(Sub)stellar companions shape the winds of evolved stars
L. Decin, M. Montargès, A.M.S. Richards, C.A. Gottlieb, W. Homan, I. McDonald, I. El Mellah, T. Danilovich, S.H.J. Wallström, A. Zijlstra, A. Baudry, J. Bolte, E. Cannon, E. De Beck, F. De Ceuster, A. de Koter, J. De Ridder, S. Etoka, D. Gobrecht, M. Gray, F. Herpin, M. Jeste, E. Lagadec, P. Kervella, T. Khouri, K. Menten, T.J. Millar, H.S.P. Müller, J.M.C. Plane, R. Sahai, H. Sana, M. Van de Sande, L.B.F.M. Waters, K.T Wong, J. Yates
SSubmitted Manuscript: Confidential 1 (Sub-)stellar companions shape the winds of evolved stars
Authors:
L. Decin , M. Montargès , A. M. S. Richards , C. A. Gottlieb , W. Homan , I. McDonald , I. El Mellah , T. Danilovich , S. H. J. Wallström , A. Zijlstra , A. Baudry , J. Bolte , E. Cannon , E. De Beck , F. De Ceuster , A. de Koter , J. De Ridder , S. Etoka , D. Gobrecht , M. Gray , F. Herpin , M. Jeste , E. Lagadec , P. Kervella , T. Khouri , K. Menten , T. J. Millar , H. S. P. Müller , J. M. C. Plane , R. Sahai , H. Sana , M. Van de Sande , L. B. F. M. Waters , K. T. Wong , J. Yates Affiliations: KU Leuven, Institute of Astronomy, 3001 Leuven, Belgium. University of Leeds, School of Chemistry, Leeds LS2 9JT, United Kingdom. The University of Manchester, Jodrell Bank Centre for Astrophysics, Manchester M13 9PL, United Kingdom. Harvard-Smithsonian Center for Astrophysics, Cambridge MA 02138, USA. Open University, Walton Hall, Milton Keynes MK7 6AA, United Kingdom. KU Leuven, Center for mathematical Plasma Astrophysics, 3001 Leuven, Belgium. University of Hong Kong, Laboratory for Space Research, Pokfulam, Hong Kong. Université de Bordeaux, Laboratoire d'Astrophysique de Bordeaux, 33615 Pessac, France. Chalmers University of Technology, Onsala Space Observatory, 43992 Onsala, Sweden. University College London, Department of Physics and Astronomy, London WC1E 6BT, United Kingdom. University of Amsterdam, Anton Pannekoek Institute for Astronomy, 1090 GE Amsterdam, The Netherlands. National Astronomical Research Institute of Thailand, Chiangmai 50180, Thailand. Max-Planck-Institut für Radioastronomie, 53121 Bonn, Germany. Université Côte d’Azur, Laboratoire Lagrange, Observatoire de la Côte d’Azur, F-06304 Nice Cedex 4, France. Laboratoire d’Etudes Spatiales et d’Instrumentation en Astrophysique, Observatoire de Paris, Université Paris Sciences et Lettres, Centre National de la Recherche Scientifique, Sorbonne Université, Université de Paris, 92195 Meudon, France. Queen’s University Belfast, Astrophysics Research Centre, Belfast BT7 1NN, United Kingdom. Universität zu Köln, I. Physikalisches Institut, 50937 Köln, Germany. California Institute of Technology, Jet Propulsion Laboratory, Pasadena CA 91109, USA. SRON Netherlands Institute for Space Research, NL-3584 CA Utrecht , The Netherlands. Institut de Radioastronomie Millimétrique, 38406 Saint Martin d’Hères, France. ubmitted Manuscript: Confidential 2 *Corresponding author. Email: [email protected].
Abstract:
Binary interactions dominate the evolution of massive stars, but their role is less clear for low and intermediate mass stars. The evolution of a spherical wind from an Asymptotic Giant Branch (AGB) star into a non-spherical planetary nebula (PN) could be due to binary interactions. We observe a sample of AGB stars with the Atacama Large Millimeter/submillimeter Array (ALMA), finding that their winds exhibit distinct non-spherical geometries with morphological similarities to PNe. We infer that the same physics shapes both AGB winds and PNe. The morphology and AGB mass-loss rate are correlated. These characteristics can be explained by binary interaction. We propose an evolutionary scenario for AGB morphologies which is consistent with observed phenomena in AGB stars and PNe.
Main Text:
At the end of their life, low and intermediate mass (0.8 to 8 solar masses, M ☉ ) stars turn into luminous cool red giant stars when ascending the AGB. Our Sun will reach that phase in ~7.7 Gyr from now (1) . During the AGB phase, the star’s radius may become as large as one astronomical unit (au) and its luminosity may reach thousands of times that of the Sun. The AGB phase lasts between ~
Myr, the more massive stars are short-lived (2) . At the start of the AGB phase, stars are oxygen-rich with a carbon-to-oxygen (C/O) ratio lower than 1, and are referred to as M-type stars. During the AGB phase, carbon is fused in the stellar core and brought to the surface by convection. Eventually, the C/O ratio gets larger than 1, leading to a carbon star. The AGB phase is characterized by a stellar wind with mass-loss rate greater than ~10 *+ M ☉ ,- *. . The increase in luminosity while ascending the AGB induces an increase in mass-loss rate, of up to ~ */ M ☉ ,- *. (3) . For stars with mass-loss rate greater than *0 M ☉ ,- *. , the mass-loss rate exceeds the envelope’s hydrogen nuclear burning rate, and mass loss determines the further stellar evolution (4) . The wind strips away the star’s outer envelope. At the moment that the envelope is less than about 1% of the stellar mass, the star becomes a post-AGB star (5) . During this short evolutionary phase which takes a few 1000 yr, the temperature of the star increases at constant luminosity and it becomes a planetary nebula (PN), characterized by a hot central star which ionizes the gas ejected during the previous red giant phase. The lifetime of PNe is roughly 20 000 yr. The PN nebula then disperses quickly, leaving an inert white dwarf which slowly cools (6) . One puzzling aspect about PNe formation concerns the mechanism that shapes the nebulae into a wide range of morphologies, including elliptical, bipolar, and `butterfly’-shaped geometries (7) . While ~80% of the AGB stars have a wind with overall spherical symmetry (8) , less than of PNe are circularly symmetric (9,10) . Various hypotheses - including rapidly spinning or strongly magnetic single stars (11) - have been proposed to explain this morphological metamorphosis, but they have been questioned because strong asymmetries are not formed efficiently ( . More recently, short-period (orbital period ≲ 10 days) binary systems (orbital separation ) surrounded by a common gaseous envelope - referred to as the common-envelope phase - have become the favored hypothesis ( ). The proposed PN shaping mechanisms operate over a short time, either during the final few hundred years of the AGB or during the early post-AGB phase (14) . Identifying the shaping mechanism and its time of occurrence are observationally challenging owing to the short lifetime of the post-AGB and PN stages; the strong observational bias towards detecting binary post-AGB stars and PNe with short orbital periods (15) ; and the high mass-loss rates at the end of the AGB phase, which surrounds the star with high optical depth material and obscures the inner workings. ubmitted Manuscript: Confidential 3 During the last few years, observations at high spatial resolution have shown that AGB winds may exhibit small-scale structural complexity - including arcs, shells, bipolar structures, clumps, spirals, tori, and rotating disks (16, 17) - embedded in a smooth, radially outflowing wind. Only about a dozen AGB winds have been studied in detail (18) . It has not been possible to determine any systematic morphological change during the AGB evolution, and the transition from the smaller scale structures observed during the AGB to the PN morphologies is not understood . In the ALMA ATOMIUM - ALMA Tracing the Origins of Molecules In dUst-forming oxygen-rich M-type stars - Program (18) , we have observed a sample of oxygen-rich AGB stars spanning a range of (circum)stellar parameters and AGB evolutionary stages (Table S1). We study the wind morphology at a spatial resolution of ~0.24″ and ~1″ using the rotational lines of CO J =2→1, SiO J =5→4, and SiO J =6→5 in the ground vibrational state, with J the rotational angular momentum quantum number. These two molecules have large fractional abundances with respect to molecular hydrogen and yield complementary information on the density (CO) and on the morphological and dynamical properties close to the stellar surface (SiO). Figure 1 shows a gallery of the CO observations. None of the sources has a smooth, spherical geometry. The images exhibit various structures in common with post-AGB stars and PNe: bipolar morphologies with a central waist, equatorial density enhancements (EDE) and disk-like geometries, eye-like shapes, spiral-like structures, and arcs at regularly spaced intervals (18) . We infer from these images that the same physical mechanism shapes both AGB winds and PNe. These data constrain the mechanism shaping the winds while it is in operation in a sample of stars with a range of AGB properties, i.e. cover the moment in time when AGB morphologies are being transformed into aspherical geometries. Combining the CO and SiO data provides an observational criterion (Fig. S2) for classifying the prevailing wind morphologies (Table S2). We find a correlation between the AGB mass-loss rate (:)̇ and the prevailing geometry (Table 1), with a Kendall’s rank correlation coefficient = of 0.79 (Fig. S3-S4, (18) ). A dynamically complex EDE is often observed for oxygen-rich AGB stars with low mass-loss rates (which we refer to as ‘Class 1’), a bipolar structure tends to be dominant for stars with medium mass-loss rates (‘Class 2’), while the winds of high mass-loss rate stars preferentially exhibit a spiral-like structures (‘Class 3’). Other oxygen-rich AGB stars whose geometry was deduced from previous observations follow this same schematic order (Table S3). This correlation suggests that a common mechanism controls the wind morphology throughout the AGB phase, and that it depends on the mass-loss rate. Among the mechanisms proposed to explain asphericity, binary models including long-period systems ( ≳ 1 yr, ) (19-21) can explain both the morphologies and the correlation with mass-loss rate (18) . Stellar evolution models (22) show that the majority of AGB stars with a mass-loss rate above *0 M ☉ ,- *. - including all those in the ATOMIUM sample - have masses above ~1.5 M ☉ . Planet and stellar binary population statistics (23,24) indicate that stars with these masses have an average number of companions (with mass above ~5 Jupiter mass) ≳ 1 (18) . Binary interaction is known to dominate the evolution of more massive stars (25) . We conjecture that (sub-)stellar binary interaction is the dominant wind shaping agent for the majority of AGB stars with mass-loss rate exceeding the nuclear burning rate. Our conjecture is supported by the growing number of aspherical PNe detected whose binary central stars have a long-period orbit ( ≳ 1 yr) not undergoing a common-envelope evolution (18). ubmitted Manuscript: Confidential 4 On the assumption that binary interaction dominates, we derive (18) an analytical relation @ . = 10 *B @ C = 8.32 1(1 − F) G I JK L3MC : ☉ N ./P J : ∗ : ☉ N R 71 78S *P/G
T :̇10 *B M ☉ ,- *. U *. that estimates the probability of a binary system forming a (possibly rotating) EDE structure (large value of @ . ) or being dominated by a spiral-like structure (low value of @ . ) (18) . Here F denotes the eccentricity of the orbit, H I the fraction of the stellar wind mass present at a distance - = 7 , and : ∗ and K L3MC are the mass of the primary star and companion, respectively. This relation holds for a wind velocity at - = 7 that is lower than the orbital velocity, and can be easily reformulated for the case of a high wind velocity (18) . Our analytical relation supports the correlation observed in the ATOMIUM data. Higher mass-loss rate or orbital separation leads to lower injection of angular momentum into the initially spherical AGB wind by interactions with the orbiting companion, and weaker shaping of the material along the orbital plane into an EDE, a circumbinary disk or an accretion disk (18) . Wide binaries, with a separation of up to several tens of astronomical units, produce a spiral-like structure (19) . The transition of the wind morphology during the AGB phase we observe applies to oxygen-rich AGB stars, whereas carbon-rich winds most often display a (broken) spiral/arc-like structure (18) . We attribute this differentiation to stronger wind acceleration for carbon-rich than oxygen-rich stars, due to the different dust composition (Fig. S5). Stronger acceleration results in a smaller geometrical region in which the velocity field is non-radial, in a lower probability of forming an EDE, and in a smaller radius beyond which the wind shows a self-similar morphology (18) . This implies that carbon-rich AGB stars are more commonly surrounded by an expanding, self-similar, spiral structure, which is consistent with past observations (18) . Although EDEs may form in carbon-rich winds, EDEs will be most commonly found around low mass-loss rate oxygen-rich AGB stars with slowly accelerating winds (18) . Observations of binary companions around AGB progenitor stars indicate the highest fraction of binary companions are at an orbital distance greater than ~20 au (23) . We calculate (18) that those orbits will widen during the AGB evolution as the mass-loss rate increases (Fig. S6-S7). This implies that early-type low mass-loss rate AGB stars will often have an EDE, with complex flow patterns, and the wind of late-type high mass-loss rate AGB stars are mainly shaped by spiral structures (Fig. 2). Our results also imply that the effects of planets around evolved stars are more easily detected in early-type oxygen-rich AGB stars (18) . Our proposed evolutionary and chemical scheme for AGB wind morphologies can explain multiple AGB, post-AGB and PN phenomena (18), including why (i) circular detached shells are only detected around carbon-rich AGB stars (26) ; (ii) disks are mainly found around oxygen-rich post-AGB and PN binaries ( ); (iii) carbon-rich stars can be surrounded by silicate dust ( ); (iv) PNe in the bulge of the Milky Way can have a mixed carbon/oxygen chemistry ( ); (v) post-AGB envelopes can be classified according to two distinct morphological types ( ); (vi) post-AGB binaries can have non-zero eccentricities with values as high as 0.3 ( ); and (vii) the low fraction of round PNe ( ). ubmitted Manuscript: Confidential 5 Fig. 1. Gallery of AGB winds.
Emission maps of twelve stars are shown, derived from the ATOMIUM CO J =2 → ″ . Full channel maps and position-velocity diagrams for each source are shown in Figures S8-S65. Fig. 1A: S Pav, Fig. 1B: T Mic, Fig. 1C: U Del, Fig 1D: V PsA, Fig. 1E: R Hya, Fig. 1F: U Her, Fig. 1G: V . Gru, Fig. 1H: R Aql, Fig 1I: W Aql, Fig. 1J: GY Aql, Fig. 1K: IRC -10529, and Fig. 1L: IRC +10011. For two stars (RW Sco and SV Aqr) the signal-to-noise of the data was too low to produce a three-colour map, although the individual channels show clear signs of asymmetry (Fig. S20, Fig. S28 (18) ). ubmitted Manuscript: Confidential 6 Fig. 2. Schematic illustration of our inferred evolution of wind morphology during the AGB phase . Most (sub-)stellar companions have initial orbits ( WXW ) greater than 20 au (24) . These orbits widen during AGB evolution because the stellar mass decreases. Binary systems with close-by companions often have a high-density EDE and accretion disk (shown in orange) and complex inner wind dynamics. For increasingly wider orbits and higher mass-loss rates, the prevailing outflow morphology first transitions to a bipolar structure (in blue) and then to a regularly spaced spiral structure (in black). EDEs or accretion disks can be present at these later stages, but at lower density.
Table 1. Wind characteristics of the AGB stars in the ATOMIUM sample.
Columns 1-6 contain the source name, luminosity in units of solar luminosity Y ⊙ , mass-loss rate, wind velocity based on the CO J =2→1 line ([ \]^_ ) , the identification of arc morphologies in the CO J =2→1 channel map, and the SiO wind dynamics characterizing the velocity field ( `) in the vicinity of the AGB star as derived from our ALMA data (Fig. S8-S65, Table S2 (18) ). The stars are ordered by increasing mass-loss rate. The last column indicates objects with similar wind characteristics. Class 1 designates sources with multiple density arcs and dynamically complex inner wind structures, with signs of a biconical outflow and/or rotation, shaping the wind in an equatorial density enhancement (EDE). Class 2 indicates a bipolar structure, some with additional hourglass morphology in the CO channel maps. Class 3 has large density arc(s) often with a recognizable spiral-like structure. Name Lumi- nosity ( a ⊙ ) Mass-loss rate ( b ⊙ yr -1 ) ` cdef (km s -1 ) CO morphology arcs (a) SiO inner wind dynamics (b)
ATOMIUM classification S Pav 4859 *+
14 (x) Skewed rotating v -field Class 1 T Mic 4654 *+
14 x Skewed rotating v -field Class 1 U Del 4092 *0
17 c-xx Bipolar/rotating flow Class 2 RW Sco 7714 *0
19 c-xx - (low S/N) Class 2 ubmitted Manuscript: Confidential 7
V PsA 4092 *0
20 c-xx Bipolar flow Class 2 SV Aqr 4000 *0
16 c-xx - (low S/N) Class 2 R Hya 7375 *0
22 o-xx Skewed rotating v -field Class 2 U Her 8026 *0
20 a-xx Complex dynamics Class 3 j k Gru 4683 *0
65 o-xx Bipolar/rotating flow
Class 2 R Aql 4937 *B
16 xxx - Class 3 W Aql 9742 *B
25 xxx Complex dynamics Class 3 GY Aql 9637 *B
18 xxx Complex dynamics Class 3 IRC -10529 14421 *B
20 xxx Bipolar/rotating flow Class 3 IRC +10011 13914 *p
23 xxx Complex dynamics Class 3 (a) (x): faint arc, x: several arcs with extent < 180° , c-xx: circular/elliptical arc centered around the star, o-xx: arcs symmetrically offset from the central star, a-xx: pronounced asymmetric arcs, xxx: more than 1 arc with extent > 270° , linked to a (complex) spiral structure. (b) ‘Bipolar/rotating flow’ indicates a directed bipolar flow or an EDE/disk-like structure, sometimes with Keplerian rotation; ‘skewed rotating v -field’ denotes systematic, but complex, signs of rotation and the [ = 0 signature in the map of the intensity weighted velocity field (moment1-map) is skewed; ‘complex dynamics’ refers to a clear blue-shifted and red-shifted velocity structure in the moment1-map, but no obvious systematic rotation can be deduced; ‘-’ denotes that no conclusion could be drawn, sometimes owing to too low a signal-to-noise ratio of the SiO data (‘low S/N’). ubmitted Manuscript: Confidential 8 References and Notes: K. P. Schröder, R. C. Smith, Distant future of the Sun and Earth revisited.
