Astron. Astrophys. 361, 1095-1111 (2000)

5. Deuterated species

We have searched in this source for the deuterated species CH₃OD, HDO and HDCO molecules at millimetre wavelengths. HDO and HDCO molecules appeared readily whereas for CH₃OD, only a few lines are tentatively detected. We do not discuss the HDCO observations as we cannot infer the D/H ratio based on the H₂CO lines which present a self reversed saturated absorption profile. To perform a comparison with the solid phase fractionation, we have used the infrared spectra provided both by ISO and UKIRT, and put constraints on the abundance of the OD containing species based on laboratory measurements.

5.1. HDO, H₂O and CH₃OD in the gas phase

High signal to noise data obtained with ISO in the 6 µm wavelength range allow us to trace a hot gas phase water component in addition to the cold saturated one already reported (Dartois et al. 1998). This component is much less abundant than the cold one, as suggested by Dartois et al. (1998) where it was reported to be at least a factor of three lower in abundance. The derived parameters for the hot H₂O are 410¹⁷cm^-2 as shown in the Fig. 10, with a mean derived excitation temperature of 200 K.

Fig. 10. H2O ro-vibrational lines from the hot component seen in absorption with ISO towards RAFGL7009S. Modelled spectra at different temperatures are shown for comparison, assuming the same Doppler parameter as the one used in Dartois et al. (1998). We derive NHDO=41017cm-2 and a mean temperature of 150 K

With the IRAM 30m telescope we measured four HDO lines, presented in Fig. 11. We could then estimate the column density, source size and excitation temperature for this molecule using our maximum likelihood minimisation. The source size is 11" , the column density N_HDO=1.010¹⁵cm^-2 and excitation temperature 180 K. For comparison, the corresponding rotational diagram is also presented in the same figure.

Fig. 11. HDO , , and transitions. The best fit rotational diagram was obtained with a FWHM source size of 11"  leading to a column density of =1.1015cm-2 with an excitation temperature of 180 K.

Using the information obtained from these observations, and comparing with the corresponding H₂O abundances for the same hot component (they should be spatially correlated), the derived ratio for gas phase water is then of 2.510^-3. As the derived temperature is high, we think the source size is overestimated for this molecule, as it is very hard to keep a so huge volume of hot gas phase water. We therefore estimate the above D/H ratio value to be an upper limit.

To pursue our investigation on deuterated species in this source, we also observed several CH₃OD rotational lines with the 30m telescope. The result of these observations are shown in Fig. 12. We detected the 1(1)-1(0)E (110188.860 MHz), 2(1)-2(0)E (110262.640 MHz) and 5(0)-4(0)A (226538.674 MHz) lines. In the individual plots of Fig. 12, we draw vertical lines to indicate the expected position for the CH₃OD lines, as well as the energy in cm^-1 at which lies the upper level of the transitions. We thus only see low energy level lines.

Fig. 12. CH3OD lines observed toward RAFGL7009S. The expected lines positions are indicated as well as the corresponding upper state energy (in cm-1) of the transition.

Based on our minimisation analysis, the lines lying in the 3mm atmospheric window lead to a rotational temperature of 7 K, a column density of 610¹³cm^-2 in a source size of =8" .

In the 1 mm window, we find a temperature of 25K and a column density of 110¹⁴cm^-2. Taking into account these measurements and their associated uncertainties, we then derive N=0.3-1.610¹⁴cm^-2 and T=5-45 K.

For the methanol molecule, taking into account that the source size was estimated to be =8"  at 241 GHz, we obtain for the main isotope species a corrected column density of 1.210¹⁶cm^-2 (different from that shown in the Fig. 8 where no source size has been specified).

The derived ratio is then estimated to be between 10^-2 and 210^-3.

5.2. Solid phase HDO

To compare the gas phase D/H ratio to that in the solid phase, we observed the OD stretching mode in the infrared spectrum obtained both by ISO and UKIRT in the vibrational associated wavelength range. We also determined in the laboratory the absorption cross section of this mode.

