4.1. Assignment of Absorption Features
Previous studies of the spectrum of show both similarities to and significant differences from the measurements reported in this work. We do not believe that these differences are due to carelessness on the part of past researchers (or on our part) but rather are due to differences in the interpretation of the measured spectra as well as slight differences in the itself; differences in crystal structure, the relative importance of surface vs. bulk modes, or the relative proportions of amorphous, glassy or crystalline material in the samples measured. A major disagreement between our measurements and those of Begemann et al. (1996) lies in the interpretation and handling of features observed by both groups near 10 microns. Begemann et al. (1996) treat these features as a silica contaminant (as we did initially) rather than as due to SiS stretching vibrations.
4.2. The 8-13 micron features
A feature at 10 microns due to a silica stretching mode is usually accompanied by a bending feature of nearly comparable strength. In our initial attempts to obtain "pure" spectra we assumed that the features observed at 10 microns were due to SiO stretching modes in an amorphous silica contaminant and we spent several months fruitlessly trying to subtract spectra of mixtures of of varying degrees of oxidation and crystallinity from our measured spectra in order to obtain the "true" spectrum of . While no combination of silica spectra completely eliminated all of the features observed between 8-13 microns, several did lead to a significant degree of suppression of these features. Unfortunately, since silica spectra consist of both SiO stretching modes near 10 microns as well as OSiO bending modes near 22 microns, all such attempts at eliminating the 10 micron silica "contaminant" resulted in significant negative absorbance values for the "corrected" sample near 22 microns. We were unable to simultaneously minimize the 10 micron features and salvage any sensible data in the 20 micron region. We were therefore led to the conclusion that the features we observed between 8 and 13 microns were due not to but to .
We believe that the 8-13 micron and 22 micron absorption features are due to for a variety of reasons. The most important of these is that the features appear in strength in all samples measured in our lab, including a fresh sample whose XRD pattern is consistent with pure, crystalline . Furthermore, an absorption feature near 10 microns was consistently observed by Begemann et al. (1996) despite the extreme measures taken during that study to avoid contamination by atmospheric oxygen or water vapor. They could not eliminate these features from their laboratory spectrum: they chose to identify them as silica contaminants. However, since silica has features of comparable strength near both 9-10 and 22 microns, subtraction of sufficient silica contaminant spectra from that of to eliminate the 10 micron features will also remove a comparable amount of absorption at 22 microns. If the 10 micron feature were due to a silica contaminant, then its removal should still preserve the absorption near 20 microns. This was never possible with any of the spectra obtained in our laboratory: subtraction of sufficient silica "contaminant" absorption to eliminate features near 10 microns always resulted in an unphysical (negative) absorbance near 22 microns.
4.3. Possible source of the 8-13 micron features
We could find no value for the molecular force constant of the SiS bond in in the literature, so a calculation of the position of the SiS stretch from first principles is not possible. However, the energy of the CS stretch in is approximately 1500 while that of the SiO stretch in is near 1100 (Conley 1966). If we calculate a value of the force constant for the (OSi) = O stretch, assume that this force constant remains the same for the (SSi) = S stretch and calculate the new energy for the vibration due to the substitution of the heavier sulfur for the lighter oxygen atoms, the (SSi) = S stretch will occur near 800 . If we perform a similar calculation for the (SSi) = S stretch based on the known energy of the (SC) = S stretch at 1500 , we obtain a value of 1400 . Of course is a molecular liquid while is an infinite solid: we make no claim that the 8-13 micron features in the spectrum of orthorhombic, needle-like crystals must be due to the (SSi) = S stretch. However, we do argue that, based on the above, the energy of the SiS stretching vibration in solid could lie in the range from 800- 1200 , the range over which we observe significant absorption in each of the samples measured in our laboratory.
4.4. The effect of reaction with
If we examine the reaction of crystalline with humid air (shown in Fig. 2) we find additional support for the argument that the 8-13 and 22 micron features are due to vibrational modes in . First, we know from the XRD pattern shown in Fig. 1 that the original sample is more than 99% crystalline . Therefore if the 8-13 and 22 micron features are due to a "contaminant" present in this sample, the absorption strength of this contaminant must be nearly two orders of magnitude stronger than that of in order to appear at nearly comparable strength. Reaction of the crystalline sample with humid air results in the rapid loss of sharp features at 12, 17 and 20 microns, but in no significant increase in the strength of the other 8-13 or 22 micron absorptions as would be expected if the reaction of with humid air was the source of the "contaminant" in our samples. Our observations indicate that the vibrational modes responsible for the 12, 17, and 20 micron absorptions appear to be quenched by reaction with humid air and could therefore be due to surface modes in the orthorhombic crystals.