MNRAS , 155–163 (2008). 2.
A. I. Karakas. Helium enrichment and carbon-star production in metal-rich populations,
MNRAS , 347–358 (2014). 3.
H. J. Habing, H. Olofsson, Eds.,
Asymptotic Giant Branch Stars (Springer-Verlag New York, Inc., 2004). 4.
I. McDonald, E. De Beck, A. A. Zijlstra, E. Lagadec, Pulsation-triggered dust production by asymptotic giant branch stars.
MNRAS , 4984–4999 (2018). 5.
M. M. Miller Bertolami, New models for the evolution of post-asymptotic giant branch stars and central stars of planetary nebulae.
Astron. Astrophys. , A25 (2016). 6.
K. Gesicki, A. A. Zijlstra, M. M. Miller Bertolami, The mysterious age invariance of the planetary nebula luminosity function bright cut-off.
Nature Astron. , 580–584 (2018). 7. R. Sahai, J. T. Trauger, Multipolar Bubbles and Jets in Low-Excitation Planetary Nebulae: Toward a New Understanding of the Formation and Shaping of Planetary Nebulae.
Astron. J. , 1357–1366 (1998). 8.
R. Neri, C. Kahane, R. Lucas, V. Bujarrabal, C. Loup, A CO ( J =1-0) and ( J =2-1) atlas of circumstellar envelopes of AGB and post-AGB stars. Astron. Astrophys. Suppl.
Ser. , 1–64 (1998). 9.
R. Sahai, M. R. Morris, G. G. Villar, Young Planetary Nebulae: Hubble Space Telescope Imaging and a New Morphological Classification System.
Astron. J. , 134 (2011). 10.
Q. A. Parker, A. Acker, D. J. Frew, M. Hartley, A. E. J. Peyaud, F. Ochsenbein et al.,
The Macquarie/AAO/Strasbourg H t Planetary Nebula Catalogue: MASH.
MNRAS , 79–94 (2006). 11.
G. García-Segura, N. Langer, M. Różyczka, J. Franco, Shaping Bipolar and Elliptical Planetary Nebulae: Effects of Stellar Rotation, Photoionization Heating, and Magnetic Fields.
Astrophys. J . , 767–781 (1999). 12. J. Nordhaus, E. G. Blackman, A. Frank, Isolated versus common envelope dynamos in planetary nebula progenitors.
MNRAS. , 599–608 (2007). 13.
B. Miszalski, A. Acker, A. F. J. Moffat, Q. A. Parker, A. Udalski, Binary planetary nebulae nuclei towards the Galactic bulge. I. Sample discovery, period distribution, and binary fraction.
Astron. Astrophys. , 813–825 (2009). 14.
B. Balick, A. Frank, Shapes and Shaping of Planetary Nebulae.
Ann. Rev. Astron. Astrophys. , 439–486 (2002). 15. H. van Winckel, Post-AGB Stars.
Ann. Rev. Astron. Astrophys . , 391–427 (2003). 16. N. Mauron, P. J. Huggins, Imaging the circumstellar envelopes of AGB stars.
Astron. Astrophys. , 257–268 (2006). ubmitted Manuscript: Confidential 9
P. Kervella, W. Homan, A. M. S. Richards, L. Decin, I. McDonald, M. Montargès et al.,
ALMA observations of the nearby AGB star L2 Puppis. I. Mass of the central star and detection of a candidate planet.
Astron. Astrophys. , A92 (2016).
Materials and Methods are available as supplementary materials.
H. Kim, R. E. Taam, Wide Binary Effects on Asymmetries in Asymptotic Giant Branch Circumstellar Envelopes.
Astrophys. J. , 59 (2012).
Z. Chen, A. Frank, E. G. Blackman, J. Nordhaus, J. Carroll-Nellenback, Mass transfer and disc formation in AGB binary systems.
MNRAS , 4465–4477 (2017).
M. I. Saladino, O. R. Pols, E. van der Helm, I. Pelupessy, S. Portegies Zwart, Gone with the wind: the impact of wind mass transfer on the orbital evolution of AGB binary systems.
Astron. Astrophys. , A50 (2018).
P. Marigo, A. Bressan, A. Nanni, L. Girardi, M. L. Pumo, Evolution of thermally pulsing asymptotic giant branch stars - I. The COLIBRI code.
MNRAS
M. Moe, R. Di Stefano, Mind Your Ps and Qs: The Interrelation between Period (P) and Mass-ratio (Q) Distributions of Binary Stars.
Astrophys. J. Suppl. Ser. , 15 (2017).
B. J. Fulton, E. A. Petigura, The California-Kepler Survey. VII. Precise Planet Radii Leveraging Gaia DR2 Reveal the Stellar Mass Dependence of the Planet Radius Gap.
Astron. J. , 264 (2018).
H. Sana, S. E. de Mink, A. de Koter, N. Langer, C. J. Evans, M. Gieles et al.,
Binary Interaction Dominates the Evolution of Massive Stars.
Science , 444 (2012).
H. Olofsson, U. Carlstrom, K. Eriksson, B. Gustafsson, L. A.Willson, Bright carbon stars with detached circumstellar envelopes - A natural consequence of helium shell flashes
Astron. Astrophys. , L13–L16 (1990).
F. J. Molster, I. Yamamura, L. B. F. Waters, L. Å. Nyman, H. U. Kaüfl, T. de Jong, C. Loup, IRAS 09425-6040: A carbon star surrounded by highly crystalline silicate dust.
Astron. Astrophys. , 923–929 (2001).
L. Guzman-Ramirez, A. A. Zijlstra, R. Níchuimín, K. Gesicki, E. Lagadec, T. J. Millar et al. , Carbon chemistry in Galactic bulge planetary nebulae.
MNRAS , 1667–1678 (2011).
T. Ueta, M. Meixner, M. Bobrowsky, A Hubble Space Telescope Snapshot Survey of Proto-Planetary Nebula Candidates: Two Types of Axisymmetric Reflection Nebulosities.
Astrophys. J. , 861–884 (2000).
G.-M. Oomen, H. Van Winckel, O. Pols, G. Nelemans, A. Escorza, R. Manick et al. , Orbital properties of binary post-AGB stars.
Astron. Astrophys . , A85 (2018). J. L. Beuzit, A. Vigan, D. Mouillet, K. Dohlen, R. Gratton, A. Boccaletti et al.,
SPHERE: the exoplanet imager for the Very Large Telescope.
Astron. Astrophys.
B. Lopez, S. Lagarde, W. Jaffe, R. Petrov, M. Scholler, P. Antonelli et al.,
An Overview of the MATISSE Instrument — Science, Concept and Current Status.
The Messenger , 5–12 (2014).
I. McDonald, M. Trabucchi, The onset of the AGB wind tied to a transition between sequences in the period-luminosity diagram.
MNRAS , 4678–4682 (2019). ubmitted Manuscript: Confidential 10
E. Vassiliadis, P. R. Wood, Evolution of Low- and Intermediate-Mass Stars to the End of the Asymptotic Giant Branch with Mass Loss.
Astrophys. J. , 641 (1993).
E. De Beck, L. Decin, A. de Koter, K. Justtanont, T. Verhoelst, F. Kemper, K. M. Menten, Probing the mass-loss history of AGB and red supergiant stars from CO rotational line profiles. II. CO line survey of evolved stars: derivation of mass-loss rate formulae.
Astron. Astrophys. , A18 (2010).
I. McDonald, A. A. Zijlstra, Pulsation-triggered Mass Loss from AGB Stars: The 60 Day Critical Period.
Astrophys. J. , L38 (2016).
Gaia Collaboration, A. G. A. Brown, A. Vallenari, T. Prusti, J. H. J. de Bruijne, C. Babusiaux et al.,
Gaia Data Release 2. Summary of the contents and survey properties.
Astron. Astrophys. , A1 (2018).
M. A. C. Perryman, L. Lindegren, J. Kovalevsky, E. Hog, U. Bastian, P. L. Bernacca et al.,
The Hipparcos Catalogue.
Astron. Astrophys. , 501–504 (1997).
H. Olofsson, D. Gonzalez Delgado, F. Kerschbaum, F. L. Schöier, Mass loss rates of a sample of irregular and semiregular M-type AGB-variables.
Astron. Astrophys.
P. Marigo, L. Girardi, A. Bressan, M. A. T. Groenewegen, L. Silva, G. L. Granato, Evolution of asymptotic giant branch stars. II. Optical to far-infrared isochrones with improved TP-AGB models.
Astron. Astrophys. , 883–905 (2008).
I. L. Andronov, L. L. Chinarova, Two-Component Variability of the Semi-Regular Pulsating Star U Delphini.
Odessa Astronomical Publications , 148 (2012). A. Richichi, I. Percheron, M. Khristoforova, CHARM2: An updated Catalog of High Angular Resolution Measurements.
Astron. Astrophys. , 773–777 (2005).
M. A. T. Groenewegen, F. Baas, J. A. D. L. Blommaert, R. Stehle, E. Josselin, R. P. J. Tilanus, Millimeter and some near infra-red observations of short-period Miras and other AGB stars.
Astron. Astrophys. Suppl. Ser. , 197–224 (1999).
N. Vogt, A. Contreras-Quijada, I. Fuentes-Morales, S. Vogt-Geisse, C. Arcos, C. Abarca et al.,
Determination of Pulsation Periods and Other Parameters of 2875 Stars Classified as MIRA in the All Sky Automated Survey (ASAS).
Astrophys. J. Suppl. Ser. , 6 (2016).
A. A. Zijlstra, T. R. Bedding, J. A. Mattei, The evolution of the Mira variable R Hydrae.
MNRAS , 498–510 (2002).
W. H. T. Vlemmings, H. J. van Langevelde, Improved VLBI astrometry of OH maser stars.
Astron. Astrophys. , 547–553 (2007).
K. Young. A CO(3–2) Survey of Nearby Mira Variables.
Astrophys. J. , 872 (1995).
C. Paladini, F. Baron, A. Jorissen, J. B. Le Bouquin, B. Freytag, S. van Eck et al.,
Large granulation cells on the surface of the giant star V . Gruis.
Nature , 310–312 (2018). ubmitted Manuscript: Confidential 11
L. Doan, S. Ramstedt, W. H. T. Vlemmings, S. Hofner, E. De Beck, F. Kerschbaum et al.,
The extended molecular envelope of the asymptotic giant branch star V . Gruis as seen by ALMA. I. Large-scale kinematic structure and CO excitation properties.
Astron. Astrophys. , A28 (2017).
S. Ramstedt, S. Mohamed, W. H. T. Vlemmings, T. Danilovich, M. Brunner, E. De Beck et al.,
The circumstellar envelope around the S-type AGB star W Aql. Effects of an eccentric binary orbit.
Astron. Astrophys. , A126 (2017).
C. Loup, T. Forveille, A. Omont, J. F. Paul, CO and HCN observations of circumstellar envelopes. A catalogue - mass loss rates and distributions.
Astron. Astrophys. Suppl. Ser. , 291–377 (1993). E. A. Olivier, P. Whitelock, F. Marang, Dust-enshrouded asymptotic giant branch stars in the solar neighbourhood.
MNRAS , 490–514 (2001).
P. A. Whitelock, M. W. Feast, F. Van Leeuwen, AGB variables and the Mira period-luminosity relation.
MNRAS , 313–323 (2008).
W. Homan, L. Decin, A. de Koter, A. J. van Marle, R. Lombaert, W. Vlemmings, Simplified models of stellar wind anatomy for interpreting high-resolution data. Analytical approach to embedded spiral geometries.
Astron. Astrophys. , A118 (2015).
L. Decin, W. Homan, T. Danilovich, A. de Koter, D. Engels, L. B. F. M. Waters et al.,
Reduction of the maximum mass-loss rate of OH/IR stars due to unnoticed binary interaction.
Nature Astron. , 408–415 (2019). S. Ramstedt, F. L. Schöier, H. Olofsson, A. A. Lundgren, On the reliability of mass-loss-rate estimates for AGB stars.
Astron. Astrophys. , 645–657 (2008).
F. van Leeuwen, Validation of the new Hipparcos reduction.
Astron. Astrophys. , 653–664 (2007).
E. Humphreys, R. Miura, C. L. Brogan, J. Hibbard, T. R. Hunter, R. Indebetouw, The ALMA Science Pipeline: Current Status. In
Proceedings of the 2016 ALMA Conference, “Half a Decade of ALMA: Cosmic Dawn Transformed”, Sept. 20-23, 2016, Indian Wells, CA, USA, id.1 , 1 (2016).
A. Avison, Lumberjack. https://github.com/adam-avison/LumberJack (2019).
K. Willacy, T. J. Millar, Chemistry in oxygen-rich circumstellar envelopes.
Astron. Astrophys. , 237–248 (1997).
G. Winnewisser, S. P. Belov, T. Klaus, R. Schieder, Sub-Doppler Measurements on the Rotational Transitions of Carbon Monoxide.
Journal of Molecular Spectroscopy , 468–472 (1997).
F. L. Schöier, F. F. S. van der Tak, E. F. van Dishoeck, J. H. Black, An atomic and molecular database for analysis of submillimetre line observations.
Astron. Astrophys. , 369–379 (2005).
C. P. Endres, S. Schlemmer, P. Schilke, J. Stutzki, H. S. P. Müller, The Cologne Database for Molecular Spectroscopy, CDMS, in the Virtual Atomic and Molecular Data Centre, VAMDC.
J. Mol. Spectrosc. , 95–104 (2016). ubmitted Manuscript: Confidential 12
H. S. P. Müller, S. Spezzano, L. Bizzocchi, C. A. Gottlieb, C. Degli Esposti, M. C. McCarthy, Rotational Spectroscopy of Isotopologues of Silicon Monoxide, SiO, and Spectroscopic Parameters from a Combined Fit of Rotational and Rovibrational Data.
J. Phys. Chem. A , 13843–13854 (2013).
A. G. G. M. Tielens, L. B. F. M. Waters, F. J. Molster, K. Justtanont, Circumstellar Silicate Mineralogy.
Astrophys. Space Sci. , 415–426 (1998).
R. Lucas, V. Bujarrabal, S. Guilloteau, R. Bachiller, A. Baudry, J. Cernicharo et al.,
Interferometric observations of SiO v =0 thermal emission from evolved stars. Astron. Astrophys. , 491–500 (1992).
R. Sahai, J. H. Bieging, Interferometric Observations of Non-Maser SiO Emission From Circumstellar Envelopes of AGB Stars: Acceleration Regions and SiO Depletion.
Astron. J. , 595 (1993).
J. L. Verbena, V. Bujarrabal, J. Alcolea, M. Gómez-Garrido, A. Castro-Carrizo, Interferometric observations of SiO thermal emission in the inner wind of M-type AGB stars IK Tauri and IRC+10011.
Astron. Astrophys. , A107 (2019).
M. Van de Sande, C. Walsh, T. P. Mangan, L. Decin, Chemical modelling of dust-gas chemistry within AGB outflows - I. Effect on the gas-phase chemistry.
MNRAS , 2023–2041 (2019).
C. R. O’Dell, B. Balick, A. R. Hajian, W. J. Henney, A. Burkert, Knots in Nearby Planetary Nebulae.
Astron. J. , 3329–3347 (2002).
C. R. O’Dell, P. R. McCullough, M. Meixner, Unraveling the Helix Nebula: Its Structure and Knots.
Astron. J. , 2339–2356 (2004).
K. Y. L. Su, Y. H. Chu, G. H. Rieke, P. J. Huggins, R. Gruendl, R. Napiwotzki et al.,
A Debris Disk around the Central Star of the Helix Nebula?
Astrophys. J. , L41–L45 (2007).
T. Ueta, K. Murakawa, M. Meixner, Hubble Space Telescope NICMOS Imaging Polarimetry of Proto-Planetary Nebulae. II. Macromorphology of the Dust Shell Structure via Polarized Light.
Astron. J. , 1345–1360 (2007).
