DEATHSTAR: Nearby AGB stars with the Atacama Compact Array I. CO envelope sizes and asymmetries: A new hope for accurate mass-loss-rate estimates
S. Ramstedt, W.H.T. Vlemmings, L. Doan, T. Danilovich, M. Lindqvist, M. Saberi, H. Olofsson, E. De Beck, M.A.T. Groenewegen, S. Höfner, J.H. Kastner, F. Kerschbaum, T. Khouri, M. Maercker, R. Montez, G. Quintana-Lacaci, R. Sahai, D. Tafoya, A. Zijlstra
aa r X i v : . [ a s t r o - ph . S R ] A ug Astronomy & Astrophysicsmanuscript no. deathstar1 c (cid:13)
ESO 2020August 19, 2020
DEATHSTAR: Nearby AGB stars with the Atacama Compact Array
I. CO envelope sizes and asymmetries: A new hope for accurate mass-loss-rateestimates
S. Ramstedt , W. H. T. Vlemmings , L. Doan , T. Danilovich , M. Lindqvist , M. Saberi , H. Olofsson , E. De Beck ,M. A. T. Groenewegen , S. Höfner , J. H. Kastner , F. Kerschbaum , T. Khouri , M. Maercker , R. Montez ,G. Quintana-Lacaci , R. Sahai , D. Tafoya , and A. Zijlstra Theoretical Astrophysics, Division for Astronomy and Space Physics, Department of Physics and Astronomy, Uppsala University,Box 516, SE-751 20 Uppsala, Swedene-mail: [email protected] Department of Space, Earth and Environment, Chalmers University of Technology, Onsala Space Observatory, 439 92 Onsala,Sweden Department of Physics and Astronomy, Institute of Astronomy, KU Leuven, Celestijnenlaan 200D, 3001 Leuven, Belgium Koninklijke Sterrenwacht van België, Ringlaan 3, B-1180 Brussels, Belgium Rochester Institute of Technology, Rochester, NY, USA Department of Astrophysics, University of Vienna, Türkenschanzstr. 17, 1180 Vienna, Austria Smithsonian Astrophysical Observatory, 60 Garden Street, Cambridge, MA 02138, USA Instituto de Física Fundamental (IFF-CSIC), Serrano 123, Madrid, CP 28006 , Spain Jet Propulsion Laboratory, MS 183-900, California Institute of Technology, Pasadena, CA 91109, USA Jodrell Bank Centre for Astrophysics, Alan Turing Building, University of Manchester, Manchester M13 9PL, UK
ABSTRACT
Context.
This is the first publication from the DEATHSTAR project. The overall goal of the project is to reduce the uncertainties ofthe observational estimates of mass-loss rates from evolved stars on the Asymptotic Giant Branch (AGB).
Aims.
The aim in this first publication is to constrain the sizes of the CO emitting region from the circumstellar envelopes around42 mostly southern AGB stars, of which 21 are M-type and 21 are C-type, using the Atacama Compact Array (ACA) at the AtacamaLarge Millimeter / submillimeter Array (ALMA). The symmetry of the outflows is also investigated. Methods.
Line emission from CO J = → → uv -plane. A detailed radiative transfer analysis will be presented ina future publication. The major and minor axis of the best-fit Gaussian at the line center velocity of the CO J = → Results.
We find that the CO envelope sizes are, in general, larger for C-type than for M-type AGB stars, which is as expected ifthe CO / H ratio is larger in C-type stars. Furthermore, the measurements show a relation between the measured (Gaussian) CO J = → ff erences between the di ff erent stellar types. For lower mass-loss-rate irregular and semi-regular variables of bothM- and C-type AGB stars, the CO J = → CO J = → Conclusions.
These results for CO envelope radii and shapes can be used to constrain detailed radiative transfer modeling of thesame stars so as to determine mass-loss rates that are independent of photodissociation models. For a large fraction of the sources,observations at higher spatial resolution will be necessary to deduce the nature and origin of the complex circumstellar dynamicsrevealed by our ACA observations.
Key words. stars: AGB and post-AGB - stars: mass-loss - stars: winds, outflows - stars: circumstellar material
1. Introduction
Stars with zero-age-main-sequence masses in the range of ∼ ⊙ evolve into asymptotic giant branch (AGB) stars duringthe late stages of their evolution. The heavy mass loss during the AGB makes the stars major contributors of newly synthe-sized elements and dust to their surroundings. Understandingthe mass-loss process is crucial for comprehending the evolu-tion of stars in this mass range, but also for extragalactic pop- Article number, page 1 of 28 & Aproofs: manuscript no. deathstar1 ° ° Galactic Longitude -90 ° ° G a l ac ti c L a tit ud e
100 300 500 700 90005101520253035 -9 -8 -7 -6 -5 -4 -9 -8 -7 -6 -50510152025 -8 -7 -6 Fig. 1.
Sample statistics of the full ∼
180 star DEATHSTAR sample. M-type stars are blue, C-type stars are red, and S-type stars are green. Filledsymbols mark the stars included in this paper.
Upper left:
Galactic distribution of the full sample.
Upper middle:
Distance, D , distribution.These are preliminary distances from previous publications (see text and Table 1). The solid lines show the expected distributions for each spectraltype with Poisson errors (Jura 1990; Jura & Kleinmann 1992; Jura et al. 1993). Upper right:
Wind properties (mass-loss rate, ˙ M , and terminalexpansion velocity, v ∞ ) from previous publications. See text for references. Lower left:
The ˙ M / v ∞ distribution for the full sample. We note that˙ M / v ∞ is a proxy for the wind density. Lower middle: ˙ M / v ∞ as a function of the pulsational period. Lower left: CO / CO ratio for the samplesources (Ramstedt & Olofsson 2014) is plotted against the pulsational period. Both parameters are expected to increase as the stars evolve. Seetext for a further discussion. ulation studies. Mass-loss rates, ˙ M , on the AGB are found torange from ∼ − –10 − M ⊙ yr − (e.g., Höfner & Olofsson 2018,and references therein). It is challenging to find reliable ob-servational methods to measure mass-loss rates covering thiswide range (Ramstedt et al. 2008), but it is crucial since themeasurements will provide key constraints for theoretical mod-els (e.g., Eriksson et al. 2014; Marigo et al. 2016; Bladh et al.2019). Wind formation is studied using dynamical wind mod-els (e.g., Höfner 2008; Eriksson et al. 2014; Bladh et al. 2015;Höfner et al. 2016) with the ultimate goal of developing a pre-dictive theory of AGB mass loss. This will permit reliable es-timates of dust production and, for example, the intrinsic red-dening of distant galaxies (e.g., Conroy 2013). These modelsare constrained using observed wind properties, that is, ˙ M andwind velocities, v ∞ . However, with the derived ˙ M uncertaintiesreaching as high as a factor of three (within the adopted spher-ically symmetric model, Ramstedt et al. 2008), the dynamicalwind models cannot be su ffi ciently well constrained (e.g., Fig. 7in the recent paper by Bladh et al. 2019 shows how the wind pa-rameters vary with model input parameters).Observations of CO radio line emission (originating in thecircumstellar envelope, CSE, which is created by the wind), incombination with detailed radiative transfer, is considered to bethe most reliable method for determining AGB wind proper-ties (e.g., Höfner & Olofsson 2018, and references therein). Thepoorly constrained size of the CO envelope is a remaining, sig- nificant source of uncertainty for the mass-loss rates estimatedusing this method. The generally adopted strategy is to use anenvelope-sized estimate based on the photodissociation modelby Mamon et al. (1988). Major uncertainties are that this modelassumed a standard interstellar radiation field (Draine 1978), andthat numerical methods and shielding functions have been up-dated since then (e.g., Groenewegen 2017; Saberi et al. 2019).In radiative transfer models that are intended to determine windparameters from CO line observations, the CO envelope sizeis estimated using a functional fit to the Mamon et al. photodisso-ciation model results. The envelope size is a function of the windparameters, including the CO abundance, which increase withdensity, that is, ˙ M / v e . A more exact way to determine the COenvelope size (e.g., independent of our knowledge of shieldingfunctions) is to constrain it directly using interferometry. Directobservations will further improve the accuracy, since the radia-tion environment changes from source to source .A pioneering interferometric survey of AGB (and post-AGB)CO CSEs was performed by Neri et al. (1998). Forty-six sourceswere mapped in CO J = → Article number, page 2 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA significant scatter. They also concluded that about 30% of theenvelopes show significant asymmetry. This investigation waslater followed up with considerably improved capabilities at thePlateau de Bure by the COSAS program (Castro-Carrizo et al.2010). The goal was to investigate the morphologies of the en-velopes, in particular during the transition to post-AGB, and thissample contained several sources at later evolutionary stagesthan the AGB. Castro-Carrizo et al. (2010) presented detailedmaps of the CO(1-0) and (2-1) emission from 16 sources andthoroughly discussed each source. The synthesized beams weretypically of the order 3-5 ′′ and 1-2 ′′ for CO(1-0) and (2-1), re-spectively. They found that the measured envelope sizes were, onaverage, slightly larger than expected from the photodissociationmodel by Mamon et al. (1988). They further concluded that theAGB envelopes generally show round shapes and approximatelyisotropic expansion, while most later sources, that is post-AGB,exhibit axial symmetry and fast bipolar flows.We have started a new project called DEATHSTAR (DE-termining Accurate mass-loss rates for THermally pulsing AGBSTARs) in which the overall aim is to better constrain the mea-surements of AGB mass-loss rates. Observations of a large sam-ple of "typical" AGB envelopes (sources with known strong de-viations from spherical symmetry have been omitted), coveringthe full range of AGB stellar and wind parameters, will be ana-lyzed in detail in future work. The already-available data base ofCO lower- J transitions will be modeled, together with availableand new interferometric observations using updated radiativetransfer models. In this first paper, new interferometric data ofCO(2-1) and (3-2) emission obtained with the Atacama CompactArray (ACA) at the Atacama Large Millimeter / submillimeterArray (ALMA) are presented. The sample selection is explainedin Sect. 2. The observations, data reduction, and data analysis areoutlined in Sect. 3. The results with an analysis of the CO lineemission distributions (size and morphology) and an overviewof the detections of emission from other molecular species arepresented in Sect. 4. Finally, the results are discussed and sum-marized in Sects. 5 and 6.
