both absorption and emission along the line of sight, shows that the CO2 column densities needed to match the observations are up to $\sim$30% higher than those listed in Table 2 as long as the excitation temperature is less than $\sim$250-300 K. This is within the listed error bars. The emission only becomes important for $T_{\rm ex} \ga 250$-300 K, i.e. for the sources W 3 IRS5, AFGL 2136, AFGL 4176, NGC 7538 IRS1, and AFGL 2591 (see also Boonman et al. 2003). For these sources, the CO2 column density that best matches the observations can be up to a factor of $\sim$3 higher than that derived from the pure absorption models (Table 2). Note that this model does not include other radiative transfer effects, such as infrared pumping, nor does it include temperature and density gradients, so that the derived CO2 abundances within this model are not very accurate.

4.2 Temperature and density gradients: Jump models

Van der Tak et al. (2000b) show that a density and temperature gradient is present in their sample of deeply embedded massive young stars, which is a sub-set of the sample studied here. Since radiative transfer effects are expected to be largest for the warmer sources, AFGL 2591 is taken as a test case. Adopting the physical model for AFGL 2591 derived by van der Tak et al. (2000b), the level populations and radiative transfer are calculated with the Monte Carlo code of Hogerheijde & van der Tak (2000) on a grid of concentric shells, assuming spherical geometry. The calculations include energy levels up to J=40 in both the $\nu_2=0$ and $\nu_2=1$ states, the same as in the excitation model described in Sect. 4.1. The collisional rate coefficients used are taken from Allen et al. (1980). Radiative excitation through the 15 $\mu $m band due to warm dust mixed with the gas is included, using grain opacities from Ossenkopf & Henning (1994) and assuming $T_{\rm dust}=T_{\rm gas}$. No external radiation field apart from the 2.73 K cosmic background radiation was applied. The resulting level populations are used in the excitation model described above to calculate a synthetic model spectrum including both emission and absorption along the line of sight. As a central radiation source a blackbody at $T>T_{{\rm dust}}$ in the innermost shell is chosen. The resulting model spectrum is then compared with the observed absorption spectrum.

\par\includegraphics[width=8.8cm,clip]{MS3258f9.eps} \end{figure} Figure 9: Results from the detailed radiative transfer models including temperature and density gradients (see Sect. 4.2). The solid line shows the data, the dashed line the model spectrum. a) Model spectrum for AFGL 2591 using a constant CO2 abundance of n(CO2)/n(H $_2)=3\times 10^{-7}$ throughout the envelope, similar to that inferred from the pure absorption models in Sect. 3.2. b) Model spectrum for AFGL 2591 using n(CO2)/n(H $_2)=2\times 10^{-6}$ for T>300 K and n(CO2)/n(H $_2)=1\times 10^{-8}$ for T<300 K. c) Model spectrum for NGC 7538 IRS9 using a constant CO2 abundance of n(CO2)/n(H $_2)=7\times 10^{-8}$.
Open with DEXTER

For AFGL 2591 it is found that a constant CO2 abundance x(CO2)=n(CO2)/n(H2) throughout the envelope produces a synthetic spectrum with a Q-branch that is too narrow compared to the observations (Fig. 9). This indicates that the constant abundance model is dominated by absorption from gas at T<300 K. Therefore a model with a jump in the abundance at T=300 K has been tried. It is found that a model spectrum with a CO2 abundance of $\sim$1- $2 \times 10 ^{-6}$ for T>300 K and 10-8 for T<300 K can reasonably explain the observed spectrum (Fig. 9). Applying a jump at lower temperatures, e.g. at T=200 K or T=100 K produces a Q-branch that is too narrow. This indicates that the observed Q-branch represents warm CO2 gas from the inner part of the molecular envelope at $T \geq 300$ K, in agreement with that inferred from the pure absorption models. It also shows that the CO2 abundance in the outer envelope is much lower, about two orders of magnitude. Similar jumps in the abundance have been seen for other molecules toward AFGL 2591, such as HCN, H2O, and SO2 (Boonman et al. 2001; Boonman & van Dishoeck 2003; Keane et al. 2001a; van der Tak et al. 2003).