M. A. Guerrero, Y.-H. Chu, A. Manchado, K. B. Kwitter, Physical Structure of Planetary Nebulae. I. The Owl Nebula.
Astron. J. , 3213–3221 (2003).
M. Cohen, H. Van Winckel, H. E. Bond, T. R. Gull, Hubble Space Telescope Imaging of HD 44179, The Red Rectangle.
Astron. J. , 2362–2377 (2004).
R. Wesson, X. W. Liu, Physical conditions in the planetary nebula NGC 6543.
MNRAS , 1026–1042 (2004).
S. Ertel, D. Kamath, M. Hillen, H. van Winckel, J. Okumura, R. Manick et al.,
Resolved Imaging of the AR Puppis Circumbinary Disk.
Astron. J. , 110 (2019).
V. Bujarrabal, A. Castro-Carrizo, J. Alcolea, M. Santander-García, H. van Winckel, C. Sanchez Contreras, Further ALMA observations and detailed modeling of the Red´ Rectangle.
Astron. Astrophys. , A92 (2016). ubmitted Manuscript: Confidential 13
W. Homan, T. Danilovich, L. Decin, A. de Koter, J. Nuth, M. Van de Sande, ALMA detection of a tentative nearly edge-on rotating disk around the nearby AGB star R Doradus.
Astron. Astrophys. , A113 (2018).
V. Bujarrabal, A. Castro-Carrizo, J. Alcolea, H. Van Winckel, Detection of Keplerian dynamics in a disk around the post-AGB star AC Herculis.
Astron. Astrophys. , L7 (2015).
W. Homan, A. Richards, L. Decin, P. Kervella, A. de Koter, I. McDonald, K. Ohnaka, ALMA observations of the nearby AGB star L Puppis. II. Gas disk properties derived from CO and CO J = 3-2 emission.
Astron. Astrophys. , A5 (2017).
I. S. Shklovsky, The nature of planetary nebulae and their nuclei.
Astronomicheskii Zhurnal , 315–329 (1956). S. Höfner, H. Olofsson, Mass loss of stars on the asymptotic giant branch. Mechanisms, models and measurements.
Astron. Astrophys. Rev. , 1 (2018). H. J. G. L. M. Lamers, J. P. Cassinelli,
Introduction to Stellar Winds (Cambridge University Press, 1999).
W. Homan, J. Boulangier, L. Decin, A. de Koter, Simplified models of circumstellar morphologies for interpreting high-resolution data. Analytical approach to the equatorial density enhancement.
Astron. Astrophys. , A91 (2016).
A. Castro-Carrizo, R. Lucas, V. Bujarrabal, F. Colomer, J. Alcolea, SiO emission from a huge, detached shell in IRC +10420.
Astron. Astrophys. , L34–L37 (2001).
F. Hoyle, R. A. Lyttleton, The effect of interstellar matter on climatic variation.
Proceedings of the Cambridge Philosophical Society , 405 (1939). H. Bondi, F. Hoyle, On the mechanism of accretion by stars.
MNRAS , 273 (1944).
L. Decin, A. M. S. Richards, T. Danilovich, W. Homan, J. A. Nuth, ALMA spectral line and imaging survey of a low and a high mass-loss rate AGB star between 335 and 362 GHz.
Astron. Astrophys. , A28 (2018).
Agresti, A,
Analysis of Ordinal Categorical Data (New York: John Wiley & Son, 2010).
McFadden, D, Quantitative methods for analysing travel behavior of individuals: Some recent developments. In D. Hensher, P. Stopher (eds.)
Behavioural travel modelling , 279–318 (1979).
N. Mastrodemos, M. Morris, Bipolar Pre-Planetary Nebulae: Hydrodynamics of Dusty Winds in Binary Systems. II. Morphology of the Circumstellar Envelopes.
Astrophys. J. , 357–380 (1999).
B. Paczyński, Evolutionary Processes in Close Binary Systems.
Ann. Rev. Astron. Astrophys. , 183 (1971). S. Mohamed, P. Podsiadlowski, Mass Transfer in Mira-type Binaries.
Baltic Astronomy , 88–96 (2012). T. Theuns, H. M. J. Boffin, A. Jorissen, Wind accretion in binary stars - II. Accretion rates.
MNRAS , 1264–1276 (1996). ubmitted Manuscript: Confidential 14
I. El Mellah, J. O. Sundqvist, R. Keppens, Wind Roche lobe overflow in high-mass X-ray binaries. A possible mass-transfer mechanism for ultraluminous X-ray sources.
Astron. Astrophys. , L3 (2019).
Z. Chen, N. Ivanova, J. Carroll-Nellenback, A 3D Radiation Hydrodynamic AGB Binary Model.
Astrophys. J. , 110 (2020).
N. Mauron, P. J. Huggins, Multiple shells in the circumstellar envelope of IRC +10216.
Astron. Astrophys. , 203–208 (1999).
S. Ramstedt, S. Mohamed, W. H. T. Vlemmings, M. Maercker, R. Montez, A. Baudry et al.,
The wonderful complexity of the Mira AB system.
Astron. Astrophys. , L14 (2014).
K. T. Wong, T. Kamiński, K. M. Menten, F. Wyrowski, Resolving the extended atmo-sphere and the inner wind of Mira (o Ceti) with long ALMA baselines.
Astron. Astrophys. , A127 (2016).
W. Homan, A. Richards, L. Decin, A. de Koter, P. Kervella, An unusual face-on spiral in the wind of the M-type AGB star EP Aquarii.
Astron. Astrophys. , A34 (2018).
D. T. Hoai, P. T. Nhung, P. Tuan-Anh, P. Darriulat, P. N. Diep, T. Le Bertre et al.,
The morpho-kinematics of the circumstellar envelope around the AGB star EP Aqr.
MNRAS , 1865–1888 (2019).
V. Bujarrabal, J. Alcolea, J. Mikołajewska, A. Castro-Carrizo, S. Ramstedt, High resolution observations of the symbiotic system R Aqr. Direct imaging of the gravitational effects of the secondary on the stellar wind.
Astron. Astrophys. , L3 (2018).
S. Ramstedt, S. Mohamed, T. Olander, W. H. T. Vlemmings, T. Khouri, S. Liljegren, CO envelope of the symbiotic star R Aquarii seen by ALMA.
Astron. Astrophys. , A61 (2018).
H. Kim, S.-Y. Liu, N. Hirano, R. Zhao-Geisler, A. Trejo, H.-W. Yen et al.,
High-resolution CO Observation of the Carbon Star CIT 6 Revealing the Spiral Structure and a Nascent Bipolar Outflow.
Astrophys. J. , 61 (2015).
L. Decin, A. M. S. Richards, D. Neufeld, W. Steffen, G. Melnick, R. Lombaert, ALMA data suggest the presence of spiral structure in the inner wind of CW Leonis.
Astron. Astrophys.
4, A5 (2015).
M. Guélin, N. A. Patel, M. Bremer, J. Cernicharo, A. Castro-Carrizo, J. Pety et al.,
IRC +10 216 in 3D: morphology of a TP-AGB star envelope.
Astron. Astrophys. , A4 (2018).
M. Maercker, S. Mohamed, W. H. T. Vlemmings, S. Ramstedt, M. A. T. Groenewegen, E. Humphreys et al.,
Unexpectedly large mass loss during the thermal pulse cycle of the red giant star R Sculptoris.
Nature , 232–234 (2012).
M. Maercker, W. H. T. Vlemmings, M. Brunner, E. De Beck, E. M. Humphreys, F. Kerschbaum et al.,
A detailed view of the gas shell around R Sculptoris with ALMA.
Astron. Astrophys. , A5 (2016).
H. Kim, A. Trejo, S.-Y. Liu, R. Sahai, R. E. Taam, M. R. Morris et al.,
The large-scale nebular pattern of a superwind binary in an eccentric orbit.
Nature Astron. , 0060 (2017). ubmitted Manuscript: Confidential 15 S. V. Jeffers, M. Min, L. B. F. M. Waters, H. Canovas, O. R. Pols, M. Rodenhuis et al.,
Surprising detection of an equatorial dust lane on the AGB star IRC +10216.
Astron. Astrophys. , A3 (2014).
R. Sahai, M. Morris, G. R. Knapp, K. Young, C. Barnbaum, A collimated, high-speed outflow from the dying star V Hydrae.
Nature , 261–264 (2003).
L. Decin, K. Justtanont, E. De Beck, R. Lombaert, A. de Koter, L. B. F. M. Waters et al.,
Water content and wind acceleration in the envelope around the oxygen-rich AGB star IK Tauri as seen by Herschel/HIFI.
Astron. Astrophys. , L4 (2010).
T. Khouri, A. de Koter, L. Decin, L. B. F. M. Waters, R. Lombaert, P. Royer et al.,
The wind of W Hydrae as seen by Herschel. I. The CO envelope.
Astron. Astrophys. , A5 (2014).
M. Van de Sande, L. Decin, R. Lombaert, T. Khouri, A. de Koter, F. Wyrowski et al.,
Chemical content of the circumstellar envelope of the oxygen-rich AGB star R Doradus. Non-LTE abundance analysis of CO, SiO, and HCN.
Astron. Astrophys. , A63 (2018).
I. El Mellah, J. Bolte, L. Decin, W. Homan, R. Keppens, Wind morphology around cool evolved stars in binaries. The case of slowly accelerating oxygen-rich outflows.
Astron. Astrophys. , A91 (2020).
A. F. Illarionov, R. A. Sunyaev, Why the Number of Galactic X-ray Stars Is so Small?
Astron. Astrophys. , 185 (1975). P. P. Eggleton, Aproximations to the radii of Roche lobes.
Astrophys. J. , 368–369 (1983).
D. P. Hamilton, J. A. Burns, Orbital stability zones about asteroids II. The destabilizing effects of eccentric orbits and of solar radiation.
Icarus , 43–64 (1992). N. Soker, Planetary nebula progenitors that swallow binary systems.
MNRAS , 1584– 1593 (2016).
A. J. Mustill, E. Villaver, Foretellings of Ragnarök: World-engulfing Asymptotic Giants and the Inheritance of White Dwarfs.
Astrophys. J. , 121 (2012).
Z. Chen, E. G. Blackman, J. Nordhaus, A. Frank, J. Carroll-Nellenback, Wind-accelerated orbital evolution in binary systems with giant stars.
MNRAS , 747–756 (2018).
M. Trabucchi, P. R. Wood, J. Montalbán, P. Marigo, G. Pastorelli, L. Girardi. Modelling long-period variables - I. A new grid of O-rich and C-rich pulsation models.
MNRAS , 929–949 (2019).
R. De Nutte, L. Decin, H. Olofsson, R. Lombaert, A. de Koter, A. Karakas et al.,
Nucleosynthesis in AGB stars traced by oxygen isotopic ratios. I. Determining the stellar initial mass by means of the O/ O ratio.
Astron. Astrophys. , A71 (2017).
K. H. Hinkle, T. Lebzelter, O. Straniero, Carbon and Oxygen Isotopic Ratios for Nearby Miras.
Astrophys. J. , 38 (2016).
A. Weiss, J. W. Ferguson, New asymptotic giant branch models for a range of metallicities.
Astron. Astrophys. , 1343–1358 (2009). ubmitted Manuscript: Confidential 16
M. A. T. Groenewegen, J. A. D. L. Blommaert, Mira variables in the OGLE bulge fields.
Astron. Astrophys. , 143–156 (2005).
E. L. Nielsen, R. J. De Rosa, B. Macintosh, J. J. Wang, J.-B. Ruffio, E. Chiang et al.,
The Gemini Planet Imager Exoplanet Survey: Giant Planet and Brown Dwarf Demographics from 10 to 100 au.
Astron. J. , 13 (2019).
B. Fulton, Frequency of Gaseous Planets Beyond the Ice Line. Paper presented at the meeting of Extreme Solar Systems IV, August 19-23, 2019, Reykjavik, Iceland (2019).
N. Soker, Extrasolar planets and the rotation and axisymmetric mass-loss of evolved stars.
MNRAS , 699–704 (2001).
P. Mróz, A. Udalski, J. Skowron, R. Poleski, S. Kozłowski, M. K. Szymański et al.,
No large population of unbound or wide-orbit Jupiter-mass planets.
Nature , 183–186 (2017).
B. T. Gänsicke, M. R. Schreiber, O. Toloza, N. P. G. Fusillo, D. Koester, C. J. Manser, Accretion of a giant planet onto a white dwarf star.
Nature , 61–64 (2019).
P. Kervella, F. Arenou, F. Mignard, F. Thévenin, Stellar and substellar companions of nearby stars from Gaia DR2. Binarity from proper motion anomaly.
Astron. Astrophys. , A72 (2019).
M. W. Feast, I. S. Glass, P. A. Whitelock, R. M. Catchpole, A period-luminosity-colour relation for Mira variables.
MNRAS , 375–392 (1989).
M. I. Saladino, O. R. Pols, C. Abate, Slowly, slowly in the wind. 3D hydrodynamical simulations of wind mass transfer and angular-momentum loss in AGB binary systems.
Astron. Astrophys. , A68 (2019).
J. D. Cummings, J. S. Kalirai, P. E. Tremblay, E. Ramirez-Ruiz, J. Choi, The White Dwarf Initial-Final Mass Relation for Progenitor Stars from 0.85 to 7.5 : ⨀ Astrophys. J. , 21 (2018).
F. L. Schöier, M. Lindqvist, H. Olofsson, Properties of detached shells around carbon stars. Evidence of interacting winds.
Astron. Astrophys. , 633–646 (2005).
T. M. Gledhill, K. P. Forde, Imaging the transition between pre-planetary and planetary nebulae: integral field spectroscopy of hot post-AGB stars with NIFS.
MNRAS , 1080–1095 (2015).
H. van Winckel, Post-AGB binaries as tracers of stellar evolution. In X. Liu, L. Stanghellini, A. Karakas (eds.)
Planetary Nebulae: Multi-Wavelength Probes of Stellar and Galactic Evolution , vol. 323 of
IAU Symposium , 231–234 (2017).
M. Meixner, T. Ueta, M. Bobrowsky, A. Speck, Two Subclasses of Proto-Planetary Nebulae: Model Calculations.
Astrophys. J. , 936–946 (2002).
S. R. Pottasch, J. Bernard-Salas, Planetary nebulae abundances and stellar evolution II.
Astron. Astrophys. , A95 (2010).
C. Sánchez Contreras, A. Báez-Rubio, J. Alcolea, V. Bujarrabal, J. Martín-Pintado, A pilot search for mm-wavelength recombination lines from emerging ionized winds in pre-planetary nebulae candidates.
Astron. Astrophys. , A67 (2017). ubmitted Manuscript: Confidential 17
T. Dermine, R. G. Izzard, A. Jorissen, H. Van Winckel, Eccentricity-pumping in postAGB stars with circumbinary discs.
Astron. Astrophys. , A50 (2013).
R. R. Rafikov, On the Eccentricity Excitation in Post-main-sequence Binaries.
Astrophys. J. , 8 (2016).
N. Soker, Binary Progenitor Models for Bipolar Planetary Nebulae.
Astrophys. J. , 833–841 (1998).
B. Paczynski, Common Envelope Binaries. In P. Eggleton, S. Mitton, J. Whelan (eds.)
Structure and Evolution of Close Binary Systems , vol. 73 of
IAU Symposium , 75 (1976).
H. M. J. Boffin, B. Miszalski, T. Rauch, D. Jones, R. L. M. Corradi, R. Napiwotzki et al.,
An Interacting Binary System Powers Precessing Outflows of an Evolved Star.
Science , 773 (2012).
D. Jones, H. M. J. Boffin, Binary stars as the key to understanding planetary nebulae.
Nature Astron. , 0117 (2017). H. M. J. Boffin, D. Jones,
The Importance of Binaries in the Formation and Evolution of Planetary Nebulae (SpringerBriefs in Astronomy, 2019).
B. Miszalski, H. M. J. Boffin, D. J. Frew, A. Acker, J. Köppen, A. F. J. Moffat, Q. A. Parker, A barium central star binary in the Type I diamond ring planetary nebula Abell 70.
MNRAS , 39–49 (2012).
A. A. Tyndall, D. Jones, H. M. J. Boffin, B. Miszalski, F. Faedi, M. Lloyd et al.,
Two rings but no fellowship: LoTr 1 and its relation to planetary nebulae possessing barium central stars.
MNRAS , 2082–2095 (2013).
H. Van Winckel, A. Jorissen, K. Exter, G. Raskin, S. Prins, J. Perez Padilla et al.,
Binary central stars of planetary nebulae with long orbits: the radial velocity orbit of BD+33 2642 (PN G052.7+50.7) and the orbital motion of HD112313 (PN LoTr5).
Astron. Astrophys. , L10 (2014).
D. Jones, H. Van Winckel, A. Aller, K. Exter, O. De Marco, The long-period binary central stars of the planetary nebulae NGC 1514 and LoTr 5.
Astron. Astrophys. , L9 (2017).
M. Moe, O. De Marco, Do Most Planetary Nebulae Derive from Binaries? I. Population Synthesis Model of the Galactic Planetary Nebula Population Produced by Single Stars and Binaries.