4.5. Composite spectra
Examination of the spectra in Figs. 2 and 3 reveal a range in intensities of various features, but three distinct groups of spectral features. Each of the spectra measured at room temperature after some exposure to humid air contains a feature at 10.7 microns that is missing from the spectra of samples heated in vacuo to 600 K. In addition, the spectrum of "surface modes" is distinctly different from any other spectra in Figs. 2 and 3. For ease of comparison with astronomical observations and with the spectra of possible "contaminants" we constructed the following composite spectra. First the spectrum of "surface modes" was obtained by subtracting Fig. 2b (the sample exposed to moist air and scaled such that the height of the 9.5 micron feature equaled that in 2a) from Fig. 2a (the unexposed sample) and displaying this difference as Fig. 4b. Second, a 300 K composite was constructed by scaling Figs. 3a-3d to a constant area of the 10.7 micron feature and averaging these spectra to yield Fig. 4c. Finally, a 600 K composite was constructed by first scaling the spectra in Figs. 3e-h as was done for the 300 K composite, then averaging the four spectra to obtain Fig. 4d. For comparison with these spectra we have included the spectrum of an amorphous, underoxidized silica smoke sample produced via the combustion of silane in our laboratory (Fig. 4e), the spectrum of monodisperse silica spheres kindly provided to us by Dr. Thomas Henning (Fig. 4f) and the spectrum of elemental sulfur (Fig. 4g) all measured using the Harrick Diffuse Reflectance Attachment in our laboratory. Note that the spectrum of a freshly-condensed, underoxidized silica smoke tends to maximize the relative contribution of the 10 micron SiO stretch compared to the OSiO bending mode near 20 microns (Nuth and Hecht, 1990). Finally an astronomical spectrum of 07134+1005 (SAO96709, HD56126, DM+10 1470), a proto planetary nebula candidate star of spectral Class F5, obtained from Dr. Sun Kwok (1996) is shown for comparison as Fig. 4a.
The most obvious aspect of Fig. 4 is the very great difference between "Surface Modes" (4b) and any of the other or silica spectra. Initial comparison of this spectrum (4b) with astronomical observations (4a) looks quite promising in fitting the sharp peak seen at 20 microns, but falls short of perfection for two reasons. First, the 20 micron "Surface Mode" peak is insufficient by itself to explain the astronomical observation, additional absorption at longer wavelengths and much less absorption near 19 microns is required. Second there is no evidence for a sharp and reasonably strong feature in the astronomical spectrum at 17 microns, although there is some evidence for the presence of a minor feature at a slightly shorter wavelength. Neither of the composite spectra (4c and 4d) can explain the sharp peak at 20 microns seen in the astronomical spectra, although these spectra can help to explain the long wavelength wing and do contain features near 11 and 12 microns as seen in the spectrum of 07134 + 1005.
Comparison of the "Surface Modes" to the spectrum of elemental sulfur (4g) easily demonstrates that elemental sulfur is not an important contaminant in this sample. Similarly elemental sulfur does not seem to contribute to the more amorphous composite sample spectra (4c or 4d). However, comparison of spectra 4a and 4g does indicate that elemental sulfur could be responsible for the minor spectral components observed in 07134 + 1005 in emission near 21 and 17 microns and in absorption at 12 microns. The spectra of both amorphous silica smoke and the monodisperse silica spheres are also interesting: there are significant similarities between these spectra and both composite spectra, supporting the interpretation of Begemann et al. (1996) that the features between 8-13 microns could be due to a silica contaminant. We have previously argued that the XRD pattern (Fig. 1) of the sample that yielded Fig. 2a conclusively demonstrates that crystalline can produce features in the 8-13 micron region. We have also argued that the spectrum of solid (consisting of edge-sharing, tetrahedral units) should resemble the spectrum of solid silica (made up of corner-sharing units) as is the case in Fig. 4. Amorphous silica smokes show strong absorption at 9.2 microns with a shoulder near 8.5 microns, small features near 11.4 and 12.5 microns, and a second absorption at 22 microns. The monodisperse silica spheres show a similar set of more distinctive features and a much stronger feature near 22 microns. The silica and sample spectra do differ in the relative widths, and exact positions of the features. Another difference is the presence of the 10.7 micron feature in cold that is not seen in any silica sample. This comparison illustrates the difficulty of subtracting a significant silica "contaminant" spectrum from that of . Subtraction of sufficient silica intensity to eliminate the 8.5, 9.5 and 12.5 micron features in spectra more than eliminates the 21-23 micron absorption in and still leaves residual features in the 8-13 micron region. Nevertheless, it is obvious from Fig. 4 that spectra of bulk are similar to those of , except in the case of the surface modes. The spectrum of this particular component of is quite distinctive.
