HIGH-RESOLUTION MID-INFRARED SPECTROSCOPY OF NGC 7538 IRS 1: PROBING CHEMISTRY IN A MASSIVE YOUNG STELLAR OBJECT

The observations were made with TEXES, the Texas Echelon Cross Echelle Spectrograph (Lacy et al. 2002), on the NASA Infrared Telescope Facility (IRTF) in 2001 June and November, 2002 September and December, and 2005 December. TEXES is a high-resolution cross-dispersed spectrograph operating at mid-infrared wavelengths between 5 and 25 μm. It achieves a spectral resolution of R = 75,000–100,000 (Δv = 3–4 km s⁻¹) shortward of 14 μm with a slit width of ∼15. (Of the observations presented here, only the spectra near 11 μm, which used a 1'' slit, achieved the higher resolution.) The spatial resolution along the North–South oriented slit is ∼1'', slightly larger than the diffraction limit of the IRTF. The spectral and spatial sampling by the 256 × 256 pixel² Si:As detector array are 1.0 km s⁻¹ and 035, respectively. At each spectral setting, 5–10 orders of the high-resolution echelon grating are recorded, giving a spectral coverage of Δλ_tot/λ ∼ 0.5%. At wavelengths shortward of 11  μm full order widths are observed, giving continuous spectral coverage, but longward of 11 μm the order width exceeds the array width, leaving gaps in the observed spectra. The location of the gaps can be shifted in order to optimize line coverage. However, since the gaps are small and, at 13 μm, there are many telluric lines, it would be impractical to repeat observations with small shifts in order to get complete spectral coverage for each setting.

Much of the data were taken in high water vapor conditions, but the observing conditions did not seriously affect the spectra except for increased noise, especially on atmospheric lines. Observations of NGC 7538 IRS 1 were interspersed with observations of a bright asteroid (usually Ceres (greater than 500 Jy), but also Hygeia (∼150 Jy)), which served as a comparison source for removal of telluric absorption features. The telescope was nodded by 3–5'', moving the source along the spectrograph entrance slit to allow subtraction of background emission. At the beginning of each set of nodded observations an ambient temperature blackbody was observed for flat-fielding. The first nods were used to peak up on the source by maximizing the throughput of the slit. The nod pairs taken during peak up were not included in the final sum.

Data reduction followed the standard TEXES procedure (Lacy et al. 2002) of first subtracting the two nod position spectra, flat-fielding with the blackbody-sky difference echellogram (Lacy et al. 1989), and interpolating over spikes and bad pixels. Optimally weighted point-source spectra were then extracted from the echellograms. The spectra were linearized in ln(ν) based on the known optical distortions, and absolute wavenumber calibration was obtained from telluric emission features and corrected for the Earth's motion relative to LSR. The resulting wavenumber scale is correct to within 1 km s⁻¹. The same procedure was used for the asteroid spectra, and then the IRS 1 spectra were divided by the asteroid spectra with correction for differences in airmass. With this procedure, telluric absorption lines as deep as 90% can be removed, although of course the noise increases on lines. However, broad response variations often remained, which were removed by fitting the continuum in each echelon order to a quartic polynomial and dividing. This procedure also sets the continuum in each order to one.

