EXTRAGALACTIC CS SURVEY

The observations have been performed using the IRAM-30m telescope for the CS(2–1), CS(3–2), and CS(4–3) lines (ν = 97.980 GHz, 146.969 GHz, and 195.954 GHz, respectively) and the James Clerk Maxwell Telescope (JCMT) for the detection of the CS(7–6) transition (ν = 342.883 GHz). Every 2–3 hr, when using both telescopes, the pointing, focus and calibration were performed carefully on planets (Mars and Jupiter) and on evolved stars. The pointing error was estimated to be ⩽3'' in both cases. We have observed the CS(5–4) line with both telescopes but we have not averaged the signals because the weather conditions were significantly different. Between the two telescope observational sessions, for all the sources but NGC 4038, we have kept the less noisy spectrum (i.e., the spectrum showing the highest signal-to-noise ratio (S/N)). These (reduced) spectra are shown in Figures 1–10 and their Gaussian fitting parameters are seen in Table 1. Unfortunately, for NGC 4038, the reduced spectra obtained with the IRAM-30m and with the JCMT are equally noisy (i.e., they have similar S/N) and thus we present both spectra in Figure 7.

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

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 IRAM-30m telescope observations were carried out in 2008 June under various weather conditions (≈1.5–6 mm of water vapor). Scans were taken using the wobbling secondary with a throw of 120'' for the two positions in M 82 (NE and SW components), while for the Antennae positions (NGC 4038, NGC 4039, and Overlap) we have used the position switching mode with a larger throw of 1' (reference positions are indicated in Columns 12 and 13 of Table 1). The 100 GHz, 150 GHz, 230 GHz, and 270 GHz receivers (A/B/C/D) were used under several configurations, optimizing the allocated time. The 1 MHz and 4 MHz backends were used providing velocity resolution between 1.2 km s⁻¹ (at 244 GHz) and 3.0 km s⁻¹ (at 98 GHz), and 4.9 km s⁻¹ (at 244 GHz) and 12.3 km s⁻¹ (at 98 GHz), respectively. At the observed frequencies of 97.980 GHz, 146.969 GHz, 195.954 GHz, and 244.936 GHz, the IRAM-30m has an HPBW of ≈ 25'', 17'', 13'', and 10'' with main beam efficiencies of 0.75, 0.69, 0.65, and 0.52,⁵ respectively. The system temperature ranged from ≈ 200 K to ≈ 1800 K, depending on the frequency and the weather conditions. The data reduction was performed using the CLASS program of the GILDAS package developed at IRAM.

We have used the JCMT in 2007 March, 2008 February, 2008 July, and 2009 February under medium weather conditions (τ₂₂₅≈ 0.15). We have either used a beam switch mode with a throw of 180'' or a position switch mode with hard off-source reference, depending on the source. We have used the RxA3 and the 16-pixel camera HARP receivers for the detection of the CS(5–4) and the CS(7–6) transitions, respectively. The ACSIS digital autocorrelation spectrometer with a bandwidth of 1000 MHz was used at both frequencies and in all the sources because the lines were expected to be broad (see Figures 1–10). The HPBW and the main beam efficiencies of the JCMT at ν = 244 GHz and ν = 343 GHz are 20'', 14'' and 0.69, 0.63,⁶ respectively (see Table 1). The system temperatures ranged between 200 K and 400 K. The data pre-reduction was done using Starlink softwares (KAPPA, SMURF, and STLCONVERT packages) and subsequently translated to CLASS format for final reduction.

All the spectra obtained are shown in Figures 1–10, while the Gaussian fit parameters are listed in Table 1. In this table, we have also included detections of CS lines from the literature (see references mentioned in the last column of the table). Hence this table contains the most complete compilation of CS transitions observed in external galaxies.

For each source, we find a general good agreement between the Gaussian fits and the observations (see Figures 1–10). Table 1 also shows that the velocity positions of the lines and the line widths of the CS transitions are consistent among each other and with previous data from the literature.

