A SURVEY OF H₂O, CO₂, AND CO ICE FEATURES TOWARD BACKGROUND STARS AND LOW-MASS YOUNG STELLAR OBJECTS USING AKARI^*

The three most abundant solid phase molecular species in the interstellar medium (ISM) are H₂O, CO₂, and CO. The strongest absorptions of all three species are seen in the 2.5–5 μm region of the spectrum, associated with the stretching mode vibrations of each molecule. AKARI was the first space telescope since the Infrared Space Observatory (ISO) with which it was possible to make spectroscopic observations of the full near-infrared (NIR) range. Compared with Spitzer or ground-based observations, AKARI spectroscopy offers two distinct advantages: simultaneous observation of the stretching mode absorption bands of H₂O, CO₂, and CO, and observation of the blue wing of the H₂O stretching mode absorption. Two further benefits of AKARI are slitless spectroscopy, with spectra simultaneously captured for all objects in the field of view, and exceptional sensitivity, allowing observation of ice spectra on lines of sight toward objects with fluxes approaching 1 mJy.

We present a study of the 2.5–5 μm ice absorption spectra toward background stars and young stellar objects (YSOs) with AKARI. To date, very few spectroscopic ice studies with AKARI have been published (see, e.g., Shimonishi et al. 2008, 2010; Burgdorf et al. 2010; Yamagishi et al. 2011; Aikawa et al. 2012; Sorahana & Yamamura 2012), reflecting the difficulty of reducing AKARI spectral data. In all previous observations, ice spectra were derived from a single relatively bright object in the field of view. To study background star ice spectra, we have extracted data from multiple faint objects in a single observation, requiring the development of a unique data reduction pipeline.

The ultimate aim of our AKARI observing program is to map the spatial distribution of the key ice species H₂O, CO₂, and CO. A prerequisite, however, is a robust data reduction and analysis method to ensure the reliable determination of molecular ice column densities over hundreds of objects concurrently. The first aim of this article is to detail this analysis process. These data enhance the sample of ice spectra toward background stars and low-mass YSOs; our second aim is to test how the chemical processes governing ice formation and evolution are related.

The chemistry of CO, CO₂, and H₂O in molecular clouds is intrinsically linked. Solid H₂O is the dominant interstellar ice species, seemingly ubiquitous in the dense, molecular ISM (Leger et al. 1979; Whittet et al. 1983; Boogert et al. 2011). This H₂O forms on the surface of silicate dust grains (Tielens & Hagen 1982; Boogert & Ehrenfreund 2004; Miyauchi et al. 2008; Ioppolo et al. 2008; Dulieu et al. 2010), most likely through a combination of reactions involving H atoms, O atoms, and OH radicals.

Gas phase CO is highly abundant in molecular clouds and is commonly used to trace astrophysical conditions (e.g., van Kempen et al. 2009). However, unlike H₂O, CO accretes onto icy grain surfaces at a rate proportional to the gas density (e.g., Aikawa et al. 2001; Pontoppidan et al. 2003b), resulting in a CO-rich ice layer. Subsequently, as a function of time, temperature, or other energetic inputs to the ice mantle, the H₂O and CO layers can mix, producing a second CO ice environment, as illustrated by previous experimental (Collings et al. 2003b) and observational (Pontoppidan et al. 2003b) work. Both of these CO ice environments have been detected in interstellar observations (Tielens et al. 1991; Whittet & Duley 1991; Chiar et al. 1995; Pontoppidan et al. 2003b).

CO₂ is believed to form in the solid phase, based on its low gas phase abundances (van Dishoeck et al. 1996). CO₂ ice has been observed toward numerous interstellar environments, including low-mass YSOs (Nummelin et al. 2001; Pontoppidan et al. 2008) and background stars (Knez et al. 2005). Laboratory experiments prove that CO₂ formation is effective via several energetic (Gerakines et al. 1996; Palumbo et al. 1998; Jamieson et al. 2006; Mennella et al. 2004) and purely thermal (Goumans et al. 2008; Oba et al. 2010; Ioppolo et al. 2011; Noble et al. 2011; Zins et al. 2011) mechanisms. Recent results imply that the reaction CO + OH occurs in competition with H₂O formation (via H + OH), and that some CO molecules freeze out from the gas phase long before CO ice is formed. Such postulations can only be tested through concurrent observations of CO, CO₂, and H₂O ices.

Any sample of ice spectra observed to date has been, necessarily, dictated by both the spectral range and sensitivity of IR telescopes. Low-mass YSOs have therefore been rather more extensively studied than background stars, beginning with limited ISO observations (Guertler et al. 1996; de Graauw et al. 1996; Gerakines et al. 1999; Alexander et al. 2003). Spitzer, which is more sensitive than ISO, observed many more low-mass YSOs (e.g., Gibb et al. 2004; Zasowski et al. 2009; Boogert et al. 2008). The Spitzer Legacy program "From Molecular Cores to Planet-Forming Disks" (c2d) extensively detailed the abundances of CO₂ (Pontoppidan et al. 2008), H₂O (Boogert et al. 2008), CH₄ (Öberg et al. 2008), and NH₃ (Bottinelli et al. 2010) ices toward a sample of 41 low-mass YSOs, providing the largest data set of column density values for these key interstellar molecules (Öberg et al. 2011). An important conclusion of this survey is that the abundances of the major ice species are remarkably similar in different star-forming regions, although, for reasons not yet fully determined, the ice composition of minor and volatile species varies extensively. This suggests that ices form in the quiescent ISM, a region probed by ice spectra toward background stars.

Limited background star observations were first made in the 2.4–5 μm wavelength range by ISO; more recently, Spitzer has observed many more background stars, but not in the 2.5–5 μm range covering the stretching mode absorptions of H₂O, CO₂, and CO. Observing ices in this region is vital in benchmarking the inventory of ice species present at the earliest stages of star formation. Two background stars were observed by ISO: Elias 13 was observed in the range of 4.20–4.34 μm (focusing on the CO₂ stretching mode at 4.25 μm; Nummelin et al. 2001) while Elias 16 was observed in the range of 2.4–5 μm (Whittet et al. 1998; Gibb et al. 2004) and 14–19 μm, with additional detailed studies of the CO₂ stretching mode (Whittet et al. 1998; Gerakines et al. 1999; Nummelin et al. 2001), ¹³CO₂ (Boogert et al. 2000), and potentially the N₂ stretching mode (Sandford et al. 2001). With Spitzer, mid-IR observations were made of the same two background stars (Elias 13 (5–20 μm) and Elias 16 (5–14 μm)) and a third background object (CK 2 (5–20 μm); Knez et al. 2005). The CO₂ bending mode region around 15.2 μm was investigated for a further 12 background stars across Taurus, Serpens, and IC 5146 (Bergin et al. 2005; Whittet et al. 2007, 2009). CO₂ was found to exist in H₂O-poor ices in quiescent regions for the first time (Bergin et al. 2005), pointing to a formation mechanism involving CO ice. Successful ground-based studies have been carried out to observe the 4.67 μm CO stretch toward 12 background stars (Whittet et al. 1985, 1989; Eiroa & Hodapp 1989) and the observable part of the 3 μm water band toward 73 background stars (Whittet et al. 1988; Eiroa & Hodapp 1989; Murakawa et al. 2000). By combining ground-based and Spitzer observations, Boogert et al. (2011) quantified the ice abundances of H₂O, NH$_4^+$, CH₃OH, HCOOH, CH₄, and NH₃ toward 31 background stars (also including CO₂ abundances toward eight objects) and confirmed that little difference in the abundances of major ice species is observed between background stars and YSOs.