Ten spectral settings were observed in the 728–820 cm⁻¹ (13.7–12.2 μm) region, two in the 860–930 cm⁻¹ (11.5–10.7 μm) region (see Figure 1), and three in the 1240–1312 cm⁻¹ (8.0–7.6 μm) region (see Figure 2). A total of 110 echelon orders, or ∼70 cm⁻¹, were observed. The original observations were meant to study the line profiles of C₂H₂ and HCN, which had previously been detected. In addition to the prominent fundamental band lines of C₂H₂ and HCN, many C₂H₂ hot band lines were detected in our spectra including lines from the ν₄ + ν₅ − ν₄ and the 2ν₅ − ν₅ bands. After the hot bands were identified, unidentified lines still remained, some of which were double dipped. The double lines turned out to be CH₃ lines which are split due to spin-rotation interactions (see the Appendix). This is the first detection of CH₃ toward dense gas and the first ground-based detection of CH₃. Previously, CH₃ had only been observed with ISO toward the Galactic center (Feuchtgruber et al. 2000). In efforts to detect ethane, C₂H₆, at 810–820 cm⁻¹, many lines were detected, though none corresponded to C₂H₆. Some were due to NH₃, but many lines remained unidentified (see Figures 3 and 4). Based on their separation we were able to identify the lines as due to HNCO. Interstellar HNCO is a well-known hot core molecule at radio wavelengths (Zinchenko et al. 2000), but it had never been seen at infrared wavelengths. In total, five molecules were observed in the 728–820 cm⁻¹ region: C₂H₂ (including ¹³C¹²CH₂), HCN, HNCO, CH₃, and NH₃. At higher frequencies, CH₄, C₂H₂, and CS were observed in the 1240–1320 cm⁻¹ region. CH₄ observations from the ground are possible for this source because of the large Doppler shift with respect to the telluric lines.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 1 gives a summary of the detected lines. At least five lines were observed for each molecule, with some molecules like C₂H₂ (including isotopic lines and hot bands), and NH₃ having more than 40 lines detected. However, the number of HNCO lines dominates over lines from the other molecules, with over 100 lines detected. Table 2 (full version is available online only) lists all the individual lines detected in our spectra. A selection of spectra including all observed molecules are shown in Figures 1 and 2, along with fits described in Section 3. For most molecules the line strengths were taken from the GEISA03 database.⁶ The CH₃ band strength was obtained from Wormhoudt & McCurdy (1989) and the HNCO oscillator strength was obtained from Lowenthal et al. (2002). The CS oscillator strength was taken from Botschwina & Sebald (1985).

Before discussing the results of our measurements of individual molecules, we attempt to derive H₂ and CO column densities from previous measurements. The H₂ column toward NGC 7538 IRS 1 has not been directly measured, but it may be estimated from the extinction measurements of Willner (1976) with the assumption of a normal interstellar gas-to-dust ratio. This method gives ${\rm N}_{\rm H_2} = 6\hbox{--}9 \times 10^{22}\,{\rm cm}^{-2}$. If the continuum source is nonuniformly covered, this column density represents an average over the source. Hence, it should probably be compared to the values we derive for average column densities, N, of the molecules we observe. However, the silicate optical depth includes extinction from all material along the line of sight to NGC 7538, most of which is unlikely to contain our molecules. This suggests that 6–9 × 10²² cm⁻² is an overestimate of ${\rm N}_{\rm H_2}$ in the relevant gas. On the other hand, we suggest below that we may be observing absorption by gas in the atmosphere of a protoplanetary disk, and dust in such a region could be depleted by settling toward the disk midplane. This effect would result in an underestimate of ${\rm N}_{\rm H_2}$ from the extinction. We give the ratios of our derived column densities to an H₂ column density of 7.5 × 10²² cm⁻² in Table 4.

It may be preferable to compare our column densities to the column density of CO, which has been observed in absorption toward NGC 7538 IRS 1 by Mitchell et al. (1990). They obtain a column density of 1.5 ± 0.3 × 10¹⁷ cm⁻² for ¹³CO in cold gas at 25 K and 1.4 ± 0.1 × 10¹⁷ cm⁻² in warm gas at 176 K. We would not be sensitive to their cold gas. Presumably their warm gas includes both of our components, although our temperatures are generally higher than theirs. If we assume ¹²CO/¹³CO = 72, using the ¹²C/¹³C ratio we derived from our C₂H₂ observations, we derive the abundances of our molecules relative to CO listed in Table 4. (Note that Mitchell et al. use ¹²C/¹³C = 89.) However, we should note that we have measured one spectral setting including CO lines, while searching for OCS absorption near 5 μm. The noise is rather high, but it is apparent that the ¹³CO lines are not well described by moderately saturated Gaussians, as assumed by Mitchell et al. The ¹³CO lines are deeper and broader than the lines of any of our other molecules, and have prominent blue-shifted shoulders probably tracing an outflow. We suspect that the lines are more saturated, and the CO column density larger than Mitchell et al. concluded. The different shapes and greater depths of the CO lines than our C₂H₂ lines also indicates that CO probes different gas. Likely, C₂H₂ and the other molecules we observe in the mid-infrared are found only in unusual regions, whereas CO is distributed along the entire line of sight.

