"ICE CUBES" IN THE CENTER OF THE MILKY WAY: WATER-ICE AND HYDROCARBONS IN THE CENTRAL PARSEC^*

In a previous paper (Moultaka et al. 2004) we derived a spectrum of the foreground extinction in the wavelength range of the SL filter of ISAAC. To that end, we used the L-band spectrum of a late-type star showing CO bandheads in its K-band spectrum. In the following we will refer to this object as the "CO-star." Owing to the shape of its near-infrared spectrum and to the presence of this molecular feature, it can be safely assumed to have an effective temperature, of typically 3600 K, corresponding to an M0-type star. Moreover, it is close to the studied region at an angular distance of 13'' (i.e., ∼0.5pc) from Sgr A* and is located at the edge of Sgr A* West, outside the mini-spiral and the CND structures. Its L-band spectrum does not show excess emission. Hence, we can assume that it is not affected by local extinction and that its L-band spectrum, shown in Figure 2(a), is only extincted by the line-of-sight material. One can notice that the observed CO-star spectrum shows two absorption bands at $\sim 3\;\mu {\rm m}$ and $\sim 3.4-3.48\;\mu {\rm m}$ due to the presence of water-ice and hydrocarbons along the line of sight.

Zoom In Zoom Out Reset image size

In order to derive the line-of-sight extinction spectrum, we created a blackbody spectrum of 3600 K temperature and shifted it by a multiplicative constant to match that of the observed CO-star around $4.2\;$ μm. At this wavelength, we can safely assume that the continuum is free of absorption features and therefore matches the continuum of a nonabsorbed spectrum. Nevertheless, this assumption does not mean that the spectrum is not absorbed by a constant continuum extinction over the whole wavelength range (see explanation hereafter). We then divided the observed spectrum by that of the shifted blackbody. Consequently, the resulting spectrum represents the wavelength-dependent extinction along the line of sight in the L-band domain that accounts only for the extinction due to the absorption features (see Figure 2(b)). Indeed, if we call ${I}_{\mathrm{CO}-\mathrm{star}\;\mathrm{obs}\lambda }$ the flux at wavelength λ of the observed spectrum of the CO-star and ${I}_{\mathrm{CO}-\mathrm{star}\;\mathrm{intr}\lambda }$ its intrinsic flux at the same wavelength (that is, ideally a blackbody emission of about 3600 K temperature), then we have

$$\begin{eqnarray}&&{I}_{\mathrm{CO}-\mathrm{star}\;\mathrm{obs}\lambda }={I}_{\mathrm{CO}-\mathrm{star}\;\mathrm{intr}\lambda }\ast [{E}_{\lambda }]\ast k\end{eqnarray} \tag{ 1 }$$

where ${E}_{\lambda }$ is the line-of-sight extinction spectrum and k a constant accounting for an additional constant continuum due to dust extinction. Here k is assumed to be equal to 1 since its value does not influence the results we present in the following sections.

The optical depth spectrum $\tau (\lambda )$ corresponding to the line-of-sight extinction is shown in Figure 8 of Moultaka et al. (2004) and in Figure 2(c). It is obtained via the definition $\tau (\lambda )=-{ln}\left({\displaystyle \frac{{I}_{\mathrm{CO}-\mathrm{star}\;\mathrm{obs}\lambda }}{{I}_{\mathrm{CO}-\mathrm{star}\;\mathrm{intr}\lambda }}}\right)=-{ln}([{E}_{\lambda }]\ast k)$.

From the L-band line-of-sight extinction spectrum shown in Figure 2(c), we derive a mean optical depth of about 0.55 (or a mean extinction E_L of 0.57). This corresponds to an extinction ${A}_{L}\sim 0.6$ mag. If we consider the extinction law by Martin & Whittet (1990) (obtained from observations of stars in the solar neighborhood and in the ρ Oph cluster), the above value of A_L will be consistent with an extinction in the K band of ${A}_{K}\sim 1.38$ mag. On the other hand, it will be consistent with a K-band extinction of ${A}_{K}\sim 1.16$ mag if we use the extinction law of Rieke & Lebofsky (1985) resulting from observations of stars in the GC. This is less than the mean value of about 2.46 mag obtained toward the GC by Schödel et al. (2010). We conclude that the absorption features alone do not account for the foreground extinction, but an important contribution comes from the dust continuum absorption as well. This component is represented by the constant k in Equation (1). As a matter of fact, the continuum extinction by dust was not taken into account since we assumed a constant k = 1 in Section 3.1 when deriving the line-of-sight extinction spectrum.