Most previous background star studies have been made toward Taurus and Serpens molecular clouds, with the exception of two studies of the Cocoon Nebula (IC 5146; Whittet et al. 2009; Chiar et al. 2011), one of ρ Ophiuchi (Tanaka et al. 1990), and the survey by Boogert et al. (2011), which observed 31 background stars toward 16 isolated dense cores. Our study adds 22 background stars across a further 15 dense molecular cores to these statistics. Boogert et al. (2011) highlight that the existing studies of background stars focus on cores and cloud regions well separated from lines of sight where YSOs have been studied. To understand the subtle changes in ice chemistry as a region evolves throughout the star formation process would require comparisons between ice spectra of these different evolutionary objects toward the same core. In three of our cores, it has been possible to compare ice column densities from background and YSO sources separated by only a few 1000 AU.

In this paper, the column densities of H₂O, CO₂, and CO are calculated from the 2.5–5 μm AKARI spectra of 22 background stars and 8 low-mass YSOs. As the only complete spectrum of a background star in this spectral region to date is that of Elias 16 taken by ISO, this represents a significant increase in data on background stars in the NIR. In addition to H₂O, CO₂, and CO, it was possible in certain cases to calculate the column densities of OCN⁻ and CO gas grain (Fraser et al. 2005). This large increase in data, including low flux background stars, offers much evidence about the initial conditions in quiescent molecular clouds before star formation begins. In particular, it offers clues to where ices form, and how the chemistry of different molecular species is linked in prestellar regions.

Observations were carried out between 2006 December and 2007 August using the spectroscopic mode of the Infrared Camera on AKARI (Onaka et al. 2007; Ohyama et al. 2007). We observed fields of view toward 19 dense molecular clouds. Of these, 11 were part of the AKARI European Users Open Time Observing Programme IMAPE, a selection identified from the c2d Spitzer Legacy program (Evans et al. 2003); a further 8 fields of view are included here from the Japan/Korea Users Open Time Observing Programme ISICE. Details of the 19 cores are given in Table 1.

In the IMAPE program, clouds were selected by identifying those containing dense prestellar cores, a statistically large number of identifiable field stars behind the cloud, and with the proviso that no previously identified YSO candidates were in the line of sight. Two pointings were made toward each cloud, affording an improvement in the spectral signal to noise, and allowing for better removal of hot pixel and cosmic ray effects from the data. Details of the cloud selections in the ISICE program, for which only single pointings were made of each cloud, are given in Aikawa et al. (2012).

The observations presented here focus specifically on data obtained with the IRC04 AKARI astronomical observing template (AOT) using the NIR grism disperser (NG). The AOT produced eight spectral frames, an imaging frame, and a pre- and post-dark frame per field of view per pointing. The spectra were extracted from the Np aperture, a 1' × 1' field of view with a spectral range of 2.5–5.0 μm (R = 120 at 3.6 μm). Imaging frames for 16 of the clouds are shown in Figure 1. The images were taken with the N3 filter (2.7–3.8 μm) at a reference wavelength of 3.2 μm. As is clear from Figure 1, some fields of view contain multiple objects, and as the AKARI spectroscopic mode is slitless, the spectral traces of these objects often overlap. Nevertheless, 30 individual objects were identified across these fields of view (highlighted in Figure 1) for which spectra were successfully extracted using our data reduction pipeline. The selection and classification of these individual objects are discussed later in Section 3.2.1. As far as we can ascertain from the literature, ice spectra have not previously been published for any of the objects in the lines of sight observed here. Three clouds listed in Table 1 do not feature again in this paper. The field of view in DC 300.2−03.5 was too confused to enable us to extract any spectra, even though ice features may have been present; data from L 1082A and DC 269.4+03.0 were below the lower flux limit at which our pipeline functioned effectively, as discussed in Section 3.2.1.

Zoom In Zoom Out Reset image size

3.1. AKARI Reduction Facility

The AKARI 1' × 1' aperture was designed for point-source spectroscopy of single sources, and the data reduction toolkit written by the IRC team (Ohyama et al. 2007, hereafter referred to as "the toolkit") was intended to analyze such sources. In the majority of the observations in this work, there was more than one object in the 1' × 1' aperture; the dispersed spectra of these objects overlapped and produced confused spectra when reduced using the toolkit. A purposely written data reduction pipeline—the AKARI Reduction Facility, hereafter referred to as "ARF"—was therefore developed by the authors to analyze AKARI NIR spectroscopic data. ARF was written in IDL, incorporating procedures taken from the IDL Astronomy User's Library (Landsman 1993) and the Markwardt Library (Markwardt 2009). Full details of ARF are presented in Noble (2011).

The basic function of ARF is illustrated in Figure 2. The eight raw spectral data frames (Figure 2(a)) are dark-subtracted (using the pre-dark frame) and corrected for bad and hot pixels before being shifted and stacked. As with previously published AKARI spectroscopic studies, no flat-fielding of the spectral frames is performed, as the calibration data provided in the toolkit did not include flat fields for this region of the detector. The shift between frames is calculated by cross-correlation, for the direction perpendicular to dispersion only (X). Frames are shifted by non-integer values, as appropriate, with linear interpolation of flux to maintain absolute flux values upon stacking. Background correction is performed on the stacked frame, using a region of dark sky picked manually from the spectral aperture (Figure 2(b)). Objects in the field of view are identified according to their coordinates, taking account of the distortion in the frame. The spectrum of each object is then extracted by fitting the grism point-spread function (PSF) to the dispersed flux (Figure 2(c)). Extracted data are divided by the spectral response function (SRF) of the NIR detector to produce spectra on a flux–wavelength scale (Figure 2(d)). In general, this is a standard reduction technique, the main exception being the extraction method, described in detail below.

Zoom In Zoom Out Reset image size

3.2. Extraction of Spectra

3.2.1. Target Object Specification

Objects for extraction are introduced to ARF from target tables. Pointings that were part of the IMAPE program are a subset of the c2d Spitzer Legacy program and had target tables available from that program (Evans et al. 2003), including photometric data points from the Two Micron All Sky Survey (2MASS) and Spitzer InfraRed Array Camera (IRAC). In ARF, the c2d target lists are edited to exclude objects outside the 1' × 1' aperture and those below a threshold 2MASS K_S-band (centered at 2.2 μm) magnitude of 1.19 mJy, the magnitude of the brightest isolated object (in DC 269.4+03.0) which could not be fitted with the PSF using ARF (discussed in Section 3.2.3). Target tables were prepared by hand, from the 2MASS and Spitzer project databases, for pointings in the ISICE program.

3.2.2. Dispersion Modeling

Individual spectra are extracted by transforming the source positions from the target files to the dispersed frames, accounting for (1) the distortion in the telescope optics and (2) the displacement of the first-order dispersion relative to the direct light position by approximately 60 pixels in the dispersion direction. The dispersion is subsequently modeled using the linear relation of 0.0097 μm pixel⁻¹ with a vertical (non-dispersion, X, direction) correction of 0.00746929 pixel row⁻¹ to take into account the deviation of the peak position of the flux (as can be seen by inspection of Figure 2(a)).