In general, we conclude that the ratios of column densities to those of either H₂ or CO given in Table 4 are quite uncertain because the different column density tracers may be sensitive to different components along the line of sight. The column densities of the different molecules we observe should be much more comparable.

Since our initial search for molecules began with C₂H₂ and HCN, the first models only included those two molecules. In these models, we derived similar temperatures and column densities as derived from ISO observations (Lahuis & van Dishoeck 2000; Boonman et al. 2003). Once we include absorption from other molecules, HNCO in particular, the temperatures for HCN and especially for C₂H₂ are lowered significantly. Some HNCO lines overlap high-J lines of C₂H₂, which means that more hot C₂H₂ is necessary to produce the observed absorption when HNCO is not included in the model. In our final model including all of the observed molecules, our derived column densities of C₂H₂, N = 5.8 × 10¹⁶ cm⁻², and HCN, N = 6.9 × 10¹⁶ cm⁻², (see Table 3) are larger than those derived in Lahuis & van Dishoeck, $N_{\rm C_2H_2} = 8 \times 10^{15}$ cm⁻² and N_HCN = 1.0 × 10¹⁶ cm⁻². Part of the difference is due to the different Doppler parameter used: b = 5 km s⁻¹ is used in Lahuis & van Dishoeck, whereas, we derive b ∼ 1 km s⁻¹. We derive lower temperatures, $T_{\rm C_2H_2} = 190\hbox{--}230$ K and T_HCN = 250–450 K than derived by Lahuis & van Dishoeck, $T_{\rm C_2H_2} = 800$ K and T_HCN = 600 K, and Boonman et al. (2003), $T_{\rm C_2H_2} = 500$ K. We also find that we need a small covering factor to account for the saturated yet shallow lines. The level of saturation is difficult to determine from the ISO data. In order to verify that our fit to the TEXES data is also consistent with the ISO observations, we take our model at the ISO-SWS resolution and compare to the data (Figure 6). The model matches the fundamental Q-branches for the two molecules as well as a hot band Q-branch of C₂H₂.

Zoom In Zoom Out Reset image size

C₂H₂ is expected to be in rotational local thermal equilibrium (LTE) since it has no permanent dipole moment. The fact that the two excited vibrational states of C₂H₂ from which absorption was observed, ν₄ and ν₅, are found to be populated near LTE at the C₂H₂ rotational temperature, or even slightly higher, provides information about the physical conditions of the gas. For both bands, the hot bands indicate they are slightly out of LTE since the vibrational temperatures (T_v) are higher than the rotational temperatures (T_r). For the ν₄ + ν₅ − ν₄ bands, the T_v temperatures for the two components were 346 and 271 K while for the 2ν₅ − ν₅ bands, they were 277 and 227 K. The rotational temperatures for the ν₅ fundamental are 225 and 191 K for the two components. In LTE, the vibrational temperatures for the hot bands would equal the rotational temperature measured by the fundamental band lines.