5.2.1. Laboratory experiments

Since the HDO molecule is rather unstable towards substitution, we decided to produce it in-situ in a controlled experiment in order to derive a reliable integrated absorption cross section of the OD stretching mode. Similar setups have been used successfully to obtain infrared spectra which are directly comparable to astronomical spectra (d'Hendecourt et al. 1996). The setup used to simulate interstellar spectra has been described elsewhere (Allamandola 1987). It consists of a liquid helium flow cryostat cooled to temperatures of 4K-300K, and a CsI window onto which our gas mixtures are slowly condensed. Mixtures are prepared separately in an adjoining stainless steel vacuum line. Condensed mixtures may be UV irradiated using a microwave discharge H₂ lamp. The cryostat is coupled to an FTS infrared spectrometer (Bruker IFS 66v). Spectra presented here were recorded at a resolution of 1 cm^-1, comparable to the ISO SWS resolution. We have performed two different experiments. The first involved deuterated methane and oxygen, and was aimed at determining the integrated absorption cross-section of the OD stretch. The second, based on water ice mixed with molecular deuterium, was used to directly compare with astronomical spectra.

5.2.2. Determination of the OD stretch absorption cross-section

We have condensed a mixture of deuterated methane (CH₃D) mixed with an equal proportion of O₂ at 10K. After UV irradiation with the microwave H₂ discharge lamp that produces hard UV photons (Okabe 1980), we principally formed H₂O, HDO, CO and CO₂ as can be seen in the vibrational spectrum presented in Fig. 13. On this figure the upper curve is the spectrum of the condensed mixture (CH₃D:O₂_1:1) before irradiation, where the observed lines are all attributed to transitions arising from the CH and CD in CH₃D. The lower curve is the resultant spectrum after photolysis. The assignments of the vibrations are indicated below the curve. H₂O and HDO are identified via their OH and OD stretch at around 3300 cm^-1 and 2450 cm^-1, as the mass difference between D relative to H induces a shift of the OH stretching mode such that /. HDO is then responsible for the transition observed around 2450 cm^-1. Note that the OH stretch pertains both to H₂O and HDO and is indistinguishable in the broad amorphous band observed around 3300 cm^-1. Finally, the broad nature of the two observed bands, due to hydrogen bonding of the clustered molecules, ensures that the clusters are themselves amorphous. This is essential to determine the integrated absorption cross-section to be used in the comparison with astronomical data.

Fig. 13. Spectra of a CH3D:O2_1:1 ice mixture which were condensed at 10K (top curve), and after irradiation by UV photons (lower curve). The irradiation mainly produces H2O as well as HDO and D2O, giving rise to the two broad bands around 3300 cm-1 and 2500 cm-1. The different vibrational modes observed are labelled.

To a first approximation, we assume in the photolysis process that the formation rate of H₂O, HDO and D₂O is ruled by the statistics of the D/H ratio from the deuterated methane molecule. We can then proceed to evaluate the oscillator strength for the OD stretching mode in the laboratory spectra. Let us denote by r the initial isotopic ratio in deuterated methane:

During photolysis, we then produce H₂O molecules, 2r(1-r) HDO and (1-r)² D₂O (9:6:1). If we write the integrated absorbance ratio observed for the OH and OD stretching modes we then obtain, regardless of the molecules from which they are part of:

where and are the number of OH and OD groups participating in the respective features, and are the integrated absorption cross-sections. Remembering that , we deduce:

Measurements of the integrated absorbances in the spectrum after the photolysis lead us to:

The half width at half maximum of the OD stretching line in the amorphous state measured in this spectrum is about 170 cm^-1. The optical depth at the line centre can be approximated by:

For astronomical purposes, this number implies that with a signal-to-noise ratio of 30, one can determine with a 3 accuracy an optical depth of 0.1 corresponding to a column density of 4.710¹⁷ cm^-2.

We emphasise that the line centre absorption cross-section will change when one goes to the crystalline form of water ice. In this case, the line will be sharper as shown in the next experiment, rendering a tentative detection easier.

5.3. Production of an H₂O_HDO mixture

This first experiment was dedicated to the evaluation of the integrated absorption cross section of the OD stretching mode in HDO. We can now move onto the interpretation of the astronomical spectra using a more realistic mantle composition i.e. a water dominated one.

As is well known, astronomical ices are largely dominated by water ice. We have chosen to obtain a pure H₂O environment in which the minor species HDO is embbeded. We have thus condensed a mixture of H₂O and deuterated molecular deuterium (D₂) at 10K that was irradiated using the same UV lamp discussed above. We then converted part of the H₂O molecules into HDO. After slight warming, the extremely volatile D₂, HD and H₂ molecules that were present or created in the process, migrated into the ice matrix and escaped the sample. Thus we ended up with a mixture essentially containing H₂O, HDO and D₂O, the latter being much less abundant.

