Revealing the transition from post-AGB stars to planetary nebulae: non-thermal and thermal radio continuum observations

Abstract

Midcourse Space eXperiment and Infrared Astronomical Satellite colour diagnostics as well as OH maser profile characteristics were used to select a sample of post-asymptotic giant branch (pAGB) candidates for a radio continuum detection experiment with the Australia Telescope Compact Array. Seven out of 28 sources, six of which are new detections, show a continuum. A planetary nebula serendipitously detected in the field of an undetected pAGB candidate also reveals radio continuum. The radio continuum properties of these eight sources are described. Almost half have non-thermal emission. dusty modelling of the infrared spectral energy distributions (SEDs) of the three strongest detections reveals that they all have central stars with temperatures substantially lower than that required for significant photoionization, leading us to infer that the radio continuum has arisen from wind–shock interactions. This hypothesis is consistent with the detection of non-thermal radio emission in one of these three objects.

masers, stars: AGB and post-AGB, stars: late-type, stars: winds, outflows, radio continuum: stars

1 INTRODUCTION

1.1 Evolution to planetary nebula

Planetary nebulae (PN) form from ∼0.8 to 8 M_⊙ main sequence (MS) mass stars. After leaving the MS, the star ascends the asymptotic giant branch (AGB), a phase of a few hundred thousand years during which typically 30 per cent of the central object's mass is shed, powered by pulsations that are in turn driven by nuclear shell burning, and resulting in the formation of the circumstellar envelope (CSE). On leaving the AGB, in the post-AGB (pAGB) or protoplanetary nebula1 (PPN) phase, the CSE detaches from the progenitor and a fast wind is launched that ionizes the surrounding, remnant AGB ejecta, either through uv-photoionization (when the central star has evolved to be hot enough) or though wind–shock interactions. This fast wind sculpts the material that will eventually be the PN, but the mechanism driving the fast wind remains unclear, as does the evolutionary phase of its onset.

The shape eventually displayed by the ionized PN shell, which may be spherical, elliptical, multipolar/axisymmetric or irregular, is a product of the unique nuances of its evolutionary path. While many of the details have yet to be observationally verified, the available models demonstrate the potential of magnetic fields (e.g. Frank 2005) to replicate axisymmetry, by sculpting the outflowing winds. Near infrared (NIR) imaging (e.g. Gledhill et al. 2001), and OH and H₂O maser studies (e.g. Deacon et al. 2007) indicate that the shaping commences at the latest in the immediate pAGB phase. In a limited number of pAGB objects, milli-Gauss magnetic fields have been measured from OH and H₂O maser observations (e.g. Bains et al. 2003, 2004; Vlemmings, Diamond & Imai 2006). For some PNe, aspherical geometries may be due to binary interactions. However, the fraction of evolved stars in close binary systems is probably too small to explain the high percentage of sources with strong asymmetries. For pAGB stars with OH masers, a binary origin of the asymmetries appears unlikely since the maser emission would be strongly disrupted (Deacon et al. 2007).

In the last decade or so, an increase in studies of pAGB stars has arisen due to both the realization of the significance of this evolutionary phase and the implementation of high-resolution observations courtesy of instruments such as the Hubble Space Telescope (e.g. Ueta, Murakawa & Meixner 2007), Very Large Telescope (e.g. Lagadec et al. 2007) and Very Long Baseline Array (e.g. Desmurs et al. 2007). However, we may narrow the evolutionary phase of interest still further. The key to elucidating the nature and onset of the PN shaping mechanism is potentially to be found in the study of the hybrid ‘transition’ objects; the subset of rapidly evolving stars found within the PPN group that display characteristics commonly found in the AGB and PN phases, but are usually deemed to be mutually exclusive – such as, in the O-rich case, OH and/or H₂O maser emission alongside ionized (continuum and recombination line) emission. Even if the masers can survive the disruption to their coherence length wrought by the onslaught of the fast wind, once the central star becomes hot enough to provide a source of ionizing photons, it is only a matter of time before the masing molecules suffer dissociation. The concurrent existence of both maser emission and radio continuum emission narrows the window of the evolutionary time frame within which chemistry permits such objects to exist. Given the rapid evolution associated with the transition phase, it is unsurprising that only a few such objects are known to inhabit this window (e.g. Vy 2–2, NGC 6302, K3-35, V1018 Sco; Seaquist & Davis 1983; Payne, Phillips & Terzian 1988; Aaquist 1993; Cohen, Parker & Chapman 2005) and the ‘OHPN’-classified objects of Zijlstra et al. (1989), yet these form a crucial link in the evolutionary sequence of evolved stars. Further to these OH-emitting pAGB stars and PN, recent detection of H₂O masers in PN has been found by, for example, Suárez, Gómez & Morata (2007) and Gómez et al. (2008).

1.2 PPN candidate diagnostics

As follow-up to the Australia Telescope Compact Array/Very Large Array (ATCA/VLA) 1612-MHz OH maser survey for Galactic OH/IR sources by Sevenster et al. (1997a,b, 2001), Sevenster (2002a,b) showed that the infrared (IR) colours, courtesy of Infrared Astronomical Satellite (IRAS) and Midcourse Space eXperiment (MSX) data, could be used to identify pAGB stars within the survey. In the ‘original’IRAS colour–colour diagram, van der Veen & Habing (1988) showed that AGB stars follow a clear path through the [12–25], [25–60] plane.2 The right-hand IRAS (RI) sources, found to the right of the AGB path, are traditionally associated with pAGB stars whose thick CSEs have a 25-μm excess; this is the ‘standard’ pAGB region. Sevenster (2002b) determined that the IRAS characteristics of the ATCA/VLA survey objects also produced a left-hand IRAS (LI) region to the left of the AGB track, of sources with a 60-μm excess such as that which may arise due to large amounts of cool dust created by objects with high mass-loss rates. Sevenster (2002b) found that statistically, the LI sample represents higher mass stars with high mass-loss rates, whilst the RIs represent lower mass and lower mass-loss objects. Furthermore, by plotting an MSX colour–colour diagram using MSX fluxes at 8, 12, 15 and 21 μm, this time with colours [8–12] and [15–21] defining the plane, she defined four diagnostic MSX quadrants: Quad I containing late pAGB stars, Quad II containing star-forming regions (SFRs), Quad III containing AGB and LI sources and Quad IV containing early pAGB stars. Sevenster (2002b) postulated that the LI sources may be the precursors to bipolar PN, with high outflow velocities, although this has not been observationally verified.

With the aim of investigating the onset and development of pAGB stellar wind asymmetries, Deacon, Chapman & Green (2004) observed the OH 1612-, 1665- and 1667-MHz maser profiles of 85 objects that comprise the subset of the ATCA/VLA survey sources which correspond to likely pAGB candidates. The selection criteria for the sample of Deacon et al. (2004) were those objects found in the IRAS RI or LI groups and/or the MSX Quads I and IV (as shown in Fig. 1). They characterized the OH maser profiles they obtained according to whether they show signs of originating in spherical shells or from asymmetric winds and/or bipolar outflows. For example, a double-peaked 1612-MHz OH profile with a peak-to-peak flux density ratio <8 likely indicates an AGB star with a maser shell that is still largely spherical, whereas broader double-winged profiles likely indicate bipolar outflows (see also e.g. Zijlstra et al. 2001). Furthermore, Deacon et al. (2004) found that irregular profiles, particularly those evident in the OH mainlines that exist at smaller stellar radii than the 1612-MHz masers (e.g. Chapman & Cohen 1986; Chapman, Cohen & Saikia 1991; Zijlstra et al. 2001), can indicate early departures from spherical symmetry in the inner CSEs. They found that, based on MSX colours, three times more late-post AGB sources (Quad I) show signs of aspherical outflows in their OH maser spectra compared with early pAGB sources (Quad IV).

Figure 1

Adapted from Deacon et al. (2004). Top: new IRAS two-colour diagram for 85 sources in the ATCA/VLA survey. The solid line is the evolutionary sequence for AGB stars defined by van der Veen & Habing (1988). The regions marked ‘LI’ and ‘RI’ are associated with pAGB stars (Sevenster 2002a,b). Three sources originally included in the full Deacon sample of 88 sources, but later identified as SFRs, are excluded. The small circles show sources not searched for radio continuum emission; open squares denote stars searched unsuccessfully; filled squares show stars detected in the continuum; filled triangle represents the location of V1018 Sco. The six pAGB stars we detected in the continuum are labelled. Upper limits to flux densities are indicated by arrowheads in appropriate directions, including the situation when both wavelengths have only upper limits to flux densities. Bottom: new MSX two-colour diagram for sources in the ATCA/VLA survey. Following Sevenster (2002a), this is divided into four quadrants according to minima which are seen in bimodal distributions of the MSX IR colours [8–12] and [15–21]. Quadrants I–IV are thought to contain in turn: late pAGB stars, SFRs, AGB stars and early pAGB stars. Symbols are as in Fig. 1(a).

This maser profile classification, taken with the MSX and IRAS colour–colour diagram diagnostics, provides evidence that some of the pAGB sources in the sample of Deacon et al. (2004) are experiencing the onset of the pAGB fast winds. In this paper, we report the results of using the ATCA to observe a subsample of the Deacon et al. sources at radio continuum wavelengths, to search for evidence of ionization and to verify and further constrain the diagnostic tools. For three sources where ionization is detected, we model the IR SED to infer whether the central star is sufficiently hot to provide a source of ionizing photons or if wind interactions are the more likely cause of the ionization. Radio continuum observations provide an excellent means of detecting such sources during a phase of evolution characterized by a small angular size and hence a high intrinsic dust density, resulting in heavy extinction at optical wavelengths.