2. The sample
The full sample for which the circumstellar CO line emis-sion will be modeled consists of the ∼
180 C-, M-, andS-type AGB stars analyzed in Schöier & Olofsson (2001),González Delgado et al. (2003), and Ramstedt et al. (2006) to-gether with additional sources presented in Danilovich et al.(2015). In this initial paper, the new interferometric data forthe southern M- and C-type stars are presented. Some of theavailable sample statistics for the full ∼
180 star DEATHSTARsample are shown in Fig. 1. The distance distribution (Fig. 1,middle) is compared with the estimated distribution in the so-lar neighborhood (Jura 1990; Jura & Kleinmann 1992; Jura et al.1993). The estimated distribution is derived from 2MASS andground-based observations (Jura & Kleinmann 1990) and as-sumes a smooth distribution of 40 C-type stars kpc − , a scaleheight of 200 pc, and that there are a third as many S-type asC-type stars. For the full ∼
180 star DEATHSTAR sample, theC-type stars from Schöier & Olofsson (2001) are all brighterthan K = µ mbands and 60 µ m flux & / deathstar sists of stars that have good quality flux measurements in theIRAS 12, 25, and 60 µ m bands, that are found in the GeneralCatalog of Galactic S stars, and that are detected in Tc and arehence intrinsic. The completeness of the S-type sample is dis-cussed in Ramstedt et al. (2009) and it is thought to be completeout to 600 pc. Furthermore, stars of all three spectral types areonly included in the sample if they are detected in CO radioline emission, which could be reproduced under the assumptionof spherical symmetry. Sources that show strongly asymmetricline profiles when observed with single-dish telescopes, or withknown CO-detached shells, are hence not included (e.g., R Scl,U Ant, EP Aqr, and π Gru). In this paper, we also exclude starsthat have previously been observed with ALMA; however, theywill be included in the future analysis. The lower panel of Fig. 1shows that the stellar and wind parameters of the full ∼
180 starsample cover the ranges expected for AGB stars. As expected,fewer stars are found at the high end. The sample is biased tomass-losing stars since only stars that are previously detectedin CO radio line emission are included. It is also likely that thefull range of AGB masses is not covered simply because highermass stars are rare. It is our assessment that the full ∼
180 starDEATHSTAR sample is representative of Galactic mass-losingAGB stars and covers the relevant ranges of wind and stellarparameters to provide the necessary constraints for theoreticalmodels.The 42 stars (21 M-type, 21 C-type), which were observedwith ALMA ACA in Cycle 4 and for which the data are pre-sented in this paper, are listed in Table 1 with the variability type(as listed in the GCVS), mass-loss rate, final wind velocity, dis-tance according to the previous analysis in Schöier & Olofsson(2001), González Delgado et al. (2003), and Danilovich et al.(2015), and Gaia data release 2 (DR2) (Gaia Collaboration et al.2016, 2018) distance from Bailer-Jones et al. (2018). The mass-loss rates and wind velocities were estimated by reproducingseveral low- J CO lines for each source using the non-LTE,nonlocal, spherically symmetric radiative transfer code that wasfirst presented in Schöier & Olofsson (2001). The CO / H abun-dance ratio, which is necessary to derive the total gas mass-loss rate, is assumed to be 2 × − for the M-type stars and1 × − for the C-type stars. The results from the photodis-sociation model by Mamon et al. (1988) was used in the ra-diative transfer modeling. In order to give consistent mass-lossrates and distances, the distances adopted here (and listed in Ta-ble 1) are the same as those used in the papers listed above. Thedistances will be updated as part of the future radiative trans-fer analysis planned for the full sample. For the semi-regularM-type stars, a stellar bolometric luminosity of 4000 L ⊙ wasadopted. For M-type Mira variables and some C-type stars,the period-luminosity relations from Whitelock et al. (1994) andGroenewegen & Whitelock (1996) were used to estimate the lu-minosity, respectively. From the luminosity, the distances weredetermined by either using two blackbodies or by using the dustradiative transfer code DUSTY (Ivezic et al. 1999) to model thespectral energy distribution (SED). For the remaining C-typestars, the distance was estimated directly from the original Hip-parcos parallax (The HIPPARCOS and TYCHO catalogs 1997)or adopted from Menzies et al. (2006). The method used for eachC-type star is noted in Table 1. Figure 2 shows the comparisonbetween the adopted distances from a previous analysis and thenew Gaia DR2 distances for the full sample (Bailer-Jones et al.2018). The spread is very large and the uncertainties a ff ecting theGaia DR2 distances for these types of stars are further discussedin Appendix A. The Gaia DR2 distances are typically larger thanthe adopted distances. The mass-loss rates given in Table 1 scale Article number, page 3 of 28 & Aproofs: manuscript no. deathstar1
Fig. 2.
Comparison between distances from Gaia DR2 parallaxes andthe adopted distances from the previous analysis (see text for details)for the full ∼
180 star DEATHSTAR sample. Red symbols are carbonstars, blue symbols are M-type stars, and green symbols are S-type stars.Filled symbols are the sources included in this paper. The solid, blackline marks a one-to-one correlation, while the dashed lines show therange of ± with distance as D n where 1.4 . n .
3. Observations, data reduction, and analysis
The 42 sample sources listed in Table 1 were observed with theACA in stand-alone mode in Cycle 4 in Bands 6 and 7 (projectcodes: 2016.2.00025.S and 2017.1.00595.S). The correlator wasset up with four spectral windows in each band. In Band 6 thewindows were centered on 216.4, 218.3, 230.7, and 232.1 GHz.In Band 7 they were centered on 330.75, 332.25, 343.52, and345.6 GHz. Line emission from CO J = →
1, 3 →
2, and CO J = → − in the CO and COwindows and to 1.0 and 1.5 km s − in the other spectral windowsin Band 6 and 7, respectively.All data were calibrated using the standard pipeline scriptsand imaged using the Common Astronomy Software Applica-tions package (CASA; McMullin et al. 2007). Self-calibrationwas performed using a small number of channels, typically two,across the peak CO line emission and applied to all sources inboth bands. This improved the signal-to-noise ratio by 10-15%on average. All data were cleaned initially using 10000 itera-tions. For sources brighter than ∼
20 Jy, we found that furthercleaning iterations was necessary to recover the signal and fi-nally 20000 iterations were applied to all bright ( > / deathstar).The full width at half-maximum (FWHM) beam widths androot mean square (rms) noise levels at a velocity resolution of0.75 km s − , which were measured in the emission-free channelsin the CO line window of the final images in both bands, aregiven in Table B.1 for all sources. The science goal beam widthsand rms values at 1.5 km s − were 5.5 ′′ and 40 mJy beam − aswell as 3-4 ′′ and 50 mJy beam − in Bands 6 and 7, respectively.For a significant fraction of the data sets, the beams are largerthan what was aimed for as well as elongated; this is an e ff ect ofthe project being partly observed as a filler program with someof the sources at low elevation. The maximum recoverable scalewill cover a range depending on the exact antenna configura-tion and frequency, but on average it was 25 ′′ ± ′′ and 18 ′′ ± ′′ for Band 6 and 7, respectively. The data quality is su ffi cient forthe DEATHSTAR project goals; however, as is discussed below,there are compelling reasons to follow up on a majority of thesources with higher-spatial-resolution observations. uv -plane The first step is to fit the visibility data in the CO(2-1) and (3-2) measurement sets with Gaussian distributions for all sourcesusing the CASA task UVMULTIFIT (Martí-Vidal et al. 2014).The least-square fit gives the center position coordinates, the ma-jor axis and the axis ratio of the best-fit Gaussian, and the posi-tion angle of the major axis for each channel over which the fithas been performed. This provides initial estimates for the sizesof the emitting regions and indications of deviations from sym-metry (as described below). The next step will be to performdetailed radiative transfer modeling that is constrained by theavailable multitransition single-dish, including the CO(1-0) line,and interferometry data for each source to determine the size ofthe CO envelope and to fit the emission distributions. This willbe presented in a future publication.