Astrophys. J. , 916–932 (2006).
N. Soker, Why Magnetic Fields Cannot Be the Main Agent Shaping Planetary Nebulae.
Publ. Astron. Soc. Pac. , 260–269 (2006).
V. Bujarrabal, J. Alcolea, R. Sahai, J. Zamorano, A. A. Zijlstra, The shock structure in the protoplanetary nebula M1-92: imaging of atomic and H line emission. Astron.
Astrophys. , 361–371 (1998).
N. Smith, B. Balick, R. D. Gehrz, Kinematic Structure of H and [Fe II] in the Bipolar Planetary Nebula M2-9. Astron. J. , 853–861 (2005).
R. Sahai, N. A. Patel, An Extreme High-velocity Bipolar Outflow in the Pre-planetary Nebula IRAS 08005-2356.
Astrophys. J. , L8 (2015).
N. Soker, M. Livio, Disks and jets in planetary nebulae.
Astrophys. J. , 219 (1994). ubmitted Manuscript: Confidential 18
B. Balick, A. Frank, B. Liu, Models of the Mass-ejection Histories of Pre-planetary Nebulae. IV. Magnetized Winds and the Origins of Jets, Bullets, and FLIERs.
Astrophys. J. , 13 (2020).
P. Kervella, L. Decin, A. M. S. Richards, G. M. Harper, I. McDonald, E. O’Gorman et al.,
The close circumstellar environment of Betelgeuse. V. Rotation velocity and molecular envelope properties from ALMA.
Astron. Astrophys. , A67 (2018).
H. Kim, S.-Y. Liu, R. E. Taam, Templates of Binary-induced Spiral-shell Patterns around Mass-losing Post-main-sequence Stars.
Astrophys. J. Suppl. Ser. , 35 (2019).
M. I. Saladino, O. R. Pols, The eccentric behaviour of windy binary stars.
Astron. Astrophys. , A103 (2019).
Z.-W. Liu, R. J. Stancliffe, C. Abate, E. Matrozis, Three-dimensional Hydrodynamical Simulations of Mass Transfer in Binary Systems by a Free Wind.
Astrophys. J. , 117 (2017).
T. Danilovich, G. Olofsson, J. H. Black, K. Justtanont, H. Olofsson, Classifying the secondary component of the binary star W Aquilae.
Astron. Astrophys. , A23 (2015).
M. W. Feast, The absolute magnitude and spectrum of the class S star V . Gruis.
MNRAS , 510 (1953).
A. Mayer, A. Jorissen, C. Paladini, F. Kerschbaum, D. Pourbaix, C. Siopis et al.,
Large-scale environments of binary AGB stars probed by Herschel. II. Two companions interacting with the wind of V . Gruis.
Astron. Astrophys. , A113 (2014).
B. D. Mason, G. L. Wycoff, W. I. Hartkopf, G. G. Douglass, C. E. Worley, The 2001 US Naval Observatory Double Star CD-ROM. I. The Washington Double Star Catalog.
Astron. J. , 3466–3471 (2001).
Acknowledgments:
ALMA is a partnership of ESO (representing its member states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC and ASIAA (Taiwan), in cooperation with the Republic of Chile. The Joint ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. The CASA data reduction package was developed by an international consortium of scientists based at the National Radio Astronomical Observatory (NRAO), the European Southern Observatory (ESO), the National Astronomical Observatory of Japan (NAOJ), the CSIRO Australia Telescope National Facility (CSIRO/ATNF), and the Netherlands Institute for Radio Astronomy (ASTRON) under the guidance of NRAO. We thank the Data Reduction team at ESO for customizing the imaging pipeline. We thank the UK Science and Technology Facilities Council (STFC) IRIS for provision of high-performance computing facilities allowing UK's Radio and mm/sub-mm Interferometry Services to improve the data quality and much more data to be processed. We thank the ALMA Archive scientist Felix Stoehr for providing the technical information;
Funding:
L.D., D.G., W.H., J.B., J.M.C.P., and S.H.J.W. acknowledge support from the ERC consolidator grant 646758 AEROSOL, L.D., H.S., and E.C. acknowledge support from the KU Leuven under the C1 MAESTRO grant C16/17/007, H.S. acknowledges support from the European Research Council (ERC) under the European Union’s DLV-772225-MULTIPLES Horizon 2020 research and innovation programme, W.H. acknowledges support from the FWO Flemish Fund of Scientific Research under grant G086217N, F. H. acknowledges support from the "Programme National de Physique Stellaire" ubmitted Manuscript: Confidential 19 (PNPS) of CNRS/INSU co-funded by CEA and CNES, F.D.C. is supported by the EPSRC iCASE studentship programme, Intel Corporation and Cray Inc., J.M.C.P. acknowledges support from the UK STFC grant ST/P00041X/1, J.Y. acknowledges support from the UK STFC grant ST/R001049/1, M.V.d.S. acknowledges support from the FWO through grant 12X6419N, T.D. acknowledges support from the FWO through grants 12N9917N & 12N9920N, M.M. acknowledges support from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie Grant agreement No. 665501 with the FWO ([PEGASUS] Marie Curie fellowship 12U2717N, A.B. acknowledges support from the "Programme National de Physique Stellaire" (PNPS), I.M. acknowledges funding by the UK STFC grant ST/P000649/1, EDB acknowledges financial support from the Swedish National Space Agency, I.E.M. acknowledges support from the FWO and the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No 665501, S.E. acknowledges funding from the UK STFC as part of the consolidated grant ST/P000649/1 to the Jodrell Bank Centre for Astrophysics at the University of Manchester, P.K. acknowledges support from the French PNPS of CNRS/INSU, C.A.G. acknowledges support from NSF grant AST-1615847, T.J.M. is grateful to the STFC for support under grant ST/P000321/1, A.A.Z. was supported by the STFC under grants ST/T000414/1 and ST/P000649/1, M.D.G. thanks the STFC for support under consolidated grant ST/P000649/1 to the JBCA;
Author contributions:
L.D. is principal investigator (PI) of the ALMA Large Program ATOMIUM, leads the ATOMIUM team and supervises the ATOMIUM project, developed the methodology, analyzed the results, wrote the draft manuscript, contributed to Fig. 1, made Fig. 2, Fig. S1-S2, wrote the software for Fig. S5-S7 and Fig. S8-S65; M.M. developed the observational strategy and contributed to Fig. 1; A.M.S.R. led the ATOMIUM data reduction team and wrote software used in the data reduction; C.A.G. is co-PI of the ATOMIUM project; W.H., I.E.M., and J.B. performed hydrodynamical simulations of binary systems; I.McD. and H.S. investigated the (sub-)stellar binary population statistics; M.M, T.D., W.H., S.H.J.W., A.B., S.E., F.H., K.T.W. contributed to the data reduction; J.D.R. performed the statistical analysis and made Fig. S3-S4, J. Y. provided hardware for the ALMA data reduction, P.K. and F.D.C. contributed to Fig. 1. All authors discussed the interpretation of the data, contributed scientific results, and helped prepare the paper;
Competing interests:
We declare no competing interests;
Data and materials availability:
The ALMA data are available from the ALMA data archive at http://almascience.eso.org/aq/ under project code 2018.1.00659.L. Scripts for producing Figure S5-S7 are available at https://github.com/LeenDecin/Supplementary_material_for_Science_paper_abb1229.
Supplementary Materials:
Materials and Methods Supplementary Text Figures S1-S65 Tables S1-S7 Data S1-S2 References ( ) upplementary Materials for (Sub-)stellar companions shape the winds ofevolved stars L. Decin ⇤ , M. Montarg`es, A. M. S. Richards, C. A. Gottlieb, W. Homan, I. McDonald, I. ElMellah, T. Danilovich, S. H. J. Wallstr¨om, A. Zijlstra, A. Baudry, J. Bolte, E. Cannon, E. DeBeck, F. De Ceuster, A. de Koter, J. De Ridder, S. Etoka, D. Gobrecht, M. Gray, F. Herpin, M.Jeste, E. Lagadec, P. Kervella, T. Khouri, K. Menten, T. Millar, H. S. P. M¨uller, J. Plane, R.Sahai, H. Sana, M. Van de Sande, L. B. F. M. Waters, K. T. Wong, J. Yates ⇤ Corresponding author. Email: [email protected].
This PDF file includes:
Materials and MethodsSupplementary TextFigs. S1–S65Tables S1–S7Data S1–S2 1 aterials and Methods
S1 ALMA ATOMIUM Large Program and target selection
The ALMA
ATOMIUM
Large Program aims to establish the dominant physical and chemicalprocesses in the winds of oxygen-rich evolved stars over a range of initial stellar masses, pul-sation behaviours, mass-loss rates, and evolutionary phases. The
ATOMIUM observing proposalwas selected in Cycle 6 as Large Program for a total of 113.2 hr (2018.1.00659.L, PI L. Decin).To disentangle the impact of various (circum)stellar properties, a sample of seventeen targetswas selected that cover i) a range in initial stellar mass — hence Asymptotic Giant Branch(AGB) stars versus the more massive counterparts, the red supergiants (RSG); ii) various pulsa-tion characteristics (regular versus semi-regular) and AGB versus RSG (large versus small am-plitude); iii) various evolutionary phases — hence inclusion of sources with different mass-lossrates and S-type AGB stars (with C/O almost equal to 1). All selected targets have an apparentstellar angular size above 3 mas and an R -band magnitude <
11 to allow for contemporaneousobservations with the Spectro-Polarimetric High-contrast Exoplanet REsearch (SPHERE) in-strument mounted on the Very Large Telescope (VLT) and the Multi-AperTure mid-InfraredSpectroScopic Experiment (MATISSE) mounted on the Very Large Telescope Interferometer(VLTI) (31, 32) . Fourteen sources are AGB stars and three are red supergiants.Summarized in Table S1 are the (circum)stellar parameters of the fourteen AGB sources inthe ATOMIUM sample. Variability is a common feature of AGB stars and is mainly causedby pulsations. Mira variables have regular, large amplitude variations in the visible light, with V > . mag; semiregular variables of type a (SRa) are similar to regular Mira variables, buthave smaller V -band amplitude. Semiregular variables of type b (SRb) show poor regularitywith a small amplitude (3) . A source is classified as a long-period variable (LPV) if no regularpulsation period could be deduced from observations. Stellar pulsations are an important triggerfor the onset of the dust-driven AGB wind (33) by increasing the density scale height. It followsthat, in general, stars with regular long-period large amplitude pulsations have larger the mass-loss rate. For Mira-type variable stars with a luminosity L above ⇠ and pulsationperiod P between ⇠
300 – 800 days, a linear relation exists between the pulsation period and thelogarithm of the mass-loss rate ( ⇠ M yr < ˙ M < ⇥ M yr ) (34,35) . Semiregularvariables with a pulsation period less than 200 days cover essentially the same mass-loss rateregime as the Mira variables with pulsation period between 200 – 400 days, while a maximummass-loss rate of a few M yr is reached for P > ⇠ days (34, 35) . Between ⇠
60 and ⇠
300 days, an approximately constant mass-loss rate of ⇠ ⇥ M yr is found, whilefor P <
60 days the mass-loss rate is a factor of ⇠
10 smaller (33, 36) . Distance:
Obtaining accurate parallax observations for the ATOMIUM targets is challengingowing to the movements of the photocenter as the star pulsates, and the motions of the convec-tive cells around the central star. Both effects produce correlated errors in the proper motion,however the excess noise in the astrometric solution can be used to estimate the magnitude of2able S1:
Summary of (circum)stellar parameters of the ATOMIUM sample.
The first column gives the target name,the second column the AGB variability type, the third column the pulsation period P , the fourth column the distance D , thefifth column the stellar angular diameter ✓ D , the sixth column the effective temperature T e ↵ , the seventh column the stellarluminosity L , the eighth column the mass-loss rate ˙ M , the ninth column the wind velocities as determined from the ALMAATOMIUM CO J = 2 ! observations, the tenth column the local standard of rest velocity v LSR used as input for theALMA observations, and the last column an estimate of the local standard of rest velocity using the ALMA ATOMIUM data.“-” indicates there are insufficient measurements available. Targets are ordered according to increasing mass-loss rate.
Name Variability Pulsation Distance Angular T e ↵ L Mass-loss v wind v LSR v newLSR type period D diameter rate ALMA CO ALMA obs.(days) (pc) (mas) (K) (L ) (M yr ) (km s ) (km s ) (km s )S Pav SRa 381 (39) (37) ? (40) † ⇥ (39) (39) (37, 38) ? (40) † ⇥ (39)
14 25.3 25.5U Del SRb 119 & 1170 (41) (37, 38) (42) § † . ⇥ (39) (43) (37) ? (40) † . ⇥ (43) (39) (37) ? (39) † ⇥ (39) (37) ? (40) ‡ ⇥ (39)
16 8.5 6.7R Hya k Mira 366 (44) (45) (42) (35) † ⇥ (35) (44) (46) (42) § † . ⇥ (47) ⇡ Gru k , ¶ SRb 150 (35) (37, 38) (48) (35) † . ⇥ (49) (44) (37, 38) (42) § † . ⇥ (47)
16 47.0 47.2W Aql k , ¶ Mira 479 (44) (37) (42) (35) † ⇥ (50) (44) (37) ? (40) † . ⇥ (51)
18 34. 34.0IRC (35) (35) ? (35) † . ⇥ (35) (35) (52) ? (35) † . ⇥ (35)
23 10.0 10.1 ? calculated from L , T e ↵ and distance D ; † using M bol ( P, L ) -relation (35) ; ‡ assumed following Olofsson et al. (39) ; § from L and R ? ( ✓ d , D ) ; k known binary system (see Sect. S9); ¶ S-type star with carbon over oxygen ratio (C/O) slightly below 1. hese effects which should be random and average out on long timescales. The Gaia Data Re-lease 2 (DR2) and Hipparcos reductions (37, 38) are independent. If the two reductions agreeto within the estimated uncertainties, it implies the parallax is correct. Excess noise of any kindwill generally result in a larger measured parallax and will manifest itself as a closer apparentdistance.We adopt the Hipparcos parallax if its fractional uncertainty is < < (46) , and is used because it is well known thatmaser parallaxes give a better solution than optical parallaxes for AGB and red supergiant stars.If the parallax is unavailable, or if neither Hipparcos or Gaia parallax meets these criteria, thedistance is determined from the period-luminosity relationship (53) , where we used the K -bandmagnitude from the Two Micron All-Sky Survey (2MASS). The distance D was obtained froman inversion of the adopted parallax. Angular diameter:
Listed in Table S1 is the stellar angular diameter, ✓ d , measured frominterferometric observations, otherwise ✓ d was calculated from the luminosity L , the effectivetemperature T e ↵ , and distance D , where the luminosity is derived from the M bol ( P, L ) -relation (35) , with M bol the absolute bolometric magnitude and P the pulsation period. Mass-loss rate:
The wind of all ATOMIUM targets seems to be shaped by a companion,resulting in a non-spherical envelope. In these circumstances, the AGB wind mass-loss rateis obtained from low-excitation CO rotational lines observed with large (single-dish antenna)beams (54, 55) so that the presence of a geometrically compact equatorial density enhancement(EDE) or a (broken) spiral-like structure (in the case of high mass-loss rate objects; see Sect. S3)only translates into an uncertainty of the mass-loss rate smaller than < ⇠ (21) . All mass-loss rates in Table S1 are retrieved from low-excitation COlines, typical uncertainties are a factor of ⇠ (56) . Local standard of rest velocity:
The local standard of rest velocity v LSR in Table S1 is usedas input for the ATOMIUM observations (Sect. S2.1). The ATOMIUM observations yield arevised estimate of the v LSR by comparing the difference between the astronomical frequenciesand the laboratory measured rest frequencies of identified molecular lines (see below). Anestimate of v LSR obtained from the low and medium resolution ATOMIUM measurements isgiven in Table S1, the uncertainty is ⇠ . Wind velocity:
The wind velocity derived from the low and medium resolution data of therotational line CO J = 2 ! in the ground vibrational state — with J specifying the rotationalquantum number — is listed in Table S1 (see also Sect. S2). For each target, the integrated flux4s extracted for a range of apertures. The extraction aperture yielding the largest integrated fluxof the CO line is used to measure the extent of the blue and red wing (in km s ) with respect tothe v LSR (for intensities greater than 3 times the rms noise). Listed in Table S1 is the maximumvelocity obtained from the blue and red wings, and for the observations at low and mediumspatial resolution. The uncertainty in the velocity measure is taken from the spectral resolutionof the data, being ⇠ . S2 ALMA observations and data reduction
S2.1 ALMA observations
ALMA data have been secured in band 6 between 213.83 and 269.71 GHz. Fig. S1 showsthe frequency coverage. Each set of four spectral windows (spw) in each array configurationcomprises one Science Goal (SG), which together sample ⇠
30 GHz. The requested spectralresolution was ⇠ . Typical line widths range between 5–60 km s , where the smallerline widths probe the wind acceleration region.All targets have been observed with ALMA at a spatial resolution of ⇠ ⇠ –0.62 . To get the full line strength of the transitions, wecomplement these observations with the antenna configuration C43-5/C43-6 (minimum andmaximum baseline between 15 m – 2 500 m, angular resolution of 0.24 /0.13 ) to encompass aMRS of ⇠ . Even with C43-5/C43-6 the CO and SiO transitions in the ground-vibrational statemight be resolved out. Hence, for 15 sources, C43-2 observations at an angular resolution of ⇠ were requested (MRS ranging between 2 –10 , minimum and maximum baseline between15 m – 314 m). The ALMA antenna configurations during Cycle 6 allowed for most targets tobe observed at medium spatial resolution from 3 October – 23 November 2018, at low spatialresolution from 26 December 2018 – 23 January 2019 and from 3 March – 20 March 2019, andat high spatial resolution from 12 June – 13 July 2019.The positions and proper motions were taken from the ALMA Observing Tool (OT) usingits SIMBAD look-up facility, which at the time were taken from the Hipparcos catalogue (57) .Each target was observed multiple times during Cycle 6, and each observation was centered onthe position and the assumed v LSR (Table S1) which were adjusted at the start of the execution(Table S5). Three sources (RW Sco, R Aql and W Aql) lie close to the Galactic plane, with othersources in the constellations Aquila and Delphinus within 20 degrees. Some of these sufferfrom Galactic CO contamination, especially when they are observed at the lowest resolution,but this can be distinguished from circumstellar emission by velocity offsets and by its diffusenature which is unrelated to the morphology of other lines from the circumstellar envelope.The highest proper motions are ⇠
70 mas yr , with a precision better than a few mas yr .During the data reduction phase, the positions were aligned to that of the earliest execution. TheALMA astrometric accuracy is better than 1/3 synthesized beam, and although this and errors5n the assumed proper motion could cause a position offset of a few mas between executions,self-calibration (Sect. S2.2) prevents this from causing image artefacts.Fig. S1: The frequency coverage for the ATOMIUM observations in each array configu-ration.