1.3 ATCA sample source selection

We selected 30 objects from the Deacon et al. sample for follow-up radio continuum observations with the ATCA. The sources chosen belong to either the IRAS LI or MSX Quad I categories shown in Fig. 1. These categories were chosen as being perhaps the most likely to show radio continuum emission.

The LIs are thought to be massive stars at the very top of the AGB sequence and these may create localized wind–shock interactions that result in continuum emission (e.g. as in IRAS 16342−3814 and M1-92; Bujarrabal et al. 1998; Zijlstra et al. 2001). The sample included 15 LI sources selected from the 30 LI sources in Deacon et al. (2004). These authors found that LI sources with OH mainline emission were located close to the AGB evolutionary track in the IRAS two-colour diagram, whereas ‘extreme’ LI objects found further from the AGB track do not show OH mainline emission. We selected the 15 ‘extreme’ LI sources without OH mainline emission on the basis that they were likely to be more massive and more evolved, and hence may show stronger radio continuum emission. Of the 15 LI sources, 11 show standard double-peaked OH 1612-MHz profiles indicating near-spherical outflows. Two show a single-peaked OH 1612-MHz profile and one is an ‘extreme’ double with a large peak–peak ratio. One LI source, b165, has an irregular OH 1612-MHz spectrum and this is highly unusual for an LI star.

The sample also included 15 sources with classifications as MSX Quad I sources. The Quad I sources are thought to be late pAGB objects in which the pAGB fast wind may have switched on. Six of the Quad I sources are also identified as IRAS RI sources, consistent with their status as pAGB stars. One source, b262 (IRAS 17574−2403), was later found to be a well-known H ii region; we consider it worthwhile to include it in this study however, as one of its raison d'etre is to study the strength of the colour and maser diagnostics in picking out evolving late-type objects and b262 represents a test case of an interloper masquerading in our diagnostic space as a pAGB star. This source is discussed in Section 5.3.1. The remaining 14 Quad I sources are all considered likely pAGB stars. Eight of the 14 Quad I (pAGB) members have 1665-MHz emission, with three double-peaked, four irregular and one single-peaked OH 1665-MHz profiles. As discussed by Deacon et al. (2004), irregular OH 1665-MHz profiles may indicate a possible onset of asymmetry in the inner CSE. Two of the irregular Quad I sources (d46 and b11) have OH emission that is stronger at 1665 MHz than at 1612 MHz, and this is another indicator of a highly evolved object with irregular outflows.

Fig. 1 also shows the source V1018 Sco. This was not part of this sample but has IRAS colours that place it just outside the defined LI region. Cohen et al. (2005) have shown that this source is remarkable in being a still pulsating long-period variable that has produced a PN.

Table 1 summarizes the 30 sources observed. Column 1 gives the identification from the ATCA/VLA survey; Column 2 the IRAS name; Column 3 the J2000 OH position, the survey position errors are typically 0.5 arcsec; Column 4 the dates of observation; Column 5 the on-source integration time in min; Column 6 the IRAS and/or MSX IR classification; Column 7 gives the OH profile classification at 1612-, 1665- and 1667-MHz profile from Deacon et al. (2004).

Table 1

Observing details: Columns are: (1) identification from the ATCA/VLA survey; (2) IRAS name; (3) and (4) OH maser position (Sevenster et al. 1997a,b); (5) observing date; (6) integration time; (7) IRAS (LI/RI) and/or MSX (Quad n) classification; (8) OH profile classification at 1612, 1665 and 1667 MHz (Deacon et al. 2004). D = double-peaked, De = extreme double, Dw = wide double, S = single peak, I = irregular.

ATCA/VLA	IRAS name	OH position (J2000)		Date	Time	IRAS/MSX class	OH profile class
ident		(hh:mm:ss dd:mm:ss)		(dd-04-05)	(min)		1612, 1665, 1667
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
d3	14341–6211	14 38 04.961	−62 24 46.83	01,04	129	Quad I, RI	D..
d39	15367–5420	15 40 38.278	−54 30 18.59	01,04	74	LI	S..
d46	15445–5449	15 48 19.498	−54 58 21.52	01,04	76	Quad I	I I I
d117	16372–4808	16 40 55.829	−48 13 58.34	01,04	66	Quad I	I..
b11	17150–3224	17 18 19.889	−32 27 21.96	01,04	33	Quad I, RI	S I D
d197	17151–3845	17 18 34.776	−38 48 57.37	01,04	32	Quad I	D S.
b14	17164–3226	17 19 40.776	−32 29 51.69	01,04	29	Quad I, RI	D D D
b15	17162–3751	17 19 42.067	−37 54 55.16	01,04	29	Quad I	I..
d200	17188–3838	17 22 18.780	−38 41 40.13	01,04	29	LI	De..
b25	17193–3564	17 22 42.634	−35 49 31.60	01,04	29	LI	D..
b30	17205–3556	17 23 57.403	−35 58 50.31	01,04	29	LI	D..
d202	17245–3951	17 28 04.637	−39 53 44.20	01,04	29	Quad I, RI	D D D
b44	17256–3258	17 28 55.751	−33 00 41.86	01,04	29	LI	S..
b106	17367–3134	17 39 57.498	−31 35 57.28	01,04	29	Quad I	D..
b133	17392–3020	17 42 30.517	−30 22 07.96	01,04	29	LI	D..
b155	17414–3108	17 44 39.771	−31 10 05.40	01,04	29	LI	D..
b165	17426–2804	17 45 46.658	−28 05 28.46	01,04	28	LI	I..
b262 (H ii)	17574–2403	18 00 30.391	−24 04 01.29	04	31	Quad I	I I I
b263	17576–2653	18 00 49.500	−26 53 12.52	04	30	Quad I, RI	I I.
b301	18052–2016	18 08 16.376	−20 16 11.51	04	30	LI	D..
v41	18076–1853	18 10 38.665	−18 52 58.08	04	30	Quad I	Dw I I
v56	18103–1738	18 13 20.240	−17 37 17.35	04	30	LI	D..
v132	18276–1431	18 30 30.676	−14 28 57.78	04	30	Quad I, RI	D D De
v146	18310–0806	18 33 49.578	−08 04 01.38	01,04	30	Quad I	D..
v149	18314–0900	18 34 11.303	−08 58 02.55	01,04	38	LI	D..
v154	18327–0715	18 35 29.202	−07 13 11.01	01,04	30	LI	D..
v162	18342–0655	18 36 57.999	−06 53 25.07	01,04	30	Quad I	D..
v172	18361–0647	18 38 50.529	−06 44 49.85	01,04	43	LI	D..
v212	18434–0202	18 46 05.779	−01 59 17.62	01,04	30	LI	D..
v268	18588+0428	19 01 20.032	+04 32 31.25	01,04	32	LI	D..

2 OBSERVATIONS AND DATA REDUCTION

The 30 PPN candidates were observed with the six telescopes of the ATCA during two ∼13-h observing runs on 2005 April 1 and 4. The observing details are given in Table 1. The 6.0A configuration of the array gave the maximum available resolutions of 2 and 1 arcsec, respectively, at the 4800- and 8640-MHz observing frequencies. The 3/6-cm receiver was used which permits simultaneous dual frequency observations; the observing bandwidth was 128 MHz for each. The target pointings used were those of the OH maser positions of Sevenster et al. (1997a), with an associated precision of typically 0.5 arcsec, displaced by −30 arcsec in declination to avoid any artefacts at the phase centre. The sources were position-switched, to optimize uv-coverage, with typically 15 min spent on-source per cut. The observations were phase-referenced to suitable nearby calibrators, with a phase reference scan typically every 15 min. The flux density scale was calibrated with reference to a scan on 1934–638 and has an absolute error of ∼10 per cent. Because of the narrow LST range of the sample, some of the later targets suffered from a poor beam shape owing to being available only at the end of the allocated run and in competition with a number of other targets (see Table 4).

Table 4

Table of non-detections. No radio emission was detected above the 3σ level at either observing frequency within a 1-arcmin radius of the OH maser position for these targets. Columns are: (1) IRAS name; (2) 1σ off-source noise level measured from the images (4800-MHz); (3) restoring beam size (4800-MHz); (4) and (5) are the same as (2) and (3) but for the 8640-MHz observations. Note that the data for IRAS 18434−0202 and IRAS 18588+0428 were unusable due to poor uv-coverage.