Emission distributions for the CO(2-1) and (3-2) lines were cal-culated from our previous best-fit radiative transfer model resultsfor a subsample of sources. This is a first step toward full radia-tive transfer modeling in order to reproduce the data availablefor each source (planned for a future publication). The subsam-ple consists of six stars: three M-type and three C-type stars ofwhich there is one low-, one intermediate-, and one high-mass-loss-rate source for each type. The models were directly adaptedfrom the previous best-fit models (Danilovich et al. 2015, seealso Table 1) and no attempt has been made to improve the fit tothe ALMA ACA data. Instead, the purpose is to start evaluatingthe validity of the photodissociation model used to estimate thesize of the circumstellar CO envelopes in the models. For thesesources, the original radiative transfer used the photodissociationradius from Mamon et al. (1988). The output from the best-fit ra-diative transfer models of the six sources was used to create im-age cubes with an imaging ray-tracing program that is part of theradiative transfer package developed by Schöier (2000). The im-age cubes were created with 32 ×
32 pixels with sizes of 0 ′′ . ′′ . Article number, page 4 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA
4. Results
The CO J = → → ′′ was used for both transitions becausethis covered the emitting region without reaching lower-fidelityregions that are close to the image edges. Table B.1 gives thepeak flux, the center velocity, and the total velocity width ofeach line. The peak value was measured at the maximum pointacross the line. The center velocity was measured at the cen-ter between the two points where the flux reaches 5% of thepeak value at the extreme velocities. The total width is the veloc-ity width between the two 5%-peak-flux-value points. The lineprofiles show an interesting structure that is likely indicative ofcircumstellar dynamics and which was revealed due to the veryhigh signal-to-noise ratio attained. For example, some lines aredistinctly asymmetric (see e.g., R Hya and SS Vir). Furthermore,some lines seem to show extended wings (see e.g., L Pup andTW Hor). Indications of circumstellar inhomogeneities and non-isotropic kinematics are further discussed below (Sect. 4.3).
The results of fitting the data using Gaussian visibility distri-butions are given in Table 2. The second column gives twotimes the predicted photodissociation radius, R p , over the dis-tance, D (from Table 1), as a measure of the expected sizeof the CO line emitting region. The photodissociation radius,which was measured as the radius where the CO abundance hasdropped to half of its initial value, was calculated using Eqns.10 and 11 from Schöier & Olofsson (2001), which were derived(Stanek et al. 1995) by fitting the results of the CO photodis-sociation model by Mamon et al. (1988). The parameter valuesgiven in Table 1 were used for the calculation. Table 2 also liststhe beam-deconvolved FWHM major axis and the axis ratio ofthe best-fit Gaussian at the center velocity (from Table B.1). Theerror is the average error of the fits in the two channels adjacentto the center velocity. When the error is large when the centerchannels are strongly a ff ected by resolved-out flux, for example(see discussion in Sect. 4.3 and Figs E.1–E.6), the results aregiven without errors and marked by a colon (:).There are good reasons (from previous radiative transfermodels, e.g., Ramstedt & Olofsson 2014) to expect that theCO(2-1) line will be excited almost throughout the entire CSE.However, without detailed models for individual sources, wecannot estimate how well the fitted Gaussian semi-major axiswill correspond to the photodissociation radius, or by what factorit may deviate (see Sect. 4.2.2). In Fig. 3, we illustrate the relia-bility of using the measured Gaussian size as a proxy for the sizeof the emitting region and for the photodissociation radius, R p ,based on the recent models by Saberi et al. (2019). These mod-els include a more complete treatment of CO self-shielding andyield radii that are mostly similar to Mamon et al. (1988), butthey can be up to ∼
40% smaller for some combinations of mass-loss rates and initial CO abundances. We used the LIME 1.9.5radiative transfer code (Brinch & Hogerheijde 2010) to produceradiative transfer models of the photodissociation models fromSaberi et al. (2019), using the CSE parameters that were adoptedin their chemical models. Subsequently, we fit a Gaussian profileto the CO(2-1) emission at the systemic velocity using a 1 km s − channel width using the same method as in Sect. 3.2. Figure 3 shows the CO(2-1) line intensity distributions for three M-starmodels with di ff erent mass-loss rates. Each model intensity dis-tribution has been scaled to the fitted Gaussian. For intermediate-to high-mass-loss-rate sources in particular, the intensity distri-bution is significantly non-Gaussian, and the determination of aGaussian FWHM is strongly dependent on the exact shape ofthe intensity distribution in the inner region. Since low-angular-resolution observations lack the uv-coverage to extract this in-formation and as irregular structures are especially common inthese areas, the Gaussian radii determined for the CSEs withthe highest density, ˙ M / v ∞ & − M ⊙ km s − yr − , are lessreliable. The figure also shows the calculated photodissociationradius from Saberi et al. (2019) for each model (vertical lines)scaled with the Gaussian FWHM radius. For the densest CSEs,the CO(2-1) line does not extend to the photodissociation radiusand the Gaussian radii are therefore not a measure for R p .Based on the formula and the distance estimates fromStanek et al. (1995) in Table 1, the CO(2-1) major axis is a factorof two smaller than 2 × R p / D for the semi- and irregular variables,and a factor of three smaller for the Mira variables, regardless ofchemical type. This is consistent with the expectations from themodeling (both chemical and radiative transfer). Furthermore,for a small majority of the sources (26 / ≤
10% from one, meaning that slightly more thanhalf of the sources are close to being circular. For the more as-pherical sources, the photodissociation calculations that assumea standard spherically expanding CSE introduce further uncer-tainties.The major axes of the Gaussian fits to the CO(2-1) emis-sion at the center velocity channels (third column of Table 2multiplied with the distance from Table 1) are plotted against˙ M / v ∞ in Fig. 4. The uncertainty of the major axis is very smallin arcseconds (see Figs E.1–E.6). Instead, the error bars in the y-direction are dominated by the distance uncertainty, which can-not be evaluated easily. The figure also shows the photodisso-ciation diameter for the M-type and C-type stars, respectively.The photodissociation radii were calculated across the range of˙ M / v ∞ for v ∞ = . − . We used the LIME 1.9.5 radiativetransfer code again (Brinch & Hogerheijde 2010) to produce ra-diative transfer models of the Saberi et al. (2019) photodissoci-ation model grid, and we used the same Gaussian fitting pro-cedure as mentioned above to measure the predicted CO(2-1)size. A spline fit to the expected CO(2-1) size is shown by thesolid lines in Fig. 4, which thus indicates the dependence of themeasured size of the CO(2-1) emitting region on the photodisso-ciation radius as well as on the circumstellar density. The di ff er-ent results for higher- and lower-mass-loss-rate sources are ap-parent in the figure: The measured size of the CO(2-1) emittingregion for lower-mass-loss-rate M-type stars appears to show arather weak dependence on the circumstellar density, which isalso seen in the dependence of photodissociation diameter andthe modeled Gaussian diameter on the density. For higher-mass-loss rate M-type Mira sources, the dependence on the density ismuch steeper, with a slope similar to what is expected from thephotodissociation models. In almost all cases, however, the mea-sured diameter is smaller than the expected CO(2-1) diameterbased on the chemical and radiative transfer models. The lower-mass-loss-rate C-type stars show no significant dependence oncircumstellar density, or at least a large scatter around the ex-pected relation. Essentially all of the observed low-mass-loss-rate C-type stars have CO(2-1)-emitting major axes of 1000–2000 AU. The higher-mass-loss-rate C-type Mira stars, on theother hand, show a dependence on density that appears to be Article number, page 5 of 28 & Aproofs: manuscript no. deathstar1 somewhat steeper than for the M-type Miras. For all C-type Mi-ras, the measured diameters are larger than expected.Although the measured size of the CO(2-1) region is smallerthan the calculated photodissociation diameter for essentially allsources, it is apparent from Fig. 4 that the correlation betweenthe two is not straightforward. There are a variety of possibleexplanations for the deviations between the observations andthe photodissociation models. The most obvious one is the un-certainty of the distance, which introduces significant scatter inthe size determinations. Furthermore, as shown in Saberi et al.(2019), changes in initial CO abundance, CSE temperature pro-files, and the interstellar radiation field can have significant ef-fects. Finally, di ff erent density profiles, dust properties, or dust-to-gas ratios from what is assumed in the photodissociation mod-els can all change the estimated radius.The full results of the fitting of Gaussian emission distribu-tions to the ALMA ACA data across all of the CO line chan-nels are displayed in Figs E.1–E.6. The major and minor axis ofeach channel, as well as the RA and Dec o ff sets of the center ofthe best-fit Gaussian relative to the SIMBAD J2000 coordinatesof each source, are plotted against lsr-velocity for each line andsource as denoted in each plot. The interpretation of these plotsis further discussed in Sect. 4.3. Fig. 3.
CO(2-1) intensity distribution at the systemic velocity as a func-tion of radius for three M-star models from Saberi et al. (2019). Themodels were normalized to the peak of the fitted Gaussian and scaledto the FWHM radius of that Gaussian for each model separately. Thenormalized Gaussian is indicated by the solid red line in the figure. Foreach model, the vertical lines indicate the photodissociation radius R p ,which was also scaled to the Gaussian FWHM radius for each model.See text for a further explanation. Here we present the comparison between the emission distribu-tions from ALMA and those calculated from our previous best-fit radiative transfer model results (Sect. 3.3). The same analysiswas performed on the model results as on the observational data.
Fig. 4.