The ν₄ level has no allowed radiative transitions to the ground vibrational level, and none of the rotational levels of the ground vibrational level are coupled radiatively, in both cases due to the symmetry of the C₂H₂ molecule. This suggests that the populations of the ground and ν₄ states should be set by collisions, and so should be in LTE at the same temperature. However, the ν₄ level can be radiatively pumped by absorption in the 7.6 μm ν₄ + ν₅ band followed by emission of photons in the ν₄ + ν₅ − ν₄ band. The ν₅ state must be populated predominantly radiatively for any plausible density, since the critical density is >10¹⁰ cm⁻³. From these considerations, we conclude that the brightness temperature of the radiation field seen by the observed gas must be close to the kinetic temperature of the gas. In addition, the fact that the measured rotational temperature for HCN is similar to that of C₂H₂ indicates that the rotational levels of HCN are kept populated either by collisions, which would require a gas density >10⁷ cm⁻³, or by infrared pumping through the ν₂ transitions. Lahuis & van Dishoeck (2000) conclude that either excitation mode is possible. The presence of C₂H₂ 2ν₅ − ν₅ absorption requires infrared pumping, although ν₄ + ν₅ − ν₄ absorption does not.

We can attempt to estimate the extent to which emission in our lines fills in the absorption, at least for C₂H₂, by using the 2ν₅ − ν₅ lines as probes of the ν₅ population. Our spectra require a ν₅ vibrational temperature ∼300 K. We need to compare this number to the blackbody brightness temperature of the background continuum radiation. From the shape of the 8–13 μm spectrum, Willner derived a dust temperature of 330 K assuming the emitting dust is optically thick, or 370 K assuming optically thin silicate emission. He preferred the latter model, which gives a brightness temperature only when combined with an assumption about the solid angle subtended by the source. In either case, our vibrational temperature is only ∼10% less than the background temperature, so that the source function is ∼30% below the background intensity, and lines would be expected to saturate at ∼70% of the continuum. However, the continuum spectrum from which Willner derived a dust temperature is the average over the continuum source. If the gas we observe covers only a small fraction of the source, the brightness temperature of the radiation passing through that gas could be much larger than the average.

Given the many assumptions and uncertainties in this calculation, it is difficult to make a strong statement. The level at which our C₂H₂ lines saturate, ∼85% of the continuum, which we model with a small covering factor, could instead be due partly or entirely to emission by the absorbing gas. It is possible for the lines to saturate at 85% of the continuum even with small covering factors (e.g., 24% and 6% for C₂H₂) since the line widths are narrower than our spectral resolution. Presumably the same statement applies to the other observed molecules.

From our infrared observations, we find that NH₃ has two components: one at −57.3 km s⁻¹ and the other at −60.1 km s⁻¹. Most of the NH₃ is found in the first component, for which the temperature is ∼278 K and the column density is 5.2 × 10¹⁶ cm⁻². The second component has a lower temperature (∼248 K) and the column density is 2.8 × 10¹⁶ cm⁻²(see Table 3). NH₃ absorption toward IRS 1 has also been studied in the radio by Wilson et al. (1983) and Henkel et al. (1984). They found $N_{\rm NH_3} = 2 \times 10^{18}$ cm⁻² in gas with T = 170–220 K and V_LSR = −60 km s⁻¹. Although their temperature and Doppler shift is in reasonable agreement with our bluer absorption component, their column density is nearly 20 times ours. Without constraining the ¹⁵N/¹⁴N isotopic ratio, our data allow an NH₃ column as large as that from radio observations, but with this NH₃ column ¹⁵NH₃ lines should have been apparent in our spectra. It is possible that we sample only a fraction of the column observed at cm wavelengths as a result of larger dust opacity at mid-infrared wavelengths preventing us from observing lines of sight through the densest regions of a knotty gas distribution. It may be notable that Wilson et al. found, as did we, that although relative depths of lines (in their case hyperfine components) require large optical depths, line shapes do not appear flat-bottomed as would be expected if they are optically thick. They suggest that the absorbing gas may be in small spectrally and spatially unresolved knots, each of which is optically thick, and which combine to make a line that does not have the shape of a thick line. Our use of two components in our fit may be a way of approximating this situation.