Name	4800-MHz		8640-MHz
	1σ (mJy/b)	beam size (arcsec × arcsec)	1σ (mJy/b)	beam size (arcsec × arcsec)
(1)	(2)	(3)	(4)	(5)
17151–3845	0.20	5.8 × 2.5	0.15	2.9 × 1.4
17164–3226	0.13	4.4 × 2.8	0.11	2.4 × 1.5
17162–3751	0.30	4.1 × 2.7	0.11	2.3 × 1.5
17188–3838	0.10	4.2 × 2.7	0.10	2.3 × 1.4
17193–3564	1.89	4.6 × 2.6	0.15	2.5 × 1.4
17205–3556	0.36	4.7 × 2.5	0.12	2.6 × 1.4
17245–3951	0.15	4.5 × 2.5	0.11	2.4 × 1.4
17256–3258	0.15	5.2 × 2.4	0.12	2.9 × 1.3
17367–3134	0.22	5.6 × 2.3	0.10	3.1 × 1.3
17392–3020	0.53	5.9 × 2.3	0.16	3.2 × 1.2
17414–3108	0.51	5.8 × 2.2	0.13	3.2 × 1.2
17426–2804	0.30	6.5 × 2.2	0.09	3.6 × 1.2
17576–2653	0.22	5.6 × 2.5	0.11	3.1 × 1.4
18052–2016	0.57	6.8 × 2.6	0.11	3.7 × 1.4
18076–1853	0.14	7.3 × 2.5	0.10	4.0 × 1.4
18103–1738	0.36	7.9 × 2.4	0.12	4.3 × 1.3
18276–1431	0.14	8.7 × 2.5	0.10	4.8 × 1.4
18310–0806	10.64	12.7 × 2.4	0.98	6.9 × 1.3
18314–0900	2.36	14.2 × 2.1	0.21	7.7 × 1.2
18327–0715	0.47	16.2 × 2.2	0.12	8.7 × 1.2
18342–0655	1.65	18.4 × 2.1	0.63	9.9 × 1.2
18434–0202	–	–	–	–
18588+0428	–	–	–	–

The data were reduced using the miriad package (Sault & Killeen 1998). Bad visibilities were edited out and phase reference solutions applied to calibrate out the effects of atmospheric and instrumental gains. As this is a detection experiment, the data were gridded with natural weighting to optimize sensitivity and a 4800 (8640) MHz cell size of 0.5 arcsec (0.25 arcsec) was used. Deconvolution was implemented with the clean algorithm. The images were convolved with a restoring beam, of maximum resolution typically 2.5 arcsec at 4800 MHz and 1.5 arcsec at 8640 MHz. The rms off-source noise level in the maps depends on the calibration solutions, integration time and the number and brightness of sources contained in the field and was typically a few tenths of a mJy beam⁻¹ at both frequencies. The phase-referenced ATCA positions at 8640 and 4800 MHz have an accuracy which varies in RA and Dec. as a function of the beamsize and is typically 2.5 and 1 arcsec in each of these directions, respectively.

2.1 Updated colour–colour diagrams

We have updated the IRAS and MSX flux densities that underpin the two-colour diagnostic diagrams in order to plot as many as possible of the Deacon sample of stars in both diagrams. Many of the original 88 stars lacked IRAS and/or MSX flux densities at some key wavelengths. We sought matches with the newer MSX Point Source Catalog (PSC) ver. 2.3 (Egan 2003), replacing the older ver. 1.2 data. We examined MSX images of any remaining stars for which even the PSC2.3 has no point source above 5σ, and performed aperture photometry to derive a flux density for fainter objects or an upper limit if no detection was made. For stars missing IRAS detections in the PSC we examined the Faint Source Reject Catalog (FSR) and found a number of additional detections. These include the important source, d46, for which no PSC match is found within 30 arcsec but the FSR source Z15444−5449 lies within 4 arcsec of this star and validates the photometry of a PSC object 36 arcsec away from the stellar position. We also derived weak detections and/or upper limits for some faint objects from the IRAS Sky Survey Atlas images. The resulting new two-colour plots appear as Fig. 1. We have tailored the sample and symbols appropriately for this paper. In both the IRAS and MSX planes, we present data for 85 OH/IR stars: the original Deacon sample of 88 objects without the three known H ii regions. We distinguish between stars not searched for radio continuum emission, those searched unsuccessfully and those in which continuum emission was detected, including V1018 Sco (Cohen et al. 2005, 2006). Upper limits are shown for any star without four MSX or three observed IRAS fluxes.

3 RADIO CONTINUUM IMAGES

In searching for radio continuum emission associated with the OH maser sources, we do not expect the two to be necessarily coincident. The OH 1612-MHz maser emission arises from within the AGB CSE, typically at distances of several hundred stellar radii, corresponding to an angular separation of around 1 arcsec. The positions of the OH 1612-MHz emission peaks, which are from the front and back of the CSE (uniform expansion assumed), are a good indicator of the stellar position. The IRAS far-IR emission is associated with a cool dust shell and originates close to, and centred on, the star; thus the IRAS position also indicates the stellar position, albeit with much less precision. On the other hand, the radio continuum emission is associated with hot gas and may extend well beyond the dust shell and OH masers.

The case of V1018 Sco (Cohen et al. 2005, 2006) provides a good example of a source whose OH 1612-MHz maser emission is distinctly offset from the ionized continuum. This object has a phase-lag distance of 3.2 kpc (with an uncertainty of 20 per cent; Cohen et al. 2005). The maser lies in the centre of an optically visible ionized ring of emission. Blobs of radio continuum emission were found just inside this ring, at positions some ∼10 arcsec offset from that of the OH.

The circumstellar shells of PPN typically subtend a few arcsec (e.g. table 3 of Sahai et al. 2007) with the largest being of the order of ∼30 arcsec (e.g. AFGL2688, the ‘Cygnus Egg’ nebula, at ∼1 kpc; Skinner et al. 1997). Our objects range in distance from 0.9 to 16 kpc (table 1.2 of Deacon 2005, and references therein). We use the OH/radio emission model of V1018 Sco and the dimensions of AFGL2688 to define reasonable dimensions within which a strong association of radio continuum detected in our maps can be made to the 1612-MHz OH masers of Deacon et al. (2004). We have considered any radio source in our maps within a 30-arcsec radius of the OH position and with a flux density in excess of the 3σ off-source noise level to be a strong association. We additionally verified that no extragalactic sources were present in the NASA/IPAC Extragalactic Data base (NED) at these positions.

In this manner, we have confidently detected radio emission at both 4800 and 8640 MHz associated with IRAS 15367−5420, IRAS 15445−5449 and IRAS 16372−4808. The radio continuum maps of these sources are shown in Fig. 2. We illustrate the positions of the OH maser and IRAS sources on the maps. We consider 3σ radio continuum detections with 30 arcsec < offset ≤1 arcmin and with no other known association to be tentative associations; IRAS 14341−6221, IRAS 17150−3224 and IRAS 18361−0647 fall into this category and maps of them are shown in Fig. 3. Note that the emission in the fields of IRAS 14341−6211 and IRAS 17150−3224 was detected at 4800 MHz only and so we do not show the 8640-MHz fields for these. Finally, in Fig. 4 we show the 4800- and 8640-MHz images of the SFR, IRAS 17574−2403 and the evolved object, PN G353.9+02.7, which was serendipitously detected in the field of IRAS 17164−3226 but which is over an arcmin away from the IRAS source. These are the first radio continuum images of all the aforementioned objects and the first radio continuum detections of all bar IRAS 15445−5449.

Figure 2

Images of the detected radio continuum sources strongly associated with an OH maser source, i.e. with offset <30 arcsec; (left) at 4800 MHz and (right) at 8640 MHz. The image sizes have been chosen to encompass the positions of the radio, OH maser and IRAS sources in each field. The OH maser positions of Deacon et al. (2004) are indicated by a white or black * and the partial ellipses shown in the IRAS 15445−5449 images trace a 30-arcsec radius from the OH maser positions. IRAS source positions are indicated by a hollow ★. Contour levels are given in square brackets next to each source name in the following text. Beam sizes are indicated at the bottom right of the plots. Top: IRAS 15367−5420. [3, 6, 9, 12 × 8.8e-5 Jy beam⁻¹ (4800 MHz) and 3, 6, 9 × 1.2e-4 Jy beam⁻¹ (8640 MHz)]. Middle: IRAS 15445−5449. [3, 6, 9, 12, 15 × 1.1e-3 Jy beam⁻¹ (4800 MHz) and × 4.7e-4 Jy beam⁻¹ (8640 MHz)]. Bottom: IRAS 16372−4808 [3, 6, 12, 24, 36, 48, 60, 72, 84, 96, 108, 120, 132, 144 × 9.7e-5 Jy beam⁻¹ (4800 MHz) and 3, 6, 12, 24, 36, 48, 60, 72, 84 × 2.5e-4 Jy beam⁻¹ (8640 MHz)].

Figure 3

As Fig. 2 for the tentative radio continuum-detected associations with OH maser offsets 30 arcsec < offset <1 arcmin. In each image, the radio continuum source is indicated by an arrow. The partial ellipses shown in some of the images trace a 1-arcmin radius from the OH maser positions. Top left: IRAS 14341−6211. [3, 4, 5 × 9.6e-5 Jy beam⁻¹ (4800 MHz)]. A radio continuum source is visible to the NNW of the maser and IRAS positions at 4800 MHz; there is no corresponding 8640-MHz emission (map not shown). Top right: IRAS 17150−3224. [3, 6 × 2e-4 Jy beam⁻¹ (4800 MHz)]. A radio continuum source is visible to the NNE of the maser position at 4800 MHz; there is no corresponding 8640-MHz source (map not shown). The cross indicates the OH maser position measured by Nyman et al. (1998) and the box shows the position of the ‘Cotton Candy Nebula’ (Kwok et al. 1998). Bottom: IRAS 18361−0647 [3 × 1.1e-3 Jy beam⁻¹ (4800 MHz) and 3, 6, 9 × 1.8e-4 Jy beam⁻¹ (8640 MHz)]. The detected radio source, to the ESE of the OH maser position, is brightest at 8640 MHz and is a 4σ detection at 4800 MHz. Other sources in the field are labelled as follows: the cross indicates the position of GLMP 810 (Garcia-Lario et al. 1997); a circle with cross identifies the location of GPSR 25.489-0.287.

