Free Access
Issue
A&A
Volume 560, December 2013
Article Number A82
Number of page(s) 13
Section Interstellar and circumstellar matter
DOI https://doi.org/10.1051/0004-6361/201322187
Published online 09 December 2013

© ESO, 2013

1. Introduction

Planetary nebulae (PNe) display a variety of morphologies, from purely spherical (see, e.g., Jacoby et al. 2010) to bipolar or even multipolar (Sahai & Trauger 1998). Non-spherical PNe are of particular interest, because the explanation of the process that shapes them is still under debate. There is growing evidence of highly collimated jets in evolved stars, which are believed to play a key role in the subsequent shaping of PNe (Sahai & Trauger 1998). For low- and intermediate-mass stars (0.8–8 M), a short-lived transition from a spherical to a collimated mass-loss at some point between the asymptotic giant branch (AGB) and the PNe stages is therefore expected. Several processes have been proposed to explain the nature of collimated ejections in these evolved stars, such as the magnetic launching from a single star (García-Segura et al. 2005) or a binary system (Soker 1998).

A crucial step in this study is identifying the sources that have just made the transition from spherical to collimated mass-loss. The group of objects often referred to as water fountain stars (hereafter WF) are probably the most appropriate candidates. WFs are evolved objects (mostly late AGB and post-AGB stars) with water maser emission tracing high-velocity motions, faster than the typical velocities seen in AGB mass-loss motions of 10–30 km s-1 (see, for example, Neri et al. 1998). Candidate objects to this class are usually identified by their broad velocity spread (≳50 km s-1) in their water maser spectra. Since the first identification of IRAS 16342-3814 (Likkel & Morris 1988), this novel group of sources has grown very little in number. A total of only 14 candidate WFs have been reported so far (Engels et al. 1986; Deacon et al. 2007; Deguchi et al. 2007; Suárez et al. 2008, 2009; Day et al. 2010; Gómez et al. 2011).

When water maser emission in WFs is observed with interferometers, it seems to trace highly collimated jets, with dynamical ages as young as <100 yr (Boboltz & Marvel 2007; Imai 2007a; Day et al. 2010). In IRAS 18286-0959, the water masers are tracing a double-helix structure (Yung et al. 2011). In all cases, these recent ejections represent one of the earliest known manifestations of collimated mass-loss in evolved stars.

WFs are thought to possess a thick, expanding circumstellar envelope (CSE), that was expelled during the AGB phase. The maser emission in these sources may be produced when a newly produced jet strikes into the CSE (Imai 2007a). This scenario explains the highly collimated jets of OH and H2O with velocities higher than 100 km s-1 (Boboltz & Marvel 2007; Day et al. 2010; Gómez et al. 2011) and very short dynamical time-scales. After some decades, the tip of the jet is expected to have reached the outer regions of the CSE, where physical conditions for maser pumping are no longer met. The thick CSE in WF is evident by their high obscuration in optical wavelengths.

The modelling of the broad-band spectral energy distribution (SED) of the thermal dust emission may give important clues about the envelope in WFs and also about the presence of disks (Durán-Rojas et al., in prep.). SED modelling may, in principle, provide hints about the circumstellar mass, the luminosity of the system (e.g., including both the envelope and the disk), and the physical structure (density and temperature profiles). Unfortunately, SED modelling is still hampered by some free parameters that are poorly constrained due to the complexity of the sources (star, envelope, disks, jets) and the lack of knowledge of some global parameters.

The detection and detailed study of line emission from thermal gas is a key in providing additional valuable information about the physical characteristics of CSEs, mainly about the total mass, the mass-loss rate and the global kinematics. CO and 13CO spectra have been widely observed in AGB and post-AGB stars; modelling them has been used to determine the mass-loss rate, the total mass, and other physical parameters (see, for example, Teyssier et al. 2006). So far, IRAS 16342-3814 is the only WF where thermal line emission has been unambiguously reported (He et al. 2008; Imai et al. 2009). From the detection of the CO and (tentatively) 13CO J = 2 → 1 lines, He et al. (2008) have determined an expansion velocity of 46 km s-1, higher than that expected in an AGB envelope (typically 15 km s-1). Surprisingly, the CO J = 3 → 2 line reported recently (Imai et al. 2012) shows a different kinematics, depicting an even broader profile (>200 km s-1). These high-velocity dispersions have been interpreted by the kinematics related to the collimated ejection also traced by water masers, instead of with the expanding motions of the CSE.

Imai et al. (2009) illustrated the difficulty of properly identifying CO associated with WFs when using single-dish observations. Most of these objects are at low Galactic latitude and therefore multiple Galactic foreground and background CO components are also gathered within the telescope beam. It is necessary to take spectra offset from the target position, and identify whether there are CO components that are present only at the target position, and at a velocity close to that of the star. Imai et al. (2009) reported a tentative detection of CO J = 3 → 2 emission associated with IRAS 18286-0959, since it was only present in spectra taken towards this source. However, its large velocity offset (≃25 km s-1) from the central velocity of the jet and the narrow width of the line precluded these authors to ascertain the association of the CO emission with the WF.