We discuss the experimental results together with the astronomical comparison hereafter.

We have obtained spectra of RAFGL7009S with both the ISO satellite (SWS01 template) and the United Kingdom InfraRed Telescope (UKIRT). The ISO spectrum gives a view of the complete spectrum of the source (d'Hendecourt et al. 1996). The ground based observations are much more sensitive in the atmospheric windows. Unfortunately, with the UKIRT CGS4, we have only measured the spectrum region located just before the strong telluric CO₂ absorption, between 3.4 and 4.04 µm. With the ISO satellite, we only used the part of the spectrum in the "1e" band (from 3.4 to 4.1µm), as the next band ("2a") is less sensitive and also because in this band the spectral response possesses a large feature peaking at 4.1 µm. This feature could appear as an absorption band if badly subtracted, which would result in a spurious identification of a weak band. It is more secure to use the "1e" band as it covers half of the expected amorphous HDO absorption feature.

The main result from the UKIRT observation has been described extensively in Dartois et al. (1999). It consisted in the detection of various vibrational modes of CH₃OH that are particularly well matched through laboratory simulations and allowed us to confirm the high abundance of this molecule in the mantles toward this object. We present in Fig. 14 the continuum estimate (left panel) and the resultant baseline corrected spectrum (right panel) between 3.4 and, 4.04 µm for the UKIRT spectrum, 4.1 µm for the ISO spectrum. Both spectra clearly show a double methanol feature attributed to combination modes. As we are looking for a large feature (FWHM0.2µm at 4µm), we fitted the continuum with a first order polynomial, whose slope was derived by adjusting the data in the range shown by the arrows. We chose a first order polynomial to be sure not to introduce any artificial curvature as could be the case with higher orders. The continuum is expected to curve downwards in this range since the 3 mode of the water ice band will absorb there. This assumption can be safely verified in the numerous ice dominated spectra recorded by ISO and especially the ones where an HDO detection was reported by Teixeira et al. (1999). On the right part of the figure, we display the transmittance spectra after substraction of the polynomial. We have normalised the laboratory HDO spectra shown in Fig. 13 in accordance to the derived integrated absorption cross-section for the OD stretching mode. We display the transmittances expected for various column densities, as indicated on the curves in the Fig. 14. We do not detect the amorphous HDO absorption band, the smaller value being our estimate on the upper limit. The upper curve represents the methanol contribution to the astronomical spectrum (see Dartois et al. 1999 for a more detailed discussion).

Fig. 14. estimate of an upper limit on the solid HDO content of the ice towards RAFGL7009S using ISO and UKIRT spectra. See text

The HDO column density is estimated to be less than 710¹⁶cm^-2. Using the H₂O column density in this source (1.210¹⁹cm^-2, d'Hendecourt et al. 1996) the ratio is then 610^-3.

Contrary to gas phase millimetre wave spectra, which are linked to the rotational excitation ladder of the molecules, we probe in the infrared the vibrations of molecules. In the solid phase, the complex rovibrational spectrum merges into a broad main feature whose position will be less specific for the particular molecule involved but sensitive to the kind of atomic bonds. This implies that every molecule possessing an OD bond (such as HDO, CH₃OD, C₂H₅OD...) will absorb in the same region. This behaviour is crucial as it strengthens the derived upper limit which holds very strictly for HDO as well as for CH₃OD. On the basis of the high CH₃OH abundance detected in RAFGL7009S (Dartois et al. 1999) and assuming the integrated absorption cross section ratio of OH groups are conserved when one substitutes hydrogen by deuterium, we can determine an upper limit on the D/H ratio in methanol, given by:

which leads to: 310^-2.

5.4. Discussion

5.4.1. Solid phase

In a paper on the surface chemistry of deuterated molecules, Tielens (1983) calculates the D/H ratio expected in grain mantles. The main point addressed by surface chemistry is that whereas in the gas phase the D/H ratio in HD and to a lesser extend in D (atomic deuterium) should approach the cosmological value, it will be different on grain surfaces. Indeed, the change in mass between D and H atoms by a factor of 2 considerably influence the behaviour of the species relative to their surface binding energy. The residence time-scale on grains will therefore favour deuterated species over hydrogenated ones. This aspect as well as the difference in activation barriers for the reactions involving deuterated species was treated by Tielens (1983).