Each bar represents the frequency coverage of one spectral window, re-numbered infrequency order for convenience. Each individual Science Goal covered four spectral windows,grouped as follows: [0,1,4,5], [2,3,6,7], [8,9,12,13], [10,11,14,15]. The coverage in the highspatial resolution configuration is the same as in the medium-resolution configuration, whilefor the low-resolution observations only 8 spectral windows were requested. The exact fre-quency extent for each target depends on the correction for the assumed v LSR on the dates ofobservation (Table S5).
S2.2 ALMA data reduction
The medium resolution data are most suitable for retrieving the morphological characteristicsof the AGB stars in the ATOMIUM sample, because these data sample the stellar wind at properangular scales of 0 .
2, corresponding to ⇠
50 au for a target at 250 pc. Each fully observed Sci-ence Goal was first processed using the ALMA calibration and imaging pipelines (58) . Thepipelines apply all calibration derived instrumentally (for example from water vapour radiome-try) and from calibration and phase-reference sources (see Table S5). The line-free channels areidentified from the visibility data and subtracted, and data cubes are made for each spectral win-dow. We inspected the web logs; occasionally a few instances of over- or under-flagging wereidentified, but the former were negligible and the latter were remedied during our processing.This comprised the following steps:1. Split out two copies of the calibrated target data, one at a ‘continuum’ spectral resolution of15.625 MHz and one at a ‘line’ spectral resolution of 0.9765625 MHz, adjusted to constant v LSR in the target frame. These were then concatenated to make continuum and line datasets containing the full spectral coverage for each star and array configuration.6. Use L
UMBERJACK (59) to identify line-free channels from the pipeline image cubes. Theselection is adjusted to correspond to the channelisation of the continuum and line data sets,and checked interactively.3. Image the continuum-only channels of the continuum data set and self-calibrate, startingwith phase-only. This removes any small offsets due to differences in calibration or propermotion uncertainty, as well as improving image quality. If the signal-to-noise ratio is suf-ficient, an image using a first-order spectral index is made to provide a model for morecycles of self-calibration, including amplitude self-calibration. The dominant cause of smallamplitude scaling offsets between executions and Science Goals is the ALMA flux densityuncertainty of up to 5%, in practice only statistically significant above the noise in sourcesbright enough to remove any offset by self-calibration.4. Apply the corrections to the line data set, check the selection of line-free channels and sub-tract the continuum using a linear fit.5. Make a spectral image cube for each spectral window and configuration large enough toencompass all detectable emission. This was checked against the pipeline cubes which aremade to the 20% primary beam response radius for the lowest-resolution observations, mostsensitive to extended emission. This tends to be resolved-out at higher resolution, so smallerimages (in angular size) can be made. In making the separate per-configuration cubes, theALMA default weighting for optimum signal-to-noise was used, resulting in a restoringbeam varying slightly depending on target elevation and exact antenna positions. The cubesused for analysis are primary-beam corrected, but (unless otherwise stated) the rms is esti-mated near the center of the field avoiding any emission.The image cube and continuum image properties can be found in Tables S6–S7 (Sect. S8). S3 ATOMIUM morphological classification
S3.1 ATOMIUM wind morphologies
The lines of particular relevance for this study are the rotational transitions in the ground vi-brational state of CO J = 2 ! (230.538 GHz), SiO J = 5 ! (217.105 GHz), and SiO J = 6 ! (260.518 GHz) observed at medium spatial resolution. The high fractional abun-dance of both molecules ([CO/H ] ⇠ ⇥ , [SiO/H ] ⇠ ⇥ (60) ) facilitate their de-tection. The low-excitation CO line is predominantly collisionally excited and is a tracer ofthe density distribution in the outer wind region because the excitation energy is low (up-per state energy, E up of 16.6 K), the dipole moment µ is only 0.11 Debye, and the Einstein A u ! l coefficient (with u indicating the upper energy level, and l the lower level) is very small( A u ! l = 6 . ⇥ s , (61–63) ). The Einstein A u ! l coefficients of the two rotationallines of SiO ( E up = 31 . K and 43.76 K for the J = 5 ! and ! line) are three orders of7agnitude higher ( A ! = . ⇥ s , and A ! = 9 . ⇥ s , with µ = 3.08 De-bye; (62,64) ). As a result the lines of SiO are sensitive to radiative (de-)excitation effects and arediagnostic of the dynamics in the inner wind region. SiO molecules can be depleted in the innerwind region if silicate grains are forming (65) . No conclusive evidence can yet be drawn onthe depletion efficiency, but observation and modelling efforts indicate at least a fraction of SiOremains in gaseous form outside the main dust condensation region before photodissociationby the interstellar ultraviolet (UV) photons sets in (66–69) . The weaker isotopologue lines inthe ground vibrational state ( CO J = 2 ! at 220.399 GHz, SiO J = 5 ! at 214.386 GHz,and SiO J = 6 ! at 254.217 GHz) provide complementary information, because the lines ofthe dominant isotopic species can be prone to high optical depth effects. Owing to lost flux ofthe CO J = 2 ! line in the medium resolution observations of some sources (largest angularscale, LAS, of ⇠ ), the low resolution data were included in the analysis to encompass alarger angular scale. The resolved-out flux density question affects physical parameters derivedin a radiative transfer analysis. However, our analysis of the prevailing wind morphology is notaffected because any asymmetry detected at a given angular resolution is a real feature imprintedin our data. The filtering of extended structures provides higher dynamic range for identifyingasymmetric and clumpy structures. The derivation of the wind velocity from the CO and SiOlines (see Table S1 and Table S2) is also not affected because resolved-out flux is mainly seenas a depression of the line flux around the central velocities and not in the line wings.Figure 1 was obtained by selecting three velocity frames of the channel map of the CO J = 2 ! line: one at rest velocity, one blue shifted, and one red shifted with respect to thelocal standard of rest velocity v LSR . For two targets (RW Sco and SV Aqr) the signal-to-noiseof the data was too low to produce a three-colour map, although the individual channels showclear signs of asymmetry (Fig. S20, Fig. S28). Figure 1 shows a range of morphologies, withnone of the stars in our sample displaying a spherically symmetric wind geometry. All thesemorphologies have a counterpart in the more evolved post-AGB stars and planetary nebulae(PNe); some prominent examples are given below. – The ‘rose-like’ structure of R Aql resembles the inner structure of the Eskimo Nebula, abipolar double-shell PN (70) ; – the ‘eye’-like morphology of U Del bears a morphological resemblance with the outer re-gions of Helix Nebula (71, 72) , a bipolar planetary nebula; – the biconical shape of R Hya is seen in various post-AGB stars and PNe including the post-AGB star IRAS 17150 (73) andthe Owl Nebula (NGC 3587), a planetary nebula which has a barrel-like structure in its innershell caused by bipolar cavities (74) ; – regularly spaced arcs embedded in the bipolar outflow of ⇡ Gru (see Fig. S42) are reminis-cent of the Red Rectangle (75) , a pre-PN or to the outer halo structure of the Cat’s Eye Neb-ula (76) , a planetary nebula. A regular spiral structure (as in IRC + in the wind of the oxygen-rich AGB star OH 26.5+0.6(see Table S3, Sect. S9; (55) ) (for a distance of 1 370 pc) is similar to the regular arc spacingin the outer halo of the Cat’s eye nebula, which is ⇠ for a distance of 1 001 pc (76) ; – (stable) disks have been detected in numerous post-AGB stars and PNe, such as AR Puppis (77) ; – the kinematical behaviour of disks surrounding AGB stars and post-AGB star shows simi-larities. In particular, the Red Rectangle is an oxygen-rich post-AGB binary system whichhas a Keplerian (rotating) dosk and an outflow, the latter mainly being formed of gas leavingthe disk (78) . The velocity vector field in the disk surrounding the Red Rectangle resemblesthat of the oxygen-rich AGB star R Dor (see Table S3, Sect. S9; (79) ). Another example isthe oxygen-rich post-AGB binary system AC Her which has a disk with a Keplerian velocityfield (80) similar to the kinematical structure of the disk surrounding the oxygen-rich AGBstar L Pup (see Table S3, Sect. S9; (81) ).Even during the early stage of the AGB phase, characterized by a low mass-loss rate, thetargets quite frequently have a pronounced aspherical morphology with an axisymmetric ge-ometry. The resemblance between the morphologies of AGB stellar winds and post-AGB starsand PNe supports the idea that planetary nebulae are descendants of evolved low and interme-diate mass AGB stars (82) . We conjecture that the same formation mechanism gives rise to theplethora of morphologies observed in AGB winds, post-AGB objects, and planetary nebulae.
S3.2 Observational classification of ATOMIUM wind morphologies
The rotational emission lines of CO and SiO are the basis of our observational classificationof the prevailing wind morphologies. All data used for the morphological classification aredisplayed in Figs. S8–S65. Our morphological classification scheme is intended to elucidatethe predominant morphology, however in each source smaller scale structural features oftenenrich this picture. Our classification follows a stepwise approach and encompasses three steps(Fig. S2). – Step 1: CO morphological information: The prevalence of arc-like structures is investigatedin the medium and low-resolution CO channel maps (Table S2). Some sources display onlyone arc with an extent < , other sources are surrounded by a regularly spaced spiralstructure, while an elliptical/circular structure surrounds the central target in a fraction ofsources. – Step 2: Inner versus outer wind kinematics: In the next step, we investigate the kinematicsin the inner versus the outer wind region. Winds of AGB stars are radiation driven, so theabsorption of the outwards-directed stellar radiation by newly formed dust grains producesa net force that can overcome gravity (83) . The gas is then accelerated beyond the escape9 (")
SiO moment1-mapSiO PV diagramSTEP 3 v (CO)versus v (SiO)STEP 2STEP 1 CO J=2 → R AqlR HyaU Del v (SiO)< v (CO) v (SiO)> v (CO) V PsARW ScoIRC +10011 U DelSiO J=5 → C O C O angular extent [arcsec]/2angular extent [arcsec]/2angular extent [R * ]/2angular extent [R * ]/2 v e l o c i t y [ k m s - ] v e l o c i t y [ k m s - ] v e l o c i t y [ k m s - ] Relative R.A. [arcsec] R e l a t i v e D e c . [ a r c s e c ] Fig. S2:
Decision tree used to classify the prevailing wind morphology for the ATOM-IUM AGB stars.
Step 1 assesses the presence of arcs and spiral-like structures based on the CO J = 2 ! channel maps. Step 2 investigates the difference between the inner and outerwind kinematics. Step 3 deduces systematic, although sometimes complex, inner wind dy-namics. The images and plots at each step are illustrative of the data shown in Figs. S8–S65.Step 1 is illustrated using R Aql, R Hya, and U Del as an example for a complex spiral-likestructure, a bipolar structure, and an ‘eye’-like feature, respectively. Step 2 shows the velocitymeasurements in RW Sco and IRC +10011. All dots represent measurements of various molec-ular species in the ATOMIUM data. The measurement of the CO J = 2 ! is indicatedin blue, the SiO lines with red crosses. For RW Sco v (SiO) < v (CO) , while for IRC +10011 v (SiO) > v (CO) . Step 3 is illustrated using the SiO J = 5 ! moment1-map of U Del, showinga clear sign of bipolarity or a rotating flow, and the SiO J = 5 ! PV diagram of V PsA inwhich 2 bright blobs can be discerned.velocity with the main acceleration occurring in the first few stellar radii. The velocity profileof AGB stellar winds is parametrised using the -type law (84) v ( r ) = v + ( v v ) ⇣ r dust r ⌘ , (S1)with r the radial distance to the star, v the velocity at the dust condensation radius r dust , v the terminal wind velocity, and larger values of indicating a lower wind acceleration. IfSiO is excited closer to the central star than CO, as confirmed by the measured angular ex-tents, the width of the SiO lines should be smaller or equal to that of CO. However, in some10TOMIUM AGB sources, a different behavior is observed with v wind (CO) < v wind (SiO);see Table S2. The uncertainty in the derived fraction is ⇠ ⇠ . – Step 3: SiO morpho-dynamical information: The SiO data were then analyzed using stere-ograms, moment1-maps, and position-velocity (PV) diagrams (see Sect. S7). Stereogramsand moment1-maps are used to identify possible positional offsets between blue- and red-shifted emission, hence yielding information on the velocity vector field ( v ) and rotation (Ta-ble S2). As examples, see the data for S Pav, T Mic, U Del, and ⇡ Gru (Fig. S10, Fig. S14,Fig. S18, and Fig. S43). PV diagrams are constructed for two orthogonal slits so that theasymmetry between the two PV diagrams is maximized to unravel the morpho-kinematicalbehavior (85) (Table S2). A rotating Keplerian disk or an EDE with constant radial velocityappear in one (or both) of the PV diagram(s) as a ‘butterfly’-like signature (85) . See, forexample, the PV diagrams of CO in ⇡ Gru shown in Fig. S44.The outcomes of each of these three steps are described below and summarized in Ta-ble S2. This stepwise procedure allows the characterization and classification of 13 out of the14 ATOMIUM AGB stars. U Her seems an outsider: the CO channel maps show pronouncedasymmetric arcs although they seem not to be linked to a spiral-like structure. – Step 1: Based on the CO channel maps, we discern a spiral-like structure in five targets(R Aql, W Aql, GY Aql, IRC-10529, and IRC+10011; Fig. S46, Fig. S50, Fig. S54, Fig. S58,and Fig. S62). The CO data of both R Hya and ⇡ Gru show a clear ‘bipolar’ signature. ForR Hya ( ⇡ Gru) the blue-shifted emission is east (north) with respect to the central star, whilethe red-shifted emission is west (south) (see the CO channel maps in Fig. S32 and Fig. S41,and the CO moment1-maps in Fig. S36 and Fig. S45). In both sources, an hourglass signaturecan be discerned. An ‘eye’-like morphology is recognized in the channel maps of U Del,RW Sco, V PsA, and SV Aqr (Fig. S16, Fig. S20, Fig. S24, Fig. S28). – Step 2: Five sources have an SiO line width considerably larger than the CO line width(Table S2): S Pav, T Mic, W Aql, IRC – Step 3: The analysis of the SiO stereograms and moment1-maps reveals two sources (U Deland ⇡ Gru) with a definite offset between red-shifted and blue-shifted emission (Fig. S18,Fig. S43), due to either a rotating inner wind structure or a bipolar outflow. This is recog-nizable in the moment1-maps as an almost straight-line division between red and blue. ForU Del and V PsA two bright blobs are discerned in the PV diagrams of SiO, independent ofthe slit position angle (PA) (Fig. S19, Fig. S27). This pattern cannot be explained by a de-tached SiO shell (86) , because that would produce lower emission for the central velocities,which is not seen. The SiO signatures of U Del, ⇡ Gru, and V PsA can be understood as in-dication of ‘bipolarity’. We use the term ’bipolarity’ to indicate a directed bipolar flow or anEDE/disk-like structure, which may display Keplerian rotation, so that lower density biconi-cal poles are created perpendicular to the EDE. In five sources (S Pav, T Mic, R Hya, W Aql,11nd GY Aql) a systematic, but complex, flow can be seen. The v = 0 signature in theirmoment1-maps is strongly skewed (Fig. S10, Fig. S14, Fig. S34, Fig. S52, and Fig. S56).The signal-to-noise ratio of the SiO lines of RW Sco and SV Aqr is too low to determine theinner wind dynamics (Fig. S23, Fig. S31).A correlation emerges between the mass-loss rate (and hence pulsation properties of theAGB star) and the prevailing morphological appearance (Table S2) allowing us to define threeClasses (Table 1). ‘Class 1’ sources are low mass-loss rate AGB stars ( ˙ M < ⇠ ⇥ M yr )characterized by a systematic, but complex, inner wind dynamics with signs of distorted rota-tion. Because rotation, or in general any tangential velocity component, implies a fraction ofthe material is not radially streaming, this characteristic often implies an EDE structure will beformed (85) . We use the term ‘EDE’ instead of ‘disk’ or ‘torus’, because the data does not allowus to determine whether the material is bound to the AGB star. ‘Class 2’ sources have mediummass-loss rates between ⇠ ⇥ – ⇥ M yr and display a bipolar morphology includ-ing bipolar outflows and/or (compact) disk-like structures at smaller spatial scales. A spiral sig-nature is only seen in the ‘Class 3’ sources with mass-loss rates greater than ⇠ ⇥ M yr .These sources might also exhibit an EDE in their inner wind region, but the CO channel map(at sufficiently high angular resolution) is dominated by the spiral.The correlation is quantified by a Kendall’s rank ⌧ b correlation coefficient of 0.79, indi-cating a high correlation between the two observables (see Sect. S3.3). We also apply thesethree Classes to previous observations of other AGB sources in the literature (see Sect. S4.1and Sect. S3.3). More than twenty nearby AGB stars are then recognized as pronouncedlynon-spherical, and can be classified to first order on the basis of the mass-loss rate. This sug-gests there is a single mechanism which modifies the AGB wind morphology, and the mass-lossrate is an important parameter establishing the morphological outcome. Of the proposed mod-els aiming to explain asphericity in AGB winds, (sub-)stellar binary interaction can explainthe observed ATOMIUM morphologies (see Sect. S4.1) and the mass-loss rate relation (seeSect. S4.2). Single-star models based on rapid stellar rotation or strong magnetic fields (11) cannot explain these characteristics in a generalized way, without invoking target-specific con-cepts. Binary-induced characteristics include (see Sect. S4): – The presence of a companion can perturb the AGB wind flow into a spiral structure either by:i) the direct gravitational pull of the companion which produces a bow shock around it and atail in its wake, reminiscent of the Bondi-Hoyle-Lyttleton (BHL; (87, 88) ) configuration; orii) the orbital motion of the primary AGB star around the common center-of-mass (19) . – Binary interaction can also result in the formation of a circumbinary disk, or an accretiondisk around the companion. In each case, a systematic rotating flow will arise, axially shap-ing the disk and producing a velocity structure that is far more complex than a (simple)Keplerian flow (20) . – For five ATOMIUM sources, the low and medium-resolution ALMA observations indicatethat the kinematical structure departs from the classical velocity law and that v wind (SiO) > able S2: Description of the prevailing morphological and dynamical structure of the ATOMIUMAGB sample.