Considering the A_K value given by Schödel et al. (2010), we can estimate the value of the constant k. If we calculate the L-band extinction using the extinction law by Rieke & Lebofsky, we obtain an ${A}_{L}\sim 1.27$ mag. This is equivalent to an optical depth of ${\tau }_{L-\mathrm{Sch}}=1.17$. Doing the same calculation but using Martin & Whittet's law, we find an ${A}_{L}\sim 1.07$ mag, implying an optical depth of ${\tau }_{L-\mathrm{Sch}}=0.98$. In order to match the extinction A_K of Schödel et al. (2010), the required constant, ${\tau }_{\mathrm{add}}$, to be added to the optical depth spectrum of Figure 2(c), is then ${\tau }_{\mathrm{add}}={\tau }_{L-\mathrm{Sch}}-0.55$. We obtain ${\tau }_{\mathrm{add}}=0.62$ (for Rieke & Lebofsky's law) and ${\tau }_{\mathrm{add}}=0.43$ (for Martin & Whittet's law). The corresponding constant k of Equation (1) is then derived through $k={e}^{(-{\tau }_{\mathrm{add}})}$. In the case of Rieke & Lebofsky's law, we find $k={e}^{(-0.62)}=0.54$, and in the case of Martin & Whittet's law, we get $k={e}^{(-0.43)}=0.65$.

In Moultaka et al. (2004) we obtained spectra of a dozen bright sources located in the central parsec that we corrected for the foreground extinction. To that end, we divided the observed spectra by the calibrator extinction spectrum described previously and shown in Figure 2(b). The resulting spectra were then fitted by blackbody continua. The best fits were obtained with blackbody temperatures that agree very well with the known temperatures of the sources (or of the dust they are embedded in), obtained from NIR spectroscopy or imaging.

On the other hand, we also fitted the noncorrected spectra for the foreground extinction by blackbody continua reddened with a varying extinction A_K. The temperatures of the best fits matched very well those found previously using the calibrator extinction spectrum (see Table 1). Also, the mean value of the extinction A_K obtained from the best fits agrees with the value of about 3 mag derived from imaging data (Schödel et al. 2010; see Table 1). These results show that the calibrator extinction spectrum is a very good approximation of the foreground extinction.

Table 1.  Table Summarizing the Results of the Two Methods Used in Moultaka et al. (2004) to Estimate the Foreground Extinction

Source	Temperature 1a	Temperature 2b	${A}_{K}$ b
IRS 1W	1200	900	3.30
IRS 3	1200	800	4.20
IRS 7	1100	2100	3.55
IRS 9	2000	1900	3.10
IRS 13	5000	1300	3.00
IRS 13N	1500	1000	3.90
IRS 16C	26000	22600	3.45
IRS 16CC	20000	20200	4.00
IRS 16NE	20000	22400	3.25
IRS 16SW	20000	22500	3.25
IRS 21	1000	1200	3.95
IRS 29	1200	1600	3.70

Notes.

^a The temperature is obtained from the best fit of a blackbody spectrum to the corrected spectra for the foreground extinction using the calibrator extinction spectrum. ^bThe temperature and K-band extinction A_K are obtained from the best fit of a reddened blackbody spectrum to the observed source spectra.

Download table as:  ASCIITypeset image

In the present work, we have an L-band data cube that covers the entire field of the central half parsec. Using the calibrator extinction spectrum, we built a data cube corrected for the foreground extinction. To this end, we divided all the spectra of the field at each spatial pixel (or equivalently, all the lines in the frames corresponding to the different slit positions) by the extinction spectrum. We show in Figures 3 and 19 the maps of the integrated fluxes along the wavelength range obtained from the corrected and noncorrected data cubes for extinction, respectively. The similarity of both maps shows that there is no over-correction of the extinction in the observed field.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To test the validity of the calibrator extinction spectrum in the region, we fitted blackbody continua to a number of spectra of early- and late-type stars located all around the central half parsec and outside the mini-spiral. These stars have been classified as early-type or late-type by Buchholz et al. (2009) by means of near-infrared photometry and spectral energy distributions. The best fits were obtained with temperatures of about 10,000–40,000 K for the early-type stars and about 1000–5000 K for the late-type stars, which is in agreement with the temperature of their spectral types. This shows that our calibrator extinction spectrum of the line-of-sight extinction is also valid over the entire region. In Figure 4 we show a number of spectra for each of the stellar types presented by Buchholz et al. (2009) with the best blackbody continuum fits overlaid. We also show the spectra of the bright infrared sources IRS 1W, IRS 16C, IRS 16NE, IRS 21, and IRS 29.

