3 Results

  
3.1 Detected sources

A total of 43 sources were detected: a complete list is presented in Table 1. The 6 $\hbox{$^{\prime\prime}$ }$ pfov used here results in coordinates which are accurate only to $\pm3$- $4\hbox{$^{\prime\prime}$ }$ at best, so where sources have been identified with previous studies the coordinates presented are those from previous observations, where higher spatial resolutions or better S/N have resulted in more accurate positions. The coordinates evaluated here are only presented for sources which represent new or uncertain detections. In the 3 cases where sources have been identified as unresolved doubles the coordinates presented are those of the brighter source. Of the 43 sources detected 6 were not identified with detections from previous surveys (see Sect. 3.1.1) and 2 (Ser A 7 and Ser B 11) were determined to be different images of the same source, as the two Serpens fields overlap slightly. Consequently we were able to obtain complete 5.0-16.3 $\mu$m spectra for a total of 40 sources, along with a further 2 partial spectra: a complete atlas of these 42 spectra is presented in Appendix A.

  
3.1.1 Ghost images and new detections

As noted by Blommaert et al. (2001) ghost images and stray-light are a significant problem when using the CVF, so care must be taken when identifying sources. Ghost images arise in two ways and are found either around the true image or directly opposite it (relative to the optical axis). Potential ghost images were identified by looking at the spectra of the objects, their flux densities and their positions relative to the instrument optics. In this manner 7 objects initially identified as sources were reclassified as ghost images (4 in $\rho$ Oph E, 2 in Cha I and 1 in RCrA). This leaves 6 "new'' detections in our survey, which we now address:

RCrA:

3 sources in RCrA are identified as new detections: sources RCrA 2, 6 and 7. Source RCrA 2 lies close to RCrA 1, separated by $\sim$30 $\hbox{$^{\prime\prime}$ }$, but shows sufficiently different spectral characteristics that it cannot be a ghost of this source. It is therefore considered to be a real source, probably a YSO. Both RCrA 6 and 7 lie in a region of nebulosity between the bright sources HH100-IR (RCrA 5) and RCrA (which lies just outside the image frame). These are not considered to be ghost images, but may well represent peaks in the nebulous emission rather than true YSOs. It should also be noted that this region of the RCrA cloud was not included in the ISOCAM survey of Olofsson et al. (1999), due to detector saturation, so it is not unreasonable to expect new detections here.

$\rho$ Oph E:

A large number of ghost images were found here and after they were removed 1 unidentified source remained: $\rho$ Oph E 4. However, this is the faintest source in the survey, with a flux rising to just 0.17 Jy at 16 $\mu$m, so is probably a peak in the nebulosity rather than an "new'' YSO.

Ser B:

2 sources in Ser B were not identified with previous observations: Ser B 4 and 7. Both these sources lie near to SVS20 and so could be ring-like ghost images associated with it. However they also lie in a previously identified region of nebulous emission (Kaas 1999) and so may instead be peaks in this nebulosity. It is impossible to resolve this ambiguity completely, but in either case these 2 sources are probably not "new'' YSOs.

  
3.1.2 Background stars

Of the 36 sources identified with previous studies 3 have previously been identified as stars which are not associated with the star-forming clouds: CK2 (Ser B 12) by Chiar et al. (1994) and Casali & Eiroa (1996); Ser B 1 and Ser B 11 / Ser A 7 by Giovannetti et al. (1998).

CK2 is the only one of these 3 sources which has been identified as a background field star (probably a background supergiant, Casali & Eiroa 1996). Unfortunately it fell directly on the bad column in the array here and so no useful spectroscopic data regarding CK2 was obtained. Both of the other two sources in Serpens show extremely deep absorption features due to ices and silicates and both have very red SEDs. Taking the K-band magnitudes of Giovannetti et al. (1998) and assuming the SEDs of these stars to be Rayleigh-Jeans spectra ( $F_{\nu}\propto \lambda^{-2}$) results in predicted mid-IR flux densities of $\sim$0.1 mJy for both sources, approximately 1000 times less than what is observed. Consequently it seems unlikely that these are background sources, so we interpret them to be deeply embedded objects: with this interpretation these 2 sources are 2 of the 3 most deeply embedded objects in this survey.