Full-width half-maximum (FWHM) major axis of the best-fit Gaussian at the center velocity channel of the CO(2-1) emission(marked by triangles and in astronomical units by multiplying by thedistance from Table 1) as a function of circumstellar density (as mea-sured by ˙ M / v ∞ from Table 1). M-type stars are blue. C-type stars are red.Mira-type variables are marked with solid symbols and other variablesappear with open symbols. The dashed lines show the photodissocia-tion diameter for M-type (blue) and C-type (red) stars from the modelgrid by Saberi et al. (2019) (calculated with v ∞ = − ). The dottedlines are the parametrized fits to the results from Mamon et al. (1988).The solid lines show a spline-fit to the expected Gaussian FWHM deter-mined from radiative transfer modeling of the models from Saberi et al.(2019). See text for a further explanation. Figure 5 shows the results of the Gaussian emission distributionfitting (Sects 3.2 and 4.2.1) from the models compared with theobservational results for the CO(2-1) emission for each of thesix selected sources. For error bars on the observational results,see Figs E.1–E.6. The results are a ff ected by the di ffi culties offitting a Gaussian distribution to the weaker and smaller emis-sion distribution close to the edge of the line. It is obvious thatthe radiative transfer analysis of the full data set will be neces-sary before any firm conclusions can be drawn. For the M-typestars, it seems as if the size of the emission might be smallerthan expected for the low mass-loss-rate source (R Leo) and theopposite for the high mass-loss-rate source (IRC-10529). Theupcoming analysis will evaluate whether this is a general trendor specific to the selected sources. This trend would indicate aneven steeper dependence on the circumstellar density than pre-dicted from the photodissociation model by Mamon et al. (1988)and is contrary to the fit to the M-type Mira stars in Fig. 4. Theresults for the C-type stars are less clear, partly because the fit-ting results to the U Hya data seem strongly a ff ected by resolvedout flux (see Sect. 4.3). It is also possible that the emission fromC-type stars is subject to stronger optical depth e ff ects due to thehigher CO abundance. This will also be evaluated in the upcom-ing radiative transfer analysis. Article number, page 6 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA -10 -5 0 5 10024681012 datamodel
30 35 40 45024681012 datamodel -30 -20 -10 00246810 datamodel -40 -35 -30 -25024681012 datamodel -10 0 10 20024681012 datamodel
30 40 50024681012 datamodel
Fig. 5.
Comparison between the results from fitting a Gaussian emission distribution to the CO(2-1) data (blue) and the model results (orange)based on previous best-fit models (see text). The solid lines show the major axis and the dashed lines show the minor axis. The red and greenlines show the RA and Dec. o ff sets for the data, respectively. The top row shows three M-type stars, and the bottom row shows three C-type stars,with an increasing mass-loss rate (Table 1). The results are a ff ected by the di ffi culties of fitting a Gaussian distribution to the weaker and smalleremission distribution close to the edge of the line. For error bars on the results from the measurements, see Figs E.1–E.6. Several asymmetrical features can be identified in the ob-servational results presented thus far, that is, the line pro-files (Figs D.1–D.6) and the results from the Gaussian fitting(Figs E.1–E.6). These features can be divided into two classes:Features that are consistent with a spherically symmetric CSEand those that are not.
The features that can be explained with a spherically symmetricCSE are as follows.1. An o ff set between the center position (of the fitted Gaussian)and the source coordinates is explained the most easily byuncertainties in the adopted coordinates (taken from SIM-BAD pre-Gaia DR2), when seen consistently in both lines.Examples include V Tel and RT Cap, for instance.2. The interferometer acts as a spatial filter and since only theACA was used (and not the total-power array), large-scale,smooth emission is resolved out. Resolved-out flux results ina lower peak flux level than when measured with a (similar-sized) single-dish telescope. Furthermore, if a significantfraction of the flux is resolved out, the visibility distributioncannot be well-fitted by a simple Gaussian distribution. This is the most apparent across the center channels where theemission distribution is expected to be the largest. Therefore,it is also a more common problem for the CO(3-2) line wherethe beam is smaller. Examples of this can be seen in RR Aql(CO(3-2)), WX Psc (both lines), and R Vol (CO(3-2)).3. Self-absorption on the blue-shifted side of the line can beseen as asymmetry in the line profile where it is sometimesapparent that emission is "missing" on the blue side of theline. Examples include IRC-10529, NP Pup, and T Ind. Itis even more apparent when looking at how the sizes ofthe sources change across the channels. Sources with self-absorption show a prominent peak in the size distribution atblue-shifted velocities since these channels probe emissioncoming from cooler circumstellar layers (and hence a largerregion).These features are relatively easy to explain and related to howthe observations were performed for the first and second points. The next class of features is less straight-forward to interpret andthe features are mostly thought to be clearly inconsistent with aspherically symmetric outflow. Further modeling and / or obser-vations will be necessary to explain exactly how they arise. Themajority of sources that exhibit these features need to be fol- Article number, page 7 of 28 & Aproofs: manuscript no. deathstar1 lowed up with higher-spatial-resolution observations. Featuresthat cannot be explained by a smoothly expanding, sphericallysymmetric CSE are the following.4. Triangular line profiles are not a newly observed feature;however, a satisfactory explanation as to how they ariseis lacking. This feature seems to be more common amonghigh-mass-loss-rate sources. Clear examples are R Hor andV1259 Ori.5. High-velocity wings are not a new feature either, but theycan be detected from more sources in this data set due tothe superior sensitivity reached with the ACA observationscompared to previous single-dish observations. The line pro-files of TW Hor, for example, are very similar to those of theknown asymmetrical CSE of π Gru (e.g., Doan et al. 2020).The CO emission from π Gru can be reproduced by a CSEconsisting of an expanding equatorial torus (explaining thetwo central peaks) and a bi-polar faster outflow leading towide wings (Doan et al. 2017). Other, more tentative exam-ples of line profiles with high-velocity wings include thoseof L Pup, R Leo, and BK Vir.6. The fitting of the visibility distribution gives the center po-sition of the best-fit Gaussian in each channel and the o ff -set relative to the center position is plotted in Figs E.1–E.6.Several sources show center position gradients. A positiongradient with velocity is either indicative of expansion ina higher-density layer (e.g., an equatorial torus such as inthe case of π Gru) or of rotation (e.g., Ramstedt et al. 2018;Vlemmings et al. 2018) depending on the orientation of thesources. Examples of sources with center position gradientsare BK Vir (in RA), TW Hor (both RA and Dec), SS Vir(both RA and Dec), and W CMa (in RA).7. A spherically symmetric CSE with a constant expansion ve-locity grows monotonically in size from the edge of theline to the center velocity. The e ff ects of self-absorptionand resolved-out flux on the size distribution are discussedabove. For some sources, the source size changes in an un-predictable way that cannot be easily explained without fur-ther modeling and / or observations. Examples include T Mic,CW Cnc, RT Vir, and W Ori.8. It is directly apparent from the line profile that the emissioncannot come from a standard CSE for a handful of the sam-ple sources. Instead, these line profiles are anomalous. Someare previously known to be anomalous, for example, L Pup(e.g., Kervella et al. 2016) and R Leo (Fonfría et al. 2019),while others have not been studied with a focus on the de-tailed structure of the CSE, for example, R Hya and SS Vir. CO A large number of molecules other than CO have transitionswith frequencies within the observed bands. For all sources, thepeak fluxes of the detected lines as measured across a 10 ′′ aper-ture are listed in Tables C.1-C.4. The spectral resolution acrosseach spectral window is the same as the one used for imagingand is given in Sect. 3. The lines listed in Tables C.1-C.4 arethose for which we are confident of detection at the given spec-tral resolution and aperture. It is possible that further detectionscould be confirmed with more optimization of the data analysis.Lines from SiO, SiS, CO, CS, and H CN are detected ina majority of the sources. Unsurprisingly, oxides such as SO,SO , and H O are only detected in the M-type stars as are H Sand less common isotopologues of SiS and NaCl. In the C-typestars, CN and CS are detected, sometimes together with lines from molecules with several carbon atoms, such as HC N, C H,and SiC . A further analysis of these lines is beyond the scope ofthis paper, but an overview of the firm detections are given hereand follow-up studies are encouraged.