3.2.3. Point-spread Function

The PSF of the AKARI NG disperser was calculated by fitting various combinations of Gaussian functions or Lorentzian functions to a bright point source observed in AOT04 mode (ObsID: 5020032_001, standard star KF09T1; see Sakon et al. 2012 for further details). The best-fitting combination was a double Gaussian function:

$$\begin{eqnarray} &&f(X) =[ A_{1}\,\exp \lbrace -(\frac{X-P_{1}}{\sigma _1})^2 \rbrace\nonumber \\ &&+0.8 A_{1}\,\exp \lbrace -(\frac{X-(P_{1}+2.7)}{\sigma _2})^2\ \rbrace], \end{eqnarray} \tag{ 1 }$$

where X is X pixel, A₁ is the peak height, P₁ is the peak position of the first Gaussian function, and σ_n is the FWHM of Gaussian functions (n = 1, 2).

3.2.4. Running the Extractor

For rows containing more than one target, a linear sum of PSFs is fitted when any targets are within five pixels of each other (in X). This is the key strength of the reduction method presented here, as this approach allows the separation of spectra that are overlapping in X, which would otherwise be discarded due to confusion. For each object in each row, the flux (in ADU) of that object at that point in its dispersion is then extracted by integrating the calculated PSF. In this way, the spectrum of each object is extracted systematically, row by row.

3.3. Flux and Wavelength Calibration

3.4. Error Propagation

The flux error in each pixel is ±4 ADU (Y. Ohyama 2007, private communication). Thus the minimum error in each stacked pixel, _min, is calculated by

$$\begin{equation} \epsilon _{{\rm min}} = \sqrt{n \sqrt{2 \times 4^2}}, \end{equation} \tag{ 2 }$$

where n is the number of frames stacked, and the factor of two accounts for a contribution from subtracting the dark from each frame prior to stacking. During stacking, the standard deviation in the flux, _sd, is calculated. These errors are combined to calculate _ADU, the overall error in flux, used to weight the fit of the PSF during extraction. The error in the extracted flux, _ext, is the combination of the data error _ADU with the standard deviation of the fit. During division by the SRF, the final error on each data point, , is calculated by combination of the extracted flux error with the error on the SRF.

In total, across the 19 clouds there were 94 target objects with some 2MASS K_S-band magnitude. Twenty-seven of these objects fell below the signal-to-noise threshold of ARF, defined as K_S = 1.19 mJy, the magnitude of the brightest isolated object for which the PSF fitting failed to converge (including all objects in DC 269.4+03.0 and L 1082A.) The confusion limit for galactic cirrus and point sources is <1 μJy in N3, the 3 μm imaging filter (ASTRO-F User Support Team 2005). Four objects were only partially observed due to their position at the edge of the field of view, and were thus discounted. Thirty-three objects were too confused to extract any spectral data because of their low flux and overlap with nearby objects, including all those in DC 300.2−03.5. The objects that were successfully extracted using ARF are detailed in Table 2; those derived from only a single spectrum are denoted by a dagger symbol (†). The spectra of the 30 extracted objects are presented in Figure 3. The final spectra are shown in black, with the error on each point overplotted in gray.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Spectra were extracted across the full 2.5–5 μm region, except in cases where the dispersion reached the edge of the detector or where the flux from two neighboring objects could not be resolved. For example, due to detector limitations, in Object 9 the region of the spectrum beyond 4 μm was not obtained, while for a further four objects (Objects 1, 21, 24, and 27), the higher wavelength region of the spectrum, beyond 4.5 μm, was not extracted. Where unresolved overlap occurred between objects, it was possible to extract a partial spectrum, or a spectrum with some extra flux contribution, from which upper limits could be obtained on the molecular species present in the line of sight (Objects 6, 7, 8, 9, 11, 15, 16, 17, 18, 19, and 20). These objects are indicated by an arrow in Table 3 and are designated as upper limits in all figures (↓).

The H₂O band (3.0 μm) was fully extracted for all objects, including the blue wing, which is not observable from the ground. In many spectra, such as Objects 2 and 4, it is clear that the H₂O band is saturated. The error values calculated in the pipeline are the errors on the observed flux, not the errors on the absolute value of the spectral data. Where we suspect bands are saturated, we have not included any additional uncertainty on these data points. In at least 12 objects the water band is still clearly visible even though the flux is < 5 mJy (see Figure 3, Objects 1, 5, 6, 9, 21, 22, 24, 25, 26, 27, 28, and 30). The absorption bands of CO₂ (4.25 μm) and CO (4.67 μm) are also clearly visible for the majority of objects.

On each spectrum in Figure 3, the IRAC photometric data points are plotted as diamonds at 3.6 and 4.5 μm. For most objects the photometric data match the extracted flux very well. For some of the upper limits mentioned above, notably Objects 9, 11, and 16, the IRAC photometry does not exactly match the flux of the object, suggesting that there was excess light on the detector in the vicinity of the spectrum, vindicating the decision to only quote upper limits to the molecular column densities in these objects. In all cases where there is a slight photometry mismatch, the flux of the object has been overestimated, rather than underestimated, supporting the conjecture that there is excess light in these fields of view. The only two examples of completely isolated objects in the data set are Objects 27 and 30. The photometric data points for these two objects match the extracted flux within the errors.

4.1. Object Classification

For further spectral analysis, it was necessary to classify each object according to whether it was a field star behind the molecular cloud ("star") or a YSO embedded in the cloud ("YSO"). All objects in the IMAPE program had a classification of this type preassigned by the c2d team, based on the full Spitzer photometric data as well as 2MASS photometry. The classification of objects in the ISICE program was based on 2MASS photometric data. If objects satisfied

$$\begin{equation} 1.75(H-K_S)-0.04 \le J-H, \end{equation} \tag{ 3 }$$

they were classed as background stars; if not, they were classed as YSOs (Itoh et al. 1996; Shenoy et al. 2008). Henceforth, irrespective of the observational program from which they are derived, the objects in this paper are designated by a star symbol (⋆) for YSOs, or a circle (•) for background stars; in the text they are referred to as "YSO" and "star," respectively.

4.2. Baseline Fitting

To convert the observational data to spectra on an optical depth scale (most relevant for molecular abundance analysis), two different methods of baseline fitting were employed. As is typical in ice spectra extraction (Gibb et al. 2004; Öberg et al. 2011; Aikawa et al. 2012), this was either a second-order polynomial (in the case of YSO objects) or a NextGen model (Hauschildt et al. 1999; in the case of background stars), with the proviso that second-order polynomials were also fitted to "upper limit" spectra, i.e., where the data were not fully extractable. These baselines are shown overplotted on each individual spectrum in Figure 3, as either a (green) dotted line (NextGen models, background stars), a (blue) dashed line (second-order polynomials, YSOs), or a (red) dash-dotted line (second-order polynomials, "upper limit" spectra).

The NextGen models were fitted to the background stars based on photometric data from 2MASS, IRAC, and MIPS observations. Using the extinction law of Weingartner & Draine (2001), objects were dereddened along a reddening vector of 2.078 on a (J − H) versus (H − K_S) color–color diagram to determine their spectral type and A_V (tabulated in Table 3). Potential spectral types were defined based on where the dereddened object crossed the main sequence, giant, or dwarf branches. This approach accounted well for the reddening in the 2MASS wavelength region of the spectra, but at longer wavelengths the model baselines were further reddened by a polynomial.