On one side, this result validates our calibration of the line-of-sight extinction spectrum, and on the other side, our L-band spectra confirm the nature of the observed stars. Our results also agree with Scoville et al. (2003) and Schödel et al. (2010), who found a rather smooth distribution of the extinction toward the GC on a 1''–2'' scale in the near-infared, with a variation of A_K not exceeding 0.5 mag.

We aim at deriving the distribution of the local water-ice and hydrocarbon features in the half parsec around Sgr A*. We assume that the continua of the hydrocarbon and water-ice absorptions can both be approximated by straight lines (see Figure 4). These assumptions are justified by the following facts:

1. The hydrocarbon absorption is located in the red wing of the water-ice absorption, and the latter can be approximated by a straight line from 3.32 to 3.77 μm. This is observed, for example, in the GC spectra by Moultaka et al. (2004; see Figures 2–4 and 9–11 of that paper), in Figure 7 by Chiar et al. (2002; showing a mean GC ice feature, as well as three water-ice models), and in the spectra of YSOs and quiescent molecular clouds with background stars, shown in Figure 3 of Noble et al. (2013).

2. The relative uncertainty on the optical depths of the water-ice feature is less than 10%, if we assume a straight line continuum from 2.84 to 3.77 μm instead of a blackbody continuum or instead of fitting a laboratory water-ice spectrum. Indeed, this deviation can be estimated from the figures mentioned above. Moreover, in the Taurus molecular cloud complex Murakawa et al. (2000) estimated that due to this approximation of the continuum, the relative uncertainty in deriving the optical depths of the water-ice absorption is about 5%. Note, finally, that the wavelength interval of the hydrocarbon absorptions includes the 3.53 μm methanol ice absorption, but this feature is not clearly detected in our spectra. The optical depths are calculated through the definition

$$\begin{eqnarray}&&\tau =-\mathrm{ln}({\displaystyle \frac{{F}_{\mathrm{obs}}}{{F}_{\mathrm{cont}}}})\end{eqnarray} \tag{ 2 }$$

where F_obs is the integrated observed flux over the absorption feature and F_cont is the integrated continuum flux along the spectral feature. The map of the optical depth of the water-ice absorption is obtained after subtracting that of the hydrocarbon absorption since the latter is located in the wing of the water-ice feature (see Figure 4).

The advantage of using optical depths is that they are independent of the continuum intensity. Their values can, therefore, be directly compared, regardless of the background brightness.

The resulting maps obtained from the data cube corrected for the foreground extinction are shown in Figures 5 and 6 for the water-ice and the hydrocarbon features, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figures 20 and 21 (see the Appendix), we show the optical depth maps of the same absorption features obtained from the noncorrected data cube for extinction. The water-ice maps corrected and noncorrected for extinction are very similar as can be seen by comparing Figures 5 and 20. Also, we find similar distributions of the 3.4–3.48 μm bands in the corrected and noncorrected maps (see Figures 6 and 21). This suggests that the line-of-sight extinction is not overestimated and the data cube is not overcorrected since no negative values are encountered in the maps.

The first striking result is the presence of residual absorptions in the corrected maps. Since we corrected for the foreground effects, this is indicative of absorbing material possibly in the local medium of the GC.

On the other hand, the corrected maps show that the 3 μm and 3.4–3.48 μm features obviously trace the mini-spiral structure. This is also highlighted in Figure 7, where we overlay the water-ice corrected map with contours of the infrared image at 8.6 μm obtained with the VISIR/ESO imager (Viehmann et al. 2006). Since emission at 8.6 μm is dominated by dust, Figure 7 shows that water-ices trace the dust very well except in two regions. As a matter of fact, in the northeastern part of the field, the 8.6 μm emission is bright while water-ices are not detected. In the northwestern part we find the opposite behavior. The mismatch in the first case is due either to the absence of water-ices stuck on the dust grains or, more probably, to the absence of bright background sources resulting in a low S/N of the spectra. In the northwestern part, we detect the ice absorption since there are bright background sources (see Figure 8), but the $8.6\;$ μm emission is very faint or absent. This is probably due to the small amount of dust in the region, which is consistent with the small optical depths found in our map at the same positions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In addition, we show in Figure 9 our water-ice absorption map (right panel) in comparison with the high-pass filtered L'-band (at $3.8\;$ μm) image from Mužić et al. (2007; left panel). The latter image highlights the compact components of the L'-band image obtained with NACO (NAOS/CONICA at ESO/VLT) and reveals a large number of thin filaments in the mini-spiral. The authors interpret these features as the result of the interaction of wind from Sgr A* with dust in the mini-spiral. Here our map has a lower spatial resolution (about 1') than that of Mužić et al. (2007), which is 100 mas. Despite this discrepancy, it is clear from Figure 9 that the peaks of the optical depth values of the water-ice match very well the positions of the thin filaments of dust. Hence, water-ice seems to trace the high-density pockets of the dust.