Figure 4

As Fig. 2 for the other (non-pAGB or not associated with a OH maser) sources detected in the ATCA observations. Top: IRAS 17574−2403. [3, 6, 12, 24, 36, 48, 60 × 2.2e-2 Jy beam⁻¹ (4800 MHz) and 3, 6, 12, 24, 36, 48, 60, 72, 84, 96 × 1.2e-2 Jy beam⁻¹ (8640 MHz)]. After the observations, this was identified as an H ii region. Bottom: PN G353.9+02.7 [3, 6, 9, 12, 15, 18, 21, 24, 27, 30, 33, 36, 39, 42, 45 × 1.3e-4 Jy beam⁻¹ (4800 MHz) and 3, 6, 9, 12, 15, 18, 21 × 1.1e-4 Jy beam⁻¹ (8640 MHz)]. This source was serendipitously detected in the field of the IRAS 17164−3226 observations but is over an arcmin away and is not considered to be associated with either the OH maser or the IRAS source.

In order to measure the flux densities of the radio sources, we interactively boxed the emission above the 3σ off-source noise level using miriad task cgcurs. We measured the largest angular size of each object down to the 3σ level from its image. The three confidently detected sources are all partially resolved. In Table 2, we list their measured properties; all detections are in fact above 5σ, with the exception of IRAS 18361−0647 at 4800 MHz, which we include because of the brighter, coincident emission at 8640 MHz.

Table 2

Measured properties of detected radio continuum sources. Source properties were measured as detailed in the text. Columns are: (1) ATCA/VLA identity (2) IRAS source name; (3) and (4) position of peak of radio continuum emission; (5) offset of radio continuum source peak from OH maser position; (6) offset of radio continuum source peak from IRAS source position (for PN G353.9+02.7 the offset from IRAS 17164−3226 is given) (7) peak radio flux density; (8) total radio flux density; (9) 1σ off-source noise measured from map; (10) peak signal-to-noise ratio; (11) largest angular source size; (12) restoring beam size; (13) flag: S = strong association with OH source, offset between OH maser and continuum radio peak positions <30 arcsec; T = tentative association, 1 arcmin ≥ OH offset ≥ 30 arcsec; U = unlikely association, offset >1 arcmin.

Ident.	IRAS	Position (J2000)		Offset		Peak	f. d.	1σ	Peak snr	LAS	Beam	Flag
		(hh:mm:ss)	(dd:mm:ss)	OH (arcsec)	IRAS (arcsec)	(mJy/b)	(mJy)	(mJy/b)		(arcsec)	(arcsec × arcsec)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
4800-MHz detections
d3	14341–6221	14 38 02.94	−62 23 56.82	51.95	50.40	0.51	0.30	0.10	5	5	5.7 × 2.0	T
d39	15367–5420	15 40 37.65	−54 30 08.58	11.40	8.10	1.15	1.08	0.09	13	11.5	6.5 × 2.1	S
d46	15445–5449	15 48 19.38	−54 58 21.01	1.13	37.08	19.74	22.81	1.11	18	10	5.6 × 2.1	S
d117	16372–4808	16 40 55.83	−48 13 59.33	0.99	7.60	14.50	15.17	0.10	145	16	6.7 × 2.2	S
b11	17150–3224	17 18 20.80	−32 26 43.46	40.19	40.82	1.42	1.47	0.18	8	5.5	4.4 × 3.0	T
b262	17574–2403	18 00 30.43	−24 04 01.78	0.78	23.22	1480.86	2753.48	21.88	68	15	5.7 × 2.7	S
v172	18361–0647	18 38 52.94	−06 45 06.78	39.69	35.65	4.23	4.86	1.08	4	20	22.9 × 2.2	T
-	PN G353.9+02.7	17 19 45.44	−32 30 32.65	71.84	66.80	6.36	8.16	0.13	49	9.5	4.4 × 2.8	U
8640-MHz detections
d3	14341–6221	–	–	–	–	–	–	0.10	–	–	3.0 × 1.1	T
d39	15367–5420	15 40 37.67	−54 30 08.58	11.32	8.09	1.25	1.13	0.12	10	5	3.4 × 1.1	S
d46	15445–5449	15 48 19.38	−54 58 20.76	1.26	37.16	14.33	18.67	0.47	31	7	3.0 × 1.2	S
d117	16372–4808	16 40 55.80	−48 13 59.33	1.03	7.89	22.45	24.60	0.25	90	8.5	3.5 × 1.2	S
b11	17150–3224	–	–	–	–	–	–	0.15	–	–	2.8 × 1.7	T
b262	17574–2403	18 00 30.43	−24 04 01.28	0.60	23.26	1327.03	3882.59	11.58	115	10	3.2 × 1.5	S
v172	18361–0647	18 38 52.94	−06 45 05.28	39.07	35.13	1.82	2.50	0.18	10	20	12.4 × 1.2	T
-	PN G353.9+02.7	17 19 45.46	−32 30 32.40	71.92	66.89	2.41	4.54	0.11	22	5	2.4 × 1.5	U

In Table 3, we present derived properties for the detected sources. These include the spectral index α, defined according to S_ν∝ν^α. Uncertainties in flux densities were assigned as the larger of 10 per cent of the measured quantity and the 1σ off-source noise. We also give the distances to the sources taken from the literature, and calculate the corresponding linear offsets between the OH maser and radio continuum peak emission positions. Where radio continuum has been detected at both frequencies, we use the mean angular offset in the calculation. The distances given are mainly kinematic, except where more precise distances have been determined. The kinematic distances have an associated uncertainty that is typically in the range 20–60 per cent (e.g. Foster & MacWilliams 2006).

Table 3

Derived properties of radio continuum sources. Columns are: (1) IRAS source name; (2) RI/LI/Quad classification; (3) source type: pAGB, SFR or PN; (4) spectral index S_ν∝ν^α; (5) and (6) distance estimates, taken from the literature; kinematic distances are flagged as such in Column 9; where only one distance exists, we list it as a near distance; (7) and (8) corresponding linear offsets between the OH maser position and the peak radio continuum position for the near and far distances, respectively; (9) references for the distances.

Name	Class	Type	α	Dist.		Offset		Ref.
				Near	Far	Near	Far
				(kpc)		(m)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
14341–6211	Quad I, RI		–	1.7	10	1.3e16	7.8e16	(1) kin.
15367–5420	LI	pAGB	0.08 ± 0.21	7.0	–	1.2e16	–	(2) kin.
15445–5449	Quad I	pAGB	−0.34 ± 0.24	7.1	–	1.3e15	–	(1) kin.
16372–4808	Quad I	pAGB	0.82 ± 0.25	5.4	10	8.2e14	1.5e15	(1) kin.
17150–3224	Quad I, RI	pAGB	–	2.42	–	1.5e16	–	(3)
17574–2403	Quad I	SFR	0.58 ± 0.25	3.8	–	3.9e14	–	(4)
18361–0647	LI	pAGB	−1.13 ± 0.25	2.7	13	1.6e16	7.7e16	(5) kin.
PN G353.9+02.7	–	PN	−1.00 ± 0.25	5.5	11	5.9e16	1.2e17	(5) kin.

References in Column 9 are: (1) Ortiz et al. (2005) (2) Deacon (2005) (3)Meixner et al. (1999) (4) Fish et al. (2003) (5) Garcia-Lario et al. (1997).

In Table 4, we give the image statistics for the non-detections. The high off-source rms noise associated with IRAS 18310−0806, particularly at 4800 MHz, is due to the presence of a confusing source in the field, whose deconvolution is compromised by the poor beam shape resulting in the smearing of its emission over the field. The targets with the latest LSTs rose towards the end of the allocated run and hence all competed for limited observing time, resulting in poor beam shapes and high rms noises for these maps. IRAS 18434−0202 and IRAS 18588+0428 were particularly affected by poor uv-coverage and it was not possible to deconvolve these data.

4 DUST SHELLS

4.1 Infrared photometry

The SEDs of pAGB stars typically display a ‘double-humped’ profile across the optical and IR wavebands, as the dust shells have sufficiently detached to spectrally distinguish the shorter wavelength component due to the photosphere from the longer wavelength component due to the dust shell. In order to derive more properties of the three continuum radio-detected pAGB sources with strong associations to the OH masers (i.e. relative offset <30 arcsec) in the Deacon et al. (2004) sample, we searched the literature and the 2MASS, GLIMPSE-I, MSX and IRAS data bases for optical and IR photometric data that could be used to construct their SEDs. Table 5 details the photometry found, which is solely in the IR bands; no optical wavelength data are available for these sources. The data have not been dereddened; indeed, no extinction corrections were made. At the longest wavelengths, these are poorly known. Even worse is the difficulty in separating the contribution from the ISM along the line-of-sight to the objects from the probably dominant circumstellar obscuration (Section 4.2). The two components very likely have very different extinction properties (e.g. small grains in the ISM versus large grains in the CSE).

Table 5

Photometric data collated from the literature for the three radio continuum sources with strong associations to OH masers in the Deacon et al. (2004) sample. Columns are: (1) waveband; (2) central wavelength; (3)–(5) flux densities of the targets.