First column gives the name of the target, second column the mass-loss rate (see Table S1),third and fourth columns the information deduced from the CO J = 2 ! channel maps, column 5 theratio of the velocity deduced from the CO J = 2 ! line (see Table S1) over the maximum velocity asdeduced from one of the SiO lines in the survey, and columns 6–9 the information as deduced from theSiO stereograms, moment1-maps and position-velocity (PV) diagrams, in particular on the velocity field v . Explanation of the notation: ? (x) = faint arc, x = several clear arcs with extent < , c-xx = circular/elliptical arc centred aroundtarget, o-xx = arcs symmetrically offset from central target, a-xx = pronounced asymmetric arc, xxx =more than 1 arc with extent > , often at regular intervals. † r/b: minor red/blue shift, R/B: well-resolved red/blue shift, ‘-’: no red/blue shift. ‡ c-rb: red/blue shift with complex signature, RB: clear red/blue shift indicating rotation or bipolarity,(o) shift with spatial offset w.r.t. central target, (rb) potential red/blue shift, ‘-’: no stringent red/bluesignatures. § (B): (weak) butterfly-like signature, BB: 2 bright blobs, (HG): weak hourglass signature, BW: bluewing absorption (89) , S: almost axi-symmetric PV for axis at 0 . k ‘bipolar/rotating flow’ indicates a directed bipolar flow or an EDE/disk-like structure, with may displayKeplerian rotation; ‘skewed rotating v -field denotes that systematic, but complex, signs of rotation aredetected with the v = 0 signature in the moment1-map being skewed; ‘complex dynamics’ refers to aclear blue/red signature in the moment1-maps, but no obvious systematic rotation can be deduced; ‘-’denotes that no conclusion could be drawn, sometimes owing to too low a signal-to-noise ratio of theSiO data (‘low S/N’).Name Mass-loss Morphology Kinematics Morpho-dynamics Figuresrate CO J = 2 ! channel map SiO(M yr ) Arcs ? Description v (CO)/ v max Stereogram † Moment 1 ‡ PV § Description k S Pav ⇥ (x) red/blue shift centralemission 0.77 r/b c-rb (B) skewed rotating v -field S8–S11T Mic ⇥ x several regular arcs 0.82 r/b c-rb (B) skewed rotating v -field S12–S15U Del . ⇥ c-xx ‘eye’-like feature+arcs 1.00 R/B RB BB + (B) bipolar/rotatingflow S16–S19 Continued on next page able S2 – Continued from previous page
Name Mass-loss Morphology Kinematics Morpho-dynamics Figuresrate CO J = 2 ! channel map SiO(M yr ) Arcs ? Description v (CO)/ v max Stereogram † Moment 1 ‡ PV § Description k RW Sco . ⇥ c-xx weak ‘eye’ + arcs 1.00 - - - - (low S/N) S20–S23V PsA ⇥ c-xx ‘eye’-like feature + arcs +brighter red/blue 1.00 - - BB + S bipolar outflow S24–S27SV Aqr ⇥ c-xx weak ‘eye’+brighterred/blue 0.88 - - BW - (low S/N) S28–S31R Hya ⇥ o-xx bipolar+hourglassstructure 1.00 R/B(inner 0.8 ) c-rb(inner 0.8 ) - skewed rotating v -field (inner 0.8 ) S32–S36U Her . ⇥ a-xx pronounced asymmetricarcs 0.88 r/b - (B) +BW complex dynamics S37–S40 ⇡ Gru . ⇥ o-xx bipolar+regulararcs+weak hourglass 1.00 R/B RB (B) bipolar/rotatingflow S41–S45R Aql . ⇥ xxx rose-like spiral, brighterin red 1.00 - - S - S46–S49W Aql ⇥ xxx filamentary spiral 0.75 r/b c-rb (o) BW complex dynamics S50–S53GY Aql . ⇥ xxx spiral 0.89 - c-rb (o) - complex dynamics S54–S57IRC . ⇥ xxx weak, but regular spiral 0.74 r/b RB - bipolar/rotatingflow S58–S61IRC +10011 . ⇥ xxx weak, but regular spiral 0.72 r/b (rb) - complex dynamics S62–S65 wind (CO). This can be caused, for example, by a Keplerian disk-like structure for which thetangential velocity component is inversely proportional to the radial distance, i.e. v ? Kep . ( r ) = p G M ? /r , with G the gravitational constant, M ? the mass of the AGB star to which thedisk is gravitationally bound, and r the radial distance (17) . Another cause for this behaviormight be the presence of a stellar companion so the companion’s gravity locally enhancesthe velocity amplitude by lowering the effective gravity felt by a particle leaving the primaryAGB star.Other physical phenomena, such as stellar pulsations and a magnetic field might induce addi-tional complexities on the binary-induced morphologies. For example, pulsations can cause anadditional ripple-like structure, visible in the lower density bipolar lobes (20) . S3.3 Statistical correlation coefficient and probability
We quantify the correlation in Table 1 using the Kendall’s rank correlation coefficient ⌧ b , astatistical measure of the ordinal association between two quantities based on the ranks of thedata. The rank correlation quantifies the strength of association based on the relative occurrenceof concordant and discordant pairs. The ⌧ b statistic makes adjustments for ties (90) . The ⌧ b of0.79 between the logarithm of the mass-loss rate and our 3 Classes is high, supporting theordinal association between the mass-loss rate and the morphological Class. Adding the elevenother AGB sources whose geometry was previously deduced from observations with ALMA(see Sect. S4.1) yields a similar correlation coefficient of ⌧ b = 0 . . The value of ⌧ b = 0 . comes with a two-sided p -value of p = 2 . · under the null hypothesis of ⌧ b = 0 .The probability to belong to a given morphology class (a categorical variable) given themass-loss rate ˙ M (a continuous variable) can be modelled using a multinomial logistic regres-sion. We assume a linear model ln ✓ ⇡ n ⇡ ◆ = ✓ ,n + ✓ ,n x , (S2)where ⇡ n is the probability of belonging to class n { , } , and the explanatory variable x ⌘ log ˙ M . The resulting probability curves are shown in Fig. S3.Class 3 is distinguished from Classes 1/2 using the mass-loss rate, but there is considerableoverlap between Class 1 and Class 2. The logistic model fitting yields a pseudo- R (91) of R = 0 . , indicating the mass-loss rate is a relevant explanatory variable. This value of R increases to 0.84 if we merge Classes 1 and 2 then fit a 2-class model.This outcome is a consequence of the classification strategy and the underlying physicsgoverning binary-induced phenomena: – The most distinct outcome of Step 1 is the presence of a spiral-structure in a fraction of theATOMIUM sources. Step 1 is based on the CO channel maps, and CO is a tracer of thedensity, and hence of the mass-loss rate, of AGB stars.15ig. S3:
Probability curves for the ATOMIUM Classes.
Probability curves obtained througha linear multinomial logistic regression using the mass-loss rates and the morphology classifi-cation of 25 AGB sources (the 14 ATOMIUM AGB sources and the 11 AGB sources describedin Sect. S4.1). The black dots are the observations, the orange, blue, and green curves are theprobability curves for Classes 1, 2 and 3 respectively. – Step 3 differentiates between Class 1 and Class 2 by using SiO to diagnose the morpho-kinematical behaviour of the inner wind region. The mass-loss rate is a relevant parame-ter establishing the morphological outcome, but the wind acceleration — and in particularthe ratio of the wind speed over the orbital speed — is also involved (see Sect. S4.1 andSect. S4.2). This implies that including the wind velocity pattern as a variable would changethe multinomial logistic regression. However, the small sample size prevents us from addinganother variable. The differentiation between Class 1 and Class 2 is visible in Fig. S3 be-cause a lower mass-loss rate correlates with a lower wind acceleration (and terminal windvelocity via mass conservation), and hence the higher potential of forming an EDE withcomplex binary-induced interaction patterns (see Sect. S4.1). The differentiation betweenClass 1 and Class 2 is therefore not as strict as the differentiation between Classes 1/2 andClass 3.The high probability of high mass-loss rate targets to be Class 3 is not caused by an ob-servational bias. The CO intensity scales with ⇠ ˙ M /D (35) . The Kendall’s rank correlationcoefficient ⌧ b between log( ˙ M ) /D and the 3 Classes is 0.26, implying that both variables arenot correlated (at least not linearly), see Fig. S4.16ig. S4: Testing of an observational bias.
Morphology classes of 25 AGB sources (same asin Fig. S3) plotted as a function of the logarithm of ˙ M / D (Table S1). The Kendall’s rankcorrelation coefficient ⌧ b between both variables is 0.19, implying no linear correlation. S4 The ATOMIUM morphological classification in the context of binary-induced wind morphologies
S4.1 Binary-induced wind morphologies
In this section, we first summarize the modelling outcomes for binary systems with a mass-losing AGB star as primary star. The presence of a binary companion can give rise to a plethoraof specific morphological imprints in the AGB stellar winds. Some of these morphologicalimprints have already been detected in observations. A synopsis of the most recent ALMAresults, prior to the ATOMIUM survey, for which the morphology of other AGB winds hasbeen deduced, is given in Table S3. In a last part, we focus on the fact that all ATOMIUMsources are oxygen-rich AGB stars for which the wind acceleration might be much lower thanin the case of carbon-rich AGB stars. This difference in wind acceleration profile can have aprofound impact on the binary-induced wind morphologies, which here will be described.
Morphologies in numerical simulations:
Hydrodynamical simulations have demonstratedthe influence of a companion on a wide orbit ( a au) around a mass-losing AGB star (92) .For wide binary systems (with a au) the mass transfer does not occur through Roche-lobeoverflow (RLOF), in which the primary star transfers material to its companion once it fillsits Roche lobe (93) . Rather for these wide systems, mass transfer occurs via the stellar wind17wind Roche-lobe overflow, WRLOF) (94) . In the WRLOF situation, wind material from themass-losing AGB star fills the giant’s Roche lobe and is transferred to the companion througha narrow and compressed channel which generally does not pass through the inner Lagrangianpoint. WRLOF occurs when the Roche lobe surface is detached, while RLOF happens whenthe Roche lobe surface is connected. Simulations covering the binary parameter space haveshown the WRLOF scenario can lead to wind capture rates by the companion, which differ byup to an order of magnitude compared to the BHL wind accretion scenario (95, 96) . Complexinner wind morphologies can arise (20, 94, 97) with a strongly non-radial velocity vector fieldnear the EDE (see discussion in Sect. S8). Depending on the mass ratio, binary separation,mass-loss rate, wind velocity, eccentricity etc. a variety of morphologies arises including spiralstructures (which can be bifurcated), accretion and circumbinary disks, bipolar outflows, EDEswith a regular (Keplerian) or complex velocity vector field, ‘spider’ or ‘rose’-like structures etc. Previously observed morphologies:
The carbon-rich AGB star IRC +10216 ( ˙ M ⇠ . ⇥ M yr ) has multiple, incomplete, concentric shells in its envelope (98) . These shells wereattributed to mass-loss modulations with a time scale of ⇠ ˙ M ⇠ . ⇥ M yr ) (16) . The spiral was interpreted as being causedby binary interaction in which the mass losing AGB star undergoes reflex motion which shapesits wind with a spiral pattern (92) . Twelve AGB stellar winds have previously been observed atsufficiently high spatial resolution (see Table S3 and Sect. S9). The observations enumeratedin Table S3 are consistent with our classification scheme: the effect of EDE/disk-like structuresand bipolar outflows are more readily recognized in lower mass-loss rate targets, while highmass-loss rate AGB stars have spiral-like arcs.Observations of carbon-rich AGB stars only show spiral-like structures; each of them fallinginto the category of high mass-loss rate stars ( ˙ M > ⇥ M yr ; see Table S3). An EDE hasbeen detected in a few carbon-rich sources, including IRC +10216 (111) and V Hya (112) .Nevertheless, we regard our classification scheme as primarily applicable to oxygen-rich AGBstars. The case of low wind acceleration:
Our sample is composed of oxygen-rich AGB stars. Ob-servational studies have shown that in oxygen-rich AGB stars the wind acceleration mightbe much lower than assumed, with terminal wind velocities being reached at 50 – 200 stellarradii (89, 113–115) . Using Eq. (S1), this is quantified by greater than 1, while for carbon-richstars is around 0.5 (106) , due to the opaque carbon dust grains facilitating photon momentumtransfer. For some oxygen-rich AGB stars, (89,114) , but the relationshipbetween and the mass-loss rate is unknown.This low wind acceleration, v /dt = a , can affect the wind morphology, because the windvelocity will be lower than or similar to the orbital velocity ( v orb ( r ) ) in a geometrically extendedregion (see Fig. S5, compare red line with the two star-symbols at a distance r = a ), therebystrengthening the interaction between the wind and the companion star (116) . If the ratio of18able S3: Overview of the wind characteristics of previously-observed AGB sources.