4.6. Spectra of astronomical sources
If the spectra obtained by Chan et al. (1995) are due to , then one reasonable morphology for the astronomical sources is that of a large shell or disk of cold dust (100-200 K) surrounding the central star. The cold dust would absorb radiation near 10 microns emitted by the central star but would produce an emission maximum near 15-30 microns from the extended, cold dust shell. We have examined a simplified model of such a system as a test of this model and to suggest future observations. In our model a cool star (2000-3000 K) is surrounded by a "gray body" dust shell with a temperature of 100-200 K. The shell radius is adjusted to achieve radiative equilibrium in the dust shell. We calculated mass-absorption coefficients for as a function of energy from 80 - 2000 derived from the mass-absorption coefficients in (Nuth et al. 1985) and the spectrum of the fresh, Strem Chemical sample shown in Fig. 2a. The stellar emission as seen from outside the shell is then corrected for the absorption of a shell and the shell emission is corrected for the enhanced emissivity associated with the absorption features of . The net emission of the combination can then be determined. Fig. 5 shows the results of several such calculations.
Of particular interest is the fact that the 15 to 30 micron region always shows up in "emission" since the shell emission dominates this region. However, the features between 8.5 and 12.5 microns can show up in either absorption or emission depending on the balance between the light absorbed and that emitted by the shell. This balance is particularly sensitive to the shell temperature with these features barely appearing in absorption for a 100 K shell and in emission for a 200 K shell. Thus, the 21 micron feature is far more diagnostic of than is the 9.5 micron feature despite the latter feature's strong absorption coefficient. However it would seem prudent to obtain visible to near-infrared spectra of astronomical sources exhibiting the 21 micron emission feature so that adequate radiation-transfer models could be constructed in order to test the hypothesis that a shell of cold dust surrounds a hotter stellar source.
A more serious problem with as the source of the 21 micron feature is the poor fit of the 21 micron feature of to the astronomical spectrum. The "surface mode" fits the onset of the 21 micron feature reasonably well, but is not wide enough to account for all of the absorption. Even worse is the fact that the 20 micron surface mode feature is always accompanied by a sharp feature at 17 microns that is nearly as intense as the 20 micron peak in all available laboratory spectra (e.g., Fig. 2a and Begemann et al. 1996) but that is not observed in any of the astronomical sources. The "bulk" laboratory spectra do not exhibit a 17 micron absorption feature, but the onset of their 21 micron absorption maxima is at longer wavelength than observed in astronomical sources. Combining the 20 micron "surface mode" feature with the bulk composite spectrum (e.g. as for the fresh Strem Chemical sample in 2a) while arbitrarily excluding the "surface mode" feature at 17 microns produces a good match to the observations but is not justifiable based on the available laboratory data.
The 7-25 micron spectrum of 07134+1005 is shown on Fig. 5 for comparison with our model of an shell as is the spectrum of the fresh Strem Chemical sample from Fig. 2a. As noted above, does not provide a very good fit to the 21 micron feature. It should be noted that 07134 + 1005 may be considerably hotter than the 3000 K black body used in our model calculation, shifting the minimum in stellar flux to shorter wavelength and putting the 9.5 micron feature prominently in emission, thus increasing the discrepancy between observation and an model. An alternative material which may produce spectra consistent with the astronomical observations is nanodiamonds (Hill et al. 1996). Kwok and Bernath (1996) argued that the 21 micron feature could be due to PAHs based on the presence of strong and absorption bands in 21 micron sources (Hrivnak 1995) and the presence of the well known UIBs near 3.3, 3.4, 7.7 and 11.3 microns in many of the carbon-rich sources of the 21 micron emission band. These same arguments would be consistent with the production of nanodiamonds in such sources, especially since the infrared spectrum of such diamonds sometimes contains an absorption band at 475 (Hill et al. 1996) and because Nuth (1987) has argued that diamond is thermodynamically stable at very small particle size. More detailed spectra of unaltered microdiamonds are needed in order to confirm or refute this working hypothesis.
© European Southern Observatory (ESO) 1997
Online publication: March 24, 1998