Band	λ0 (μm)	Flux density (Jy)
		15367	15445	16372
(1)	(2)	(3)	(4)	(5)
2MASS H	1.24	–	0.0008	0.0027
2MASS J	1.66	–	0.0034	0.0015
2MASS K	2.16	0.0087	0.0294	0.0073
GLIMPSE-I	3.55	0.088	0.104	0.79
GLIMPSE-I	4.49	0.228	0.228	2.08
GLIMPSE-I	5.73	0.453	0.240	3.46
GLIMPSE-I	7.84	0.901	0.442	9.50
MSX A	8.3	0.79	0.65	–
IRAS 12	12.0	1.85	6.88	2.04
MSX C	12.1	1.82	5.39	–
MSX D	14.7	2.82	13.67	–
MSX E	21.3	3.48	45.58	–
IRAS 25	25.0	6.05	87.00	14.30
IRAS 60	60.0	13.39	<1131	<178.3
IRAS 100	100.0	<158.0	<1254.0	<354.4

4.2 Circumstellar reddening

Three of our likely pAGB continuum emitters have 7.7–22.7 μm spectra in the IRAS Low Resolution Spectrometer (LRS) data base. The most prominent features are deep silicate absorption bands near 10 and 18 μm. Their 10-μm optical depths can be converted into line-of-sight extinctions on multiplying by 18.5 (Roche & Aitken 1984). In cases of extreme extinction, the 10-μm absorption is so large that the spectra fall to zero, e.g. for 18361–0647. From comparison of the ratios of optical depths at 10 and 18 μm for a sample of IRAS sources with silicate absorptions, we find a median of 5.3 for the ratio of τ(10 μm)/τ(18 μm). For 15445–5449, 17150–3224 and 18361–0647, we calculate total extinctions, A_V, of 33, 25 and 72 mag, respectively. Deacon et al. (2007) provide distances of 7.1, 2.4, and 2.7/13 (near/far) kpc for these stars. Values of extinction near the Galactic plane have been measured as a function of longitude (Arenou, Grenon & Gomez 1992) from which the expected interstellar extinctions towards our three stars should be ∼8, 5 and 11/55 mag. Consequently, each of these objects must suffer considerable circumstellar extinction. The local material that causes this reddening is presumably in the halo of debris ejected by the AGB stars' slow, but massive, winds. It is improbable that this dust has no associated gas. Therefore, one would expect that tens of magnitudes of dust absorption would be allied with substantial free–free optical depth in the radio continuum. Perhaps this explains the dearth of thermal radio continuum detections in our sample of 30 stars. Detection of non-thermal radio emission suggests that the emitting electrons arise at collisions between the fast and slow AGB winds that are observed predominantly on the front sides of the circumstellar shells (as in the case of V1018 Sco; Cohen et al. 2005, 2006) rather than deeply embedded in the AGB ejecta.

4.3 Modelling the SEDs

The publicly available one-dimensional radiative transfer dusty code of Ivezic, Nenkova & Elitzur (1999) was used to perform calculations for model systems that were compared with the IR photometric data (Table 5). The code is based on the scaling relations appropriate for spherically symmetric dust emission, and its inputs are: the central stellar spectrum (T_eff); the chemical composition of the dust and the size distribution of its grains; the temperature at the inner shell boundary (T_d); the dust density profile and relative shell thickness in terms of the inner and outer dust shell radii (Y=r_outer/r_inner) and the optical depth τ_λ at some fiducial wavelength λ. Various analytical dust density distributions are permitted, including an option for a dynamical calculation in the case of radiatively driven winds suitable for late-type stars. Once a model SED provides a satisfactory fit to the normalized data, in addition to providing fits to properties such as T_eff and T_d, scalable properties, viz. radius and mass-loss rate are also reported.

dusty was invoked for a single spherical shell model with a radiatively driven wind, such as is suitable for AGB and pAGB stars. A chemical composition comprised of cold silicates (‘Sil–Oc’; from Ossenkopf, Henning & Mathis 1992) was assumed – as is appropriate for oxygen-rich pAGB stars in which mass loss has ceased – with a constant, standard MRN (Mathis, Rumpl & Nordsieck 1977) power-law grain size distribution: n(a) ∝a^−q for a_min≤a≤a_max, with q= 3.5, a_min= 0.005 μm and a_max= 0.25 μm. No spectral classification data are available for any of the stars, so a single blackbody radiation source at the centre of the dust distribution was adopted in each case. With no observational data available in the literature to provide constraints other than the photometry in Table 5, our methodology for the modelling commenced with an initial broad sweep of the parameter space, within plausible limits. Initial estimates on the range of likely pAGB T_eff were made by considering the limiting cases of AGB stars and hot pAGB stars. Stars on the AGB have typical minimum temperatures of T_eff= 2000 K (e.g. Dehaes et al. 2007) while hot pAGB stars typically have a maximum of 35 000 K (e.g. Fujii, Nakada & Parthasarathy 2002; Gauba et al. 2003; Gauba & Parthasarathy 2004). Similarly, a reasonable range of T_d= 50–800 K was chosen (e.g. Fujii et al. 2002; Venkata Raman & Anandarao 2008). Increments of T_eff= 1000 K and T_d= 50 K were employed. Models were computed for optical depths at a fiducial wavelength of 0.55 μm (V band), at six logarithmic steps in the interval τ_{0.55 μm}= 5–300. As such a broad parameter space was searched, to optimize the run-time of the models we invoked the option for the density distribution to be solved analytically by the code, meaning that the only density information input to the models was the relative shell thickness, of which a value of 1000 was assumed.

dusty produces a synthetic, normalized SED which is scalable to the observed data. To measure the goodness of fit, the observed photometric data were normalized and the model photometry was interpolated to the wavelengths of the observed data points. However, for each source, the wavelength range sampled in the observed photometric data appeared to stop short of a definitive turnover in the SED. To ensure against a normalization error, in calculating the best fit to the data, each model was shifted up and down in flux space by a suitable number of logarithmic steps and a best fit for each model was calculated in this fashion. The overall best fit of all the models was deemed to be the one that gave a minimum in the sum of the squares of the deviations between the observed and computed photometry. The robustness of the fits was verified by examining the statistics to ensure that a reasonable number of models with similar T_eff, T_d and τ_{0.55 μm} converged around each best-fitting minimum. The modelling then continued iteratively for each source, narrowing the searched parameter space of the three variables (T_eff, T_d and τ_{0.55 μm}) in each iteration until convergence was deemed satisfactory once the best-fitting statistic reached an overall minimum and the increments in the three variables had been narrowed such that τ_{0.55 μm} was resolved to within ∼±20, T_eff to within ±100 K and T_d to within ∼± 10 K. The fewer the available observational constraints, the larger the number of free parameters and the more possible solutions there are, so pushing the modelling beyond this was not worthwhile. The best-fitting models are shown with the data in Fig. 5.

Figure 5

IR SEDs of IRAS 15367−5420 (top), 15445−5449 (middle) and 16372−4808 (bottom). The crosses show the photometric data from the IRAS, MSX, 2MASS and GLIMPSE-I archives, which were used to constrain the best-fitting dusty model (solid line). Upper limits to the flux densities in the IRAS wavebands are indicated where relevant. IRAS LRS data were available for 15445–5449 only and these are displayed as triangles on the relevant plot.

In addition to the scatter expected due to the inherent assumptions in dusty, some additional scatter in the data with respect to the fits is to be expected considering that the data have not been dereddened and also that they span a range of observing epochs. Given that the most massive pAGB stars may spend only a few tens of years in this phase (Blöcker 1995) and that the maximum time-span over which the photometric data were collected is of the order of ∼20 yr – which passed between the observations of the IRAS and GLIMPSE programmes – it is conceivable that the stars could have significantly evolved during this period. However, inspection of Fig. 5 reveals that the fits are satisfactory.

The longest wavelength IRAS data points, particularly those at 100 μm, are upper limits only. It is obvious from the plots in Fig. 5 that the spectra do not appear to well fit the 60- and 100-μm IRAS data points. Typically, AGB stars and PN have SEDs that peak at wavelengths between ∼25 and 60 μm. The models fall short of the 60- and 100-μm data points because these are likely to be contaminated due to the large IRAS footprint. The 100-μm data point (and to a lesser extent that at 60 μm) should be regarded as a gross upper limit and is included on the plots for completeness.

dusty can only admit locally homogeneous and spherically symmetric models and it is worth briefly considering how the models might be improved if these restrictions were removed. On a global scale, the fast wind switches on in the late AGB/early pAGB phase and some observational studies suggest that the majority of pAGB dust shells are bipolar (e.g. Gledhill 2005). On a local scale, the circumstellar matter is likely to have an inhomogeneous substructure of cold dense clumps in a warmer rarefied interclump medium (e.g. Redman et al. 2003). Both of these phenomena could allow the presence of a wider range of dust temperatures (see the modelling of Indebetouw et al. 2006). The fit at intermediate wavelengths (around 1 μm) would also be improved by accounting for deviations from spherical symmetry due to inhomogeneities, the presence of a disc or the beginnings of the development of a bipolar or other non-spherical morphology.

The following scaled properties (Ivezic et al. 1999) output by dusty are of most interest here: r_i, the shell inner radius; r_c, the radius of the central star; T_o, the temperature at the outer edge of the dust shell; the mass-loss rate, and v_t, the terminal outflowing wind velocity. r_i scales in proportion to the luminosity as L^1/2 and is calculated by dusty for L= 10⁴ L_⊙. pAGB stars have typical luminosities in the range 10³–10⁴ L_⊙ (e.g. van Winckel 2003, and references therein); we use these to calculate sizes for the inner radii. We use the distances given in Table 3 to calculate angular sizes for the dust shells. The mass-loss rate reported by dusty scales according to L^3/4(r_gdρ_s)^1/2, where dusty uses a gas-to-dust ratio r_gd of 200 and a bulk grain density of ρ_s= 3 g cm⁻³, is assumed. v_t scales as L^1/4(r_gdρ_s)^−1/2. Both and v_t have an inherent uncertainty of 30 per cent (Ivezic et al. 1999). We estimate the dynamical time since the stars left the tip of the AGB as ⁠, although the core will also have evolved in this time. In Table 6, we present the results for the best-fitting models; the values for the individual sources are discussed further in Section 5.1.