Similarly one of the colder sources, NGC 7538 IRS9, is modeled for comparison using the physical model derived by van der Tak et al. (2000b). It is found that a constant CO2 abundance of $\sim$ $7 \times 10^{-8}$ can reproduce the observed spectrum. A model with a jump in the abundance at T=100 K and x(CO $_2)=2 \times 10^{-7}$ for T > 100 K and x(CO $_2) = 1 \times 10^{-9}$ for T < 100 K gives a similarly good fit in terms of $\chi ^2$. This indicates that for NGC 7538 IRS9 no evidence for a jump in the abundance at temperatures $T \geq 300$ K is found. The result that a jump in the abundance at T = 100 K fits the observed spectrum equally well as the constant abundance model may suggest that most of the CO2 is frozen-out onto the grains below this temperature.

The corresponding CO2 column densities for the best fit models for AFGL 2591 and NGC 7538 IRS9 are a factor of $\sim$2-4 and $\sim$1.5-2 higher than those listed in Table 2 respectively. On the other hand, the abundance toward NGC 7538 IRS9 is somewhat lower than that derived from the pure absorption models (Table 2).

The above results indicate that a jump in the CO2 abundance is present for the warmer, more evolved sources, but that no such jump is seen toward the colder objects. This suggests that also for AFGL 2136, AFGL 4176, and NGC 7538 IRS1 a jump in the CO2 abundance is present.

Observations of the intermediate-mass protostars AFGL 490 and AFGL 7009S show CO2 abundances of a few $\times$10-7, similar to those derived from the pure absorption models in Sect. 3.2 (Schreyer et al. 2002; Dartois et al. 1998). These CO2 abundances do not show a clear trend with temperature (Fig. 7). However the results from the detailed radiative transfer models indicate much higher CO2 abundances in the warm inner part of the envelope for the more evolved sources. Combined with the somewhat lower CO2 abundance toward the cold source NGC 7538 IRS9, this suggests that the CO2 abundance increases with temperature and evolutionary state.

5 Discussion

5.1 CO2 as an evolutionary tracer

In Sect. 3 it is shown that $T_{{\rm ex}}$(CO2) increases with $T_{{\rm ex}}$(C2H2), indicating that it is a tracer of the warm gas. The gas/solid ratio also increases with the temperature of the warm gas and the CO2 ice abundance decreases (Figs. 7 and 8). The higher ratios for the warmer sources suggest that they are in a later evolutionary stage than the sources with low gas/solid ratios, with the higher temperatures due to dispersion of a larger fraction of the molecular envelope (van der Tak et al. 2000b; van Dishoeck & van der Tak 2000).

Although the abundances derived from the pure absorption models do not show a clear trend with temperature, the results from the jump models in Sect. 4.2 suggest that the CO2 abundance increases with temperature. In addition, the same sources with a high CO2 excitation temperature and correspondingly a higher gas/solid ratio also show evidence for thermal processing of 13CO2 and CO2 ice (Boogert et al. 2000; Gerakines et al. 1999). Together, this shows that CO2 can be used as an evolutionary tracer.

5.2 Chemistry in cold sources: NGC 7538 IRS9

Envelope models by Doty et al. (2002) predict gas-phase CO2 abundances of a $\sim$10-7-10-6 at $T\la 300$ K for $t\sim3\times 10^3$-105 yrs. This indicates that pure gas-phase chemistry can explain the observed abundances toward the colder sources, such as NGC 7538 IRS9, that show no evidence for a jump in the abundance for $T\ga 300$ K (Sect. 4.2).

In the colder sources, the high CO2 ice abundances and low gas/solid ratios indicate that a large fraction of the envelope still contains cold material. This suggests that the observed gas-phase CO2 absorption is dominated by the colder outer envelope where evaporation has not yet taken place. However, these sources may still hide a small region in the inner envelope containing hot, abundant gas-phase CO2, which cannot be detected with the present observations, e.g. due to continuum optical depth effects.

5.3 Chemistry in warm sources: AFGL 2591

5.3.1 Comparison to hot core and shock models

The high inferred gas-phase CO2 abundance in the inner envelope of AFGL 2591 for T>300 K but lack of a jump at $T\sim 100$ K in Sect. 4.2 is unexpected. Such a jump is observed for simple ice constituents such as H2O and CH3OH in hot-core type objects (e.g. van der Tak et al. 2000a; Maret et al. 2002), which is due to evaporation of H2O-rich ices around $T\sim 90$-110 K (Fraser et al. 2001). Since most of the CO2 ice is embedded in H2O ice, it will evaporate around the same temperature (Fig. 10). The high gas/solid ratio and low CO2 ice abundance of $1.7 \times 10^{-6}$ (Gerakines et al. 1999) toward AFGL 2591 indicate that evaporation of CO2 ice does occur (see also Fig. 7, right panel). The lack of a jump around 100 K then indicates that CO2 must be rapidly destroyed in the gas phase after evaporation in the $T\sim 100$-300 K zone. At T > 300 K, either the CO2 is only partially destroyed or it has quickly reformed in the gas-phase.