5. Discussion and summary
The measurements presented in this paper show that, for thissample, the CO envelope sizes are, in general, larger for C-typethan for M-type stars. As explained both above and below, theenvelope size depends on many di ff erent parameters, but thistrend is expected if the CO / H ratio is larger in C-type stars.Based on chemical equilibrium models, the ratio between the C-and M-type CO-abundance is generally assumed to be equal tofive when deriving total gas mass-loss rates from observations ofCO emission lines. As part of the analysis planned for the fullDEATHSTAR sample, the di ff erence in abundance dependingon chemical type (also S-type stars) can now be observationallyconstrained for a large sample of AGB stars.Furthermore, the measurements show a relation between themeasured (Gaussian) CO(2-1) size and circumstellar densitythat, while in broad agreement with photodissociation calcula-tions, reveal large scatter and some systematic di ff erences be-tween the di ff erent stellar types. A significant amount of scatterarises due to significant distance uncertainties. Part of the dif-ferences can also be explained by excitation and optical depthe ff ects that decrease the reliability of the Gaussian size deter-mination for ˙ M / v ∞ & − M ⊙ km s − yr − . For lower-mass-loss-rate irregular and semi-regular variables of both M- andC-type stars, the CO(2-1) size seems essentially independent of˙ M / v ∞ . For the higher-mass-loss-rate Mira stars, the CO(2-1) sizeclearly increases with circumstellar density, with larger sizes forthe higher CO-abundance C-type stars. The M-type stars appearto be consistently smaller than what was predicted from pho-todissociation theory. The di ff erences between the estimates andmeasurements could (as shown in Saberi et al. (2019)) be dueto, for example, a systematically overestimated CO abundance,di ff erences in the CSE temperature profile, or di ff erences in theUV-environment. Additionally, a di ff erence in the adopted prop-erties of the dust, which is responsible for much of the shieldingin low density CSEs, and / or the adopted dust-to-gas ratio in themodels would lead to di ff erent sizes. Finally, a di ff erence in theCSE density profile due to changes in velocity and / or the massloss rate could significantly alter the photodissociation radius.With the size measurements that are now available for the largesample of sources in the DEATHSTAR project, it will be possi-ble to investigate many of these factors in more detail.In recent years there has been a strong focus on investigat-ing the isotropy of the mass-loss process in low- to intermediatemass evolved stars, which was partly expedited by the superiorimaging capabilities of ALMA. Several AGB sources with pre-viously known complex circumstellar dynamics and / or binarycompanions have been mapped in detail (e.g., Ramstedt et al.2014; Decin et al. 2015; Maercker et al. 2016; Doan et al. 2017;Kim et al. 2017; Homan et al. 2018; Fonfría et al. 2019). Withthe DEATHSTAR project, the aim is to get an overview of howwidespread and significant the detected asymmetrical featuresare, from the perspective of measuring the amount of circum-stellar wind material. A goal is to eventually be able to evaluatethe uncertainties that these features will result in when estimat-ing mass-loss rates from more distant, unresolved sources.The results presented here show that the majority of thesources have CO CSEs that are consistent with a sphericallysymmetric, smooth outflow, at least on larger scales. This is Article number, page 8 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA based on the line profile shapes and that the emission distributionacross the line channels can be well-fitted with a Gaussian visi-bility distribution with axis ratios that are close to one for a ma-jority of the sources. For about a third of the sources, indicationsof strong asymmetries are detected. This is consistent with whatwas found in previous interferometric investigations of north-ern sources (Neri et al. 1998; Castro-Carrizo et al. 2010). In theDEATHSTAR sample, some of the known asymmetric sourceshave been removed, but the higher sensitivity reached in theACA observations allowed us to detect relatively weaker fea-tures, such as extended line wings, than what was possible inprevious investigations. This o ff ers an explanation as to why thesame fraction of strong asymmetries is still found. In a large frac-tion of the sources, some deviation from spherical symmetry isdetected, often on smaller scales. Whether these smaller devia-tions can have a significant e ff ect on the estimated mass-loss rateor not cannot be evaluated without further analysis.The peak fluxes of lines from molecules other than CO de-tected within the observed bands are listed in Tables C.1-C.4.In general, lines from molecules that are known to be abundantin AGB CSEs (e.g., SiO, SiS, CS) are detected in almost allsources. Some less abundant oxides (e.g., H O and SO) are onlydetected in the M-type stars, while emission from moleculeswith several carbon atoms (e.g., SiC ) are only detected towardthe C-type stars. No further analysis has been attempted.
6. Outlook
This paper presents the first observational results from theDEATHSTAR project. The CO(2-1) and (3-2) ALMA ACAobservations of the southern M- and C-type stars observed inALMA Cycle 4 are presented and analyzed. In an upcoming pub-lication, the results on the M-, S-, and C-type sources observedin Cycle 5 will be presented. The next step is to perform detailedradiative transfer modeling that is constrained by the presentedACA data together with the previously attained single-dish ob-servations of the lower- J CO transition lines (up to J = →
5) inorder to determine mass-loss rates that are independent of pho-todissociation model results. This will also be presented in futurepublications. For a large fraction of the sources, observations athigher spatial resolution will be necessary to deduce the natureand origin of the complex circumstellar dynamics revealed bythe observations and data analysis presented in this paper.
Acknowledgements.
This paper makes use of the following ALMA data:ADS / JAO.ALMA / NRAOand NAOJ. This work has made use of data from the European SpaceAgency (ESA) mission
Gaia ( ),processed by the Gaia
Data Processing and Analysis Consortium (DPAC, ). Fundingfor the DPAC has been provided by national institutions, in particular theinstitutions participating in the
Gaia
Multilateral Agreement.
Table 1.
Forty-two sources observed with the ACA by spectral type andvariability type listed in ascending mass-loss-rate order.
Source Variability ˙ M a v ∞ a D D
Gaia type [M ⊙ yr − ] [km s − ] [pc] [pc] M-type semi-regular and irregular stars: L Pup SRb 2 × − × − × − . × − . × − . × − × − × − . × − × − × − × − × − × − M-type Mira stars:
R Leo M 2 × − × − . × − . × − . × − . × − + × − C-type semi-regular and irregular stars:
TW Oph SRb 5 × − b . × − b × − b × − b × − c . × − d . × − b . × − b . × − b . × − c -X Vel SR 1 . × − c . × − b × − c × − d C-type Mira stars:
R Lep M 7 × − b × − c . × − c . × − c × − c . × − e . × − e - Notes.
The columns give wind properties (mass-loss rate, ˙ M , and fi-nal expansion velocity, v ∞ ), and preliminary distance, D , from previouspublications (see text for details). The final column gives the distancesfrom Gaia data release 2 (Bailer-Jones et al. 2018) for comparison pur-poses (see also Fig. 2). ( a ) From the previous analysis. See text for references. ( b ) Hipparcos par-allax. ( c ) Period-Luminosity relation (Groenewegen & Whitelock 1996). ( d ) Assuming 4000 L ⊙ . ( e ) Adopted from Menzies et al. (2006).Article number, page 9 of 28 & Aproofs: manuscript no. deathstar1
Table 2.
Results from the Gaussian visibility distribution fitting. R p CO(2-1) CO(3-2) Asym.Source D Major Ratio Major Ratio Feat.[ ′′ ] [ ′′ ] [ ′′ ] M-type semi-regular and irregular stars: L Pup 10.6 6.04 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± M-type Mira stars:
R Leo 18.2 10.00 ± ± ± ± ± ± ± ± ± + ± ± ± ± C-type semi-regular and irregular stars:
TW Oph 8.0 4.68 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± C-type Mira stars:
R Lep 17.2 11.7: 0.9: 3.58 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Notes.
The second column gives the photodissociation radius two timesover the distance as a measure of the expected size of the CO emittingregion. Columns 3-6 give the major axis and the axis ratio with errorsfor the best-fit Gaussian at the center velocity of the respective line.The final column gives the asymmetrical features for each source, as isexplained in Sect. 4.3. See text for a further explanation.Article number, page 10 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA
Appendix A: Parallaxes for AGB stars from GaiaDR2
Fig. A.1.
Gaia DR2 parallaxes for our sample compared to the distancesused in the current mass-loss models. In the bottom panel, the Gaia par-allaxes are presented with their formal errors. In the top panel, the as-trometric_excess_noise value is added in quadrature to the formal errorsfollowing the empirical results from van Langevelde et al. (2018).
Since the paper presents mass-loss values based on earlieranalysis, we have presented the distances that were used in themass-loss determinations. Recently, Gaia DR2 astrometric so-lutions were released for all but one of the stars in our sam-ple. While the formal errors on the Gaia parallaxes are generallysmall, the reliability of the Gaia parallaxes for AGB stars is un-der debate (e.g., van Langevelde et al. 2018). There are severalreasons why the Gaia parallaxes to AGB stars might be wrongor have significantly underestimated the assigned errors. Firstly,AGB stars are often larger than the parallax signature itself (withsizes of order one to a few astronomical units) and they haveconvective surface motions that can cause the photo-center toshift significantly (Chiavassa et al. 2018). Secondly, AGB pul-sations can a ff ect the astrometric measurements. Finally, AGBstars are so bright that Gaia approaches saturation. The GaiaDR2 catalog contains a number of parameters that can providea measure for the reliability of the astrometry. One of these isthe astrometric_excess_noise , which represents modeling errorsfor sources that do not behave according to the adopted astro-metric model of Lindegren et al. (2012). In van Langevelde et al.(2018), an empirical analysis of very long baseline interferome-try (VLBI) parallaxes compared to the Gaia results revealed thatadding the astrometric_excess_noise to the formal parallax er-rors in quadrature was needed to reconcile the two methods. InFig. A.1 we show the di ff erence between the formal and excessnoise errors. When we adopted this procedure, only four of theAGB stars in our sample have a parallax solution that satisfies π/σ π >
5. Recently, another measure of reliability was intro-duced, namely the RUWE (renormalized unit weighted error). Lindegren et al. 2018, Gaia memo; GAIA-C3-TN-LU-LL-124-01
This represents the square-root of the reduced χ value, whichwas corrected for the strong dependence of the χ value on mag-nitude and color. For many applications, a RUWE value < . G <
12, which includes allof our AGB stars. In these cases, Lindegren et al. (2018) stressthat a RUWE threshold should be set based on empirical evi-dence and not a theoretical distribution. For this, we can use thenorthern AGB star BX Cam as an example. Recent VLBI ob-servations have revealed a parallax of π VLBI = . ± .