We conducted sensitive CO and 13CO observations of ten WFs, involving the J = 1 → 0 and 2 → 1 transitions, using the 30m IRAM radio telescope at Pico Veleta. Our goal was to survey the largest possible number of WFs to obtain new reliable detections of thermal line emission associated with these objects, and hence to derive some global physical parameters of the emitting regions.

Table 1

Sources observed.

2. Observations and strategy

The observations were carried out using the 30 m radio telescope at Pico Veleta, Spain, during two runs in June 2009 and June 2010. The new Eight MIxer Receiver (EMIR) was used for all the observations. The focal plane geometry of this instrument allowed us to gather 3 mm and 1 mm data simultaneously in two linear polarizations. In this case, we selected the simultaneous observations of CO J = 1 → 0 (115.271 GHz) and J = 2 → 1 (230.538 GHz) in one receiver configuration, and 13CO J = 1 → 0 (110.201 GHz) and J = 2 → 1 (220.399 GHz) in a second setup.

Both the VESPA and WILMA autocorrelators were connected as backends at all the observed lines. With VESPA, a frequency spacing of 312 kHz and 1.25 MHz was employed in the 3 mm and 1 mm lines. This is equivalent to 0.81 and 0.85 km s-1 at the CO and 13CO J = 1 → 0 frequencies; at the J = 2 → 1 lines, the velocity spacings are twice as high. The theoretical bandwidth employed was 160 MHz and 320 MHz at 3 mm and 1 mm, which provided an actual velocity coverage of ~360 km s-1 in all the lines. On the other hand, each module of the WILMA autocorrelator provided 1 GHz of bandwidth and a fixed spectral resolution of 2 MHz. The wide instantaneous bandwidth of EMIR (8 GHz in SSB at 3 mm) allowed us to tune the local oscillator to include the 13CO J = 1 → 0 line in the CO setup. Therefore, we used another WILMA module to cross-check the 13CO J = 1 → 0 line emission within the CO setup, albeit with a poor velocity resolution (5.4 km s-1).

The observations were made under typical winter conditions (precipitable water vapour column close to 4 mm), registering atmospheric opacities at 225 GHz between 0.1 and 0.4. Typical system temperatures were 110–450 K at 3 mm, and 160–600 K at 1 mm. The resulting rms (three-sigma) were between 8 mK and 92 mK at 3 mm, and between 26 mK and 250 mK at 1 mm.

The antenna temperature scale was calibrated every 10–12 min by the standard chopper wheel method, that is, by the sequential observation of hot and cold loads, and the blank sky; hot and cold load temperatures were room- and liquid-nitrogen temperatures, respectively. Sky attenuation was determined in real time from the values of a weather station, the measurement of the sky emissivity, and an atmosphere numerical model available at the observatory. Furthermore, we regularly observed a set of standard line calibrators (Mauersberger et al. 1989), and estimated an uncertainty lower than 20% in all the observed lines.

All the intensities throughout the paper are made with respect to a scale of main-beam temperature (TMB), assuming main-beam efficiencies of 0.75 and 0.52 at 3 mm and 1 mm. Velocities are given with respect to the LSR.

We surveyed all the WFs known in 2010 that are visible from Pico Veleta. The sample includes ten sources, detailed in Table 1. In the table we included other common names for the sources, their equatorial coordinates (J2000), distances, Galactic latitudes, and approximate velocities with respect to the LSR. The two distances quoted for IRAS 18113-2503 are kinematic distances derived in this work, using the Galactic rotation model of Reid et al. (2009).

All the observed sources are located in the Galactic disk, and most of them at Galactic latitudes as low as |    b    |  < 1°. Therefore, the CO and 13CO lines are expected to be heavily contaminated by Galactic background or foreground emission that originates in local gas and spiral arm clouds along the line of sight.

Therefore, we selected position-switching as the observing mode and were especially careful in selecting an appropriate reference position. For each source, we tried different reference positions by moving away from the Galactic plane, until we obtained clean spectra, that is, without line emission arising from the reference position.

However, a position used for reference that is too distant usually produces a worsening in the baselines, and we had to compromise. Typical distances to the reference positions were 1000′′.

To associate a velocity component to a WF, we proceeded with a strategy of five-points crosses, that is, by observing towards the star position, and another four points located 24″ away from it, in the east-west and north-south directions. This separation of 24″ corresponds approximately to the half-power beam width (HPBW) at 3 mm, and twice the HPBW at 1 mm. Due to Galactic foreground/background gas clumps, one can expect significant variations in the emission along 24″. However, if we saw a velocity component present ONLY at the star position, this was suspected to be associated to the WF, and was analysed in detail.

When a WF source had no CO emission at the star position or only a narrow component far away from the stellar velocity, it was excluded from the five-points strategy. Because of time restrictions, we were unable to observe the 13CO setup in IRAS 16552-3050 and OH12.8-0.9; in these sources, we were only able to use the J = 1 → 0 spectra from WILMA, which were observed simultaneously with the 12CO setup (as explained above).

Data reduction and analysis were made using the GILDAS software1.