Table 6

Parameters of the best-fitting dusty models to the SEDs. Scale factors have been applied to the dusty outputs to obtain the values given, as detailed in the text. All fits used a cold silicate grain composition as taken from Ossenkopf et al. (1992) and a relative shell thickness of 1000. The columns are as follows: (1) name of the IRAS source; (2) effective temperature of the central star; (3) dust temperature at the inner shell radius; (4) optical depth at a fiducial wavelength of 0.55 μm; (5) dust temperature at the outer shell radius; (6) radius of the central star; (7) inner shell radius; (8) angle subtended by the outer shell radius at the source distance given in Table 3; (9) mass-loss rate; (10) terminal velocity and (11) dynamical time since the dust shell detached (i.e. since departure from the tip of the AGB). Columns (9) and (10) have an inherent uncertainty of 30 per cent (Ivezic et al. 1999). The calculations are given for stars of L= 10³ and L= 10⁴ L_⊙ (e.g. van Winckel 2003). Note that IRAS 16372−4808 has two values for angular size because it has two kinematical distance estimates. If v_t < 5 km s⁻¹ then dusty's wind solution model is no longer reliable (Ivezic et al. 1999) and measurements of and t_dyn have been flagged with a ^★ accordingly to reflect this.

IRAS source	Teff (K)	Td (K)	τ0.55 μm	To (K)	rc (cm)	ri (cm)	θ (arcsec)	(M⊙ yr−1)	vt (km s−1)	tdyn (yr)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
L= 103 L⊙
Model inputs					Model outputs
15367–5420	3900+100−100	225+5−5	123+7−7	11	4.8e12	2.8e15	27.0	3.8e-5★	<5	1598★
15445–5449	11900+100−100	190+10−10	73+11−10	8	5.2e11	7.1e15	67.0	3.9e-5★	<5	5960★
16372–4808	14400+100−100	630+10−10	68+27−19	21	3.5e11	6.9e14	8.6,4.6	1.4e-5★	<5	151
L= 104 L⊙
Model inputs					Model outputs
15367–5420	3900+100−100	225+5−5	123+7−7	11	1.5e13	9.0e15	85.5	2.2e-4★	<5	2842★
15445–5449	11900+100−100	190+10−10	73+11−10	8	1.6e12	2.3e16	211.8	2.2e-4★	<5	10598★
16372–4808	14400+100−100	630+10−10	68+27−19	21	1.1e12	2.2e15	27.1,14.6	7.6e-5	5.2	268

Calculated values of v_t < 5 km s⁻¹ render both v_t and unreliable (Ivezic et al. 1999). Values of r_gd for AGB and pAGB stars used in the literature range from around 50 (Surendiranath, Parthasarathy & Varghese 2002) to 360 (Knapp, Jorissen & Young 1997), although the latter is for a carbon-rich object (V Hya). As we have no information on the gas-to-dust ratio in our objects, we adopt r_gd= 100 for our calculations and scale the relevant output from dusty accordingly. However, this results in all of our v_t and calculations being unreliable, bar that for IRAS 16372−4808 with L= 10⁴ L_⊙. For our stars, typical pAGB CSE bulk expansion velocities of∼10–20 km s⁻¹ (e.g. Bains et al. 2003, 2004) require unfeasibly low values of r_gd and/or ρ_s using these calculations which are based on radiatively driven expansion. It is likely that for transitional objects, these formulae for v_t and may not apply, and other mechanisms act in tandem with the radiation pressure to drive the shell expansion, such as the predicted, collimated, fast wind.

5 NOTES ON INDIVIDUAL SOURCES

In Section 5.1, we discuss the properties of the three continuum radio-detected objects strongly associated with OH maser sources in the Deacon et al. (2004) sample by virtue of their proximity (IRAS 15367−5420, 15445−5449 and 16372−4808) and in Section 5.2, we discuss the three tentative detections (IRAS 14341−6211, 17150−3224 and 18361−0647). In Section 5.3, we describe those of the two other sources (IRAS 17574−2403 and PN G353.9+02.7) detected in the observations, as their radio continuum images have not been published elsewhere. In Section 5.4, we mention sources in the sample that are elsewhere classified as radio continuum emitters but were not detected in these observations.

5.1 Radio continuum sources strongly associated with OH maser sources

5.1.1 IRAS 15367−5420

The source shows a single OH 1612-MHz emission peak at −101.5 km s⁻¹ and was not detected at 1665 or 1667 MHz (Deacon et al. 2004). Deacon et al. (2007) searched for 22-GHz H₂O maser emission in this object but found none. The 1612-MHz OH profile, whilst uncharacteristic of that expected from the expanding shell of an OH/IR star, does show some evidence of a wing on the redder side, suggesting that it could arise from the blueshifted ‘side’ of the CSE and the redshifted ‘side’ may be absorbed or simply absent due to an inhomogeneously filled shell. In this respect, it is similar to the prototypical OH-containing young PN Vy2–2 (e.g. Seaquist & Davis 1983), which also displays blueshifted emission only.

IRAS 15367−5420 was suggested to be a PN by Preite-Martinez (1988) based on its IRAS colours and flux qualities. However, this source was not detected in the 6-cm radio continuum observations of van de Steene & Pottasch (1993) down to an rms level of ∼0.5 mJy beam⁻¹ (with a beam width of ∼4.5 arcsec). The authors used the IRAS position which is offset by ∼8 arcsec south of the radio continuum source we have detected but our flux level is only about 1 mJy, less than three times their rms noise level.

The 11-arcsec offset of the radio continuum emission from the OH maser corresponds to a linear extent of 1.2 × 10¹⁶ m at the 7.0 kpc kinematic distance of Deacon (2005). This would place IRAS 15367−5420 at the large end of the size distribution expected for pAGB stars; alternatively the distance could be incorrect and it could actually be much closer.

This object has a thermal spectral index of 0.08, consistent with that expected for an inhomogeneous photoionized H ii region around an evolving progenitor. However, dusty modelling provides a best-fitting SED with a stellar temperature of T_eff= 3900 K. Photoionization of CSEs begins once a significant fraction of the photons emitted from the central star have λ < 912 Å, which corresponds to T_eff > 30 000 K. This suggests that the source of the radio continuum emission is more likely to be a localized wind–shock interaction, rather than an expanding photoionized region due to an evolving progenitor. The derived and t_dyn for this source are unreliable as v_t < 5 km s ⁻¹ for calculated models in the range L= 10³–10⁴ L_⊙. An unfeasibly low gas-to-dust ratio of 15 is required in order for the 10⁴ L_⊙ model to cross this threshold, which suggests that for a typical pAGB CSE outflow velocity to be reached, a calculation incorporating factors in addition to radiation pressure to drive the dust shell expansion needs to be invoked (Section 4.3).

5.1.2 IRAS 15445−5449

In the literature, IRAS 15445−5449 only features in the maser studies of Sevenster et al. (1997a,b) and Deacon et al. (2004, 2007). It possesses OH masers at 1612, 1665 and 1667 MHz with irregular, variable (brightness increasing) spectral profiles comprised of broad features and velocity-contiguous emission which spans up to 77 km s⁻¹, i.e. the full apparent velocity extent of the shell. In a further peculiarity, these profiles are each centred on a different velocity. The mainline and 1720-MHz OH spectra show absorption features due to intervening dust clouds (Deacon et al. 2004). Deacon et al. (2007) searched for 22-GHz H₂O and 86-GHz SiO maser emission in this object and discovered the former but not the latter. The H₂O masers are variable and their velocity extent is redshifted with respect to the OH profiles.

In their paper on the non-thermal emission in V1018 Sco, Cohen et al. (2005) report a spectral index of about −0.8 for IRAS 15445−5449 too, based on unpublished data by Sevenster & Chapman. These observations were made in 1998 November at 3, 6 and 13 cm and yielded flux densities of 11.0 ± 0.2, 17.5 ± 1, and 30.0 ± 2 mJy. These three measurements imply a spectral index of −0.77 ± 0.11. Yet our own observations in 2005 April yield −0.34 ± 0.24. Evidently, the non-thermal emission from this star varies substantially in both flux density and spectral index.

The best-fitting dusty model provides T_eff= 11 900 K, a value which again implies a star which is insufficiently evolved to be providing an adequate source of ionizing photons. This, coupled with the non-thermal spectral index, is strongly suggestive of the presence of wind–shock interactions. Based on its peculiar OH masers, the coexistence of H₂O maser and radio continuum emission and the non-thermal, variable behaviour of the latter, this appears to be a remarkable object with unique properties. Indeed, this is one of the few known evolved objects to display radio continuum emission in conjunction with all the masing ground state OH transitions, as well as that of H₂O; in this sense, it is reminiscent of the young PN, K3–35 (e.g. Gómez et al. 2006).