For some sources, some lines of CS are only marginally detected. For instance, the CS(7–6) line in NGC 1068 (Figure 2) and in IC 342 (Figure 3) have both an S/N lower than 3 and thus cannot be considered as detections. We have also obtained for NGC 4039 a hint of a marginal detection, but due to the uncertainty on the velocity position of the line we have not shown the spectrum. Except for NGC 4039, in all the marginal cases presented here, the fitted line width and velocity position match those expected either from lower-J CS lines (when available) or in the case where no other CS lines are available, from other molecular lines such as CO as in the case of the CS(7–6) line in Henize 2-10. For this source, we have derived an upper limit of the source-averaged CS column density of N(CS) = 7.3 × 10¹² cm⁻² from the CS(7–6) marginal detection, using a source size of 13'' (see Meier et al. 2001; Bayet et al. 2004). Due to the fact that there is only one CS observation so far (see Figure 4), we have excluded Henize 2-10 from the data analysis presented in Section 3.

For NGC 253, based on the literature, it is expected that some CS detections show a double peak emissions corresponding to two gas components (see Martín et al. 2005; Figure 1). Thus, in NGC 253, we have called NGC 253-1 the gas component whose line emission is located at a velocity position of about 180–200 km s⁻¹, while NGC 253-2 corresponds to the higher velocity gas component located at about 280–300 km s⁻¹. A two-components Gaussian fit has thus been applied. We have listed the fit parameters obtained in Table 1.

In some of the M 82-NE CS spectra (see Figure 5), one also notice a double peak emission. However, the nature of this emission differs significantly from the one observed in NGC 253. Indeed, after having compared Figure 5 with the data from Bayet et al. (2008a), we have concluded that the weaker peak (seen on the left side of the spectra) corresponds to a contamination from the nucleus (partially included into the telescope beam). We have thus excluded such emission from the analysis performed in Sections 3.1 and 3.2.

To obtain an estimation of the kinetic temperature and the total CS column density in the various extragalactic star-forming regions we survey here, we have used first the rotational diagram method, corrected for beam dilution effects and assuming optically thin emission (antennae temperature proportional to the column density in the upper level of the observed transition). The beam dilution effect has been removed assuming a source size of 20'', 20'', 4'', 6'', 6'', 85, 7'', 55, 20'', and 5'' for NGC 253-1, NGC 253-2, NGC 1068, IC 342, M82-NE, M 82-SW, NGC 4038, NGC 4039, Overlap, and M 83, respectively. These values have been selected from the available maps of CS published in the literature or from maps of other tracers of dense gas. When neither was available, we have used, by default, interferometric CO maps. Namely, for the NGC 253 source size, we refer to the CS(2–1) map of Peng et al. (1996), in agreement with the values used for SO₂, NS, and NO emission (tracers of relatively dense gas) presented in Martín et al. (2003). For NGC 1068, we have used the HCN maps from Helfer & Blitz (1995) and Krips et al. (2008) for deriving the CS emitting source size. For IC 342, we base our estimation on the HCN interferometric map of Schinnerer et al. (2008), whereas for M 82, we used SiO and HCO interferometric maps from García-Burillo et al. (2001, 2002), respectively. For the three positions in the Antennae Galaxies, we have used the interferometric CO map from Wilson et al. (2000) as an upper limit of the CS emitting source size, while for M 83 we have used the recent HCN interferometric map of Muraoka et al. (2009). In Figures 11–16, we show the resulting rotational diagrams for all the sources.

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 rotational diagram approach (see Goldsmith & Langer 1999 and appendix in Turner 1991; Girart et al. 2002; Martín et al. 2005) assumes the gas in LTE. If correct, then the data are well fitted by a linear regression using a single component. In fact, assuming LTE in rotational diagrams implies that the rotational temperature is equal to the excitation temperatures of all the observed transitions. As the gas may not be thermalized, the derived rotational temperature is thus a lower limit to the kinetic temperature.

The LTE solutions are shown in Figures 11–16 by dashed lines (single-component fit).⁷ As clearly seen in the figures, most of the data are not well fitted by a single-component fit. This is not surprising since it is expected that the gas contained in the nucleus of galaxies, such as NGC 253, IC 342, etc., is not in thermal equilibrium. Thus, to better reproduce the data, we have adopted a very simple approach assuming that, when the single-component fit is not applicable, at least two gas components exist in the observed regions: one responsible for the low-J CS emission and one for the high-J CS line emissions. The fits performed with two components are shown by solid lines in Figures 11–15. Note however that when a source shows two clear detections and an upper limit, our approach is crude. Whereas we simply fitted the rotation diagrams with two temperature components, it is clear that there is a continuum change of the excitation temperature.