The numbers derived in his model are very high. In Fig. 1b of his paper are plotted the expected (D/H)^ice ratios of the most abundant deuterated molecules relative to their hydrogenated counterparts in the mantle as a function of hydrogen number density. The deuterium enrichment increases steeply with density. In the source we study here, the mean density is evaluated to be about 10⁶cm^-3. Using these numbers we can exclude the very high enrichment derived from the model of Tielens, which would lead to D/H ratios 1, orders of magnitude above the upper limit we measure.

The D/H ratio in ices seems not abnormally high, if compared to the gas, as our upper limits lie just above the ratio measured in the gas for HDO and even slightly lower if we compare to IRAS16293-2422 (van Dishoeck et al. 1995). This implies that the D/H ratio, if lying just below the limits given in this study, stays almost constant in the gas after mantle evaporation.

Teixeira et al. (1999) recently reported the detection of HDO in ices toward massive young stars. Following their discussion, it remains unclear if it provides support for the assumption that the origin of high levels of deuterium fractionation in hot cores is evaporation of the deuterated species from grain mantles. Indeed, these observations can not distinguish between fractionation generated by low temperature gas phase chemistry, followed by accretion onto dust, from the production of deuteration in or on grain surfaces.

In particular, some points need to be investigated in the future if the fractionation is attributed to grain chemistry. Why should a source like W33 A, the one with the highest extinction and with the coldest grains, displays a ratio more than ten times lower than NGC 7538 IRS9, whereas the former appears colder and display stronger ice absorptions?

There is an indication of the evolution of NGC 7538 IRS9 compared to W 33A that shows the latter is cooler. It is given by the shape of the 15.2µm CO₂ bending mode which possess a triple peak structure in W 33A whereas it shows a double peak substructure in NGC 7538 IRS9. This structure is associated with a molecular complex formation in the ice (Dartois et al. 1999) and the evolution of the substructure traces the temperature evolution of the ices, showing W33A is apparently colder. The same temperature evolution behaviour is found by Boogert et al. (2000) using the ¹³CO₂ stretching mode, as well as the flux ratio at 45 and 100µm. What can then explain a so huge difference between the two sources given they share the same bolometric luminosity (9.210⁴ for NGC7538 IRS9; 1.110⁵ for W 33A, see reference in Boogert et al. 1999)?

5.4.2. Gas-solid phase models

Recently, Bergin et al. (1998) have also discussed an alternative to classical grain surface chemistry through the formation of ices behind shock waves. They show in this model a plot of the resulting relative concentration of HDO/H₂O in the grains versus that of CO₂/H₂O. In this diagram, a box is drawn and said to represent the observed values in the ISM. To draw this box, the authors are accepting the hypothesis that the gas phase measurements directly mimic the deuterium fractionation obtained on grain mantles, which still remains to be proven in the general case.

Another point to be addressed for a full discussion is the fact that all the models have been generated for dust temperatures strictly above 25K. Extrapolating from the curves presented in Fig. 2. to a 10K curve, a large fractionation is expected (above 10^-2). If the grains remain at a temperature above 25K, this model prevents a lot of deuterium fractionation. Indeed, the residence time-scale for D and HD is reduced by a huge factor as compared to a grain at 10K. The residence time-scale for a species on a grain is given by =exp(-E_B/kT), where is a characteristic vibration of the adsorbed species, where E_B is the binding energy to the surface, k the Boltzmann constant and T the grain temperature. For H and H₂ E_B/k is between 200 and 700K depending on which ice the molecule is adsorbed (Schmitt 1993). The corresponding ratio of residence time-scales at 25K and 10K is then so large that it deserves a complete discussion in gas-grain interacting models. As an example, Brown & Millar (1989), with n=310⁴cm^-3 and grains at 10K, find a ratio HDO/H₂O of 8.210^-3, well above the fractionation at 25K and with a sticking coefficient of only 0.3 for H and D atoms. Finally, the influence of higher gas densities such as the one met in RAFGL7009S ( 10⁶cm^-3) must also be investigated. It would be interesting to recalculate the same models with a density more appropriate to the source reported in this paper.