The velocities for the two components are −57.2 and −60.2 km s⁻¹, with corresponding temperatures of 319 and 171 K and column densities of 4 × 10¹⁵ and 1 × 10¹⁵ cm⁻². Since the HNCO temperatures and velocities derived from our observations agree with the values for the other molecules, we can assume that HNCO is in the same gas. It appears the region probed by HNCO is very small in angular size. Zinchenko et al. (2000) detect HNCO emission in the radio and derive a beam-averaged column density of 1.1 × 10¹⁴ cm⁻². This value is an order of magnitude lower than our observed column density. However, beam dilution in the radio observations can explain the difference in the column densities.

CH₃ is detected for the first time toward warm, dense gas. The parameters derived for this molecule do not agree as well with the other molecules, although the weakness of the observed lines makes the derived parameters rather uncertain. The separation between the centroid velocities for the two components, V_LSR = −54.2 and −62.8, is nearly twice the separation found for the other molecules. Also, the covering factor for both of the components is consistent with 100%, whereas as it is closer to 10% for the other molecules. In addition, the temperature of the second component is much hotter (greater than 900 K) than the temperatures derived for the other molecules. We note though that the fact that the observed lines are of high excitation makes them particularly sensitive to a hot gas component, which may not have been noticed for other molecules. And the weakness of the lines makes the derived covering factor (which is inferred from relative line depths indicating saturation) very uncertain.

CH₄ gas-phase absorption had been observed toward the neighboring protostar IRS 9 (Lacy et al. 1991; Boogert et al. 2004) but had not previously been seen toward IRS 1. We find centroid velocities of −56.3 and −60.1 km s⁻¹, which are similar to the other molecules. The temperatures are higher than for most molecules (∼670 K for both components) but have large uncertainties (see Table 3). The best fit seems to indicate that one of the components covers the entire source whereas the other covers about 6% of the source. Of all of the observed molecules, CH₄ has the highest column density. The implications of the high CH₄ column density the chemistry are discussed in Section 5.1.

We detect only six lines of CS (see Table 2—full table available online only). The lines are rather weak and parameters are poorly constrained. Despite uncertainties, the parameters agree with those for other molecules. The centroid velocities are at −55.2 and −59.4 km s⁻¹. The temperatures, T₁ = 224 K and T₂ = 249 K, are similar to those found other molecules (except for CH₄ and CH₃). From radio studies of the envelope around IRS 1, a CS abundance of ∼10⁻¹⁰ is found (Plume et al. 1997; Mueller et al. 2002; Shirley et al. 2003). However, also using radio observations, van der Tak et al. (2000) derive a CS abundance of ∼10⁻⁸. Our abundance is about 2 orders of magnitude larger than the largest estimate of the abundance in the envelope indicating that our observations are probing material closer to the protostar.

We now consider the implications of the observations for the chemistry of the observed material. Hot core models can be used to represent two of the scenarios discussed in Section 1: (1) material in a circumstellar disk and (2) photoevaporating knots of neutral molecular material. Hot cores are small (r < 0.1 pc), dense ($n_{\rm H_2} > 10^7$ cm⁻³) and hot (T > 100 K) regions associated with high-mass young stellar objects (Kurtz et al. 2000). Chemically, hot cores are identified by an enhanced abundance of fully hydrogenated molecules such as H₂O and NH₃, which are usually observed through rotational transitions at submillimeter and millimeter wavelengths. These abundances are enhanced with respect to the cold molecular envelope. A hot core is thought to form as a result of the protostar heating nearby material, which triggers ice sublimation from the grain mantles (Charnley et al. 1992). HNCO is another hot core molecule which has been found in the Orion hot core with an abundance ∼10⁻⁸ relative to H₂ (Zinchenko et al. 2000).