Second, the CO2 solid-state abundances toward the cold sources with little or no ice evaporation are an order of magnitude higher than the gas-phase CO2 abundance in the inner hot envelope of AFGL 2591 (Fig. 7). This suggests that CO2 is originally formed on grain surfaces, and not simply due to freeze-out of CO2 gas. The discrepancy between the observed gas-phase and solid-state CO2 abundances provides further evidence for rapid destruction of CO2 in the gas-phase after evaporation from the grains. Charnley & Kaufman (2000) propose that shocks can be responsible for this.

A study of the Orion shocked regions Peak 1 and 2 shows gas-phase CO2 abundances of $\sim$0.3- $6 \times 10^{-8}$, which are best explained by destruction of the CO2 in a shock containing a high H/H2 ratio of $\sim$0.01, and subsequent reformation in the gas-phase (Boonman et al. 2003). A similar abundance of $\sim$10-8 is found in the envelope of AFGL 2591 for T < 300 K in Sect. 4.2. This suggests that the outflow may have (partially) destroyed the CO2 in the envelope of AFGL 2591 in the past, and that it has not yet been reformed. However, such a shock model cannot explain the high abundance in the inner envelope at T > 300 K, unless destruction is much less efficient in the densest inner part due to e.g. a lower H/H2 ratio.

Doty et al. (2002) propose that a heating event can also destroy CO2 through the reaction CO $_2 + \rm H_2 \rightarrow CO + H_2$O. Recent calculations by Talbi & Herbst (2002) however show that this reaction is not likely to play a dominant role in the interstellar medium.

\par\includegraphics[angle=-90,width=8.8cm,clip]{3258f10.eps} \end{figure} Figure 10: Gas-phase CO2 abundances in the envelope of AFGL 2591 for chemical ages of $3 \times 10 ^2$, $3 \times 10 ^3$, $3 \times 10 ^4$, and $3 \times 10 ^5$ yrs (after Doty et al. 2002). a) Gas-phase chemistry including evaporation of CO2 ice for $T\sim 100$ K. The initial CO2 abundance of $3 \times 10 ^{-5}$ for $T\protect\ga 100$ K and the cosmic ray ionization rate of $\zeta _{{\rm CR}}=5.6 \times 10^{-17}$ s-1 from Doty et al. (2002) have been adopted. b) Gas-phase chemistry adopting n(CO2)/n(H2)=0 initially, $\zeta _{{\rm CR}}\sim 10^{-20}$ s-1 for T< 200 K, and an ionization rate of $\zeta =2 \times 10 ^{-16}$ s-1 for T> 200 K.
Open with DEXTER

The primary formation route of CO2 in the gas-phase is through the reaction $\rm CO + OH \rightarrow CO_2 + H$ which proceeds rapidly at $T\geq 100$ K (Charnley 1997). Above $T\sim 230$-300 K most of the OH is driven into H2O, thus reducing the formation of CO2 through gas-phase reactions. This formation route predicts gas-phase CO2 abundances of $\sim$10-6 at T=100 K to a few $\times$10-7 at T=300 K (Charnley 1997). This is much lower than the inferred CO2 abundance of 1- $2 \times 10 ^{-6}$ for T>300 K toward AFGL 2591. On the other hand, gas-phase CO2 can also be formed through the reaction of $\rm CO + H_2$O, but this reaction has an energy barrier of $\sim$ $5.2 \times 10^4$ K, preventing production of significant CO2 in the molecular envelope.