In the models by Tielens (1983) and Brown & Millar (1989) which both include surface chemistry involving deuterium, large fractionation is predicted for all grain species considered at densities greater than 10⁴ cm^-3. If we believe that observations of high enrichment in deuterated species in hot core regions are the consequence of the evaporation of grain mantles, we would then expect from these models comparable fractionation for HDO, NH₂D and HDCO. In the Brown & Millar model, the result presented in their Table 2, shows that the / ratio lies within the range 0.12-0.64. In the model of Tielens (1983) HDO, NH₂D and HDCO roughly exhibit the same deuterium enrichment. This is in contradiction with hot cores measurements giving 0.001HDO/H₂O0.004 (Jacq et al. 1990), NH₂D/NH₃0.06 and HDCO/H₂CO0.14 (Turner 1990). The inconsistency comes mainly from the fact that these ratios are not verified simultaneously, suggesting that different enrichment mechanisms apply to different species.

5.4.3. Pure gas phase

We also have to take into account the observations of high deuterium enrichment in radicals, molecular ions and molecules which never come directly from grain evaporation and are believed to result from pure gas phase reactions (Gerin et al. 1987; Guélin et al. 1982). However, these implied molecules posses formation routes that seems decoupled from the grain chemical evolution.

The equilibrium rate constant relating forward and backward reactions in the cases of pure D/H exchange like:

are classically related by:

where MH and MD are molecules, radicals or ions, Q is the partition function, which includes internal terms (electronic, vibrational, rotational) and translational terms. and are the symmetry numbers of each species included to take into account the symmetry statistic of each species. is the difference in vibrational zero point energies between the two species MH and MD. The multiplicative term is generally of the order of unity, whereas the exponential term is highly temperature dependent. Low temperature fractionation can sometimes then proceed simply because a key reaction will displace the equilibrium in the deuterium enrichment way. Thus apart from the formation reactions of the hydride and deuteride molecules by reactions involving other species (such as ), the exchange reaction favours deuteration at low temperature as the vibrational zero point energies are generally lowered due to the lower M-D vibration frequencies compared to the M-H.

Ion molecule chemistry such as the one implying H₂D⁺ and CH₂D⁺ play a major role in pure gas phase chemistry. Recently Shah & Wooten (1999) have measured NH₂D/NH₃ ratios towards low-mass star formation regions and argue that the high deuterium enrichment observed is compatible with pure ion-molecule chemistry, and that molecules observed in region of ices evaporation could in fact simply result from the reapparition in the gas of molecules enriched in the gas and subsequently depleted on grains without any further enrichment in the solid phase.

Recently, Roueff et al. (2000) reported the detection of doubly deuterated ammonia in the dense core of L134N at the level of 10% compared to singly deuterated ammonia and question whether it is possible to obtain such observed high D/H ratios with gas phase chemistry. Tiné et al. (2000) find that NH₂D/NH₃ high fractionation levels are compatible with a C and O depleted gas phase chemistry. As the species such as CO condense on grains, molecular ions are destroyed on longer timescales. In such a case, the gas phase fractionation can proceed efficiently as the abundance of ionised precursors is enhanced.

5.4.4. Summary

The observations of deuterated species presented above show that it is difficult to obtain a definite picture on the dominant route for fractionation (either low temperature gas phase chemistry or grain surface and bulk UV assisted chemistry).

In the observations presented here, the methanol molecule is orders of magnitude more abundant in the solid phase than in the gas phase. A very small evaporated fraction (1% to 0.1%) is sufficient to account for the methanol gas phase abundance observed. It suggests that the measured D/H enrichment for CH₃OH in the gas of at least a factor of 100 and perhaps a factor of 1000 above the cosmological D/H is indeed achieved by grain chemistry for this species.

These observations are not a definite proof that grains are responsible for the general enrichment in such sources but that it seems a good tracer of grain deuteration in lines of sight where it has been measured in the solid phase.

Finally we stress that a careful search for the CD stretching mode in the spectra of embedded objects is of high priority to put further constraints on the degree of deuteration of solid carbon species such as CH₃D, CH₂DOH and HDCO. Indeed the CD stretching mode for these molecules falls in the wavelength range 4.5-4.9 µm (see Fig. 13, this paper for CH₃D), a region accessible from ground based telescopes. Opportunities are opened by new generation telescopes such as the VLT and Gemini. As seen by the results presented here for HDO, we stress the necessity to obtain very high signal-to-noise ratio, up to 10³ in order to detect CD or to obtain meaningful upper limits.

© European Southern Observatory (ESO) 2000

Online publication: October 10, 2000
helpdesk.link@springer.de