03 mas(Matsuno et al. 2020). This value is significantly di ff erent fromthe Gaia value of π Gaia = . ± .
25 even when taking the as-trometric_excess_noise of 0.67 mas into account. However, theRUWE of BX Cam is 1.04, which would have qualified as agood solution in most cases. In comparing the stellar luminosity,Matsuno et al. (2020) conclude that the larger VLBI distance ismore reliable since the VLBI distance would imply a luminosityof ∼ ⊙ , while the Gaia distance would result in a lumi-nosity of only ∼
870 L ⊙ . This example illustrates that even forAGB stars with an apparent low RUWE value, we should exer-cise caution when adopting the current Gaia DR2 parallaxes. Appendix B: Imaging resultsAppendix C: Detections of emission frommolecules other than COAppendix D: Line profilesAppendix E: Results from fitting to Gaussianemission distributionsReferences arXiv:astro-ph/9910475 ]Jura, M. 1990, in From Miras to Planetary Nebulae: Which Path for Stellar Evo-lution?, ed. M. O. Mennessier & A. Omont, 41–63Jura, M. & Kleinmann, S. G. 1990, ApJ, 364, 663Jura, M. & Kleinmann, S. G. 1992, ApJS, 79, 105
Article number, page 11 of 28 & Aproofs: manuscript no. deathstar1
Table B.1.
Imaging results.
Band 6 Band 7 CO(2-1) CO(3-2)Source θ P.A. rms θ P.A. rms F peak v c ∆ v F peak v c ∆ v [”] [ ◦ ] [ mJybeam ] [”] [ ◦ ] [ mJybeam ] [Jy] [km s − ] [km s − ] [Jy] [km s − ] [km s − ] M-type semi-regular and irregular stars: L Pup 6.5 × × × × × × × × × × × × × × × × × × × × × × × × × × × × M-type Mira stars:
R Leo 7.0 × × × × × × × × × × × × + × × C-type semi-regular and irregular stars:
TW Oph 10.3 × × × × × × × × × × × × × × × × × × × × × × × × × × × × C-type Mira stars:
R Lep 7.4 × × × × × × × × × × × × × × Notes.
Columns 2-7 give the imaging results: Full width at half-maximum beam-widths (major × minor axis), θ , and position angle, P.A., at thecenter frequency of CO spectral windows, as well as the rms noise level in both band 6 and 7, respectively, at a velocity resolution of 0.75 km s − .Columns 8-13 give the CO line parameters: Peak flux, center velocity, and total velocity width of the line profiles of both CO transitions presentedin Figs D.1–D.6. Jura, M., Yamamoto, A., & Kleinmann, S. G. 1993, ApJ, 413, 298Kervella, P., Homan, W., Richards, A. M. S., et al. 2016, A&A, 596, A92Kim, H., Trejo, A., Liu, S.-Y., et al. 2017, Nature Astronomy, 1, 0060Lindegren, L., Lammers, U., Hobbs, D., et al. 2012, A&A, 538, A78Maercker, M., Vlemmings, W. H. T., Brunner, M., et al. 2016, A&A, 586, A5Mamon, G. A., Glassgold, A. E., & Huggins, P. J. 1988, ApJ, 328, 797 Marigo, P., Ripamonti, E., Nanni, A., Bressan, A., & Girardi, L. 2016, MNRAS,456, 23Martí-Vidal, I., Vlemmings, W. H. T., Muller, S., & Casey, S. 2014, A&A, 563,A136Matsuno, M., Nakagawa, A., Morita, A., et al. 2020, arXiv e-prints,arXiv:2003.03038
Article number, page 12 of 28 a m s t e d t e t a l . : D E A T H S T A R : N ea r by AG B s t a r s w it h A L M AA C A Table C.1.
Peak flux of detected molecular emission measured within a circular 10 ′′ aperture centered on the M-type semi-regular and irregular stars. The spectral resolution for each spectralwindow is given in Sect. 3. The peak flux error is on the order of 20%. Line Frequency Peak flux [Jy][GHz] L Pup W Hya T Mic Y Scl V1943 Sgr BK Vir V Tel SU Vel UY Cet SV Aqr SW Vir CW Cnc RT Vir R Crt
Band 6, spectral window 1
SiO ( v = J = SO (6 -5 ) 215.840 0.1 0.2 - - - - - - - - - - 0.1 0.2SO (22 , -22 , ) 216.643 0.7 0.6 0.1 - - - - - - - - - 0.3 0.2H S (2 , -2 , ) 216.710 - - - - - - - - - - - - - -SiS ( v = J = Band 6, spectral window 2
SiS (12-11) 217.818 0.2 - - - - - - - - - - - - -
Band 6, spectral window 4 SiS (13-12) 231.627 - - - - - - - - - - - - - -SO ( v =
1, 14 , -14 , ) 231.981 0.1 0.1 - - - - - - - - - - - -Si S (13-12) 232.629 - - - - - - - - - - - - - -H O ( v =
1, 5 , -6 , ) 232.687 0.2 0.2 - - - - - - - - - - - - Band 7, spectral window 1 CO (3-2) 330.588 6.4 1.8 0.8 1 0.9 1.8 1.1 1.6 1.6 0.5 5.5 0.7 3.0 4.0Na Cl (27-26) 330.805 - - - - - - - - - - - - - -
Band 7, spectral window 2 SO (11 , -12 , ) 331.580 0.2 - - - - - - - - - - - - -SO (21 , -21 , ) 332.091 1.4 0.7 - - - - - - - - - - 0.5 0.5SO (4 , -3 , ) 332.505 1.4 0.6 - - - - - - - - - - 0.8 0.8 SiS (19-18) 332.550 - - - - - - - - - - - - - -
Band 7, spectral window 3 SO (34 , -34 , ) 342.762 0.2 0.5 - - - - - - - - - - 0.3 0.2CS (7-6) 342.883 - 0.5 - - - - - - - - - - 0.2 - SiO (8-7) 342.981 1.3 30 6.3 1 4.4 4.7 2.5 2.6 2.8 0.2 10 2.0 7.0 8.0SiS ( v = J = Σ v =
1, 9 -8 ) 343.829 - 0.5 - - - - - - - - - - - -SO (8 -7 ) 344.311 1.4 6.5 1.9 0.2 1.2 1.4 0.4 0.4 1.2 0.1 1.8 0.5 3.7 3.0 Band 7, spectral window 4 H CN (4-3) 345.340 3.4 8.7 - - 0.4 - - - 0.3 - 0.1 - 2.0 2.2 A r ti c l e nu m b e r , p a g e f & Aproofs: manuscript no. deathstar1
Table C.2.
Peak flux of detected molecular emission measured within a circular 10 ′′ aperture centered on the M-type Mira stars. The spectralresolution for each spectral window is given in Sect. 3. The peak flux error is on the order of 20%. Line Frequency Peak flux [Jy][GHz] R Leo R Hya R Hor RR Aql IRC-10529 WX Psc IRC + Band 6, spectral window 1
SiO ( v = J = SO (6 -5 ) 215.840 0.1 - - 0.1 0.1 - 0.1SO (22 , -22 , ) 216.643 5.7 - - 0.2 - - -H S (2 , -2 , ) 216.710 - - - - 0.3 0.2 -SiS ( v = J = Band 6, spectral window 2
SiS (12-11) 217.818 - - - - 1.8 2.9 0.9
Band 6, spectral window 4 SiS (13-12) 231.627 - - - - 0.2 0.4 0.1SO ( v =
1, 14 , -14 , ) 231.981 - - - - - - -Si S (13-12) 232.629 - - - - 0.1 - -H O ( v =
1, 5 , -6 , ) 232.687 - - - - - 0.3 0.1 Band 7, spectral window 1 CO (3-2) 330.588 3.2 2.0 4.0 0.5 7.0 7.0 6.0Na Cl (27-26) 330.805 - - - - - 0.1 -
Band 7, spectral window 2 SO (11 , -12 , ) 331.580 - - - - - - -SO (21 , -21 , ) 332.091 0.4 - - 0.3 - - -SO (4 , -3 , ) 332.505 0.4 - - 1.9 1.2 1.2 0.8 SiS (19-18) 332.550 - - - - 0.3 0.5 0.2
Band 7, spectral window 3 SO (34 , -34 , ) 342.762 0.3 - - - - - -CS (7-6) 342.883 0.7 0.2 - - 0.7 1.4 1.0 SiO (8-7) 342.981 21.5 12.5 7.0 2.0 1.3 2.7 3.1SiS ( v = J = Σ v =
1, 9 -8 ) 343.829 - - - - - - -SO (8 -7 ) 344.311 4 1.4 1.4 1.4 0.4 0.4 0.8 Band 7, spectral window 4 H CN (4-3) 345.340 7.7 0.9 - 1.4 0.6 1.7 1.6
McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, in As-tronomical Society of the Pacific Conference Series, Vol. 376, AstronomicalData Analysis Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J.Bell, 127Menzies, J. W., Feast, M. W., & Whitelock, P. A. 2006, MNRAS, 369, 783Neri, R., Kahane, C., Lucas, R., Bujarrabal, V., & Loup, C. 1998, A&AS, 130, 1Ramstedt, S., Mohamed, S., Olander, T., et al. 2018, A&A, 616, A61Ramstedt, S., Mohamed, S., Vlemmings, W. H. T., et al. 2014, A&A, 570, L14Ramstedt, S. & Olofsson, H. 2014, A&A, 566, A145Ramstedt, S., Schöier, F. L., & Olofsson, H. 2009, A&A, 499, 515Ramstedt, S., Schöier, F. L., Olofsson, H., & Lundgren, A. A. 2006, A&A, 454,L103Ramstedt, S., Schöier, F. L., Olofsson, H., & Lundgren, A. A. 2008, A&A, 487,645Saberi, M., Vlemmings, W. H. T., & De Beck, E. 2019, A&A, 625, A81Samus’, N. N., Kazarovets, E. V., Durlevich, O. V., Kireeva, N. N., & Pas-tukhova, E. N. 2017, Astronomy Reports, 61, 80Schöier, F. L. 2000, PhD thesis, Stockholm Observatory, SE-133 36 Salt-sjöbaden, SwedenSchöier, F. L. & Olofsson, H. 2001, A&A, 368, 969Stanek, K. Z., Knapp, G. R., Young, K., & Phillips, T. G. 1995, ApJS, 100, 169van Langevelde, H., Quiroga-Nuñez, L. H., Vlemmings, W. H. T., et al. 2018, in14th European VLBI Network Symposium & Users Meeting (EVN 2018), 43Vlemmings, W. H. T., Khouri, T., De Beck, E., et al. 2018, A&A, 613, L4Whitelock, P., Menzies, J., Feast, M., et al. 1994, MNRAS, 267, 711
Article number, page 14 of 28 a m s t e d t e t a l . : D E A T H S T A R : N ea r by AG B s t a r s w it h A L M AA C A Table C.3.