5.3.2 Comparison to UV, cosmic- and X-ray models

Alternatively, the UV flux from the protostar may be high enough in the inner envelope to produce OH through direct photodissociation of H2O. This could maintain sufficient OH in the gas-phase to form abundant gas-phase CO2 in the interior at $T\ga 300$ K. However, photodissociation is estimated to play a role only up to $r \sim 10^{14}$ cm from the central source, much smaller than the inner radius of the physical model used in Sect. 4.2. Doty et al. (2002) suggest that cosmic-ray or X-ray ionization can also produce significant OH, which is then channelled into CO2. Using the chemical model from Doty et al. (2002) for AFGL 2591 and adopting a zero initial gas-phase CO2 abundance corresponding to the destruction of CO2 by a shock, indeed predicts enhanced CO2 abundances of $\sim$10-6-10-5 in the interior for T>200 K if a high ionization rate of $2 \times 10^{-16}$ s-1 is adopted (Fig. 10). This illustrates that a high ionization rate in the inner envelope can produce a jump in the CO2 abundance at temperatures larger than 100 K as found for AFGL 2591. This model adopts an artificially low cosmic-ray ionization rate of $\zeta _{{\rm CR}}\sim 10^{-20}$ s-1 in the outer envelope, predicting gas-phase CO2 abundances of $\sim$10-8 for $T\la 200$ K, consistent with that found in Sect. 4.2. Adopting $\zeta_{{\rm CR}} = 5.6 \times 1 ^{-17}$ s-1 as derived by van der Tak & van Dishoeck (2000) from HCO+ observations results in much higher abundances of $\sim$10-6-10-5 for $T\sim 100$-200 K and $\sim$10-7 for T < 100 K, somewhat higher than that found for AFGL 2591. However, freeze-out of CO2 onto the grains is likely to play a role for T < 100 K, which is not included in the model.

Doty et al. (2002) note that the ionization rate needs to be at least on the order of $\zeta=5.6 \times 10^{-17}$ s-1 in the interior, in order to account for the high gas-phase HCN abundances for $T\ga 300$ K. This is consistent with our proposed chemical scenario of a high ionization rate in the inner envelope, explaining the jump in the CO2 abundance for $T\ga 300$ K. Although the production of CO2 through X-ray ionization involves destruction of gas-phase H2O and CO, the predicted enhanced CO2 abundances in Fig. 10 are $\la$10% of those predicted for H2O and CO. In addition, abundant gas-phase H2O and CO are observed in the warm inner envelope from infrared absorption (e.g. Boonman et al. 2000; Mitchell et al. 1990), suggesting that these molecules are not significantly affected by X-rays. Since the cosmic-ray ionization rate is expected to be roughly constant or potentially decreases inward within the molecular envelope, the enhanced ionization rate in the warm interior seems more likely caused by X-rays from the young star than cosmic rays from outside the molecular envelope. Using the photoionization cross section from Wilms et al. (2000), it is estimated that X-rays of a few keV can affect the chemistry up to radii at which $T \sim 400$ K in the envelopes of massive protostars. Recently, X-ray emission within this energy range has been detected toward MonR2 IRS3, one of our warm sources, further suggesting that X-ray ionization may be important for the CO2 chemistry in the inner envelope of massive young stars (Preibisch et al. 2002; Kohno et al. 2002).

5.3.3 Comparison to dynamical models

Another possibility is that gas-phase CO2 in the interior results from ice evaporation at $T\sim 100$ K in the past, and is subsequently heated as the protostar evolves to higher temperatures at the same point in the envelope, without being destroyed. However, material initially at T< 100 K will consequently be heated to T> 100 K, where ice evaporation occurs almost instantaneously (Fraser et al. 2001). This predicts gas-phase CO2 abundances of $\sim$10-6-10-5 between 100 and 300 K, contrary to the inferred abundance of $\sim$10-8.

A second scenario to consider would be the dynamical transport of gas-phase CO2 formed between 100 and 300 K inward to the T > 300 K region, e.g. through infall motions. In this case, however, the CO2-ice containing material at T < 100 K would be transported inward to the T = 100-300 K region. At this point, the CO2 ice would evaporate immediately, thus maintaining a large gas-phase CO2 abundance for T = 100-300 K.

At present, a combination of the destruction of CO2 by a past shock in the outflow, and either a high X-ray ionization rate in the interior which rapidly reforms gas-phase CO2 for T> 300 K or incomplete destruction of evaporated CO2 for T> 300 K, seems the most likely explanation for the inferred jump in the gas-phase CO2 abundance for T> 300 K toward AFGL 2591. As noted by Doty et al. (2002), more experimental work on the chemistry of gas-phase CO2 is needed to obtain a better knowledge of the formation and/or destruction pathways of gas-phase CO2 in the envelopes of massive protostars.

6 Conclusions

The main conclusions of this work are:

This work was supported by the NWO grant 614-41-003, a Spinoza grant, and a grant from the Research Corporation (SDD). The authors would like to thank W. Schutte for providing the CO2 ice column density toward MonR2 IRS3 and X. Tielens for useful comments.



Copyright ESO 2003