  
3.2 Spectral fitting

The spectra of the 43 sources identified in Table 1 were fitted using the method described above: the results of this procedure are presented in Table 2 and the fits and continua obtained are presented in Appendix A. The fitting procedure is both robust and unambiguous, converging to a good fit in most cases, but a few weaknesses exist. Firstly, as discussed in Sect. 2.3, the procedure only measures the 7 spectral features observed in almost all of the sources and consequently does not fit some rarer features, such as an apparent 14 $\mu$m absorption band observed in source $\rho$ Oph A 2. Further, the profile fitted to each feature (except the CO₂) is not allowed to vary from source to source. While this is broadly valid, a few exceptions led to some poor fits, such as the broadened 6.0 and 6.8 $\mu$m bands in SVS2 (see Fig. 3), or a handful of sources which show broadened silicate features. However, given the low spectral resolution of the CVF

Figure 4: The spectrum of HH100-IR, with the fitted spectrum (solid line) and continuum (dashed line). Note the poor fit to the CO2 feature at 15 $\mu$m as a further example of why the CO2 feature was fitted independently.

and the variation in the spectra over the large number of sources detected, this was found to be the most reliable and consistent method of spectral fitting.

In order to check the instrumental calibration and the validity of this fitting method, comparisons were made with previous observations of the bright, well-studied object HH100-IR (RCrA 5, see Fig. 4). Whittet et al. (1996) find a silicate optical depth of $1.21\pm0.05$, which is somewhat less than the value of $1.35\pm0.05$ measured here: this is almost certainly due to the fact that the iterative fitting procedure used here allows the continuum to drift away from the observed spectrum slightly. As discussed in Sect. 2.4, this is due to absorption across the entire wavelength range and is a benefit made possible by the broad wavelength coverage of the CVF. Whittet et al. (1996) also find a H₂O ice column density of $2.4\times10^{18}$ cm^-2, based on observations of the 3 $\mu$m stretching mode. Adopting a band strength of $A=1.2\times10^{-17}$ cm molecule^-1 (Gerakines et al. 1995), we evaluate the column density N as:

$$N = \frac{\int \tau(\lambda) {\rm d}\lambda}{\lambda_{{\rm peak}}^2 A}$$ (3)

where $\int \tau(\lambda) {\rm d}\lambda$ is the equivalent width. This gives a value of $4.8\times10^{18}$ cm^-2, approximately double the value obtained from the stretching mode. This is consistent with the "6 $\mu$m to 3 $\mu$m paradox'' (Gibb et al. 2000; Dartois & d'Hendecourt 2001), which commonly results in observations of these two bands producing column densities which differ by a factor of $\sim$2. The recent work of Gibb & Whittet (2002) suggests that the excess depth of the 6 $\mu$m feature is the result of blending with a feature due to organic refractory matter. They also find a strong correlation between the excess absorption in the 6 $\mu$m water ice feature and that of the 4.62 $\mu$m "XCN'' feature; unfortunately the wavelength range of the CVF prevented observation of the "XCN'' feature here. Similarly, from observations of the 4.27 $\mu$m stretching mode Nummelin et al. (2001) infer a CO₂ ice column density of $6.2\pm0.6\times10^{17}$ cm^-2, which is broadly consistent with the value of $5.1\times10^{17}$ cm^-2 measured here (assuming a band strength of $A=1.1\times10^{-17}$ cm molecule^-1, Gerakines et al. 1995). Keane et al. (2001) measure optical depths of $\tau_{6.0}=0.23$ and $\tau_{6.8}=0.09$, which are consistent with our results of $0.23\pm0.01$ and $0.13\pm0.01$ respectively. It should be noted, however, that our value of the 8 $\mu$m flux density, $1.05\pm0.02\times10^{-16}$ W cm^-2 $\mu$m^-1, is somewhat less than the value of $\sim$ $1.7\times10^{-16}$ W cm^-2 $\mu$m^-1 measured by Whittet et al. (1996). This may indicate a possible error in the absolute flux calibration of the CVF data. However, the measurement of spectral features depends only on the relative flux calibration from pixel to pixel, which is considered to be accurate throughout.

  
3.2.1 Water ice: Bending and libration modes

The observed equivalent widths of the 6 $\mu$m bending and 13 $\mu$m libration modes of water ice showed no significant correlation at all. Further, the libration mode was not observed to correlate with any of the measured features. Whilst there is evidence for this in the literature (e.g. Bowey et al. 1998) the most likely explanation is that the libration mode is poorly fitted, due to the wide variation in possible profiles. The peak of this band has previously been found at $\sim$11 $\mu$m in crystalline water, $\sim$12.5 $\mu$m in amorphous water and at even longer wavelengths in mixtures with other molecules (Hagen et al. 1983; d'Hendecourt & Allamandola 1986; Cox 1989). Here a single profile was used (that of amorphous ice) and this probably resulted in poor fitting. Unfortunately this section of the spectrum is strongly blended with the silicate feature, so this problem cannot be remedied without making further ad hoc assumptions about the silicate profile(s).