The exceptions to this approach are M 83 and the Overlap position. Indeed, for these sources, the CS data appear to be well fitted by a single-component fit. This means that, in these sources, the gas might be more homogeneously distributed than in other sources. This feature has already been observed in M 83 with the CO data (Bayet et al. 2006). Bayet et al. (2006) indeed showed that the same gas component well reproduce the CO(3–2), CO(4–3), and CO(6–5) lines in the center of M 83, the requirement for another (colder) gas component appearing only when considering lower-J CO lines. The current rotational diagram for M 83 and Overlap contains only three and two CS detections, respectively. It may be that the two-components fit may be necessary once the CS(1–0), CS(2–1), and CS(4–3) lines will be observed in M 83 and when the CS(1–0), CS(4–3), CS(5–4), and CS(7–6) transitions will be added to the rotational diagram of Overlap.

In Figures 11–16, we have used different symbols to represent marginal detections (open white squares) and detections (black filled squares). For NGC 4038, two CS(5–4) (marginal) detections exist. We have shown the IRAM-30m one with an open white triangle and the JCMT with an open white square since it appears in better agreement with the rest of the data and shows a higher S/N (see Figure 7 and Table 1).

The rotational temperatures we have obtained for each source are listed in Table 2. The typical uncertainties on the rotational temperatures are of 5%–6%. We have also derived the source-averaged total column densities N(CS) for each component in each source (see Table 2). For determining such values, we have used the partition function from the CDMS Web site,⁸ extrapolating their values to the range of rotational temperatures listed in Table 2 (linear interpolation applied).

Below, we compare our results with those available in the literature.

• 1.  
  NGC 253. As expected, the CS column densities for both NGC 253-1 and NGC 253-2 sources (low- and high-velocity components of NGC 253) are in agreement with the values derived by Martín et al. (2005). Indeed, for the NGC 253-1 source, the CS column density we have computed for the single-component fit is in agreement within a factor of 0.9. For the NGC 253-2 source, the CS column density for the single-component fit is in agreement within a factor of 1.8. The differences between these two studies originate clearly from the presence of the new CS(7–6) observations. This transition reveals or confirms in most of the cases the need of at least one more component to better reproduce the data.
• 2.  
  IC 342. Mauersberger et al. (1989) have also observed in various CS transitions the center of the galaxy IC 342. They estimated the column density to be N(CS) = 1.5 × 10¹⁴ cm⁻², in agreement (with a factor of 0.5) with our estimate (single-component fit).
• 3.  
  Molecular lobes in M 82. M 82-NE and M 82-SW lobes have never been observed previously in any CS lines. However, Bayet et al. (2008a) showed that in the center of M 82 the CS column density is estimated to be N(CS) = (0.13–6.7) × 10¹⁴ cm⁻². This is of the same order of magnitude of the values presented in Table 2 (single-component fit) for the lobe positions.
• 4.  
  The Antennae Galaxy. Similarly to the lobes in M 82, no CS detection has been published so far neither in NGC 4039 nor in the Overlap position in the Antennae Galaxies. Only for NGC 4038, Bayet et al. (2008a) have obtained N(CS) = 1.8 × 10¹³ cm⁻² (from the single CS(5–4) observation). This estimate is in agreement with the values listed in Table 2 (single-component fit).
• 5.  
  M 83. The CS column density computed when adding the CS(7–6) line to the data available from the literature is in agreement with that derived by Mauersberger et al. (1989) and Martín et al. (2009) when using their respective single transition of CS.

In the Antennae Galaxies case, using CO data for representing the more compact (CS) emitting source size is misleading and can cause large uncertainties on the results as shown by Helfer & Blitz (1995) for NGC 1068. Taking NGC 4038 as a representative example, we have thus investigated the influence of the source size choice on the results (LTE and non-LTE models, see Section 3.2) in the following way.