5.1.3 IRAS 16372−4808

In the literature, this source only features in the maser surveys of Sevenster et al. (1997a,b) and Deacon et al. (2004, 2007). Like IRAS 15367−5420, it possesses a singly peaked 1612-MHz maser spectral profile with evidence of a velocity wing on the redshifted side, suggesting that this peak may be the emission from the blueshifted ‘side’ of an expanding CSE whose redshifted counterpart is absorbed or absent. Deacon et al. (2004) and Deacon et al. (2007) searched for mainline OH and 22-GHz H₂O maser emission in IRAS 16372−4808 but found none. The existence of the masers is determined by environmental (temperature, density) conditions, which are a function of stellar radius. As late-type stars evolve past the end of the AGB and the fast wind turns on, the masers are thought to turn off sequentially in order of the radius they occupy: first SiO, then H₂O, followed by the OH mainlines and finally the 1612-MHz OH satellite line. This scenario is consistent with IRAS 16372−4808, which appears to have evolved into the fast wind stage (hence the radio continuum emission) and of all the O-rich maser species, only displays an irregular 1612-MHz profile. The lack of a redshifted peak in the profile could indicate that the velocity coherence of the masers on the redshifted ‘side’ of the shell has been disrupted by an aspherical fast wind. Alternatively, they may be continuum-absorbed or the OH shell may simply have an inhomogeneous density distribution.

This object has a thermal spectral index of 0.8. The best-fitting dusty model gives T_eff= 14 400 K for the central star, which is again below the threshold 30 000 K required for significant photoionization to occur. If the star is indeed towards the high-end of the pAGB luminosity distribution (i.e. ∼10⁴ L_⊙), then the mass-loss rate of 8 × 10⁻⁵ M_⊙ yr⁻¹ is consistent with that of the late/pAGB superwind and the star departed the AGB just ∼300 yr ago.

5.2 Radio continuum sources tentatively associated with OH maser sources

5.2.1 IRAS 14341−6211

IRAS 14341−6211 is listed as a possible PN in SIMBAD based solely on a search by Preite-Martinez (1988) for new candidates with adopted IRAS colours. There have been a great many such colour searches over years but there is no unique set of IRAS colours for PNe. Two separate boxes are required to accommodate both young and mature PNe in the IRAS[12]−[25], [25]−[60] plane (Walker et al. 1989).

It was not detected in the 6-cm radio continuum observations of van de Steene & Pottasch (1993) down to an rms level of ∼0.5 mJy beam⁻¹. We detect radio emission at 4800 MHz only, at a ∼52-arcsec offset from the positions of the OH maser and the IRAS source. A search of the SIMBAD and NED data bases does not reveal any other known sources coincident with our radio detection, so we tentatively associate it with the IRAS source. The non-detection of emission at 8640 MHz suggests that this source has a non-thermal spectral index. It displays a 1612-MHz OH maser spectrum characteristic of an expanding shell, but no OH mainline nor 22-GHz H₂O maser emission was detected (Deacon et al. 2004, 2007).

The optically visible nebula associated with IRAS 14341−6211 appears in finding chart images in Hu et al. (1993) and Suárez et al. (2006) and appears to be a few (∼2–5 arcsec) in extent, which suggests it is resolved and extended even though the seeing is quoted in neither publication. Hu et al. (1993) note that the nebulosity is very red, and this is confirmed by the faint and red continuum displayed in the optical spectrum of Suárez et al. (2006), which also displays no obvious emission lines.

It is interesting to note that its NIR photometry in the JHK bands decreased over the three epochs for which there are data in the literature, specifically in 1989, 1990 (Hu et al. 1993) and at the time of the 2MASS observations in the late-1990s.

5.2.2 IRAS 17150−3224

IRAS 17150−3224 is a well-known pAGB star (e.g. Hu et al. 1993; Hrivnak et al. 2006, and references therein) with a bipolar dust shell and concentric circumstellar dust arcs (Su et al. 2003). Its optical counterpart extends over 10 arcsec by 6 arcsec (Hu et al. 1993). It has a surrounding remnant molecular envelope from which its CO (Hu et al. 1993) and polarized (Szymczak & Gérard 2004) OH maser emission arises. Also known as the ‘Cotton Candy Nebula’ (Kwok, Su & Hrivnak 1998), the OH masers detected by Sevenster et al. (1997a) and Deacon et al. (2004) and the nebula are located within a few arcsec of each other (see Fig. 3), but offset from the position of the OH maser detected by Nyman, Hall & Olofsson (1998).

Deacon et al. (2007) searched for 22-GHz H₂O and 86-GHz SiO maser emission in this object and discovered neither. Offset by ∼39 arcsec from all the other sources in the field, radio emission was detected at 4800 MHz only and does not correspond to any other known (closer) sources in the SIMBAD and NED data bases. We tentatively include it here as a possible association to the IRAS/OH sources. As with IRAS 14341−6211, the non-detection of emission from this source at 8640 MHz suggests that it has a non-thermal spectral index.

5.2.3 IRAS 18361−0647

IRAS 18361−0647 is associated with GLMP 810 (Garcia-Lario et al. 1997), OH25.5−0.3 (Te Lintel Hekkert et al. 1989, and references therein) and source 334 from the aforementioned paper, which are all located within a few arcsec of each other. This object was searched for 22-GHz H₂O maser emission by Deacon et al. (2007) but none was detected. We detected a radio source ∼40 arcsec from the former sources, which does not correspond to any other known (closer) sources in the SIMBAD and NED data bases. It was detected at both radio frequencies, although it is faint at 4800 MHz at only at the 3σ level. We consider the 4800-MHz emission to be real as it is coincident with the much stronger 8640-MHz source. Using the VLA at L-band, Zoonematkermani et al. (1990) and Becker et al. (1990) discovered the 11-arcsec source 25.489–0.287 at α_J2000= 18 38 47_J2000=−06 45 03 with a positional accuracy of ≤3 arcsec which they associate with an IRAS source likely to be IRAS 18361−0647. We detect nothing at this position, which is slightly more offset to the west of the IRAS source than our radio continuum source is to the east.

5.3 Other detected radio continuum sources

5.3.1 IRAS 17574–2403

Among the 1612, 1665 and 1667-MHz OH maser data, the 1612-MHz in particular is double-peaked and reminiscent of an expanding OH/IR CSE while the velocity structure of the mainlines is much more complex (Deacon et al. 2004). However, the great radio continuum brightness of this source and its spectral index (∼0.6) mark it as a star-forming H ii region, and so we do not discuss it further here.

5.3.2 PN G353.9+02.7

The object PN G353.9+02.7 (also known as RZPM2-12) was detected in the field of the IRAS 17164−3226 observations. It is ∼72 arcsec away from the OH maser position of Deacon et al. (2004) and ∼60 arcsec away from IRAS 17164−3226. The distance of the maser position and the fact that the maser is itself within 6 arcsec of both the IRAS source and the position of the near-IR source GLMP 546 (Garcia-Lario et al. 1997), suggests that those three are associated with each other, while PN G353.9+02.7 is an unrelated object. However, we include its radio detection and properties here as a serendipitous result as, apart from measurements of coordinates of its optical counterpart (Kerber et al. 2003), no images of this object exist in the literature. Its optical emission-line spectrum (Suárez et al. 2006) and Two-Micron All-Sky Survey (2MASS) colours are consistent with those of a PN. The strength of the [O iii] lines and the weakness of [N ii] suggest a medium excitation nebula. The presence of a strong optical continuum is somewhat surprising. There are no indications of any molecular absorption bands such as a symbiotic star would show, nor hints of the strong broad lines of a Wolf–Rayet star. We cannot account for the spectrum and the non-thermal radio emission in any other way except to identify this object as a PN, by analogy with V1018 Sco (Cohen et al. 2005).

According to the literature, radio emission from PN G353.9+02.7 was first detected by Ratag et al. (1990) in a 4860-MHz C-array VLA snapshot search for PN associated with IRAS sources, at a position offset by ∼10 arcsec from our radio continuum position. With a bandwidth of 100 MHz, their observing frequency range overlaps that of ours centred on 4800 MHz. They measure a flux density of 9.0 mJy ±10 per cent, which is in agreement with our 4800-MHz measurement within the errors, although their size measurement of < 2.1 arcsec is ∼5 times smaller than ours. Although they acknowledge the >1-arcmin offset of their radio continuum detection from the IRAS source, they still associate the PN with IRAS 17164−3224, which is a common misclassification of the latter object and which we discuss further in Section 5.4.1.

5.4 Listed radio continuum sources not detected in our survey

In the following subsections, we present notes of those sources quoted in the literature as having radio continuum emission but which we did not detect. It is likely that IRAS 17164−3226 (Sections 5.3.2 and 5.4.1) has been misidentified with a more evolved object and IRAS 18076−1853 was an erroneous detection. The published detections of IRAS 18314−0900 and 18588+0428 are also questionable. IRAS 18327−0715 and 18342−0655 appear to be more robust detections which eluded our observations due to the poor beam shape afforded to us by the unfortunate confluence of their LST range and the telescope schedule (Section 3).

5.4.1 IRAS 17164−3226

There is much confusion in the literature about this object. Acker et al. (1992) include it in their table of possible PNe at a position that is appropriate for PN G353.9+02.7 (see Section 5.3.2). They list Galactic coordinates of l= 353.97, b=+02.73 and correctly associate it with RZPM2-12 but they also erroneously match it with IRAS 17164−3226. Ratag et al. (1990) attributed their radio continuum detection to it (Section 5.3.2). Based on its near-IR colours, this IRAS source was classified as a likely PN by Garcia-Lario et al. (1997), under the designation of GLMP 546. Their IRAS coordinates match those of our OH position and the JHK photometry by Garcia-Lario et al. (1997) is quite similar to that of 2MASS J17194122-3229553. However, there is only one PN in this area, PN G353.9+02.7, and its location is over an arcmin away from IRAS 17164−3226 which itself is either a young PN whose brightness is below our 0.39 mJy beam⁻¹ 3σ detection threshold, or it is misclassified and is actually in an earlier phase (pAGB?) of evolution.