3. Results

3.1. Overview

In Figs. 1 to 4 the spectra towards the star position are depicted, corresponding to CO J = 1 → 0, 13CO J = 1 → 0, CO J = 2 → 1, and 13CO J = 2 → 1. In all cases but the 13CO J = 1 → 0 line of IRAS 16552-3050 and OH12.8-0.9 (Fig. 2), spectra are taken from the VESPA autocorrelator; for non-detections, a smoothing of three channels was applied to the spectra. Line emission, with multiple velocity components, is present in most of the spectra, but it mainly arises in foreground and background Galactic clouds. Notable exceptions are IRAS 16552-3050, IRAS 18113-2503 (both without CO emission), and IRAS 19134+2131 (with only a narrow component), which are the sources with the highest Galactic latitude in our sample. Table 2 provides the upper limits of the CO and 13CO velocity-integrated temperatures in these cases, assuming a line width of 40 km s-1.

Table 2

Upper limits of the undetected sources.

thumbnail Fig. 1

CO J = 1 → 0 spectra towards the ten WFs surveyed. The (shortened) source name is indicated at the top of each spectrum. Note the different intensity ranges in the panels. All the spectra span 200 km s-1 of coverage to facilitate line width comparisons among the sources. Spectra corresponding to IRAS 16552-3050 and IRAS 18113-2503 have been smoothed to three times the original velocity spacing. Red vertical bars are located at approximately the stellar velocity to facilitate additional association. Most of these velocity components arise from background or foreground Galactic clouds, as discussed in the text.

Open with DEXTER

thumbnail Fig. 2

Same as Fig. 1 for the 13CO J = 1 → 0 line. Spectra from IRAS 16552-3050 and OH12.8-0.9 correspond to the WILMA autocorrelator. Spectra corresponding to IRAS 18113-2503 and IRAS 19134+2131 have been smoothed to three times the original velocity spacing.

Open with DEXTER

thumbnail Fig. 3

Same as Figs. 1 and 2 for the CO J = 2 → 1 line. Spectra corresponding to IRAS 16552-3050 and IRAS 18113-2503 have been smoothed to three times the original velocity spacing.

Open with DEXTER

thumbnail Fig. 4

Same as Figs. 1–3 for the 13CO J = 2 → 1 line. The sources IRAS 16552-3050 and OH 12.8-0.9 have not been observed in this transition. Spectra corresponding to IRAS 18113-2503 and IRAS 19134+2131 have been smoothed to three times the original velocity spacing.

Open with DEXTER

The foreground and background molecular emission makes it difficult to ascertain the presence of emission associated with our target sources. Therefore, special care must be taken in finding this association. We propose a set of three requirements that a particular velocity component must fulfil for it to be considered as probably associated with the WFs: (1) it is present only at the star position; (2) it persists at least in two of the observed lines; and (3) it is centred as closely as possible at the stellar velocity.

These three criteria are independent of any additional interpretation of the geometry and kinematics of the CO-emitting region in WFs, and we consider them as indispensable requirements. In addition to these considerations, another reliability restriction can be imposed if we assume that CO emission arises in an expanding circumstellar shell, or in the region where this shell interacts with the stellar jet, or even in the inner disk. Therefore, we expect lines to have widths consistent with expanding motions found in AGB envelopes, or larger (≃15 km s-1).

After a thorough analysis of the five-points pattern of all sources and all emission lines, we identified two strong candidates that might possess thermal gas associated with them: IRAS 18460-0151 and IRAS 18596+0315; as we see in the next section, broad components are detected towards these sources. In a third case, IRAS 18286-0959, we discovered a pair of narrow CO velocity components, symmetrically located with respect to the mean OH and H2O maser velocities (Sevenster et al. 1997; Imai et al. 2013a). While this source does not meet the three detection criteria we imposed, we cannot discard that the CO emission is associated with this WF, although with circumstellar characteristics different from those of the other two detections. We therefore include this source as a possible (albeit very tentative) detection in the following discussion.

3.2. Individual sources

3.2.1. IRAS 18460-0151

This is the clearest detection reported in this work and, if confirmed, would be the second WF known to show associated thermal line emission, after IRAS 16342-3814 (He et al. 2008; Imai et al. 2009, 2012). Figure 5 depicts all the final spectra towards this source; for each of the observed lines, a cross of panels shows the spectra at the five positions observed. To facilitate the identification of the associated feature, a zoom in intensity and velocity was applied.

thumbnail Fig. 5

All spectra observed towards IRAS 18460+0151. Each cross of five spectra corresponds to one of the CO lines. The relative locations of the spectra indicate the relative positions of antenna pointings; the centre towards the WF and four offset positions, 24′′ away from the WF. The lines are indicated on top of each cross. The velocity range from 115 to 145 km s-1 has been shaded to facilitate additional comparison. As discussed in the text, this velocity range depicts significant emission only at the star position.

Open with DEXTER

A component of about 40 km s-1 width was detected in the four lines observed, and only towards the star position. In addition, this component is centred close to the stellar velocity, inferred from the double-peaked OH spectrum (te Lintel Hekkert et al. 1989; Engels & Jiménez-Esteban 2007) and also the H2O emission (Deguchi et al. 2007). To emphasize the feature, the velocity range (115,145) km s-1has been shaded in Fig. 5. This is the only velocity range where the above mentioned characteristics are found for this source.