The advantage of fitting baselines with NextGen models is that the models account for emission and absorption in the stellar atmosphere. However, these models cannot be used to fit YSOs as the NextGen models only include main sequence and giant stars. The continua of YSOs (and all upper limits on background stars and YSOs) were therefore fitted by second-order polynomials. Fitting regions varied slightly between the spectra based on the position of bad pixels, but were generally <2.65 μm, 3.85–4.18 μm, and >4.75 μm. The fits were weighted in favor of the blue end of the spectrum, as this region had a higher signal-to-noise ratio than the red end.

Spectra were converted to optical depth scale by

$$\begin{equation} \tau =-\ln (F/F_{{\rm con}}), \end{equation} \tag{ 4 }$$

where τ is optical depth, F is the spectral flux, and F_con is the flux of the continuum (fitted in Section 4.2). The optical depth scale spectral features are shown in Figure 4 for the 3 μm H₂O band, Figure 5 for the 4.67 μm CO band, and Figure 6 for the 4.25 μm CO₂ band.

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

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The column densities of molecular species can be determined by integrating the area under their absorption bands on an optical depth scale and correcting by the optical constant for each species, using

$$\begin{equation} N=\int \frac{\tau d\nu }{A}, \end{equation} \tag{ 5 }$$

where N is the column density in molecules cm⁻², ν is the wavenumber in cm⁻¹, and A is the band strength in cm molecule⁻¹ (Gerakines et al. 1995). The specific analysis methods for H₂O, CO₂, and CO are described in detail below. All calculated column density values are presented in Table 3.

The absorption bands of interstellar ice species are altered, with respect to those measured in the laboratory, by interaction of the incident light with the dust grains on which the absorbing ices are present. This is particularly important in the NIR where the average grain radii and absorption wavelengths are similar. Therefore for an abundant ice species, if a laboratory spectrum is to be used to derive the column density of the ice by fitting it to the observational data, a grain shape correction must be applied. In this work, a continuous distribution of ellipsoids (CDE) model was employed. CDE is a reliable model of fractal aggregates (the likely structure of interstellar grains) and it has been successfully applied to many astronomical spectra (Tielens et al. 1991; Ehrenfreund et al. 1997; Pontoppidan et al. 2003b).

Due to the asymmetric shape of the PSF of AKARI IRC data (described above in Section 3.2.3), the instrumental line profile was determined by analysis of an observation of the standard star KF09T1 in the imaging mode (IRC03). The imaging frame N3 (3.2 μm; see Onaka et al. 2007) was chosen, and the central four pixels in the X-direction were summed to produce the profile of a PSF along the Y-direction, which was assumed to be the instrumental line profile.

5.1. H₂O

H₂O was detected in all 30 extracted spectra, as shown in Figure 4. These spectra include a broad, smooth band extending from ∼3600 to 2600 cm⁻¹ with a steep blue wing and, in almost all cases, an extended red wing. In order to fit this H₂O absorption feature at ∼3 μm (3333 cm⁻¹), a CDE-corrected spectrum of H₂O was generated from a laboratory spectrum of pure H₂O at 15 K (Fraser & van Dishoeck 2004), then corrected for the instrumental line profile and resolution of AKARI. This spectrum was fitted exclusively to the blue wing of the observed H₂O data, from approximately 3600 to 3400 cm⁻¹. Unlike the red wing, which is discussed in detail below, this region is unaffected by grain interaction effects, and is accessible here for the first time since ISO. The blue wing is also unaffected by saturation. As can be seen in Figure 4, the pure H₂O spectrum fitted the H₂O bands well in all cases.

Crystalline H₂O has a characteristic peak shape and position, and has been observed in the circumstellar disks around YSOs, where material is subject to intense irradiation from the newly formed star (Malfait et al. 1998; Honda et al. 2009). For all objects, test fits were made to examine the possibility of a crystalline H₂O component toward each line of sight, but no crystalline ice was found. A single, amorphous H₂O lab spectrum at 15 K was used to fit every H₂O absorption feature in the 30 AKARI spectra. Column densities of H₂O were calculated from the fitted laboratory spectrum using Equation (5), with A = 2 × 10⁻¹⁶ cm molecule⁻¹ (Gerakines et al. 1995), and are presented in Table 3.

Excluding those spectra yielding only upper limits, the H₂O profiles of all objects can be divided into three types, as illustrated in Figure 7. In this figure, all spectra are normalized to the maximum of the fitted H₂O band at 3300 cm⁻¹. The three distinct H₂O profiles identified were: Type 1, in which the red wing is weak compared to the absorption peak at 3 μm; Type 2, in which the long wavelength feature is relatively strong, i.e., a separate shoulder; and Type 3, containing no (or very little) red contribution at all. Similar effects were noted by Thi et al. (2006), who found two H₂O band profiles in the spectra of intermediate-mass YSOs in Vela: one where the extended red wing was weak relative to the H₂O ice band, and a second where it was stronger. As in Thi et al. (2006), within each subset of objects the H₂O bands show remarkably similar profiles. The question is, what are the likely origins of these profile differences? Do these profiles represent distinct physical environments, or have we, rather, sampled a continuum of red wing profiles?

Zoom In Zoom Out Reset image size

Type 3 is seen only in the spectra of Objects 10 and 12, both of which are background stars and have low spectral flux (< 12 mJy at all wavelengths observed). No previous observations exist of H₂O bands without an extended red wing emission. For these bands, a CDE-corrected pure H₂O spectrum appears to fully fit the observed data. Type 2—present in the spectra of Objects 1, 13, 21, 22, 25, and 30, a sample of only background stars—appears to contain a distinct red shoulder, as opposed to the usual extended red wing. Type 1—present in the spectra of Objects 2, 3, 4, 5, 14, 23, 24, 26, 27, 28, and 29, a mixture of YSOs and background stars—resembles the "standard" water band observed toward many lines of sight (Gibb et al. 2004; Boogert et al. 2011). The origins of the extended long wavelength feature on the water ice band have often been debated. A number of suggestions have been made as to its origin, including the interaction of light with dust grains (as discussed in Section 5), with the depth and profile of the band changing with grain size, shape, or even as a function of ice mantle thickness (Smith et al. 1989); an absorption band at 3.54 μm, attributed to the C–H stretch of methanol, CH₃OH (Hagen et al. 1980; Dartois et al. 2003); and aliphatic and aromatic C–H stretches associated with various molecular sources, including polyaromatic hydrocarbons (Onaka et al. 2011), larger organic molecules (Brooke et al. 1999), and the grains themselves (Jones 2012).