5.4.2 IRAS 18076−1853

Zijlstra et al. (1989) report only an upper limit of 1.8 mJy at 2 cm for this source. In Garwood et al. (1988), a source of flux density 422 mJy is noted at α_J2000= 18 10 38.6, δ_J2000=−18 52 58 from their VLA B-array continuum snapshot survey at 1.5 GHz. Zoonematkermani et al. (1990), again using the VLA in B-array, measured 214 mJy at 1.4 GHz from α_B1950= 18 07 42.2 δ_B1950= 18 53 37.4. It is remarkable that the latter two authors should measure such different flux densities. Given that the same array was used with only 0.1-GHz difference in observing frequency, we can rule out spectral index or resolution effects as being the cause. However, both Garwood et al. (1988) and Zoonematkermani et al. (1990) used an observing set-up with a sideband which would have been sensitive to the 1612-MHz OH emission from IRAS 18076−1853, so perhaps the difference is due to the detection and intrinsic variability of the maser. Indeed, given that the integrated flux density of the OH 1612-MHz emission is 274 Jy km s⁻¹ (Deacon et al. 2004), it is possible that no radio continuum at all was detected and the emission is solely due to a variable maser. This would explain our lack of detection.

5.4.3 IRAS 18314−0900

Garwood et al. (1988) found a 27 mJy 1.5-GHz continuum source at α_J2000= 18 34 11.3, δ_J2000=−08 57 59 and Zoonematkermani et al. (1990) detected 55 mJy from a 1.4-GHz source of mean full width at half-maximum diameter 4.6 arcsec at α_B1950= 18 28 50.82, δ_B1950=−08 40 12.0. We deem these detections questionable due to the sideband frequencies used in both the observations, one of which was sensitive to OH maser emission (Section 5.4.2).

5.4.4 IRAS 18327−0715

Zoonematkermani et al. (1990) detected 73 mJy at 1.4 GHz from α_B1950= 18 32 46.98, δ_B1950=−07 15 36.5 while the PMN (Parkes-MIT-NRAO; Griffith & Wright 1993) 4850-MHz survey equatorial catalogue lists a 31 mJy detection from α_J2000= 18 35 26.4, δ_J2000=−07 12 59. The Zoonematkermani et al. (1990) measurement is suspect due to the unfortunate choice of sideband frequency used in the observations (Section 5.4.2). The PMN detection is more robust in terms of detecting ‘real’ continuum emission, albeit with a 4.2-arcmin beamwidth. Assuming it is a compact source, it should have been above our detection threshold and the most likely reason for our non-detection is because of the poor beam shape (Section 3).

5.4.5 IRAS 18342−0655

Blommaert, van Langevelde & Michiels (1994) confirmed the 1612-MHz OH maser in this star and detected the second peak in the profile. Using the VLA with ∼4 arcsec resolution, Becker et al. (1994) found continuum detections of 26 mJy at 1.4 GHz and 111 mJy at 5 GHz from a location of α_B1950= 18 34 16.75, δ_B1950=−06 55 42.8, with a 5-GHz source size of 33.3 arcsec. 4.85-GHz continuum is reported in the equatorial PMN survey (Griffith & Wright 1993) at a position of α_J2000= 18 36 55.1, δ_J2000=−06 52 12 at a level of 412 mJy, with a 4.2-arcmin beam which encompasses the OH maser position. The discrepancy in C-band flux densities may be due to confusing sources in the Parkes (PMN survey) beam. These detections appear to be robust and our non-detection is again likely to be due to a poor beam shape (Section 3).

5.4.6 IRAS 18588+0428

Using the VLA, Zoonematkermani et al. (1990) detected a 2.7-arcsec continuum source of 33 mJy at 1.4 GHz from α_B1950= 18 58 46.82, δ_B1950=+04 25 17.7. This is ∼190-arcsec offset from the OH maser position of Deacon et al. (2004) and given that the position accuracy of the Zoonematkermani et al. (1990) detection is quoted as ≲3 arcsec, it is unlikely to be related to the OH source.

6 DISCUSSION

Here, we provide a brief overview of the IR and radio continuum properties of the ATCA-detected sources. With small-number statistics, it is not possible here to draw strong correlations between maser and IR typology and the likelihood of radio continuum detection, and we will attempt to counter this in future work by presenting ATCA observations of the remainder of the Deacon 2004 sample (a further 55 objects).

Of the 30 objects observed, two (IRAS 18434−0202 and 18588+0428) had data which were unusable due to poor uv-coverage. Of the remaining 28 sources, the colour–colour and maser diagnostics have enabled us to predict radio continuum emission in seven (i.e. 25 per cent of the sample) to varying degrees of confidence, although one of detected sources was later discovered to be an H ii region. It is possible that this is a lower limit to the number of potentially detectable radio continuum sources in the sample, as some of the later LST OH maser targets are mentioned in the literature as having associated radio continuum, but their detection was potentially hampered in our observations by poor uv-coverage (IRAS 18372−0715 and 18342−0655).

The three tentative detections (IRAS 14341−6211, 17150−3224 and 18361−0647) warrant further investigation to clarify their status. While these sources have the largest angular offsets between their respective OH maser and continuum emissions, they are also the three closest sources (Table 3). Their near-kinematic distances are all of the order of ∼2 kpc, which translates to projected offsets of ∼2000 au (Table 3), which is a reasonable length scale for a pAGB system.

We have detected and partially resolved 3/28 radio continuum sources (11 per cent of the sample; IRAS 15367−5420, 15445−5449 and 16372−4808) that we confidently associate with a pAGB OH maser source by virtue of their mutual proximity. These comprise two Quad Is and an LI. Further, all three display irregular or single-peaked OH 1612-MHz emission and irregular or absent mainline OH emission. These maser properties are consistent with evolving objects whose maser shells are losing their coherence length due to the effects of interacting winds. We tentatively detect radio continuum emission from a further three pAGB OH-emitting objects (IRAS 14341−6211, 17150−3224 and 18361−0647); these are two Quad I RIs and an LI. The three radio sources with the smallest offsets from their associated OH maser, whose radio continuum brightness distribution actually encompasses the maser position, are all Quad I sources (IRAS 15445−5449, 16372−4808 and 17574−2403), and this includes the SFR. The remaining radio continuum sources whose OH maser is more offset and lies outside of the radio contours are either LIs or Quad I RIs.

Of the three radio continuum sources we confidently associate with a pAGB OH maser, one has a non-thermal spectral index. Non-thermal behaviour is also exhibited by PN G353.9+02.7 and the tentatively OH-associated source IRAS 18361−0647. Therefore, the non-thermal emitters comprise a Quad I, an LI and a source whose colour class is undefined. The spectral indices suggest that the emission from these three objects is dominated by a non-thermal component, such as that which is expected in evolving objects whose fast winds have recently switched on and which are subjecting the remnant AGB ejecta to shock interactions, resulting in a dominant synchrotron component.

dusty modelling of our three confident detections, IRAS 15367−5420, 15445–5449 and 16372–4808, has revealed they all have central star temperatures much less than that which is required to produce a significant source of ionizing photons. It is therefore likely that their radio continuum emission arises due to localized wind–shock interactions. This is consistent with the non-thermal spectral index measured in IRAS 15445−5449, although the question of why the indices of IRAS 15367−5420 and 16372−4808 are consistent with thermal emission then arises. There is some indication of radio continuum variability in these stars (IRAS 15445−5449) which may also be linked to their transitionary phase; perhaps the fast wind which is inducing the shocks is episodic on a time-scale of years. If we have indeed detected ionization due to interacting winds then it is interesting to note that the three stars which are the strongest candidates for displaying this phenomenon have quite different effective temperatures. This suggests that the fast wind has switched on at quite different evolutionary stages within these objects, and/or they are of different core masses, so that the time-scale for the onset of the fast wind would be expected to vary between them.

7 SUMMARY

Using far- and mid-IR colour diagnostics, along with OH maser profile characterization, we selected a sample of 30 pAGB stars as candidates for a radio continuum detection experiment with the ATCA. The data from 28 of the observed fields were viable. We detected radio continuum emission within 1 arcmin of the OH source position in seven fields, one of which was later discovered to be an H ii region. We also serendipitously detected a PN in an eighth field. We consider the three radio continuum objects detected <30 arcsec from the OH source position to be the strongest associations and have modelled their IR SEDs with dusty. In each case, we find best-fitting models with T_eff≪ 30 000 K, the threshold for significant photoionization. It therefore seems likely that the source of the radio continuum is wind–shock interactions between the remnant AGB ejecta and the postulated pAGB fast wind. This is consistent with the detection of a non-thermal radio spectral index in one of the modelled objects.

1

Note that the terms ‘PPN’ and ‘pAGB’ are used interchangeably in the literature and here we favour ‘pAGB’.

2

The square bracketed numbers denote the colours measured from the source flux densities, in the form of logarithmic flux density ratios rather than in Vega-based magnitudes, at those wavelengths in microns.

The Australia Telescope Compact Array is part of the Australia Telescope National Facility. IB acknowledges funding from Swinburne University of Technology. MC is grateful for support from the Distinguished Visitor program at the Australia Telescope National Facility in Marsfield. This research has made use of the SIMBAD data base, operated at CDS, Strasbourg, France and also NASA's Astrophysics Data System. This work made use of data products from the MSX. Processing of the data was funded by the Ballistic Missile Defense Organization with additional support from NASA's Office of Space Science. This research has made use of the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.