Columns 1–5 contain the source name, spectral type (‘M’ indicating an O-rich AGB star, and‘C’ a carbon-rich AGB star), pulsation variability type, mass-loss rate, and wind velocity. InColumn 6, a description of the wind morphology as published in the literature is given. Column7 indicates the ATOMIUM classification and references are given in Column 8.Name Spectral Varia- Mass v wind Description ATOMIUM Referencestype bility loss classificationtype (M yr ) km s Mira M Mira ⇥ † (99, 100) R Dor M SRb ⇥ ? differentially ro-tating disk Class 2 (79) EP Aqr M SRb . ⇥
11 biconical + nar-row spiral indifferentiallyrotation disk Class 2 (101, 102)
R Aqr M Mira ⇥
14 EDE+two plumesperpendicular toEDE Class 1 (103, 104) L Pup M SRb ⇥ ? (sub-)Kepleriandisk + bipolar Class 2 (17) OH 26.5+0.6 M Mira . ⇥
16 spiral+EDE Class 3 (55)
CIT 6 C SRa ⇥
21 spiral Class 3 (105)
IRC +10216 C LPV . ⇥ (106, 107) OH 30.1 . ⇥
23 spiral+EDE Class 3 (55)
R Scl C SRb ⇥ (108, 109) AFGL 3068 C Mira . ⇥
14 (bifurcated) spi-ral Class 3 (16, 110) ? if two values are given, this last value between parenthesis reflects the Keplerian velocity atthe inner boundary; † ALMA observations indicate a Class 1 object, but the CO channel map exhibits a complexmorphology due to WRLOF and a low wind velocity (see Sect. S9). We therefore exclude Mira( o Cet) in the analysis performed in Sect. S3.3.the wind over the orbital velocity at the orbital distance of the companion is <
1, any binary-induced morphological feature will be more complex compared to the case of a high windacceleration with v wind v orb . For example, if an Archimedean spiral forms for a high windacceleration scenario, the same binary setup applied to a low wind acceleration scenario can19ig. S5: Velocity fields of importance for the binary interaction . Given are the escape veloc-ity (in black), the wind velocity of an oxygen-rich star assuming = 5 (red), and of a carbon-richstar assuming = 0 . (blue) for a wind acceleration starting at 2.75 au. The central target isassumed to have a mass of 1.02 M , its Roche-lobe R R, is indicated with the black arrow andits stellar size of 267 R with the shaded region. For a companion mass of 0.51 M at an or-bital separation, a , of 15 au, the common center-of-mass (CoM) is at 5 au from the primary star(indicated with the dotted magenta line); its Roche-lobe R R, is indicated with the gray arrow.The star-symbols indicate the orbital speeds of both stars with respect to the CoM, and of thecompanion with respect to the primary. The x -axis is in units of the binary separation a . Inthis example, the two Roche lobes are detached, but WRLOF can occur. In the particular caseof a low velocity wind (red curve), the material has not yet reached the escape velocity at adistance r = a and a strong interaction with the companion’s gravity field can occur, resultingin complex wind morphologies (116) .yield a ‘broken/complex’ spiral structure that will potentially only attain a self-similar regularstructure far out in the wind. The ‘rose-like’ morphology of R Aql (Figure 1) appears similarto this [ (116) , their figure C1]. In addition, a lower wind velocity increases the capture radius(Eq. S3). This induces a difference between oxygen-rich and carbon-rich AGB stars and impliesthat (compact) EDE in the orbital plane of the binary system can be more pronounced foroxygen-rich stars, provided the filling factor of the dust condensation radius by the Roche loberadius is large enough. The velocity vector fields in these EDEs strongly depart from a radialpattern. Moreover, the more progressive wind acceleration around oxygen-rich stars (i.e. withhigher -values) will also smear out the (spiral) shock compared to the wind around a carbon-rich star. Because the features are more dilute around an oxygen-rich star, we expect they will20e more difficult to identify [e.g. (116) , their figure 5] .Hence, carbon-rich stars and oxygen-rich higher mass-loss rate AGBs with higher windacceleration will have a lower v orb ( r ) /v wind ( r ) ratio, resulting in a smaller geometrical region inwhich the dominant wind streaming lines are non-radial. For those systems, we expect the windswill more readily exhibit (Archimedean) spiral structures that expand radially in a self-similarway. An EDE can also form (depending on the system’s properties), but the spiral will dominatethe wind’s appearance. An oxygen-rich AGB star with a low wind acceleration will be moreprone to complex wind morphologies, such as ‘rose’-like morphologies. These systems have ahigher chance of forming a pronounced EDE with a large equator-to-pole density contrast, andhence the formation of bipolar cavities. Spiral structures could form as well, but a low windvelocity implies the spiral features be more broken/complex in structure and more dilute andhence more difficult to detect. S4.2 Analytical estimates of binary-induced AGB wind morphology
We consider a simplified binary set-up to produce an analytical relation between binary param-eters and the departure from radial motion in the AGB stellar wind. We begin by defining theregion where the gravitational field of the companion deflects the flow from its trajectory. Theradius of the sphere of gravitational influence of the companion is denoted as R H . Wind accretion scenario:
When the wind speed at the orbital distance a , v wind ( a ) , is largecompared to the orbital velocity, v orb ( v wind ( a ) v orb ), R H can be approximated by the captureradius (sometimes referred to as gravitational or accretion radius): R H ⇡ R capt = 2 Gm comp v , (S3)with m comp the mass of the companion. This describes the critical impact parameter belowwhich a planar flow coming from infinity with a certain kinetic energy per unit mass (and hencecertain wind speed) is likely to be captured (117) . It relies on the classical BHL formalism (87, 88) . To derive the capture radius, we compare the kinetic energy per unit mass ( v / )to the gravitational potential of the companion ( G m comp /r ) and find the value of r where bothare equal. For r < R capt the mechanical energy is negative, while in the opposite situationit is positive. For decreasing wind speeds and larger companion masses, the effective crosssection set by the accretion radius increases, thereby enhancing the impact of the companion ondeflecting the flow from the primary AGB star. Roche lobe overflow scenario:
The wind-accretion scenario is not valid for low wind speedscompared to the orbital speed ( v wind ( a ) ⌧ v orb ). In that case, the accretion radius can be largerthan the Roche Lobe radius of the companion, R R, . We approximate the region of gravitational21nfluence of the companion as its Roche lobe (96) following Eggleton’s formula (118) : R H ⇡ R R, = " ( q ) a = 0 . q / . q / + ln(1 + q / ) a , (S4)where q is the ratio of the mass of the companion star to the mass of the AGB star, m comp /M ? .The function " only depends on the mass ratio. Accounting for the eccentricity of the orbit onthe assumption of the mass of the companion being much smaller than the primary mass, theradius can be approximated by the Hill radius of the companion (119) R H ⇡ R HL = a (1 e ) ✓ m comp M ? ◆ / , (S5)where e is the orbital eccentricity and a (1 e ) is the pericenter distance. This approach is validwhen the amount of kinetic energy per unit mass given to the flow by mechanisms other thangravity (for example radiative pressure on dust grains or resonant line absorption of UV pho-tons) is negligible compared with the amplitude of the Roche potential ( ⇠ G ( M ? + m comp ) /a ).The RLOF and wind accretion scenario are the two extreme situations. In between, WRLOFoccurs (94, 96) , where both the mass ratio (as in the RLOF case) and the wind speed (as in thewind accretion case) are involved (see Sect. S4.1).We introduce a dimensionless parameter Q p to predict the morphology of the outflow dueto companion motion, defined as Q p ⇡ p comp p wind = m comp v orb m wind v wind , (S6)in which p wind is the radial momentum of wind material the companion encounters in one orbit, p comp the tangential momentum of the companion, m wind the wind mass the companion encoun-ters in one orbit, v wind the local wind speed of the material lost by the AGB star at the location ofthe companion ( v wind ( r = a ) ), m comp the mass of the companion, and v orb the orbital speed ofthe companion at an orbital distance a from the primary given by p G ( M ? + m comp ) /a . Largevalues of Q p reflect a strong departure from radial motion, such as that caused by the formationof a dense, potentially rotating, EDE or a (Keplerian) disk-like structure. Intermediate valuesof Q p reflect a situation in which an EDE is formed with not too high a density contrast be-tween the equator and poles. Little departure from the radial flow will result in small valuesof Q p , in the case of a quasi unperturbed spherical outflow or when a spiral structure expandsself-similarly. The parameter Q p is defined for r ⇡ a . The formation of a circumbinary oraccretion disk — reflected by high values of Q p — does not imply that no spiral shock forms,but the expanding spiral will only dominate at larger distances from the mass-losing AGB star(see Sect. S4.1); vice versa if the prevailing wind geometry is dominated by a spiral this doesnot imply that no compact EDE can be formed (see also Sect. S4.1, (55) ).In one orbit, the companion encounters a wind mass m wind ' f w ⇡ R H ⇡ a ⇢ wind ( r = a ) + ⇡ R ⇢ wind ( r = a ) v wind ( r = a ) 2 ⇡ av orb , (S7)22ith ⇢ wind the density of the wind defined by the equation of mass conservation ⇢ wind ( r = a ) = ˙ M ⇡ a v wind ( r = a ) . (S8)The first term on the right in Eq. (S7) describes the mass initially present at r = a from theongoing AGB stellar wind with f w a fraction < , while the second reflects the additional masswhich can be captured by the companion. For the wind accretion scenario ( v wind v orb ),Eq. (S7) reduces to m wind ' ⇡ a R ⇢ wind ( r = a ) v wind ( r = a ) v orb , (S9)while in the case that v wind ⌧ v orb , we can write that m wind ' f w ⇡ a R R, ⇢ wind ( r = a ) . (S10)For a low wind velocity ( v wind ⌧ v orb ) and using Eq. (S5), we can write Q p as Q p = 8 . ⇥ e ) f w ✓ m comp M ◆ / ✓ M ? M ◆ / ⇣ a ⌘ / ˙ M M / yr ! . (S11)Eq. (S11) shows that high values for Q p will be obtained for a high mass of the companion m comp , a small orbital separation a , and a low wind mass-loss rate ˙ M . The companion massonly enters as m / , so Q p is not strongly dependent on the companion mass. This implies thatif low mass stellar companions are effective, massive planets might also be effective. Using thecapture radius (Eq. (S3)) to define R H , we obtain Q p = 0 . ⇥ ✓ m comp M ◆ ✓ q ◆ ⇣ a ⌘ ˙ M M / yr ! ⇣ v wind
10 km s ⌘ ✓ v wind v orb ◆ . (S12)Eq. (S12) shows similar dependence on binary parameters to Eq. (S11), but there is an explicitdependence on the orbital and wind speed. However, a lower wind velocity implies a smallervalue for Q p , which would indicate a smaller departure from radial motion. The smaller value of Q p arises because lower wind velocities imply a larger capture radius therefore m wind / v .Accounting for this difference in total intercepted mass, we estimate that more particles per unitmass will be deflected by the gravitational field of the companion in the case of a lower windspeed (see also Sect. S.4.4).For convenience, we set Q = 10 Q p (S13)to give a quantity close to unity. For the case of Jupiter orbiting around the Sun, and on theassumption the wind mass-loss rate of the Sun in its AGB stage will be around ⇥ M yr
23 i.e., similar to what has been deduced for L Pup, which has a near solar main-sequencemass (17) — Q is around 0.35 using Eq. (S11) for f w = 1 . Accounting for the fact that theSun will have lost part of its mass by the time it enters the regularly pulsating Mira AGB stage, Q will be even lower (around 0.2) suggesting that the presence of Jupiter will leave the solarwind morphology nearly unperturbed and only a very weak BHL type spiral will form.In summary, these simplified analytical estimates can explain the main morphological clas-sification we have identified in the ATOMIUM sample (Sect. S3.2): a decrease in mass-lossrate increases Q p and reflects the conditions in which a (compact) EDE can be formed withpronounced contrast in density between the poles and the equator. The EDE can harbor a cir-cumbinary disk or accretion disk around the companion (see Sect. S4.1). Higher mass-lossrates yield low values of Q p with minimal departure from radial motion. A spiral structure thatis expanding in a self-similar way would fall into this latter category. The observed correlationbetween the mass loss and the morphology is caused by the ability of orbiting companions toinject angular momentum into the (mostly) spherical mass loss from the star. If the mass loss istoo strong or fast, or if the companion is too far away (and hence has a low orbital velocity), thematerial in the circumstellar envelope is ‘overwhelmed’ by the radially streaming AGB windand cannot shape the material into an EDE. S5 (sub)-Stellar binary population statistics
Here we summarize the occurrence of (sub-)stellar companions around stars with an initialmass between 0.8–8 M . We differentiate between stars with initial mass less than < ⇠ ,i.e progenitor spectral types F, G, and K; and those with higher initial mass that represent A andB spectral types on the main sequence. Orbital period:
We first assess the minimum and maximum orbital radius a and orbital period P orb where companions can influence the shaping of the wind. A conservative upper limit is a < au (120) , or log P orb (days) < . . ( ⇠ ⇠ , applyingKepler’s third law). This is similar to the maximum recoverable scale of our medium resolutionALMA data ( ⇠ ) of ⇠
600 au at 300 pc; and is also the maximum orbital period where anorbiting body is likely to create arc-like structures. A conservative lower boundary would bean orbit on the radius of the AGB star ( > ⇠ . au), and hence log P orb (days) > ⇠ . (or ⇠ (21, 121, 122) . Because this will affect stars more than planets(see Sect. S6), we assume log P orb (days) > for stars and log P orb (days) > . for planets. Initial mass:
Next, we investigate the initial masses of the ATOMIUM AGB stars, becausei) the binary fraction is strongly mass-dependent (23) and ii) we need to determine how repre-sentative the ATOMIUM sample is of AGB stars in general. A first indication can be retrievedfrom the luminosity distribution of the ATOMIUM stars with . < ⇠ log L ( L ) < ⇠ . . Using24elations between the pulsation period and the luminosity, the implied progenitor stars are atleast ⇠ for a star to reach log L = 3 . during the interpulse period (22, 123) . However,stars whose mass is about a solar mass, will only seldom have a luminosity above that threshold,implying stars should have an initial mass > ⇠ so that log L . for an extended period.For six ATOMIUM stars we can determine the initial mass by using the O/ O isotopicratio which stellar evolution models predict is a sensitive function of initial mass (124, 125) .The derived initial mass is ⇠ for R Aql ( O/ O = 1.77 (125) ), ⇠ for R Hya( O/ O = 0.54 (125) ), ⇠ for U Her ( O/ O = 2.53 (125) ), ⇠ for W Aql( O/ O = 1.17, (124) ), ⇠ for IRC +10011 ( O/ O = 0.26, (124) ), and ⇠ forIRC O/ O = 3.1 (ALMA proposal ADS/JAO.ALMA2019.1.00187.S, PI T. Dani-lovich.). Most ATOMIUM stars are fundamental mode pulsators, however U Del is a first-overtone pulsator and long secondary period pulsator (41) , suggesting it is a higher-mass starwith
M > ⇠ M (123) . Therefore five out of the seven stars with mass measurements have anestimated initial mass greater than 1.5 M .There is a selection bias towards stars with high initial mass. One of our selection criteriawas the mass-loss rate, which we required be greater than ⇠ ⇥ M yr to improve thesignal-to-noise ratio. However, a star with an initial mass of ⇠ will only have a very shortperiod in its AGB phase during which the mass-loss rate is greater than M yr beforeit transforms into a post-AGB star and PN. More massive stars have a longer period duringwhich the luminosity is above log L = 3 . and the mass-loss rate is M yr . Foran initial mass of 2 M , the luminosity exceeds log L = 3 . for 1–2 Myr and the mass-lossrates exceeds M yr for just over 1 Myr, while the C/O ratio only exceeds unity for ⇠
300 000 yr [ (126) , their figure 5]. Using evolutionary tracks (22) , the criteria of log
L > . , T e ↵ < K, and ˙ M > ⇥ M yr yield a 68% confidence interval that stars with mass . . M were selected, and a 95% confidence interval of . . M . Consequently,we expect the majority of the ATOMIUM AGB stars to be more massive than ⇠ .This bias in the ATOMIUM AGB sample selection indicates our targets have a higher initialmass than typical AGB stars ( ⇠ for bulge Miras of intermediate age (1–3 Gyr) (127) ). Main-sequence stellar multiplicity factor:
We determined the main-sequence multiplicityfraction in the . . < ⇠ log P orb (days) < ⇠ . period range (see Table S4). The stellar binaryfraction of F/G/K stars (0.8 < ⇠ M < ⇠ ) is well studied (23) , but the lack of spectral featuresat optical wavelengths makes the search for radial velocity features from companions to A stars( ⇠ ) difficult. The fraction of F/G/K stars that are part of binaries with mass ratio q = m comp /M ? > . is ⇠ ⇠
85% for A/late-B stars (23) .We multiply both fractions by 1.1, because about 10% have white dwarfs companions [ (23) ,their figure 29]. Stellar companion binaries with log P orb (days) will tighten their orbit,evolve to a common envelope situation and merge; and binaries with log P orb (days) > . . will not influence the wind shaping, so the multiplicity of F/G/K stars reduces to ⇠ ⇠ (23) , their figure 37] In addition, about 10% of solar-typemain-sequence primaries are in triple/quadruple systems, while the triple and quadruple star25able S4: Main-sequence (sub-)stellar multiplicity fraction . The first and second columnsgive the main-sequence initial mass and related spectral type on the main-sequence. Columns 3–6 list the (sub)-stellar multiplicity fraction for companions with log P orb (days) < . and log P orb (days) > . for planets (or > for stars).Stellar Brown Planets M > M Jup
Planets
M > M
Jup companions dwarfs a = 10 au a = 2 au M ini < . M FGK ⇠ ⇠ ⇠ ⇠ M ini > . M AB ⇠ ⇠ ⇠ ⇠ is around 37% (23) . Main sequence sub-stellar multiplicity factor:
For planets and brown dwarfs statistics, weused the results of the Gemini Planet Imager survey (GPIES) (128) and California-Kepler survey (24, 129) . First order estimates (130) indicate a conservative minimum mass (at a given radius)for which planets can influence the shaping of the wind is a few Jupiter masses ( M Jup ), a numberconfirmed by hydrodynamical simulations. The GPIES results show that ⇠ < M < M Jup ) between 10 – 100 au, and ⇠
9% host a planet (5–13 M Jup ) inthe same orbital distance range. However, these results are strongly mass-dependent with ⇠ ⇠ having a planet (129) . Similarly, the quoted percentage forvery low-mass stellar companions around the higher mass A/B stars is probably a conservativelower limit. The frequency of 0.1–19 M Jup planets in the range of 2–10 au is ⇠ (129) . AGB (sub-)stellar multiplicity factor:
We next assess how this (sub-)stellar binary fractionevolves from the main-sequence to the AGB evolutionary phase. Tidal inspiral and stellar massloss will reduce the binary fraction by dragging nearest companions inwards and spiraling outthe farthest ones. On the assumption our target stars are a part of their way through the AGBevolution (because they are strongly mass-losing) their masses are probably half their birthmasses, hence the orbits of the majority of companions to the ATOMIUM stars will have ex-panded by a factor of ⇠ < a < au (2.22 dex),losing ⇠ ⇠
14% reduction in parameter space and ⇠ (131) show this is likely to cause a few percentreduction of the planet fraction in Table S4. Depending on the amount of photo-dissociatingUV irradiation from the host star, a fraction of the Jupiter mass gas giants can evaporate beforestars reach the AGB phase. This depletion factor is not well known, but is thought to be thecause for the absence of Neptune-mass planets at very close orbital radii (132) . We conclude26hat as stars evolve toward and on the AGB, the main-sequence (sub-)stellar binary fraction inTable S4 will decrease by roughly 10–20%.Accounting for all these effects, it is evident that the accumulated multiplicity fraction in-creases by a factor 2–3 over the range of 0.8–8 M with a (sub-)stellar multiplicity fraction forhigher mass AGB stars ( M > ⇠ . M ) of > ⇠ . This dependence of the multiplicity fractionon initial mass is also evident in the Gaia DR2 data (133) . As a result, i) the ATOMIUM datacomplemented with the additional ALMA data presented in Sect. S4.1 (Table S3) and Sect. S9,ii) the outcome of theoretical models for binary systems summarized in Sect. S4.1 and Sect. S8,and iii) the population statistics here described all indicate that binary interaction is the domi-nant wind shaping mechanism for these AGB stars. Our conclusion applies to all AGB stars forwhich observations indicate a mass-loss rate greater than ⇠ ⇥ M yr — i.e., those witha mass-loss rate exceeding the nuclear burning rate (4) so the mass loss determines the furtherstellar evolution. A high fraction of these targets will have an initial mass greater than ⇠ (see above) for which (sub-)stellar population statistics indicate the multiplicity fraction is S6 Predicting the evolution of wind morphology
S6.1 Change of orbital separation throughout the AGB evolution
We predict how the prevailing wind morphology might change throughout the AGB evolution.For a star of size 267 R (see also Fig. S5), we follow its evolution on the AGB track using therelation between the pulsation period and the luminosity, here expressed in terms of the absolutebolometric magnitude (134) M bol = .