Thi et al. (2006) suggest that, although their sample is statistically limited, a strong correlation is evident between the water ice column density and the optical depth of the extended red wing at 3.25 μm, concluding that the carriers of the red wing are directly related to the ice mantle. A previous study by Smith et al. (1993) also finds a strong correlation between A_V and the optical depth of the extended red wing at 3.47 μm. Neither of these correlations is evident in our data. However, in comparing the average H₂O ice column densities across Types 1, 2, and 3, a clear trend emerges: the column density of H₂O ice in Type 3 sources is ∼2.6 times lower than Type 1; Type 2 is ∼1.5 times lower than Type 1. Given that Types 3 and 2 only contain background stars whereas Type 1 also includes YSOs it appears that the profile and depth of the extended red wing could potentially prove to be a reliable evolutionary marker. Within Type 1, while the band profiles are identical, the average H₂O ice column densities differ by a factor of 1.7 between the subset of YSO objects and the subset of background stars. In comparison, the ratio of average column densities between the YSO subset in Type 1 and the Type 2 background stars is 1.9. The objects of Thi et al. (2006) were all class I YSOs; the average H₂O column densities associated with their Type 1 profile (i.e., the "standard" water band) is identical to ours (2.5 ± 0.9 × 10¹⁸ and 2.1 ± 0.5 × 10¹⁸ molecules cm⁻², respectively). This provides statistical evidence suggesting that the H₂O band profile also changes with the column density of ice in a particular line of sight.

It has previously been observed that the water ice abundance (H₂O/H₂) in molecular cores increases with dust density at low A_V (Whittet et al. 1988), before reaching a plateau (Pontoppidan et al. 2004). Some evidence exists of a further jump in water ice abundance at densities of a few × 10⁵ cm⁻³ H₂, which has been attributed to the removal of one or more water destruction pathways (Pontoppidan et al. 2004). From our calculated column densities, it is likely that the YSOs observed in our survey probe lines of sight with densities near the plateau, i.e., regions where grain growth and accelerated freezeout have occurred, whereas the lines of sight toward background stars probe the density-abundance slope at lower densities, similar to previous observations (e.g., Whittet et al. 1988). On the evolutionary path from background stars (probing low density molecular clouds or prestellar cores) to class I low-mass YSOs, it appears that both the grain radii and the ice mantle thickness increase. However, previous attempts to fully account for the extended red wing with this postulate have failed (e.g., Smith et al. 1989; Thi et al. 2006). Observational evidence from other ice species also points toward fundamental differences in the dust present in diffuse and dense ISM regions (Boogert et al. 2011).

As mentioned above, the H₂O band of Type 3 objects is fully fitted with a CDE-corrected H₂O ice profile, suggesting that only H₂O contributes to this absorption feature. The column densities of H₂O toward these lines of sight are relatively low, with values of 0.7 × 10¹⁸ and 0.9 × 10¹⁸ molecules cm⁻² for Objects 10 and 12, respectively. In Type 2 objects, the main H₂O band centered at ∼3 μm is equally well accounted for by a CDE model as the H₂O observed toward Type 3 objects, but the column densities of H₂O are, on average, 1.8 times greater. The water ice band does not have an extended red wing, but rather a distinct shoulder redward of the H₂O band, attributed to additional molecular species present in the ice. Type 2 objects thus likely probe lines of sight toward more evolved regions of molecular cores where surface chemistry has resulted in the formation of species such as CH₃OH, or the hydrogenation of carbon on the grain surface, producing C–H bonds that absorb in this wavelength region. One further alternative not widely discussed in the literature is that on such lines of sight the nitrogen chemistry has evolved, forming NH₃ or at least NH₂ functional groups. The stretching modes of these transitions would appear close to the H₂O absorption (e.g., N–H stretching mode of NH₃ at 2.96 μm and amides in the region 2.70–3.30 μm). However, for NH₃ (at least), such transitions would also be accompanied by features associated with the bending and umbrella modes of NH₃ at 6.16 and 9.00 μm, respectively (Bottinelli et al. 2010); these wavelengths lie outside the NIR region probed by these AKARI observations, so no further analysis is possible here. Finally, Type 1 objects display the "standard" H₂O band profile with an extended red wing; the distinct shoulder seen in Type 2 is still likely to be present (and thus the additional carriers and molecular species remain) but is now subsumed by effects in the broad red wing of the OH stretching absorption band. The average H₂O ice column density on these lines of sight is greater than those of Types 3 and 2. However, as the fits show, this increase in column density alone cannot account for the changes in the red wing profile between Types 1, 2, and 3. Some change in the scattering properties of the ice must also occur, e.g., grain growth and mantle growth occur concurrently.

The profile of the red wing on lines of sight toward YSOs is usually attributed to grain growth, accompanied by accelerated freezeout, which is known to occur in dense cores and regions surrounding YSOs, where the density is high enough for the freezeout timescale to become shorter than the age of the core (e.g., Kandori et al. 2003; Román-Zúñiga et al. 2007; Chapman & Mundy 2009). It is very interesting, therefore, that Type 1 includes both YSOs and background stars, although the reasons for this have not yet been fully ascertained. One possibility is that background stars exhibiting this H₂O band profile probe lines of sight through regions of a core that are close to an existing YSO, and are thus probing ices and grains that are at a later evolutionary stage than those probed by background stars exhibiting Type 2 profiles. However, from our data this does not seem to be the case. Most of the background star subset of Type 1 are either isolated objects or lie in fields of view containing only background stars. In the only fields of view containing a YSO and a background star (L 1165, Objects 28 (YSO) and 29, and LM226, Objects 26 (YSO) and 25), the pattern is unclear, as Object 29 is of Type 1, but Object 25 is of Type 2. Perhaps, rather, the background stars in Type 1 also probe regions that are more evolved and are about to undergo star formation, suggesting that proximity to a YSO is not the only observational indicator of evolutionary state.

A number of studies have tried to reproduce the extended red wing profile of the H₂O band using Mie scattering with an encapsulated grain, changing grain size, and mantle thickness (Smith et al. 1989; Thi et al. 2006; Dartois et al. 2002). Given that our Type 3 and 2 profiles do fit with a CDE grain shape model, we opted to continue using the CDE fitting with the Type 1 profile. Perhaps, a scattering model that took into account the evolution of the ice mantle as well as grain growth would be able to more accurately model the changing profile of H₂O bands during the evolution from a molecular cloud to a YSO.

The most obvious and widely published candidate molecule to account for the H₂O band shoulders in Types 2 and 1 is CH₃OH. It has previously been observed toward multiple lines of sight including both low-mass YSOs and background stars (Brooke et al. 1999; Pontoppidan et al. 2003a; Boogert et al. 2011). As the focus of this article is the analysis and quantification of the H₂O, CO₂, and CO ice absorption features, further analysis of the H₂O red wing was not performed. This issue will be addressed in a future article (A. Suutarinen et al., in preparation).

5.2. CO

The original aim of these observations with AKARI was to determine the relative column densities of H₂O, CO₂, and CO ices, with a view to mapping ices across molecular clouds. Given the low spectral resolution of AKARI, it was not anticipated that the absorption features would be fully resolved, and thus a component fitting method was not envisioned. However, the profiles of the CO bands observed are well enough resolved that fitting components is viable.