Peak flux of detected molecular emission measured within a circular 10 ′′ aperture centered on the C-type semi-regular and irregular stars. The spectral resolution for each spectral windowis given in Sect. 3. The peak flux error is on the order of 20%. Line Frequency Peak flux [Jy][GHz] TW Oph NP Pup TW Hor T Ind RT Cap AQ Sgr U Hya W Ori V Aql Y Pav X Vel Y Hya SS Vir W CMa
Band 6, spectral window 1
SiO (5-4) 217.105 0.3 - - - 0.1 0.1 0.1 3.6 0.4 0.1 0.3 0.4 - - CN ( N = a CN ( N = b Band 6, spectral window 2 CN ( N = c N (24-23) 218.325 0.1 - - - - - - 0.2 0.1 - 0.1 0.2 - -C H ( N = d H ( N = e Band 6, spectral window 4 CS (5-4) 231.221 - - - - - - - 0.1 0.1 - 0.1 0.1 - - SiS (13-12) 231.627 - - - - - - - - - - - - - -SiC (10 , -9 , ) 232.534 0.2 - - - - - - 0.3 0.4 - 0.2 0.6 - - Band 7, spectral window 1 CO (3-2) 330.588 - - 0.8 - - 0.2 4.5 0.4 0.4 0.7 - - - 0.3SiC (14 , -13 , ) 330.870 0.4 - - - - - - 0.4 0.5 - 0.3 0.7 - - Band 7, spectral window 3
SiC (15 , -14 , ) 342.805 0.3 - - - - - - 0.1 0.5 - 0.4 0.3 - -CS (7-6) 342.883 3.4 - - - 1.6 0.3 1.1 5.0 6.2 - 3.6 4.5 - - SiO (8-7) 342.981 - - - - - - - - 0.1 - - - - -
Band 7, spectral window 4 H CN (4-3) 345.340 0.8 - - - 0.8 0.3 4.0 1.9 1.5 0.5 0.3 1.2 0.1 -
Notes. ( a ) J = / /
2, F1 = = ( b ) J = / /
2, F1 = = ( c ) J = / /
2, F1 = = ( d ) J = / /
2, F = ( e ) J = / /
2, F = A r ti c l e nu m b e r , p a g e f & Aproofs: manuscript no. deathstar1
Table C.4.
Peak flux of detected molecular emission measured within a circular 10 ′′ aperture centered on the C-type Mira stars. The spectralresolution for each spectral window is given in Sect. 3. The peak flux error is on the order of 20%. Line Frequency Peak flux [Jy][GHz] R Lep CZ Hya R For R Vol RV Aqr V688 Mon V1259 Ori
Band 6, spectral window 1
SiO (5-4) 217.105 1.9 0.3 2 1.9 3.2 1.0 0.8 CN ( N = a CN ( N = b Band 6, spectral window 2 CN ( N = c N (24-23) 218.325 0.2 - 0.4 0.1 0.1 0.4 0.1C H ( N = d H ( N = e Band 6, spectral window 4 CS (5-4) 231.221 0.2 - 0.2 0.2 0.3 0.1 0.2 SiS (13-12) 231.627 - - - - 0.1 0.1 0.1SiC (10 , -9 , ) 232.534 0.1 - 0.3 0.2 0.3 0.4 0.4 Band 7, spectral window 1 CO (3-2) 330.588 0.9 0.6 0.8 1.4 2.5 1.6 -SiC (14 , -13 , ) 330.870 - - 0.3 0.2 0.3 0.5 - Band 7, spectral window 3
SiC (15 , -14 , ) 342.805 - - 0.1 - - - -CS (7-6) 342.883 5.6 0.5 6.8 6.5 11 5.6 - SiO (8-7) 342.981 0.1 - 0.5 0.3 0.5 - -
Band 7, spectral window 4 H CN (4-3) 345.340 3.1 0.4 2.3 2.5 5 4 -
Notes. ( a ) J = / /
2, F1 = = ( b ) J = / /
2, F1 = = ( c ) J = / /
2, F1 = = ( d ) J = / /
2, F = ( e ) J = / / =
20 25 30 35 40 4505101520 10 20 30 40 5001020304050 30 40 50 60051015202530
30 40 50 60010203040506070
10 20 30 40051015 10 20 30 4005101520
20 25 30 35 40012345 20 25 30 35 4002468 -25 -20 -15 -10 -5 0-20246810121416 -25 -20 -15 -10 -5 005101520 -45 -40 -35 -30 -25 -20-202468101214 -45 -40 -35 -30 -25 -2005101520 -5 0 5 10 15 20051015 -5 0 5 10 15 200510152025-10 0 10 200246810 -10 0 10 2005101520 -10 0 10 2001234 -10 0 10 200123456
Fig. D.1. CO J = → → & Aproofs: manuscript no. deathstar1 -25 -20 -15 -10 -5 0 50510152025303540 -25 -20 -15 -10 -5 0 5010203040506070 0 10 20 300246810 -10 0 10 20 300510152025 -10 0 10 20 30010203040 -15 -10 -5 0 5 100510152025 -15 -10 -5 0 5 100102030405060 -30 -20 -10 0 100510152025 -30 -20 -10 0 10010203040506070
25 30 35 40 45 500102030405060
25 30 35 40 45 5001020304050607080
15 20 25 30 35 40 4505101520
15 20 25 30 35 40 45051015202530 -40 -30 -20 -10 0 1005101520253035 -40 -30 -20 -10 0 10051015202530354045 -20 0 20 40051015202530 -20 0 20 400510152025303540
Fig. D.2. CO J = → → -60 -40 -20 00510152025 -60 -40 -20 005101520253035 Fig. D.3. CO J = → → & Aproofs: manuscript no. deathstar1
10 20 30 4001234567
10 20 30 400246810 -10 0 10 20 30-0.500.511.522.533.5 -10 0 10 20 30-0.500.511.522.533.54 -10 -5 0 5 10 15051015 -10 -5 0 5 10 150510152025 5 10 15 20 250123456 -40 -30 -20051015202530354045 -40 -30 -200102030405060 -20 -10 0 10 20024681012 -20 -10 0 10 2005101520
40 50 60 70-202468101214
40 50 60 700510152025 -20 -10 0 10051015 -20 -10 0 1005101520
Fig. D.4. CO J = → → -40 -30 -20 -10 002468 -40 -30 -20 -10 0051015 -20 -10 0 10024681012 -20 -10 0 10051015 -10 0 10 20 3001234 -10 0 10 20 30-101234567 -20 -10 0 10 200246810 -20 -10 0 10 20024681012 -20 0 20 400510152025 -20 0 20 400510152025 -10 0 10 20 30-101234567 -10 0 10 20 300246810 -30 -20 -10 0 10 20024681012 -30 -20 -10 0 10 2005101520 -40 -20 0 200246810121416 -40 -20 0 200510152025 -20 -10 0 10 20 30051015202530 -20 -10 0 10 20 300510152025303540 -10 0 10 200510152025 -10 0 10 2005101520253035 Fig. D.5. CO J = → → & Aproofs: manuscript no. deathstar1
20 30 40 50 60 70051015202530 20 30 40 50 60 7001020304050
Fig. D.6. CO J = → →
30 32 34 36 38v lsr [km/s]02468 a x i s w i d t h [ a r c s ec ] L2 Pup CO(2-1)
30 32 34 36 38v lsr [km/s]02468 L2 Pup CO(3-2) 35 40 45 50v lsr [km/s]024681012 W Hya CO(2-1) 35 40 45 50v lsr [km/s]024681012 W Hya CO(3-2)
20 22 24 26 28 30v lsr [km/s]02468 a x i s w i d t h [ a r c s ec ] T Mic CO(2-1) 20 22 24 26 28 30v lsr [km/s]02468 T Mic CO(3-2)
24 26 28 30 32 34v lsr [km/s]-1012345 Y Scl CO(2-1)
24 26 28 30 32 34v lsr [km/s]-1012345 Y Scl CO(3-2)-20 -18 -16 -14 -12 -10v lsr [km/s]02468 a x i s w i d t h [ a r c s ec ] V1943 Sgr CO(2-1) -20 -18 -16 -14 -12 -10v lsr [km/s]02468 V1943 Sgr CO(3-2)
12 14 16 18 20 22v lsr [km/s]-10123456 BK Vir CO(2-1) 12 14 16 18 20 22v lsr [km/s]-10123456 BK Vir CO(3-2)-40 -35 -30 -25v lsr [km/s]-1012345 a x i s w i d t h [ a r c s ec ] V Tel CO(2-1) -40 -35 -30 -25v lsr [km/s]-1012345 V Tel CO(3-2) lsr [km/s]-1012345 SU Vel CO(2-1) 2 4 6 8 10 12 14v lsr [km/s]-1012345 SU Vel CO(3-2)0 5 10v lsr [km/s]-10123456 a x i s w i d t h [ a r c s ec ] UY Cet CO(2-1) 0 5 10v lsr [km/s]-10123456 UY Cet CO(3-2) 0 5 10 15v lsr [km/s]-1012345 SV Aqr CO(2-1) 0 5 10 15v lsr [km/s]-1012345 SV Aqr CO(3-2)
Fig. E.1.