• 1.  
  We have compared the physical parameters of column density and temperature derived assuming a source extent of 7'', with those obtained by assuming sizes of 75%, 50%, and 25% smaller (i.e., 525, 35, and 175) similar to all transitions. We have rebuilt consistently the rotational diagrams and re-calculated in the same way as previously described, the rotational temperature and the total CS column density values. When these values were compared to those obtained in the case of a 7'' source size, we have obtained rotational temperatures difference less than 7% and column densities differences within a factor of <14.8 (for the smallest size considered), whatever the number of components considered (see Table 2).
• 2.  
  We compared the result of a common 7'' source size for all transitions with the case of a gradually smaller size for the higher transitions. Starting with a source size of 7'' for the CS(2–1) transition, we assumed a 20% smaller size for the 3–2 and the 4–3 (i.e., 56) and an additionally 20% decrease in the size for the 5–4 and the 7–6 (i.e., 45). When comparing these cases with the case of a 7'' source size applied homogeneously to all the CS lines, it resulted a 15%–30% difference in rotational temperatures and less than a factor of 2 in column density.

As shown in both cases (see Table 2), though the column densities might be significantly affected by the assumed source extent, both the derived rotational temperatures and therefore the different temperature components described in this paper will be basically unaffected.

In all the cases dealt with above, the CS line ratios used for constraining the LVG model do change by no more than a factor of 1.5 (for the CS(7–6)/CS(4–3) line ratio). Considering the degeneracy of the LVG models, as noted in Section 3.2, we thus believe that these variations of line ratios are negligible and that the results presented in Table 3 still provide qualitative estimates of the properties of the gas emitting the CS lines.

It is well known that rotational diagrams give only approximate estimates of the kinetic temperature and the column density. To get more accurate values, we have ran LVG models (Goldreich & Kwan 1974; de Jong et al. 1975). These models are described in various papers and the version we use in this study is the one presented in Bayet et al. (2004, 2006, 2008a). The LVG model runs with three free parameters: the gas density n(H₂), the kinetic temperature (T_k), and the CS column density divided by the line width (N(¹²CS)/Δv). Using as model constraints the integrated line intensity ratios computed from Table 2, and corrected for beam dilution effect (as done in Section 3.1), we have investigated the following range of LVG input parameters: 5 K <T_k< 150 K, 1 × 10¹¹ cm⁻²(km s⁻¹)⁻¹ < N(¹²CS)/Δv ⁹<1 × 10¹⁶ cm⁻²(km s⁻¹)⁻¹ and 1 × 10⁴ cm⁻³ < n(H₂) < 1 × 10⁷ cm⁻³.

For determining the set of parameters (T_k, n(H₂), N(¹²CS)) able to reproduce best the observations, we have constrained the predicted CS line intensity ratios with the observed values via a reduced χ² method as performed in Bayet et al. (2004, 2006, 2008a), including 20% uncertainties on the observed integrated line intensities. Except for M 83 and Overlap, we have separated the model results into two: the low-temperature component constrained by, for example, line intensity ratios such as CS(3–2)/CS(2–1) or CS(4–3)/CS(2–1), and the high-temperature component characterized by, for example, line intensity ratios such as CS(5–4)/CS(4–3) or CS(7–6)/CS(4–3). The choice of the CS line intensity ratios used for calculating the χ² depends on the source and the available data set for CS (see Table 1). For M 83 and Overlap, since the CS data are well arranged alongside a line in the rotational diagram, only a single-temperature component model has been investigated using the available CS detections. The physical properties which reproduce best the observations are summarized in Table 3.

Unfortunately, as already shown by, e.g., Bayet et al. (2004, 2006) and Martín et al. (2005, 2006a), there is no unique solution for reproducing the observational line ratios of each gas component, in each galaxy. Indeed, it is well known that degeneracies in the physical parameter predictions exist in LVG models. To remove this degeneracy and obtain more accurate estimations of the physical parameters, one should correlate LVG model solutions obtained for various molecular species. However, this methodology is only effective if all species are emitted from the same gas phase (see, e.g., Martín et al. 2005, 2006a). For most of the sources observed here, no other tracers of very dense gas (n(H₂) ⩾10^5–6) have been detected so far. We thus did not cross correlate our LVG results with those from other molecular emission.