On the other hand, the distribution of the water-ice and the hydrocarbon absorptions also trace the dusty sources like IRS 3, IRS 7, and the north of IRS 13 complex (see Figures 8 and 10). Thus, these features are clearly associated with the dust components in the region.

Zoom In Zoom Out Reset image size

All these results provide an additional validation of our calibration of the foreground extinction in the L band. Indeed, it is unlikely that concentrations of dust along the line of sight happen to be located at the same projected distances to Sgr A* as the local dust concentrations in the GC region.

Furthermore, our resulting maps suggest that the water-ice and the hydrocarbon features can be associated with the neighboring infrared sources. Indeed, in Figures 8 and 10, we overlay contours of the integrated L-band emission map on the optical depth maps; these figures show that peaks of absorptions are almost systematically offset by a few arcseconds (equivalent to a few angular resolution elements) from the bright sources of the L-band continuum map, except, maybe, for IRS 7 and in the southern part of the field. We interpret these observations as being a structure of mass-losing stars resulting from the interaction with the local ISM. This interaction may be enhancing their surrounding dust emission, possibly by piling up the local dust. In the few cases where the peaks of the optical depth maps overlap those of the integrated emission map, the ISM is probably squeezed in the direction of the line of sight.

We show in Figures 11 and 12 the K-band extinction map derived by Schödel et al. (2010) overlaid on our optical depth maps. These figures witness a full absence of correlation between the K-band map and our L-band extinction maps. This result is a necessary condition if we claim that the absorption features occur in the local environment. Indeed, the extinction as traced by the K-band images is mainly due to continuum absorption due to the presence of dust along the line of sight. If the absorption features in the L band were correlated with the K-band continuum absorption, this would imply that these absorptions are directly linked to the dust all along the line of sight. The fact that there is no correlation implies either that not all dust contains water-ices/hydrocarbons or that the absorption features are local to the GC and therefore diluted in the much higher global extinction due to the foreground and the local dust .

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In order to quantify the local extinction along the entire spectrum, we created the optical depth map of all the material absorbing over the full wavelength range. This map cannot be derived from the previous 3 μm and 3.4–3.48 μm maps because there is no correlation between the individual absorptions and the overall extinction of the spectrum. To this end, we calculated the optical depths from the corrected data cube for the foreground extinction, as in Equation (2), but in this case, the observed flux, F_obs, and the continuum flux, F_cont, are the observed and continuum/intrinsic fluxes integrated over the full wavelength range, respectively. The smoothed version of this map is shown in Figure 13. We find that the optical depths vary from 0 to 1 in the whole region. This means that A_L varies spatially from 0 to 1.08 mag, corresponding to a variation of A_K from 0 to 2.1 mag (for the case of the Rieke & Lebofsky extinction law) and from 0 to 2.5 mag (for the case of the Martin & Whittet extinction law). This is much higher than the 0.6–0.7 mag variation along the line of sight derived by Schödel et al. (2010), implying that probably most—but at least a significant part—of the residual absorbing material we observe in the L band is located in the local medium of the central parsec.

Zoom In Zoom Out Reset image size

The presence of water-ices that we find in the local medium of the GC implies temperatures of the order of 10–80 K (Chiar et al. 2002). This is consistent with our finding in Moultaka et al. (2009) and Moultaka et al. (2015) that CO ices are probably located in the region as well. We showed in Moultaka et al. (2009) that, given the complexity of the region with the presence of bow-shock sources (e.g., Tanner et al. 2002, 2005), narrow dust filaments (e.g., Mužić et al. 2007, 2010), and dust-embedded YSOs (Eckart et al. 2004, 2013), one cannot discard the presence of high-density pockets and high optical depths where material at very low temperatures can also be present. These compact dusty structures can persist and survive while traveling through the central parsec since we showed that the travel time is shorter than the evaporation time of molecular clumps and disks in similarly harsh environments.