The fitting methodology for CO was developed by consideration of various possible methods, the most crucial of which are illustrated in Figure 8. The primary CO feature (2160–2120 cm⁻¹) contains only 9 or 10 data points, while the AKARI instrumental profile contains 10 data points. These data are thus below the Nyquist limit for data sampling, and we expect the CO feature to be undersampled. A single component fit of pure CO to the primary CO feature using a lab spectrum (with an applied CDE model, AKARI instrumental profile, and resolution correction; Figure 8(a)) was discounted as the wings were poorly fitted. A two-component fit of pure CO and CO in a water environment (middle component, CO_mc, and red component, CO_rc; Pontoppidan et al. 2003b), similar to that first described by Tielens et al. (1991), well described the main CO feature, especially the wings (Figure 8(b)), with the reduced chi-squared statistic improving for 74% of spectra. However, it did not account for the 2160–2180 cm⁻¹ feature, nor the slight excess in the red wing of the CO feature. Previous AKARI observations showed that gas phase CO rovibrational lines might also be present concurrently with the CO ice feature (Aikawa et al. 2012) at least on lines of sight toward the edge-on disks they observed. At the AKARI resolution, these gas phase lines are not resolved, but the possibility of major gas phase CO features in absorption was considered, and rejected, based on a series of model gas phase spectra, one of which is shown in Figure 8(c). Clearly the gas phase features peak where the ice features do not, even at the AKARI resolution, and they cannot account for the extended red wing on the AKARI CO ice features. We conclude that there is very little, if any, gas phase CO contribution in our spectra, and do not include a gas phase component in our fit. The feature between 2160 and 2180 cm⁻¹, which is usually attributed to OCN⁻ (e.g., Pontoppidan et al. 2003b), is included in the overall fit for completeness, although the minor components contributing to this feature are not discussed further in this work. In our spectra, the feature can usually only be fully accounted for by including both an OCN⁻ component and the so-called CO_gg, CO gas grain, feature (Fraser et al. 2005), as illustrated by van Broekhuizen et al. (2005). CO_gg represents gas phase CO that chemisorbs directly to the bare grain surface during or before formation of the underlying H₂O ice mantle (Fraser et al. 2005). As can be seen from Figure 8(d), the OCN⁻ feature alone is not able to fully complete the fit and so, as illustrated in Figure 8(e), we use both components here. The inclusion of the secondary feature (2160–2180 cm⁻¹) to the fit makes no difference to the calculated CO column densities for all spectra from which CO values were calculated (nor for 80% of upper limit values).

Zoom In Zoom Out Reset image size

Ultimately, our CO ice band fitting was based on a modified version of the well-established four-component analysis of Pontoppidan et al. (2003b). In our analysis, CO_rc is replaced by a Gaussian calculated from laboratory data (Cottin et al. 2003), but CO_mc is retained as in Pontoppidan et al. (2003b; a CDE-corrected Lorentzian at 2139 cm⁻¹). The third CO component from Pontoppidan et al. (2003b) is omitted, not because it is not necessarily present, but because, given the AKARI resolution, there is no need to add additional components to a fit where two components suffice. That only two components are required to fit the main CO feature (CO_mc and CO_rc) supports our assertions that ARF, and in particular the stacking method, was appropriate for analysis of these data. The 2160–2180 cm⁻¹ feature (the fourth component in the Pontoppidan et al. 2003b method) was fitted by a Gaussian derived from laboratory OCN⁻ spectra by van Broekhuizen et al. (2005). In addition, a laboratory spectrum of CO gas chemisorbed on a zeolite surface (CO_gg; Fraser et al. 2005) was added to the model, to fit the outlying absorption feature centered at 2175 cm⁻¹. CO_mc is CDE-corrected to account for grain shape effects, and the sum of all components is corrected for the AKARI instrumental line profile. One key element of the multi-component fitting in Pontoppidan et al. (2003b) is that all the components are used to fit all the observed spectra, changing only the relative intensities of each component from one line of sight to another. This is the premise we also adopt here.

The fits of CO are presented in Figure 5. Where there are no data in a plot, it is because no data were extracted in this region of the spectrum (e.g., Object 1). Where there is only a partial data extraction, the plot is shown but no fit was made to the data (e.g., Object 11). Interestingly, it is very clear from Figure 5 that in at least two objects—Objects 4 and 20—the CO ice main peak is fully described by a single pure CO component (CO_mc). In fact, as Table 3 shows, up to nine objects fit into this category. Conversely, in Objects 8 and 18, the CO ice is fully fitted by only the CO_rc. However both these objects provide only upper limits on the CO ice column densities, and the identification of a single CO ice component is much less compelling than for Objects 4 and 20. Given that CO ice is formed in low temperature cores, where CO gas critically freezes out as densities increase, it is perhaps not surprising to observe objects with only pure CO ice present. As the temperature of a core increases (usually due to the formation of a YSO), this pure CO ice is expected to either desorb or migrate into the pores of the underlying water ice, giving rise to the CO_rc (Collings et al. 2003b). In the laboratory it is only possible to detect the spectral feature associated with the CO_rc at elevated temperatures (Collings et al. 2003a), but it seems unlikely that the lines of sight probed here are particularly warm. The origins of the CO_rc therefore remain debatable, as illustrated recently by Cuppen et al. (2011).

The column density of CO for the pure component, CO_mc, is calculated by

$$\begin{equation} N({\rm CO}_{mc}) = 6.03\ {\rm cm}^{-1} \times \tau _{{\rm max},mc} \times A^{-1}_{{\rm bulk}}, \end{equation} \tag{ 6 }$$

where A_bulk = 1.1 × 10⁻¹⁷ cm molecule⁻¹ is the band strength of the bulk material, and τ_{max, mc} is the optical depth at maximum absorption (Equation (3), Pontoppidan et al. 2003b). N(CO_rc) and N(OCN⁻) are calculated using Equation (5), with band strengths A_rc = 1.1 × 10⁻¹⁷ cm molecule⁻¹ (Gerakines et al. 1995) and $A_{{\rm OCN}^-}$ = 1.3 × 10⁻¹⁶ cm molecule⁻¹ (van Broekhuizen et al. 2004). The column density of the CO gas-grain population is calculated by

$$\begin{equation} N({\rm CO}_{gg}) = 1.07\ {\rm cm^{-1}} \times \tau _{{\rm max},gg} \times A^{-1}_{{\rm bulk}}, \end{equation} \tag{ 7 }$$

where A_bulk = 4.0 × 10⁻¹⁹ cm molecule⁻¹ is the band strength of the bulk material (Fraser et al. 2005), and τ_{max, gg} is the optical depth at maximum absorption. All column densities calculated for CO_rc, CO_mc, OCN⁻, and CO_gg are presented in Table 3. It should be noted that A_bulk is calculated based upon laboratory spectra of a particular ice, using estimations for the parameters of ice thickness and number of molecules on the surface. If astrophysical ices differ in mixture or concentration, the relationship between optical depth and molecular abundance will change, and thus the relative concentration of one component to another will differ with respect to the values calculated here. This caveat holds for any interstellar ice column density analysis.

OCN⁻ was believed to be formed around YSOs, based on upper limits established by studies of background stars (Whittet et al. 2001) that suggested that OCN⁻ is not found in quiescent clouds. However, more recent studies have revealed stricter upper limits on OCN⁻ toward background stars probing quiescent regions (Knez et al. 2005). As is evident with reference to Table 3, an OCN⁻ component was fitted toward very few lines of sight (five YSOs and two background stars, all but one of which were upper limits).

5.3. CO₂

As explained in Section 1, a key aim of these observations was to explore the links between H₂O, CO₂, and CO ice column densities, and to test the chemistry relating these species. Figure 9(a) shows the calculated column densities of the three most abundant solid phase molecular species, H₂O, CO₂, and CO, plotted against visual extinction, A_V. The lowest A_V of any object in these AKARI data is 5.68.