The contrast between the star and the off-source positions is more notable in the two J = 2 → 1 lines. In contrast, the 13CO J = 1 → 0 line shows a velocity component significantly narrower than the other lines.

The line ratio of CO (2–1)/(1–0) is higher than 2. This is an unusually high value for extended insterstellar clouds (see, e.g., Sakamoto et al. 1995), but it is more typical of a compact, unresolved source, which reinforces the association of this component to the CSE of the WF. The CO(2–1)/13CO(2–1) ratio is remarkably low (between 2 and 3). This is also uncommon for an interstellar cloud, but similar values have been found in the WF IRAS 16342-3814 (He et al. 2008; Imai et al. 2012) and some AGB and post-AGB envelopes (see, for example, Bujarrabal et al. 2001, 2005; Teyssier et al. 2006). This point is discussed in Sect. 4, where we analyse the emission and estimates of physical parameters.

3.2.2. IRAS 18596+0315

This case is quite similar to IRAS 18460-0151, although the associated CO and 13CO components are weaker than those of that source. Figure 6 displays all the final spectra for this source, similar to the previous figure. Again, a zoom in intensity and velocity was applied to emphasize the feature of interest.

thumbnail Fig. 6

Same as Fig. 5 for IRAS 18596+0315. This is similar to IRAS 18460+0151 and presents a wide component only at the WF position. The shaded area reaches from 90 to 105 km s-1.

Open with DEXTER

In the velocity range (90,105) km s-1 (shaded in the figure), there is significant emission only towards the source, and none elsewhere. As in the previous case, this is noted in all the spectral lines observed in the survey. The whole range of emission of this feature is indeed wider, as we show in the discussion. As in IRAS 18460-0151, the 13CO J = 1 → 0 line is the less evident, while the two J = 2 → 1 lines are the clearest.

The line ratio of CO (2–1)/(1–0) is unusually high (around 2), as in IRAS 18460-0151. The CO/13CO observed ratios are not as low as in IRAS 18460-0151; they vary between 3 and 5, depending on the J → (J − 1) used for the computation. The implications of this difference is also discussed below.

3.2.3. IRAS 18286-0959

This case is clearly different from the previous ones, because we did not find a wide velocity component. However, we did detect a pair of narrow lines, symmetrically located with respect to the stellar velocity. Figure 7 depicts all the final spectra towards this source, similar to Figs. 5 and 6. Unfortunately, we dealt with strong contamination in the offset positions of the two 13CO lines at one of the velocity ranges of interest. For this reason, Fig. 7 contains only the CO lines at the five positions observed.

thumbnail Fig. 7

Same as Figs. 5 and 6 for IRAS 18286-959. In this case, a different pattern is proposed in the CO and 13CO associated with the WF. Two narrow velocity components, symmetrically located with respect to the stellar velocity, are found with a higher intensity towards the star position.

Open with DEXTER

As was previously pointed out by Imai et al. (2009), the CO spectrum of IRAS 18286-0959 is rich in narrow components, due surely to its location in the Galactic plane. Nonetheless, two of these narrow components are particularly intense towards the star position and are weaker elsewhere. These two components lie at ~7 km s-1 and 66 km s-1, and are partly shaded in Fig. 7 to facilitate comparison of the intensities among the five positions. The association of these narrow components to the WF is very tentative, but worth analysing to the extent that these data allow (see also the discussion by Imai et al. 2009).

Stellar velocities inferred from maser lines – due to their own nature and circumstellar origin – may have large uncertainties of up to 20 km s-1 (te Lintel Hekkert et al. 1989). Even taking into account this potentially large uncertainty, the mean velocities of the OH and H2O masers are probably not far from the stellar velocity, and the CO mean velocity is also close to it. Another important factor is that the line profile is contaminated by the contribution of foreground and background clouds, especially the blueshifted component.

Imai et al. (2009, their Fig. 2c) detected three velocity components in the CO J = 3 → 2 line towards the star position, after subtracting the contribution of the off-source positions; the approximate velocities of these components are 10 km s-1, 40 km s-1, and 65 km s-1. In our data, the 40 km s-1 component is not particularly intense at the star position, being noted only in the J = 2 → 1 line.

In the other two components, the observed (2–1)/(1–0) line ratio is lower than one. As we discuss below, this difference with respect to IRAS 18460-0151 and IRAS 18596+0315 is a signature of different physical and excitation conditions.

4. Analysis and discussion

4.1. Possible origin of the detected CO

To analyse and characterize the molecular gas associated with the three WFs that may show circumstellar CO emission, we isolated the intrinsic CO velocity components by subtracting synthesized off-source spectra (obtained, in turn, as the average of the four outer spectra of each source). The results are presented in Fig. 8 for the three sources and for all the observed lines.

thumbnail Fig. 8

Final spectra in the three cases reported as probable detections of CO and 13CO gas in WFs. These spectra are the result of subtracting synthesized off-source from on-source spectra for each source and transition. The superimposed red lines are fits as circumstellar shells (for IRAS 18460+0151 and IRAS 18596+0315) or Gaussian profiles (for IRAS 18286-0959). Each isotope and transition has been fitted independently (see text).