Gas-phase HNCO is possibly coming from evaporation of grain mantles. The 4.62 μm feature observed toward some protostars has been attributed to OCN⁻ frozen in grain mantles (e.g., Pendleton et al. 1999; van Broekhuizen et al. 2004). Recent ISO results of solid state features toward protostars show an OCN⁻ upper limit of 10¹⁶ cm⁻² toward IRS 1 (Gibb et al. 2004). The total HNCO column derived in this work is 5.4 × 10¹⁵ cm⁻², a factor of 2 lower than the solid OCN⁻ limit. In general, IRS 1 does not show many ice features compared to the colder nearby source IRS 9 (see Gibb et al. 2004), which has a column density N(OCN⁻) = 1.2 × 10¹⁷ cm⁻². If solid OCN⁻ sublimates and captures a proton to form HNCO, the expected column density for HNCO would be comparable to the observed OCN⁻ ice toward cold lines of sight (like IRS 9). However, the gas-phase column observed is at least an order of magnitude smaller than that of the solid state ion toward sources like IRS 9. This indicates that when OCN⁻ sublimates, a fraction of the ices goes into forming HNCO, while most of HNCO is destroyed in the warm gas-phase chemistry on short time scales. Further studies of how OCN⁻ reacts in the gas-phase are needed in order to understand the difference between ice and gas phase column densities.

Similarly, CH₄ is probably also evaporating from the grain mantles, and thus the gas-phase abundance is highly enhanced compared with the cold molecular envelope. Studies of the solid CH₄ content toward IRS 1 show that the column density is 14 times less than our observed gas-phase column (Gibb et al. 2004), suggesting that the solid CH₄ on grains has sublimated. In comparison, IRS 9 has a higher content of solid CH₄ than gas-phase CH₄ (Boogert et al. 2004; Gibb et al. 2004). If we consider that IRS 1 is likely a more evolved object than IRS 9 (Elmegreen & Lada 1977), the higher gas-phase CH₄ toward IRS 1 also points to grain mantle evaporation. The high abundance of CH₄ means that daughter products such as CH₃ and CH₂ should be abundant. While the photodissociation of CH₄ preferentially forms CH₂ rather than CH₃, our observations of CH₃ agree with this branching ratio. We predict that CH₂ is also present toward IRS 1 with a higher abundance than CH₃. It is uncertain where C₂H₂ is formed in the gas-phase chemistry or is evaporating from grain mantles. However, most models do not predict the observed column densities from formation in gas-phase chemistry only.

From our observations we see a high abundance of the parent molecules NH₃ and CH₄. High abundances of the other molecules are also observed. Nomura & Millar (2004) present a time dependent chemical evolution of molecules in a hot core. The models begin at the time when the protostar turns on (t = 0 years). Table 5 shows the abundances with respect to CO at an age of 10⁴ years. The hot core model seems to underpredict abundances with respect to CO of all our observed molecules. However, as mentioned in Section 4.2, the CO column density found by Mitchell et al. (1990) may not be the appropriate number with which to compare our observations. If the CO column along the lines of sight containing our molecules is larger than the column measured by Mitchell et al., then perhaps the abundances from the model will more closely resemble the observations. We can also compare the relative abundances between our observed molecules, avoiding the uncertainties in whether the CO and dust absorption trace the lines of sight that our molecules are found in. The observed and predicted values for N(CH₃)/N(CH₄) are similar, ∼0.01. For the other molecules the ratios from the models at an age of 10⁴ years are very different from the observed values. The values for the models assume that we can see all the material and comparing the model results directly with the observations without accounting for line-of-sight does not make for a good comparison. Figure 7 shows the column density evolution for six of the detected molecules as determined by Nomura & Millar (2004) corrected for the observed beam (H. Nomura 2008, private communication). For each molecule, the observed range of column densities is shown in the horizontal shaded area. This hot core model can predict the observed column densities for four of the six molecules plotted. However, the ages indicated by each molecule do not give a consistent overall age. For CH₄ and HCN the model overpredicts the observed column densities. The other molecules suggest ages between 2 × 10³ and 2 × 10⁶ years. While ages vary by three orders of magnitude, it should be noted that the model has not been optimized to be an accurate representation of this source. With chemical models that more closely represent the source, it may be possible to determine a more consistent age.