Zoom In Zoom Out Reset image size

Our results clearly follow the trends established by previous observations. Both the H₂O and CO₂ column densities increase with A_V, commensurate with the previously calculated critical A_V values of ∼3 and ∼4, respectively, although such low extinctions were not probed in this subset of AKARI observations. For both H₂O and CO₂ we would expect an increase in ice column density with A_V; more dust equals, on average, more ice.

Conversely, in Figure 9(a), there is no clear relationship between CO and A_V. There is a much greater spread in CO values compared to those seen for H₂O and CO₂ in the panels above. CO is known to form in the gas phase and freeze out onto the grain surface above a critical A_V (which has been estimated, variously, as ∼3 in Serpens (Chiar et al. 1994), ∼5 in Taurus (Whittet et al. 1989), and ∼11 in ρ Ophiuchi (Shuping et al. 2000)), but not necessarily as a function of dust density, as the CO gas density is more critical (Pontoppidan et al. 2003b). Toward one line of sight, CO is observed in the solid phase at a relatively high column density for a low extinction (Object 10; A_V = 5.75, N(CO) = 1.6 × 10¹⁷ molecules cm⁻²), which may suggest that, with sufficient sensitivity, ices can be detected at lower A_V values than previously reported in the literature.

Figure 9(b) shows the column densities of CO₂ and CO plotted against the column density of H₂O. Overall, the data calculated in this study agree well with the literature values; the relationships between H₂O, CO₂, and CO in molecular clouds and around YSOs are well established and it is important that new data continue to reinforce previous results. As expected, there is no clear relation between the H₂O and CO column densities, as the freezeout and formation mechanisms of each species differ vastly. However, a linear correlation between the H₂O and CO₂ column densities is evident, indicating a potential link in the formation chemistry of these two species. It is possible that they both share a common chemical starting point; one route to H₂O formation involves H + OH (e.g., Cuppen & Herbst 2007), one route to CO₂ formation involves CO + OH (Oba et al. 2010; Ioppolo et al. 2011; Noble et al. 2011; Zins et al. 2011) and could account for the CO₂ ice observed in quiescent regions (Nummelin et al. 2001; Pontoppidan 2006). Recent experimental results show that the relative rates of these two reactions, if occurring concurrently, and if they were the only reactions producing H₂O and CO₂, would generate around 20 times more H₂O than CO₂, assuming that all the reagents were equally available (Noble et al. 2011). Obviously, these conditions may not be satisfied even in a quiescent line of sight. Figure 9(b) shows that the H₂O:CO₂ ratio is approximately 10:1, suggesting that other mechanisms must produce the remainder of the observed CO₂. Unlike the subset of background stars and low-luminosity YSOs surveyed here, column density values for low- and high-mass YSOs tend to be much higher (e.g., Gibb et al. 2004; Pontoppidan et al. 2008); corroborating evidence that UV processing or longer timescales promote CO₂ formation via more complex CO₂ formation mechanisms.

Figure 10 is a ternary plot showing the relative column densities of all three species on a single plot. It includes fewer data points than Figure 9, simply because far fewer studies have been able to ascertain the abundances of all three species toward individual lines of sight. This figure clearly illustrates the range of background star data and column densities that were probed by this survey.

Zoom In Zoom Out Reset image size

Looking more closely at the three major ice species, more can be learned about their interlinking chemistries by considering the individual components of each. During this process, it is important to consider all available literature data in order to ascertain general correlations, rather than potentially drawing conclusions based on survey-specific trends. We first consider the relationship between H₂O column density and the column densities of pure CO (CO_mc) and CO in an H₂O-rich ice environment (CO_rc) in Figure 11.

Zoom In Zoom Out Reset image size

It is postulated from laboratory experiments that these two CO ice environments promote very different ice chemistries: in the pure CO ice the reactions are dominated by successive H atom, O atom, and radical additions to the CO molecule to form CO₂, HCOOH, H₂CO, and CH₃OH (e.g., Watanabe et al. 2004; Fuchs et al. 2009); in the water-rich CO ice, photon- and electron-induced chemical processes become more important, first because the CO molecule is trapped, leading to a CO chemistry at temperatures above its desorption energy, and second because (unlike the CO molecule) H₂O can be photodissociated, allowing reaction with CO to form CO₂, HCOOH, H₂CO, and CH₃OH as well as many more complex molecules (e.g., Gerakines et al. 1996; Jamieson et al. 2006). It is only through observational evidence that we can start to discriminate between these processes and refine gas-grain astrochemical models.

It is clear from analysis of Figure 11 that there is no correlation of H₂O with pure CO, which condenses out of the gas phase as a layer on top of the H₂O. However, for the H₂O-rich phase of the CO, as expected, N(CO_rc) increases with N(H₂O). Collings et al. (2003b) offer evidence for the porosity of H₂O in the ISM, as their experiments suggest that the population of CO in an H₂O-rich environment develops upon heating of a layered system. A higher column density of H₂O ice suggests a larger number of pores for the CO to migrate into with time or temperature to produce a mixed phase (Pontoppidan et al. 2003b). As Figure 11 shows, toward YSOs there is generally a higher column density of CO in an H₂O-rich environment. The ice has been present in these regions for longer, so it is therefore more likely that the CO has had the time (or the temperature change) necessary to migrate into a mixed H₂O layer. Otherwise, background stars would be expected to probe lines of sight with equally high column densities of CO_rc. It is important to note that CO_rc is a derived feature, as it cannot be measured in isolation in the laboratory.

Figure 12 illustrates the relationship between the H₂O column density and that of the two components of CO₂—CO₂ in a CO-rich ice (CO_{2, non polar}) and CO₂ in an H₂O-rich ice (CO_{2, polar}). When the component column densities from the broader literature, primarily from low- and intermediate-mass objects, are included in the plot (open symbols), two different correlation gradients are clearly evident, which both scale directly with ice abundance (albeit at different rates) suggesting that the dominant formation mechanisms for CO₂ in water-rich and water-poor environments are independent of one another. Whittet et al. (2009) generated similar gradient profiles by comparing the total elemental O atom distribution, derived by summing the CO, CO₂, and H₂O ice column densities, in the H₂O-rich (polar) and CO-rich (apolar) ice phases on extincted lines of sight toward a sample of around 10 bright background stars and YSOs. They show that the elemental O is uniformly distributed between the water-rich and water-poor ice components, and is well correlated with the dust density, suggesting that the final fraction of O atoms accumulating in interstellar ices is a constant reservoir from cloud to cloud, but that the absolute source-to-source O atom distribution between the three major ice molecules—CO, H₂O, and CO₂—can vary significantly. They attribute most of this variance to the CO₂ population in the water-rich ice component, suggesting that the CO₂ ice column densities are invariant with the efficiency of surface chemical pathways, instead reflecting variations in gas phase chemistry (related to CO gas densities) or photodesorption of the CO ice.