Open with DEXTER

The wide component found in IRAS 18460-0151 and IRAS 18596+0315 strongly resembles those of the circumstellar envelopes found in different stages of late-type stars, such as planetary nebulae (see, for instance, Bujarrabal et al. 2005) and post-AGB stars (Bujarrabal et al. 2001; Teyssier et al. 2006). Wide components of 30–40 km s-1 are commonlly associated with the previous AGB envelopes of these objects.

In the case of IRAS 18286-0959 the velocity components are narrower than in the other two cases, and the emission (if it is associated with the source) probably arises from a smaller, more concentrated region; there are also some examples of circumstellar envelopes that show this double-peaked profile, although – to the best of our knowledge – in C-rich evolved stars only (Yamamura et al. 1993; Bujarrabal & Cernicharo 1994). By increasing the degree of speculation, we can explore whether the CO arises from a thin dense shell that surrounds a bipolar cavity, as in the model of IRAS 16342-3814 made by Imai et al. (2012). The water jets recently studied by Yung et al. (2011) are organized in a double-helix structure, which may produce a pair of symmetric bullets, which are detected in CO.

The weaknesses of the lines (except IRAS 18286-0959), the unresolved nature of the emission, and the lack of detailed information about the geometry and physics of the envelopes prevent a more detailed modelling. Nonetheless, first-order approximations were made by fitting the spectra. The shell method available in GILDAS was employed for IRAS 18460-0151 and IRAS 18596+0315. This method fits horn-type profiles for circumstellar envelopes, providing as outputs (1) the area of the spectrum; (2) Vpk, the central velocity; (3) the expansion velocity (Vexp), deduced from the full width at zero-level; and (4) sτ, a shape parameter that is measured as the horn-to-centre ratio, which depends on the line opacity and takes values from − 1 (for optically thick lines, parabolic shape) to infinity (for double-peaked, optically thin lines); values close to zero are interpreted as optically thin, flat-topped lines. The method assumes uniform physical conditions and does not take into account possible asymmetries due to excitation or clumping. For IRAS 18286-0959, this fitting method did not converge and we performed Gaussian fitting of the two components independently.

Table 3

CO and 13CO parameters.

Table 3 shows the results of the fitting, and the thick red lines of Fig. 8 plot them superposed on the observed spectra. The fittings are sufficient as a first approximation. The mean velocities agree very well with the assumed systemic stellar velocity deduced from the masers. In IRAS 18460-0151 and IRAS 18596+0315, the two CO lines and the J = 2 → 1 line of 13CO are optically thick, while the J = 1 → 0 line of 13CO is optically thin in both sources. In any case, the fitting for the last line is the least reliable of all, particularly for IRAS 18596+0315. Some hints of possible asymmetries are also noted and are worth to be studied in detail during follow-up observations.

Average expansion velocities (weighted by the inverse square of the errors) are 20.1 km s-1 and 17.7 km s-1 for IRAS 18460-0151 and IRAS 18596+0315. These velocities are consistent with an AGB origin of the molecular gas, as mentioned before.

From Table 3 we can also quantify some important line ratios. The CO and 13CO (2–1)/(1–0) ratio varies between 2 and 3 in IRAS 18460-0151 and IRAS 18596+0315, except for 13CO in IRAS 18460-0151; this unusually high line ratio is due to the low velocity-integrated temperature measured in the 13CO J = 1 → 0 line. For IRAS 18286-0959, the CO (2–1)/ (1–0) ratios are 0.4 and 0.6 for the blue and red components; these values, more typical of diffuse clouds, indicate different excitation conditions in this source.

Another interesting reading of the Table 3 may be made by computing the CO/13CO line ratio in IRAS 18460-0151 and IRAS 18596+0315 independently for each J transition. In all cases except for the J = 1 → 0 case of IRAS 18460-0151 these ratios are extremely low, between 2.6 and 3.3. It is worth noting that Imai et al. (2012) have measured similar values in IRAS 16342-3814. This low line ratio may be interpreted in terms of high opacity of the CO line, or a real decrease of the isotopic ratio produced in specific events of the star evolution, such as the first dredge-up at the beginning of the RGB (Karakas & Lattanzio 2007), or the deep mixing at the end of the same phase (Eggleton et al. 2008). This last hypothesis, first discussed by He et al. (2008), is supported by the fact that a low 12CO/13CO line ratio is very common in PNe (Bujarrabal et al. 2005) and proto-PNe (Bujarrabal et al. 2001).

The question arises why CO is detected in some WFs (IRAS 18460-0151, IRAS 18596+0315, and probably IRAS 18286-0959), while in others (e.g., IRAS 16552-3050, IRAS 18113-2503, or IRAS 19134+2131) it is clearly absent. In the remaining four sources (IRAS 18043-2116, OH12.8-0.9, W43A, and IRAS 19190+1102), Galactic contamination prevents any conclusive association or non-association of CO to the WFs, because in these cases there are intense spectral components in the stellar velocity range. Unfortunately, the few detections and unambiguous non-detections preclude any statistically significant discussion, and all suggestions must needs be highly speculative. We do not see any correlation of detectability with obvious physical parameters, such as the source distance or the envelope masses (estimated by Durán-Rojas et al., in prep.). It is possible that CO is detected in WFs where the abundance of this molecule is enhanced.