Zoom In Zoom Out Reset image size

If we consider the disk scenario, we can also compare our observations to a chemical model of a photoevaporating disk. Nguyen et al. (2002) present a chemical model for a disk around a 10 M_☉ star. The model consists of a flared disk with a photoevaporating layer on the surface. Table 5 compares our derived abundances with respect to CO for the various molecules to the predicted values from disk models. The summarized results of models taking the midplane temperature for the temperature of the layers below the photodissociation region (PDR) are discussed here. The difference among the three models is the variation in the X-ray ionization rate, ζ (1ζ = 1.3 × 10⁻¹⁷s⁻¹), from 1 to 10⁵ζ. Table 5 shows that the abundance for C₂H₂ is underpredicted. In the chemical model, C₂H₂ is produced only by gas phase reactions. One possible explanation is that C₂H₂ is formed as ice on grain mantles. The abundance of NH₃ is also underpredicted. The model with the highest X-ray ionization rate produces CH₃ abundances close to the observed value. For that model, N(CH₃)/N(CH₄) is 4 times larger than the observed ratio. These models do not fit our observations very well but can serve as a guide to the expected abundances in such a scenario. The model by Nguyen et al. has not been modified to try to fit data. So, adapting the model for this object may give predictions closer to the observations. The star in NGC 7538 IRS 1 is believed to be a 30 M_☉ star. The different radiation field may affect the resulting chemistry.

Doty et al. (2002) present a model of chemical evolution of the envelopes of massive protostars using AFGL 2591 as an example. From that model, they conclude that enhancements of C₂H₂, HCN, and CH₄ are possible at late times in the evolution (greater than 10⁵ years) and at high temperature (T∼ 800 K). Our derived temperatures indicate cooler material, yet the molecules seem to be enhanced in the line of sight. Alternatively, these species could be produced at high temperature in the inner disk, and subsequently brought outward by the disk wind. Including UV and X-ray radiation in the model affects the chemistry especially relating to HCN. Stäuber et al. (2004, 2005) find that increasing the UV and X-ray flux can lead to enhanced HCN starting by the ionization of N₂. The column densities for HCN and CS from the Stäuber et al. model of protostar AFGL 2591 agree with our observed abundances while their column densities for C₂H₂, CH₄, and NH₃ are lower than our observed values.

Hot core and disk chemistry models predict the enhancement of molecules such as C₂H₂, CH₄, and NH₃. However, the observed abundance of C₂H₂ is higher than what models predict from warm gas-phase chemistry (see Table 5). This indicates that C₂H₂ is probably frozen on dust grains and sublimates along with molecules like CH₄. Boudin et al. (1998) studied the solid features of C₂H₂, especially when mixed with H₂O or CO. They found that the C₂H₂ features broaden substantially when mixed with H₂O and to a lesser extent when mixed with CO. They compare the laboratory data to ISO data for the colder neighbor, IRS 9, which resulted in an upper limit of 8 × 10¹⁷ cm⁻² for the column density of solid C₂H₂. This upper limit is consistent with the observed gas-phase column density seen toward IRS 1. So, it is possible that solid C₂H₂ is sublimating from grain mantles as protostars heat the environment. The observed CH₃/CH₄ ratio agrees with the branching ratio for the destruction of CH₄. Based on these results CH₂ should also be very abundant.

We now attempt to construct a physical and geometrical model of NGC 7538 IRS 1. Within the possible scenarios presented from various radio and infrared observations (e.g., Minier et al. 2001; Lugo et al. 2004; De Buizer & Minier 2005; Kraus et al. 2006), we propose a scenario in which the molecular absorption presented here comes from a circumstellar disk. We will examine other possibilities first. Figure 8 depicts the possible scenarios allowed by the available observations.