Zoom In Zoom Out Reset image size

Our data do not entirely corroborate this conclusion. The component column densities of CO₂ in a CO-rich ice (CO_{2, non polar}), derived from our two-component fitting, match very well with previously reported values (as is the case for the CO ice components in Figure 11), providing reassurance in our analysis methods, despite the fitting limitations imposed by the AKARI resolution discussed in Section 5.3. It is therefore interesting to note that we have sampled a very specific population of CO₂ in an H₂O-rich ice (CO_{2, polar}), which clearly occupies the lower left-hand region of Figure 12, commensurate with a handful of previous data points. It seems therefore that the formation of CO₂ in H₂O-rich ices is actually bimodal. The most likely scenario is that CO₂ ices first form via atom and radical surface reactions of CO in competition with H₂O formation, producing a limited concentration of CO₂ in an H₂O-rich ice (CO_{2, polar}). Later, a combination of CO freezeout and CO migration to the water-rich ice layer, followed by energetic processing of the ices, leads to a second phase of CO₂ formation, which then dominates the CO₂ in an H₂O-rich ice (CO_{2, polar}) column density. This is entirely in agreement with the "early" and "late" formation mechanism model proposed by Öberg et al. (2011), and implies that the CO₂:H₂O ratio will vary significantly from source to source (as indicated by Whittet et al. 2009), but that this ratio will initially be dependent on competing surface reaction rates between CO, H, and OH to form CO₂ and H₂O, then subsequently on the degree to which ice mantles are thermally or energetically processed, rather than the prevailing gas phase conditions. Poteet et al. (2013) have very recently illustrated that crystalline CO₂-rich ice layers can dominate grain mantles under the unusual circumstances where significant thermal processing occurs in the vicinity of a protostar (in this case HOPS-68). On the one hand, therefore, Figure 12 corroborates the Whittet et al. (2009) hypothesis: the component of CO₂ ice in the H₂O-rich layer dominates the correlations between the major interstellar ice components. However, while our overall CO₂ column densities match well with existing literature values, the fraction of CO₂ ice we observe in water-rich environments is relatively small.

Previous studies have illustrated that there is not a direct link between CO freezeout and all CO₂ formation, as there is no overall correlation in the abundances of total CO and total CO₂ (e.g., Gerakines et al. 1999). In Figure 13, the plot of the pure component of CO against the CO-rich component of CO₂ shows no correlation, suggesting that the formation route to CO₂ formed in the CO layer is not a simple one, and probably not driven only by CO reacting with O, H, or OH from the gas phase; a similar conclusion was reached from our analysis of Figure 9(b). It is likely that the relative CO and CO₂ populations in the CO-rich ice environments are additionally regulated by competing photodesorption processes (e.g., Öberg et al. 2007, 2009; Fayolle et al. 2011).

Zoom In Zoom Out Reset image size

Figure 14 shows plots of the column density of the water-rich CO₂ component versus the column density of the CO components CO_mc and CO_rc. It is very clear from these plots that there is a strong link between the column densities of CO₂ and CO in an H₂O-rich environment (CO_rc). Thus, some CO₂ must form from CO in an H₂O environment. Mirroring the relationship established between H₂O and CO₂ in an H₂O-rich environment (Figure 12), background stars probing quiescent lines of sight have the lowest column densities of CO in an H₂O-rich environment, while in the more evolved YSOs there is a higher column density, suggesting that CO₂ forms from CO that has migrated into the H₂O upon formation of a star in the core. This is key evidence supporting the conclusions of the ice mapping in Pontoppidan (2006).

Zoom In Zoom Out Reset image size

These data illustrate the links between all of the components present in the CO and CO₂ absorption features, and the sequence of CO₂ formation in molecular clouds and star-forming cores. The proposed sequence is outlined here. Initially, CO₂ forms concurrently with H₂O, producing an abundance of CO₂ in H₂O. Upon critical freezeout of CO, further CO₂ is formed by reactions of CO with O, H, and OH, producing a CO₂ component in a CO-rich ice. With time and, potentially, processing, some CO migrates into the H₂O ice, where it can react to form large column densities of CO₂ in an H₂O-rich ice, likely by an energetic route enhanced by UV radiation from a newly formed YSO.

In this paper, the 2.5–5 μm spectra of 22 background stars and 8 YSOs have been analyzed to determine the column densities of key solid phase molecular species toward those lines of sight. Prior to this study, only one complete spectrum of a background star had been obtained in this wavelength region, a region containing absorption bands for the three most abundant molecules present in interstellar ice. Thus, the addition of 22 spectra toward background stars to the literature provides valuable data to the community, which could help to benchmark the initial conditions in star-forming regions before the onset of star formation.

The analysis of this large sample of data by a rigorous fitting approach reinforces the argument that the methods employed here are possible for large data sets (as previously shown by, e.g., Pontoppidan et al. 2003b, 2008; Boogert et al. 2008; Öberg et al. 2011). It is not necessary to employ a "mix and match" approach using multiple laboratory spectra in order to define the ice characteristics of a wide range of objects.

In summary, the key findings of this survey are as follows.

• 1.  
  The sensitivity of AKARI enabled us to probe lines of sight toward objects with fluxes < 5 mJy, allowing observation of spectra in a previously inaccessible parameter space.
• 2.  
  CO₂ and H₂O were found to be ubiquitous where ice was observed, at column densities commensurate with previously reported literature values.
• 3.  
  From the profile of the H₂O ice band, the extended red wing seems to be an evolutionary indicator of both dust properties and the extent to which the ice mantle covers the grain.
• 4.  
  New observations of the CO₂ stretching absorption band profiles and ice column densities for lines of sight toward 21 background stars have been obtained.
• 5.  
  CO ice absorption features were extracted from the spectra of 23 of the 30 lines of sight probed. In at least two objects—Objects 4 and 20—the CO ice absorption band profile is fully described by a single pure CO component only.
• 6.  
  On 15 lines of sight toward background stars, the column densities of all three species H₂O, CO₂, and CO were extracted.
• 7.  
  By considering the column densities of CO and CO₂ in different ice components, it is possible to postulate a CO₂ formation scenario involving two or three distinct stages. The proposed timeline is: early formation concurrent with H₂O; late formation by energetic routes when CO has frozen out and migrated into an H₂O environment; and the formation of CO₂ in the CO layer at an uncertain time and by uncertain mechanisms.

As our knowledge of interstellar solid state ices increases, many common factors are continuing to emerge in our understanding of their formation and evolution. By surveying ice features in the spectra of faint YSOs and background sources, we have sampled a further region of parameter space in which ices can be detected. Clearly, even where correlations are evident between one ice species and another, the range of molecular column densities remains vast. The local environment must, therefore, have a significant effect on ice column density, composition, and growth. Even the individual components of a species potentially vary across a cloud. In order to use solid phase molecular species as probes of ice and gas chemistry, it will be necessary to map their distributions on a two-dimensional spatial scale, as well as make many more observations of both solid phase species and gas phase molecules.

The authors are very grateful to Y. Ohyama for fruitful discussions, and thank M. Tamura, M. Ueno, R. Kandori, and A. Kawamura for their help with the initial target selection of the ISICE program. J.A.N. is a Royal Commission for the Exhibition of 1851 Research Fellow, and acknowledges the financial support of the University of Strathclyde, the Scottish Universities Physics Alliance, and the Japan Society for the Promotion of Science. The research leading to these results has received funding from the European Community's Seventh Framework Programme FP7/2007-2013 under grant agreement No. 238258 (LASSIE). H.J.F. and J.A.N. are grateful to EU funded COST Action CM0805 "The Chemical Cosmos: Understanding Chemistry in Astronomical Environments" for a funding contribution toward the final production of this work.

Facility: AKARI -