To explore this possibility, we investigated the morphology of the water maser emission where high angular resolution was available. In the sources without a clear detection of CO emission, the water masers trace only bipolar jets, with well-separated redshifted and blueshifted components (Imai et al. 2007b; Suárez et al. 2008; Gómez et al. 2011), while in IRAS 18460-0151 there is a central maser structure tracing low-collimating mass loss. If shocks enhance CO abundance in the gas phase (due to chemical reactions or evaporation of ices), the detection of CO lines would be favoured in sources with current episodes of low-collimation mass-loss (in addition to the collimated jets), where shocked regions would occupy a large solid angle. However, the proper evaluation of this possibility would require interferometric observations, to determine whether CO emission is co-located with water masers.

4.2. Physical parameters

Aiming to shed more light on the origin of the detected CO and 13CO, and also to estimate the physical parameters of the envelopes, we have proceeded with non-LTE radiative modelling, using our own code of the Large Velocity Gradient (LVG) approach. For a given set of physical conditions – mainly the kinetic temperature Tk, H2 volume density n(H2), line width and CO column density N(CO) – this code iterates and computes the population of the rotational levels, and predicts some line intensities. More details of the LVG methodology is provided in Appendix A.

The code did not provide satisfactory results for IRAS 18286-0959 for either of the two components. The low (2–1)/(1–0) line ratio that is observed could only be reproduced at n(H2) values well below 102 cm-3, typical of interstellar clouds, and N(CO) abnormally high. Nonetheless, and regarding our observational findings and those of Imai et al. (2009), this case should not be ruled out because the CO is clearly contaminated by foreground or background Galactic emission.

For the other two sources, a summary of the best-fit results is shown in Table 4. For IRAS 18460-0151, the LVG approach does not provide any solution for Tk above 50 K. For Tk = 10 K the solutions are optically thick, and for the range 20–50 K the line emission is optically thin (or moderately thick for Tk = 20 K). The case of IRAS 18596+0315, for a distance of 4.6 kpc, is remarkably similar to IRAS 18460-0151. In both cases, the total envelope masses are between ~0.2 and 0.4 M, the mean densities around 104 cm-3, and the mass-loss rates of the order of 10-4   M yr-1. The CO emission in IRAS 18596+0315 is more opaque than in IRAS 18460-0151. The 12C/13C ratio – computed as the ratio between the column densities of 12CO and 13CO – is low in both cases, although in IRAS 18460-0151 it is not exceptionally low.

Table 4

Results from the LVG modelling of IRAS 18460-0151 and IRAS 18596+0315.

The line profiles (Fig. 8) and their corresponding shell fitting (Table 3) show that the envelope is characterized by moderate opacities. Taking this into account, the lower values of Tk should be favoured. Another strong argument that favours low Tk is provided after comparing the CO opacities of different transitions; the (2–1)/(1–0) opacity ratios are unacceptably high for the highest Tk, because they exceed the observed line ratios by far. The low values of Tk are consistent with a scenario in which CO arises from the outer parts of the envelope.

Low temperatures also characterize the CO emission in most of the known post-AGB stars (Bujarrabal et al. 2001, and references therein). Furthermore, the values of the mass envelopes and the mass-loss rates agree very well with those derived previously (Knapp & Morris 1985; Bujarrabal et al. 2001, 2005; De Beck et al. 2010). The high mass-loss rates derived from the LVG fitting are compatible with the mass-loss experienced by the AGB stars with the highest masses (Bloecker 1995). Thus, the original masses of these sources are estimated to be in the range of 4–8 M.

The resulting parameters derived for the far kinematic distance of IRAS 18596+0315 (8.8 kpc) is, however, dramatically different with respect to the others. Firstly, the range of possible values of Tk is more restrictive, in the range 30–50 K. Secondly, due to a higher N(CO), the mass grows to 2 M, n(H2) is close to 105 cm-3, and the mass-loss rate is on the order of 10-3   M yr-1. This case is indeed puzzling. The envelope mass is very large (although compatible with that obtained by Durán-Rojas et al., in prep.), and the mass-loss rate reaches a value that is too high within current models (Bloecker 1995), and has not been observed in evolved stars with a low- and intermediate-mass progenitor. Unless something really unique occurs in this source, this result favour the nearest kinematic distance for this source, around 4.6 kpc.

Finally, a word of caution is given with respect to the LVG estimates. We made strong assumptions about the size of the envelope and CO abundance. These assumptions and the uncertainty of the distances may produce variations of the parameters of a factor of two or even larger. The structure of CSEs in WFs is indeed more complex, due to the presence of shocks, bipolar outflows, or episodic mass-loss events. The complex heating and cooling processes due to excitation of molecules contribute to its complexity.