Results from the visibility fitting to the data measured toward the M-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ff set relative to thecenter position, respectively. Article number, page 23 of 28 & Aproofs: manuscript no. deathstar1 -20 -15 -10 -5v lsr [km/s]024681012 a x i s w i d t h [ a r c s ec ] SW Vir CO(2-1) -20 -15 -10 -5v lsr [km/s]024681012 SW Vir CO(3-2) lsr [km/s]0246810 CW Cnc CO(2-1) 5 10 15 20 25v lsr [km/s]0246810 CW Cnc CO(3-2)10 15 20 25v lsr [km/s]0246810 a x i s w i d t h [ a r c s ec ] RT Vir CO(2-1) 10 15 20 25v lsr [km/s]0246810 RT Vir CO(3-2) lsr [km/s]024681012 R Crt CO(2-1) 0 5 10 15 20v lsr [km/s]024681012 R Crt CO(3-2) -5 0 5v lsr [km/s]0246810 a x i s w i d t h [ a r c s ec ] R Leo CO(2-1) -5 0 5v lsr [km/s]0246810 R Leo CO(3-2) -20 -15 -10 -5 0v lsr [km/s]02468 R Hya CO(2-1) -20 -15 -10 -5 0v lsr [km/s]02468 R Hya CO(3-2)
32 34 36 38 40 42v lsr [km/s]024681012 a x i s w i d t h [ a r c s ec ] R Hor CO(2-1) 32 34 36 38 40 42v lsr [km/s]024681012 R Hor CO(3-2) 20 25 30 35v lsr [km/s]0246810 RR Aql CO(2-1) 20 25 30 35v lsr [km/s]0246810 RR Aql CO(3-2)-30 -25 -20 -15 -10 -5v lsr [km/s]0246810 a x i s w i d t h [ a r c s ec ] IRC-10529 CO(2-1) -30 -25 -20 -15 -10 -5v lsr [km/s]0246810 IRC-10529 CO(3-2) -10 0 10 20v lsr [km/s]024681012 WX Psc CO(2-1) -10 0 10 20v lsr [km/s]024681012 WX Psc CO(3-2)
Fig. E.2.
Results from the visibility fitting to the data measured toward the M-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ff set relative to thecenter position, respectively.Article number, page 24 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA -40 -30 -20v lsr [km/s]0246810 a x i s w i d t h [ a r c s ec ] IRC+10365 CO(2-1) -40 -30 -20v lsr [km/s]0246810 IRC+10365 CO(3-2)
Fig. E.3.
Results from the visibility fitting to the data measured toward the M-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ffff
Results from the visibility fitting to the data measured toward the M-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ffff set relative to thecenter position, respectively. Article number, page 25 of 28 & Aproofs: manuscript no. deathstar1
20 25 30 35v lsr [km/s]-10123456 a x i s w i d t h [ a r c s ec ] TW Oph CO(2-1) 20 25 30 35v lsr [km/s]-10123456 TW Oph CO(3-2) lsr [km/s]-10123456 NP Pup CO(2-1) lsr [km/s]-10123456 NP Pup CO(3-2)-4 -2 0 2 4 6 8v lsr [km/s]-10123456 a x i s w i d t h [ a r c s ec ] TW Hor CO(2-1) -4 -2 0 2 4 6 8v lsr [km/s]-10123456 TW Hor CO(3-2)
10 12 14 16 18 20 22v lsr [km/s]01234 T Ind CO(2-1) 10 12 14 16 18 20 22v lsr [km/s]01234 T Ind CO(3-2) -24 -22 -20 -18 -16 -14 -12v lsr [km/s]-1012345 a x i s w i d t h [ a r c s ec ] RT Cap CO(2-1) -24 -22 -20 -18 -16 -14 -12v lsr [km/s]-1012345 RT Cap CO(3-2) 10 15 20 25 30v lsr [km/s]-10123456 AQ Sgr CO(2-1) 10 15 20 25 30v lsr [km/s]-10123456 AQ Sgr CO(3-2) -35 -30 -25v lsr [km/s]024681012 a x i s w i d t h [ a r c s ec ] U Hya CO(2-1) -35 -30 -25v lsr [km/s]024681012 U Hya CO(3-2) -10 -5 0 5 10v lsr [km/s]-1012345 W Ori CO(2-1) -10 -5 0 5 10v lsr [km/s]-1012345 W Ori CO(3-2)
45 50 55 60v lsr [km/s]-10123456 a x i s w i d t h [ a r c s ec ] V Aql CO(2-1) 45 50 55 60v lsr [km/s]-10123456 V Aql CO(3-2) -10 -5 0v lsr [km/s]-10123456 Y Pav CO(2-1) -10 -5 0v lsr [km/s]-10123456 Y Pav CO(3-2)
Fig. E.4.
Results from the visibility fitting to the data measured toward the C-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ff set relative to thecenter position, respectively.Article number, page 26 of 28amstedt et al.: DEATHSTAR: Nearby AGB stars with ALMA ACA -30 -25 -20 -15 -10v lsr [km/s]-1012345 a x i s w i d t h [ a r c s ec ] X Vel CO(2-1) -30 -25 -20 -15 -10v lsr [km/s]-1012345 X Vel CO(3-2) -15 -10 -5 0v lsr [km/s]-1012345678 Y Hya CO(2-1) -15 -10 -5 0v lsr [km/s]-1012345678 Y Hya CO(3-2)-5 0 5 10 15 20v lsr [km/s]-1012345678 a x i s w i d t h [ a r c s ec ] SS Vir CO(2-1) -5 0 5 10 15 20v lsr [km/s]-1012345678 SS Vir CO(3-2) -10 -5 0 5 10v lsr [km/s]-1012345 W CMa CO(2-1) -10 -5 0 5 10v lsr [km/s]-1012345 W CMa CO(3-2)0 10 20 30v lsr [km/s]051015 a x i s w i d t h [ a r c s ec ] R Lep CO(2-1) 0 10 20 30v lsr [km/s]051015 R Lep CO(3-2) 5 10 15 20 25v lsr [km/s]-1012345 CZ Hya CO(2-1) 5 10 15 20 25v lsr [km/s]-1012345 CZ Hya CO(3-2) -15 -10 -5 0 5 10 15v lsr [km/s]-10123456 a x i s w i d t h [ a r c s ec ] R For CO(2-1) -15 -10 -5 0 5 10 15v lsr [km/s]-10123456 R For CO(3-2) -30 -20 -10 0v lsr [km/s]-101234567 R Vol CO(2-1) -30 -20 -10 0v lsr [km/s]-101234567 R Vol CO(3-2)-10 -5 0 5 10 15v lsr [km/s]-101234567 a x i s w i d t h [ a r c s ec ] RV Aqr CO(2-1) -10 -5 0 5 10 15v lsr [km/s]-101234567 RV Aqr CO(3-2) -10 -5 0 5 10 15v lsr [km/s]02468 V688 Mon CO(2-1) -10 -5 0 5 10 15v lsr [km/s]02468 V688 Mon CO(3-2)
Fig. E.5.
Results from the visibility fitting to the data measured toward the C-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ff set relative to thecenter position, respectively. Article number, page 27 of 28 & Aproofs: manuscript no. deathstar1
30 35 40 45 50 55v lsr [km/s]0246810 a x i s w i d t h [ a r c s ec ] V1259 Ori CO(2-1) 30 35 40 45 50 55v lsr [km/s]0246810 V1259 Ori CO(3-2)
Fig. E.6.
Results from the visibility fitting to the data measured toward the C-type AGB stars of the sample discussed in this paper. The sourcename is given in the upper left corner and the transition is in the upper right corner of each plot. The upper blue and orange lines show the majorand minor axis of the best-fit Gaussian in each channel, respectively. The lower red and green lines show the RA and Dec o ffff