  
3.2.2 The unidentified feature at 6.8 $\mu$m

Figure 5: The strength of the unidentified feature plotted against the bending modes of both water (top) and CO2 (bottom) ices.

As seen in Fig. 5, the strength of the unidentified 6.8 $\mu$m feature is observed to correlate far more strongly with the neighbouring 6 $\mu$m band of water ice than with the 15.2 $\mu$m band of CO₂ ice. This is consistent with previous observations, where the 6 and 6.8 $\mu$m features usually correlate well (see Schutte et al. 1996 and the review by Schutte 1997). This suggests that the carrier of the unidentified feature is probably a strongly polar ice, as its presence matches far more closely the strongly polar H₂O ice than the markedly less polar CO₂. The more recent work of Keane et al. (2001) found the peak position of this feature at different wavelengths towards different sources, and proposed that it consists of two (related) components. Further detailed study of the 6.8 $\mu$m profiles could yield a great deal more information about their chemistry. However, at the low spectral resolution of the CVF the 6.8 $\mu$m profiles are essentially constant across all the sources, and so such an investigation was not possible here. One individual source is also worthy of note: GY262 ($\rho$ Oph E 3) shows a deep 6.8 $\mu$m feature with no corresponding 6.0 $\mu$m feature (the "rogue'' point to the upper left in Fig. 5).

  
3.2.3 11.2 $\mu$m feature

The depth of the measured 11.2 $\mu$m feature was found to correlate strongly with the depth of the silicate feature, with a direct (negative) proportionality providing a good fit to the data. Consequently the 11.2 $\mu$m feature was interpreted to be an emissive shoulder on the silicate feature, rather than an independent feature due to another species: the presence of this feature narrows the silicate absorption profile slightly. The silicate profile has previously been found to vary depending on the composition and structure of the silicate grains (e.g. Demyk et al. 2000), so such a shoulder is not unexpected. However, 3 of the sources (RCrA 1, Ser A 6 and Cha I 3) show a significant 11.2 $\mu$m emission feature without any significant silicate absorption feature: these features may be attributable to emission from crystalline silicates (Bregman et al. 1987; Campins & Ryan 1989).

   
3.2.4 CO₂ ice profiles

As mentioned in Sect. 2.4, the CO₂ ice profile is a very sensitive diagnostic of the ice environment. Whilst the spectral resolution of the CVF was not sufficient to study the profiles in great detail, it is worth noting that a significant long wavelength wing was present in almost all the observed absorption profiles. This would seem to indicate that a large fraction of the CO₂ ice observed exists in a polar (H₂O-rich) phase, which is characterised by this long wavelength wing (Gerakines et al. 1999). However quantifying the relative abundances of the different phases was not possible at this low spectral resolution.

  
3.2.5 Spectral classification and silicate profiles

Figure 6: Examples of observed silicate profiles: emission profiles are generally broader than those seen in absorption and composite emission/absorption profiles are also seen. "Type'' refers to the classification scheme described in Sect. 3.2.5.

If we look at the shapes of the spectra, it appears natural to divide the sources into 3 distinct groups based on the strengths of the spectral features observed. An example of each type is shown in Fig. 6 and the classifications are included in Table 2.

a)

The first group show deep absorption features due to all of the ices and also due to the silicates (typically these have $\tau_{{\rm Si}} \gtrsim 0.6$). These objects are interpreted to be heavily embedded objects, showing strong absorption due to cold foreground material.

b)

The second group shows weaker ice absorption features and weak silicate features ( $-0.8\lesssim\tau_{{\rm Si}}\lesssim0.6$) which are, in general, not especially well fitted. These are interpreted as less heavily embedded objects, showing less foreground absorption. The poor silicate fits are probably due to complications caused by combined silicate emission and absorption.

c)

The third group show strong silicate emission ( $\tau_{{\rm Si}}\lesssim-1.0$) and little or no ice absorption. It should be noted that in the 2 good S/N spectra showing strong silicate emission, the observed emission profiles were significantly broader than both the observed absorption profiles and the profile used to fit the data (see Fig. 6). (However it should also be noted that one of these 2 sources, SVS2, is an unresolved double.) These are probably young stars which have shed most of their circumstellar matter, retaining only a little hot, optically thin silicate dust around them (possibly in a disc). It therefore seems plausible that these three groups might represent an approximate evolutionary sequence, with the YSOs shedding circumstellar material as they evolve.