5. Conclusions

From this survey, we have discovered CO emission associated with two new WF sources: IRAS 18460-0151 and IRAS 18596+0315. A third case, IRAS 18286-0959, is reported as tentative. These CO detections in WFs add new cases to IRAS 16342-3814, the only one previously known (He et al. 2008), and are the most promising candidates in which to study the molecular gas around WFs.

The wide component found in IRAS 18460-0151 was interpreted as the envelope of the former AGB stage. By non-LTE radiative modelling, we computed some physical parameters. The total envelope masses are around 0.2 M, the mass-loss rates are of the order of 10-4   M yr-1, and the CO emission is moderately thick. The kinetic temperatures derived are rather low, in the range from 10 K to 50 K. Taking into account the moderate opacities derived from the line profiles, we favour the low values of Tk, which is also consistent with a scenario where the CO arises from the cold, outer parts of the envelopes.

For IRAS 18596+0315 we also discovered a wide velocity component. For the near kinematic distance (4.6 kpc), the LVG results were similar to those of IRAS 18460-0151. For the far kinematic distance (8.8 kpc), however, we predict higher kinetic temperatures (30–50 K), large envelope masses (≃2 M), and mass-loss rates one order of magnitude higher. These last values are hard to explain with the usual parameters found in other AGB envelopes, or with theoretical considerations (Bloecker 1995).

For IRAS 18286-0959, we found two narrow velocity components, symmetrically located with respect to the stellar velocity, ~30  km s-1 apart. Previous observations of the same source at the J = 3 → 2 transition (Imai et al. 2009) failed to detect these components, but reported another narrow component close to the stellar velocity. Line contamination by Galactic background affects the computation of parameters and prevents further conclusions. Nonetheless, it is possible that the CO detected arises from somewhere in the jets traced by the water maser at 22 GHz, and this case deserve a more detailed analysis in the future. We speculate whether CO is detected in WFs in which low-collimated mass-loss enhances the CO abundance in a relatively large area of the envelope.

The detection and study of circumstellar envelopes around WFs are the key to understanding how the mass ejections took place during the last stages of evolution of these objects. The high mass-loss rates derived from our CO data indicate that we are seeing the mass ejected at the end of the AGB phase, and that these objects derive from relatively massive (4–8 M) progenitors. Sensitive, high-resolution observations are the natural follow-up of this work, and they may show how mass is ejected at the very end of the AGB phase, and its influence in shaping multipolar planetary nebulae.


1

GILDAS is a radio-astronomy software developed by IRAM. See http://www.iram.fr/IRAMFR/GILDAS/

Acknowledgments

J.R.R. acknowledges support from MICINN (Spain) grants CSD2009-00038, AYA2009-07304, and AYA2012-32032. J.F.G., M.O., O.S. and CD-R acknowledge support from MICINN grants AYA2008-06189-C03-01 and AYA2011-30228-C03-01, co-funded with FEDER funds. L.F.M. acknowledges support from MICINN grant AYA2011-30228-C03-01, also co-funded with FEDER funds. J.F.G. and M.O. are also supported by Junta de Andalucía. The authors wish to thank Pico Veleta’s staff for their kind and professional support during the observations. The careful reading and useful comments of the anonymous referee are also acknowledged, which certainly improved the paper.

References

Appendix A: LVG modelling

The LVG approach is a method that solves for different geometries the radiative transfer equations and the level populations iteratively and without assuming local thermodynamical equilibrium. It is based on the escape probability method first introduced by Sobolev (1960). This methodology has been and is widely used for determining the physical parameters of regions observed through their molecular rotational lines. An increasingly popular online version of this method is RADEX (van der Tak et al. 2007), which we also used in this work for cross-checking.

The methodology consists of decoupling the radiative transfer from the level population trough introducing a parameter β that measures the probability of a photon to escape the cloud. For an uniform sphere, it is computed by (A.1)where τ is the optical depth of a given transition. For an optically thin cloud, τ → 0 and therefore β → 1, and all the photons escape the cloud. In contrast, β becomes lower for an optically thick cloud, with a limit value of zero.

We modelled a uniform spherical cloud with an arbitrary radius rout = 104 AU, typical of AGB envelopes. A sphere of this size, located at the distances quoted in Table 1, would subtend at most some arcsec in the sky at the observed frequencies; therefore, we ran the code for a single point that concentrated all the emission from the envelope. The crude geometry that we had to introduce is, however, sufficient to provide first-order estimates of some global parameters, and is useful to constrain the theoretical models under development (Durán-Rojas et al., in prep.).

We corrected the observed line intensities by the beam-filling factor (which is different for each source and frequency). A CO abundance with respect to H2 (XCO) of 10-4 and a H2 mass abundance of 90% were assumed. The collision partners are H2 and He, and the collisional coefficients used were taken from Green & Thaddeus (1976). We note that, although reasonable, these assumptions certainly constitute an oversimplification and would translate into significant uncertainties in our results. The uncertainties quoted in Table 4 (one-sigma) for N(CO) and n(H2) are the result of the LVG model, considering only the uncertainties in the fitting (Table 3). For the total mass and the mass-loss rate, the uncertainties were computed by the standard formula of propagation errors.