Zoom In Zoom Out Reset image size

We can quickly rule out a simple model in which the absorbing molecular gas is in the foreground molecular cloud and not closely associated with IRS 1. Our observations require the gas to be much hotter (T∼ 300 K) and denser ($n_{\rm H_2} \sim 10^7$ cm⁻³ to maintain rotational LTE of HCN out to J = 21) than is found away from luminous sources in molecular clouds. According to van der Tak et al. (2000), temperatures do not reach the observed values until you get to within 400 AU of the central star. In addition, the infrared radiation field must have a brightness temperature ∼300 K to populate the C₂H₂ ν₅ level sufficiently to account for the observed 2ν₅–ν₅ absorption. Undoubtedly, the absorbing molecular gas is in close proximity to the infrared continuum source IRS 1.

It is not quite so easy to rule out a model in which the absorbing molecules are in boundary region between the envelope around the IRS 1 hypercompact H ii region and the outflow. This gas could be compressed by the ionized wind, and if it is as close as 01, or 280 AU, from a 10⁵ L_☉ source it would have a temperature near 300 K. The temperature structure determined by van der Tak et al. (2000) indicates that indeed temperatures range from 200 to 400 K at radii of 220–380 AU. However, interaction with the ionized wind, which has a velocity ∼100 km s⁻¹, would be expected to accelerate the gas, causing broad, blue-shifted absorption (see van der Tak et al. 2000). In contrast, the observed lines have widths of <8 km s⁻¹ and have centroids within a few km s⁻¹ of those seen in surrounding molecular gas. Another argument against the presence of the observed gas in an envelope around IRS 1 is the fact that net absorption is seen. Since the vibrational temperature is comparable to the brightness temperature of the continuum radiation, emission lines would be seen if the molecular gas had a larger extent than the continuum source. If the lines arose in molecular shell surrounding the hypercompact H ii region, this probably would be the case.

On the other hand, Campbell (1984) proposes that the centimeter continuum from IRS 1 results from partially ionized material in an outflow. The centimeter continuum emission has been spatially resolved into knots by Gaume et al. (1995), who suggest that the emission comes from photoevaporation of knots of neutral molecular material (see Figure 8). The CH₃OH masers seen by Minier et al. (2000) also seem to trace the knotty structure (in addition to the disk described in Section 1) probed by the cm continuum emission. The knots may be material stripped from the disk. The stellar wind would then blow on these knots. However, if our molecules are in these knots we would expect large velocity difference between the knots resulting in broad lines, which are not observed. If we compare the velocities we derive for the two components of our molecules to those observed for the CH₃OH masers, we find that they match the two velocities found for the A cluster of masers. This indicates that the infrared observations may be probing the same gas as the CH₃OH masers. This scenario is further supported by Kraus et al. (2006). However, Kraus et al. also find that the interpretation of the masers tracing a disk also agrees with their observations.

We propose that the absorbing material is in a circumstellar disk. There is evidence for a disk close to edge-on from other observations and models (e.g., Pestalozzi et al. 2004; Lugo et al. 2004; De Buizer & Minier 2005). In this situation, hot dust in the surface layers or the inner rim of a flared disk could provide the continuum source. With a nearly edge-on orientation, radiation from the inner region could pass through the outer disk atmosphere, where absorption lines could be formed. Emission from some of the dust on the surface of a slightly inclined disk or in an outflow could be observed directly, accounting for the dilution of the spectrum that we modeled with a small covering factor. This scenario is similar to the one described by Lahuis et al. (2006) to describe C₂H₂ and HCN absorption toward the low-mass protostar IRS 46 in Ophiuchus (YLW 16B). Detailed modeling of the disk structure of IRS 46 showed that it was possible to have the absorbing material in the disk illuminated by the continuum from the surface. Figure 8 is a cartoon of our proposed picture, where the orange rays represent the continuum. Some of the continuum radiation passes through the absorbing outer disk. In this scenario, dust settling is likely to occur, allowing for observations of large columns of gas without too much dust.