In order to test this the spectrum of SVS20 (Ser B 6) was re-fitted using a composite silicate profile, consisting of the standard absorption profile used above superimposed on the broader emission profile of SVS2 (Ser B 3 - see Fig. 6). As seen in Fig. 7 this provided a good fit to the observed silicate profile, indicating that both silicate emission and absorption probably do occur, with different profiles seen in absorption and emission. The physical interpretation of this is that these objects have a hot region of optically thin silicates around them but remain embedded in the cloud, which results in foreground absorption due to silicates and ices being superimposed on the "intrinsic'' silicate emission feature in the spectra.

Figure 7: Composite silicate profiles fitted to the spectrum of SVS20. The solid line shows a composite fit with $\tau _{{\rm abs}}=0.59$ and $\tau _{{\rm em}}=-0.33$, the dashed line a composite fit with $\tau _{{\rm abs}}=0.88$ and $\tau _{{\rm em}}=-0.48$ (and a different continuum). The single profile fit reported in the results table, with $\tau _{\rm Si}=-0.33$, is shown as the dotted line.

It should be noted that while it is possible to fit such profiles to these sources, they are not well constrained. As can be seen by comparing the 2 composite profiles in Fig. 7 there is a degeneracy between the strength of this "flat-topped'' silicate profile and the continuum strength, and a great deal of information is assumed about the silicate profiles themselves. Consequently this investigation was not pursued for all the sources. Where $-0.8\lesssim\tau_{{\rm Si}}\lesssim0.6$ the values in Table 2 represent the correct optical depth at 9.7 $\mu$m, but are usually in error towards the wings of the silicate feature. Such values are useful in studies of general trends, but they do not accurately describe the entire silicate feature: a single number cannot fully describe the complicated nature of these features.

This also provides more direct evidence, similar to that of Whittet et al. (1988), that the extensive scatter in measurements of the $A_V/\tau_{{\rm Si}}$ ratio towards YSOs may be caused by complications due to silicate emission. Direct measurements of the silicate optical depth will significantly under-estimate the depth of foreground material in cases (such as SVS20) where optically thin silicate emission is also present. Rieke & Lebofsky (1985) derive $A_V/\tau_{\rm Si} \simeq 17$, whilst noting the problem posed by "intrinsic'' silicate emission, and such emission does indeed lead to errors when evaluating A_V from the silicate depth. As an example, the ISOCAM survey of Bontemps et al. (2001) derives values of $A_V \simeq 28$ for GY252 ($\rho$ Oph E 1) and $\simeq$24 for GY262 ($\rho$ Oph E 3); the X-ray survey of Imanishi et al. (2001) derives slightly lower values of $\sim$24 and $\sim$18 respectively. However GY252, as seen in Fig. 6, clearly shows a "composite'' silicate profile whereas GY262 does not, and the respective silicate optical depths of 0.52 and 1.84 are clearly inconsistent with a single $A_V/\tau_{{\rm Si}}$ ratio. Unless the emitting and absorbing components of such silicate features can be separated unambiguously obtaining a single $A_V/\tau_{{\rm Si}}$ ratio will not be possible.

  
3.2.6 Spectral Index

Of the 41 objects to which spectral indices were assigned 20 were found to be class I objects ( $\alpha_{{\rm cont}}>0$) and 17 to be class II ( $-1.5<\alpha_{{\rm cont}}<0$). In addition to this 3 objects ($\rho$ Oph E 1 and 3, Ser B 2) were found to lie very close to the class I/II border (so-called transition objects: André & Montmerle 1994; Greene et al. 1994). The lack of any class III objects is a selection effect, but it seems to indicate that all of the observed sources are within the clouds, as field stars would show little or no infrared excess and would appear as class III objects in this survey. The distribution of class I and II objects is approximately constant throughout all four regions, and no significant trends involving spectral class were found. Evaluation of mid-IR spectral indices appears to support the finding of Bontemps et al. (2001) that mid-IR spectral indices do not discriminate strongly between class I and class II objects, and a classification scheme similar to that in Sect. 3.2.5 seems preferable when considering spectroscopic data.

Comparing the index-based spectral classes to the spectral feature grouping above, we see that the majority of the class I sources belong to group a) and the majority of the class II sources to group b). However it is notable that a significant number (6) of the 23 group a) objects are either class II or transition objects. This in keeping with the convention that class I objects are the youngest (Lada 1987), but it may be significant that there are two distinct types of spectrum which both result in class II SEDs. Objects showing strong foreground absorption and objects showing little foreground absorption can both present class II continuum SEDs. This could be due to geometry of the YSOs relative to the absorbing clouds, or could be an effect intrinsic to the sources: spectroscopy along a pencil-beam cannot distinguish these. Whilst a larger sample size would constrain this problem further, it does seem that mid-IR observations alone do not provide strong constraints on the "true'' spectral class of YSOs.