We ran the LVG code independently for the CO and 13CO species and for different values of Tk (from 10 K to 100 K, in steps of 10 K). For each temperature, we computed the radiative transfer for a grid of n(H2) and N(CO), trying to find the best fit to the observed line intensities and the (2–1)/(1–0) line ratios. After determining the column densities, the opacities and envelope masses were obtained. Figure A.1 shows an example of the fitting, where the contours represent the observed J = 1 → 0 line intensity and the (2–1)/(1–0) line ratio; the position where the two sets of contours intersect determines the most probable value of N(CO) and n(H2) for this value of Tk. The uncertainties in N(CO) and n(H2) are also obtained from this fitting.

thumbnail Fig. A.1

Example of the fitting using the LVG approach. The contours correspond to the observed values of the J = 1 → 0 line intensity (red) and the (2–1)/(1–0) line ratio (green), together with their corresponding errors.

Open with DEXTER

For IRAS 18596+0315, the procedure was applied twice, once for each of the possible kinematic distances (Table 1); for IRAS 18286-0959, we also ran the code twice, once for each velocity component.

The mass was computed by integrating N(CO) over the projected disk: (A.2)where mH is the mass of the hydrogen atom, and the factor 1.1 accounts for the mentioned H2 mass abundance.

Finally, the mass-loss rate was computed directly from the derived n(H2), with the only assumption of a stationary mass loss, and following the equation of continuity: (A.3)We remark that the mass-loss rate computed here is not the current mass-loss of the sources (that are probably in the post-AGB phase), but are the mass lost in the former AGB phase, during which the envelope was expelled.

All Tables

Table 1

Sources observed.

Table 2

Upper limits of the undetected sources.

Table 3

CO and 13CO parameters.

Table 4

Results from the LVG modelling of IRAS 18460-0151 and IRAS 18596+0315.

All Figures

thumbnail Fig. 1

CO J = 1 → 0 spectra towards the ten WFs surveyed. The (shortened) source name is indicated at the top of each spectrum. Note the different intensity ranges in the panels. All the spectra span 200 km s-1 of coverage to facilitate line width comparisons among the sources. Spectra corresponding to IRAS 16552-3050 and IRAS 18113-2503 have been smoothed to three times the original velocity spacing. Red vertical bars are located at approximately the stellar velocity to facilitate additional association. Most of these velocity components arise from background or foreground Galactic clouds, as discussed in the text.

Open with DEXTER
In the text
thumbnail Fig. 2

Same as Fig. 1 for the 13CO J = 1 → 0 line. Spectra from IRAS 16552-3050 and OH12.8-0.9 correspond to the WILMA autocorrelator. Spectra corresponding to IRAS 18113-2503 and IRAS 19134+2131 have been smoothed to three times the original velocity spacing.

Open with DEXTER
In the text
thumbnail Fig. 3

Same as Figs. 1 and 2 for the CO J = 2 → 1 line. Spectra corresponding to IRAS 16552-3050 and IRAS 18113-2503 have been smoothed to three times the original velocity spacing.

Open with DEXTER
In the text
thumbnail Fig. 4

Same as Figs. 1–3 for the 13CO J = 2 → 1 line. The sources IRAS 16552-3050 and OH 12.8-0.9 have not been observed in this transition. Spectra corresponding to IRAS 18113-2503 and IRAS 19134+2131 have been smoothed to three times the original velocity spacing.

Open with DEXTER
In the text
thumbnail Fig. 5

All spectra observed towards IRAS 18460+0151. Each cross of five spectra corresponds to one of the CO lines. The relative locations of the spectra indicate the relative positions of antenna pointings; the centre towards the WF and four offset positions, 24′′ away from the WF. The lines are indicated on top of each cross. The velocity range from 115 to 145 km s-1 has been shaded to facilitate additional comparison. As discussed in the text, this velocity range depicts significant emission only at the star position.

Open with DEXTER
In the text
thumbnail Fig. 6

Same as Fig. 5 for IRAS 18596+0315. This is similar to IRAS 18460+0151 and presents a wide component only at the WF position. The shaded area reaches from 90 to 105 km s-1.

Open with DEXTER
In the text
thumbnail Fig. 7

Same as Figs. 5 and 6 for IRAS 18286-959. In this case, a different pattern is proposed in the CO and 13CO associated with the WF. Two narrow velocity components, symmetrically located with respect to the stellar velocity, are found with a higher intensity towards the star position.

Open with DEXTER
In the text
thumbnail Fig. 8

Final spectra in the three cases reported as probable detections of CO and 13CO gas in WFs. These spectra are the result of subtracting synthesized off-source from on-source spectra for each source and transition. The superimposed red lines are fits as circumstellar shells (for IRAS 18460+0151 and IRAS 18596+0315) or Gaussian profiles (for IRAS 18286-0959). Each isotope and transition has been fitted independently (see text).

Open with DEXTER
In the text
thumbnail Fig. A.1

Example of the fitting using the LVG approach. The contours correspond to the observed values of the J = 1 → 0 line intensity (red) and the (2–1)/(1–0) line ratio (green), together with their corresponding errors.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.