The hydrocarbon and water-ice corrected optical depth maps of Figures 5 and 6 show a certain resemblance, suggesting that these features are similarly distributed in the field. This idea was also invoked in our previous papers (Moultaka et al. 2004, 2005). In Figure 14, we plot contours of the hydrocarbon absorption map over that of the water-ice to highlight this trend. In order to quantify the correlation between the two absorptions, we derived the optical depths at each pixel of the maps corrected for foreground extinction, where the S/N in the integrated L-band corrected map is larger than 10. Then we calculated the mean optical depth of the hydrocarbon absorption of all pixels for which the water-ice optical depth is in bins of 0.1 width.

Zoom In Zoom Out Reset image size

We plot in Figure 15 the mean optical depths of the hydrocarbon feature as a function of the water-ice optical depths in bin widths of 0.1. With a correlation coefficient of 0.85, this figure reveals a clear correlation between the water-ice and the hydrocarbon optical depths. The best linear fit is shown in blue in Figure 15. This correlation suggests that the ISM presents itself as a—possibly clumpy—mixture of a dense and a diffuse medium.

Zoom In Zoom Out Reset image size

From the data cubes described previously, we also derive maps of the emission line strength for the ${\mathrm{Pf}}_{\gamma }$ and ${\mathrm{Br}}_{\alpha }$ lines located at 3.729 and 4.051 μm, respectively (see an example of a spectrum in Figures 3 and 13 of Moultaka et al. 2004). We define the line strength via $\frac{{\int }_{\lambda }{I}_{\mathrm{line}}(\lambda )-{I}_{\mathrm{cont}}(\lambda )}{{\int }_{\lambda }{I}_{\mathrm{cont}}(\lambda )}$ , where the integral along the emission line is calculated from $3.722\;$ to $3.744\;$ μm for the ${\mathrm{Pf}}_{\gamma }$ line and from $4.028\;$ to $4.061\;$ μm for ${\mathrm{Br}}_{\alpha }$; ${I}_{\mathrm{line}}(\lambda )$ is the intensity of the line at each wavelength and ${I}_{\mathrm{cont}}(\lambda )$ the intensity of the assumed continuum at the same wavelength. The continuum is approximated by a straight line connecting the mean continuum fluxes on the red and the blue sides of the line. They are calculated over 16 pixels (i.e., $\sim 22.8\;\mathrm{nm}$) for the ${\mathrm{Pf}}_{\gamma }$ line and over 24 pixels (i.e., $\sim 34.3\;\mathrm{nm}$) for ${\mathrm{Br}}_{\alpha }$.

The hydrogen emission line maps obtained from the corrected and noncorrected data cubes do not differ from each other (the extinction-corrected maps are shown in Figures 16 and 17, respectively). This result is expected since the emission lines are not present in the spectrum of the foreground extinction. Peaks of the hydrogen emission line maps are due either to the presence of a broad emission line star or to a high emission of the lines. Some peaks may be due to a very low continuum (i.e., with a very low S/N) producing artifacts in the emission-line maps. These regions are covered by black ellipses in Figures 16–18. We checked all other peaks and found that except the Wolf–Rayet stars WR1, WR2, and WR3 that we identified in Moultaka et al. (2005), only one peak is due to the Wolf–Rayet star already identified by Buchholz et al. (2009) located at the southwestern corner of the maps in Figures 16–18.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

From these maps, we see clearly that the hydrogen emission lines trace the shape of the mini-spiral. They are obviously produced by the gas in the mini-spiral that is ionized by the hot massive stars. On the other hand, the ${\mathrm{Pf}}_{\gamma }$ and ${\mathrm{Br}}_{\alpha }$ line maps appear to be correlated. Also, in Figure 18, we overlay the contours of the ${\mathrm{Br}}_{\alpha }$ line map over the ${\mathrm{Pf}}_{\gamma }$ map where the correlation is clear. This implies that both emissions are produced by the same gas with the same physical conditions.

Zoom In Zoom Out Reset image size

Finally, we find no correlation between the hydrogen emission line maps and the absorption maps. This implies that the two media of the gas and dust are not tightly linked. Dust is more confined in the locations of the bright sources, while the gas is more spread all over the region. Notice that gas is almost absent in the IRS 3–IRS 7 region, while molecular absorption is prominent there.