00 log P + 2 . (S14)with the pulsation period P in days. The pulsation period is derived using the period-mass-radius relation (34) log P (days) = .
07 + 1 .
94 log R ? / R . M ? / M , (S15)and the mass-loss rate for pulsation periods P > days is given by (34) log ˙ M ( M yr ) = . . P (days) , (S16)but cannot exceed the single-scattering limit (34, 55) . The mass-loss rate is assumed to be ⇥ M yr for thermally pulsating AGB stars with lower pulsation periods (4) .For an AGB star of initial mass 1.5 M , the initial luminosity is ⇠ , P = 300 days,and ˙ M = ⇥ M yr . For various companion masses (1.2, 0.6, 0.3, 0.01 M ) we computethe change in orbital separation due to angular-momentum loss [ (135) , their Eqs. (6)–(13)].These analytical relations were derived for binary population synthesis codes, but cannot cap-ture all details of the complex wind mass-transfer physics. The rate of orbital change, ˙ a/a , isdependent on the current mass-loss rate, mass of the primary, mass-ratio, capture rate by the27ompanion, and ratio of the orbital speed to the wind speed. The calculation ends when theAGB mass is less than 0.6 M (136) — by which time the AGB star has reached a luminosityof 14 500 L , a pulsation period of 685 days, and a mass-loss rate of ⇥ M yr — orwhen both stars merge. Fig. S6 shows the result for four initial orbital separations ( a = 2, 4,10, 25 au). For initial orbital separations greater than 25 au, the results are similar to the onewith a = 25 au. If the companion has a mass ⇠ M Jup , the orbit always widens. The moremassive the companion and the smaller the initial separation, the greater the chance for the or-bit to shrink. For very small initial separations ( < ⇠ > ⇠
25 au, the orbital separation always increasesindependent of the mass of the companion. However, the maximum factors of the widening(1.12) and shrinking (0.76) are quite modest.A more pronounced change occurs for primary AGB stars with higher initial mass, andhence more extreme values of the mass ratio (see Fig. S7). Given Eqs. (S14)–(S16), the lu-minosity, pulsation period, and mass-loss rate at the start of the calculations are L ⇠ , P = 161 days, and ˙ M = ⇥ M yr , respectively. In general, the more massive the AGBstar the more extreme the orbit’s expansion (if occurring) that can be attained with maximumderived widening factors of . . M ? . A more massive AGB star also implies a largeinitial orbital separation for which all companions (independent of mass ratio) increase theirseparation during the AGB evolution. Carbon-rich AGB stars also develop larger orbital sepa-rations compared with oxygen-rich AGB stars during the AGB evolution.The frequency distribution for q > . is roughly flat in log P orb -space for main-sequencestars with mass between 2–5 M and log P orb = 3–6.5 [ (23) , their figure 37]. This implies themajority of companions reside at larger orbital separation ( a > ⇠ au). Not accounting forany change in orbital separation between the main-sequence and start of the AGB phase (1) ,Figs. S6–S7 show that at an initial separation of > ⇠ au the majority of the stellar and sub-stellar companions will widen their orbit during the AGB phase as stellar mass reduces due tomass loss, eventually reaching the BHL wind-accretion regime (see Fig. 2). Our results implythat early-type oxygen-rich AGB stars with a low mass-loss rate are more likely to host planet-mass companions. S6.2 Morphological characteristics on and beyond the AGB phase
Compact EDE/disk-like structures have a higher tendency to be formed around early-type oxygen-rich AGB stars with a slow wind acceleration and spiral structures are readily formed in late-type AGB stars with high mass-loss rate (Figure 2). Our results show that the aspherical ge-ometries arise during the AGB phase when mass-loss rates exceed > ⇠ ⇥ M yr . Thisdoes not preclude an earlier timing of the shaping event, but this cannot be determined withthe current data. The prevailing AGB wind morphology and companion properties specify theinitial aspherical conditions when transitioning into the post-AGB and PN phase. Our modelhas implications for the morphology during the AGB, post-AGB (or pre-PN) and PN phases.Detached shells are detected around carbon-rich stars (137) . These shells are thought to be28ig. S6: Evolution of the orbital separation in function of time for a primary AGB star ofmass 1.5 M . The x -axis represents the time on the AGB, with the zero point defined as whenthe luminosity rises above 5 400 L . The mass of the companion is 1.2 M ( q = /q = 1.25 inblack), 0.6 M (gold), 0.3 M (blue), or 0.01 M (magenta). The initial orbital separation isFig. S6A: 2 au, Fig. S6B: 4 au, Fig. S6C: 10 au, or Fig. S6D: 25 au. Simulations for a C-richAGB star with a wind acceleration given by a -profile with =0.5 are shown with full line,while the dashed line shows an O-rich AGB star with =5. For q = 1.25, the evolution of theorbital separation for situations with a low and high -value coincide.caused by a thermal pulse yielding a peak in luminosity and hence in mass-loss rate ˙ M ( t ) ordensity. The chance for this initially spherical shell to be deformed is smaller for carbon-richbinary stars than for oxygen-rich ones because the wind streaming lines are more radial (seeSect. S4.1). These detached shells might be the precursors of ring-like features detected incarbon-rich post-AGB stars, such as IRAS 19700+3457 (138) .Our proposed scenario explains the formation of a silicate-rich EDE around carbon-richstars (so-called silicate carbon stars; (27) ). During the early AGB phase, when the mass-loss29ig. S7: Same as Fig. S6, but for a primary AGB star of mass 3 M . The zero point fortime on the AGB is when the luminosity rises above 2 500 L . The hatched region indicatesthe stellar radius, the gray line the evolution of the Roche lobe radius of the AGB star for m comp = 0.01 M .rate was still modest and the C/O ratio lower than 1, a stable disk-like structure can be formedwhich favours the formation of crystalline silicate dust (55) . Later, as mass-loss rate and C/Oratio increase due to dredge-ups, the star will gradually become carbon-rich implying a ve-locity vector field which is predominantly radial. Moreover, our scenario is consistent withthe occurrence of mixed chemistry in post-AGB stars and galactic bulge PNe (28) . Carbondredge-up cannot explain the mixed chemistry because the bulge stars are of too low mass.Ultraviolet irradiation of the dense torus (EDE in our scenario) induces the formation of hy-drocarbon chains (28) . In addition, our scenario predicts that disks are predominantly foundaround oxygen-rich post-AGB stars. This is compatible with disk detections around post-AGBstars (139) .The impact of the progenitor mass on the orbital decay or widening has implications for the30orphological classification of post-AGB stars, which are divided in two classes: i) the ‘sole’nebulae, for which the emission is star-dominated with faint extended nebulosity, and ii) the‘duplex’ nebulae, with dust-dominated emission and a faint or completely obscured centralstar (29) . ‘Duplex’ sources tend to have a higher equator-to-pole density ratio, attributed toa more intense axisymmetric (equatorial) superwind mass loss at the end of the AGB, andhence formation of an EDE, and have been suggested to have higher-mass progenitors (140) . Itfollows from the calculations in Sect. S6.1 that the more massive the AGB progenitor star, themore likely the orbit of a stellar companion will tighten during the AGB evolution favouringthe formation of a compact EDE, thereby producing the post-AGB morphological dichotomy.This progenitor-mass dependent evolution of the orbit also suggests that bipolar PNe are thedescendants of high mass stars, so ‘duplex’ sources might be the precursors of bipolar PNe.This prediction is consistent with observations that most bipolar PNe have a central star ofhigh initial mass (73, 141, 142) . ‘Sole’ post-AGB stars have been suggested to be precursorsof elliptical PNe (29) . The lower tendency of carbon-rich AGB stars to form EDEs implies ahigher fraction carbon-rich elliptical PNe, as observed (141) .The orbits of binary post-AGB stars are often non-circular. Over 70% of the post-AGBbinaries have nonzero eccentricities as high as 0.3, despite Roche-lobe radii that are smallerthan the AGB stellar radius, and tidal circularization which should have been strong when theprimary was on the AGB (30) . Tidal coupling to the circumbinary disk present during the post-AGB phase has been proposed as a mechanism to increase the eccentricities (143) . However,this scenario has difficulties explaining the eccentricities of post-AGB stars, and requires mas-sive ( > ⇠ M ) long-lived ( > ⇠ years) circumbinary disks which do not accrete (144) —a nontrivial combination, especially in the latter aspect. Our results show that circumbinarydisk-like structures can be present in the early AGB phase, thereby lengthening considerablythe timescale during which eccentricity pumping can occur as a result of gravitational interac-tion with the disk. Using parameters resembling the well-studied L Puppis system (17) , resultsin e = 0 . [using equation 5 of (144) ]. Bifurcations in spiral-like patterns suggest there is acompanion residing at an eccentric orbit around the carbon-rich AGB star CIT 6 (105) and sothe eccentricity might already be (highly) nonzero at the beginning of the post-AGB phase.The progenitor mass of PNe is greater than ⇠ (6) , a population for which the (sub-)stellar multiplicity rate approaches 100% (Sect. S5). We expect round PNe to be rare which isconsistent with the small fraction of < ⇠
20% observed (9, 10) . Aspherical PNe will not only arisedue to the action of short-period binaries within a common-envelope evolution (orbital period P orb days; (12, 145–149) ), but also the transitioning of wider AGB binaries into the post-AGB and PN phase can result in the formation of aspherical PNe. In particular the formationof an EDE will favour a polar ejection of material. This argument is supported by the growingnumber of aspherical PNe detected whose binary central stars have a long-period orbit (1 yr P orb
10 yr) so that the common-envelope phase has been avoided (150–153) . Our resultssupport the claim that binarity may be a prerequisite for the formation of observable asphericalPNe (154) . While magnetic fields and stellar rotation can play a role in shaping PNe (as wellas AGB winds), the angular momentum from a binary is thought to be required to sustain the31lobal magnetic field and rotation long enough to affect the geometry of the mass loss and theshaping of the PN (13, 155) .Although none of the ATOMIUM sources displays a smooth, spherical wind, many nebulaesurrounding post-AGB and PN stars exhibit even stronger asymmetries, notable examples in-clude M1-92 (156) , the ‘Butterfly’ nebula M2 (157) , or HD 101584 (158) . Any asphericitypresent in the AGB wind will impact the post-AGB morphology. The development of an EDEduring the AGB phase can lead to the creation of bipolar/multipolar shapes, due to the actionof a collimated fast wind (CFW) during the late AGB or early post-AGB phase (9) . This CFWcarves the dense AGB shell from the inside out. As the central star evolves further towardsthe PN phase, the existing aspherical post-AGB morphology will be further shaped due to theaction of spherical, radiatively driven, fast wind from the hot PN central star (9) . Additionalmechanisms such as jets launched from the accretion disk around the companion (159) , or thecombination of a fast collimated wind with toroidal magnetic field (160) , could also carve thepost-AGB and PNe morphology. S7 ALMA ATOMIUM observational results
In this section, we present all the data used in our morphological characterization of the ATOM-IUM AGB stars. These include: channel maps of the CO J = 2 ! line at medium and lowspatial resolution; stereograms of the SiO J = 5 ! line; and moment1-maps and position-velocity diagrams of the CO J = 2 ! and SiO J = 5 ! lines. Channel maps:
Fig. S8, Fig. S9, Fig. S12, Fig. S13, Fig. S16, Fig. S17, Fig. S20, Fig. S21,Fig. S24, Fig. S25, Fig. S28, Fig. S29, Fig. S32, Fig. S33, Fig. S37, Fig. S38, Fig. S41, Fig. S42,Fig. S46, Fig. S47, Fig. S50, Fig. S51, Fig. S54, Fig. S55, Fig. S58, Fig. S59, Fig. S62, Fig. S63:The minimum and maximum velocity offset with respect to the local standard-of-rest velocityare determined from the line profile, so that the integrated line intensities are greater than 3times the rms noise (see Table S6). We set the minimum intensity in the color scale is set to ( ⇥ rms . The maximum of the color scale is chosen to better visualize the weaker features,and is set to ⇠ ⇡ Gru have a biconical / hourglass structure(Fig. S32 and Fig. S41).
Stereograms:
Fig. S10A, Fig. S14A, Fig. S18A, Fig. S22A, Fig. S26A, Fig. S30A, Fig. S34A,Fig. S39A, Fig. S43A, Fig. S48A, Fig. S52A, Fig. S56A, Fig. S60A, Fig. S64A: Stereogramplots of the SiO J = 5 ! line show the offset between blue and red shifted emission. Theblue and red contours are constructed from 49% of the blue and red-shifted emission of themolecular lines, respectively. The black dotted contours are constructed with the remaining 2%32f the velocity channels which are centered on the local velocity of rest v LSR . The contours areevenly spaced between ⇥ rms in the (red, blue, black) velocity channel maps and the maximumof the moment0-map of the respective (red, blue, black) channels, where the moment0-map isthe emission map integrated over all frequency channels with measurable emission in the line.A distinct difference in the position of the red and blue shifted emission is observed in U Del(Fig. S18A) and ⇡ Gru (Fig. S43A); and a small systematic difference is also noted for S Pav(Fig. S10A) and T Mic (Fig. S14A), and the inner 0.8 region of R Hya (Fig. S34A). First moment maps:
Fig. S10B – Fig. S10C, Fig. S14B – Fig. S14C, Fig. S18B – Fig. S18C,Fig. S22B – Fig. S22C, Fig. S26B – Fig. S26C, Fig. S30B – Fig. S30C, Fig. S34B – Fig. S34C,Fig. S36, Fig. S39B – Fig. S39C, Fig. S43B – Fig. S43C, Fig. S48B – Fig. S48C, Fig. S52B –Fig. S52C, Fig. S56B – Fig. S56C, Fig. S60B – Fig. S60C, Fig. S64B – Fig. S64C: First moment(or moment1) maps of two SiO isotopologue lines are used as a tool for visualizing structuresin the velocity fields in the inner wind region. The maps are obtained by taking P ⌫ red ⌫ blue I ⌫ v ⌫ d ⌫ P ⌫ red ⌫ blue I ⌫ d ⌫ , (S17)with v blue and v red indicating the minimum and maximum velocity offset with respect to the lo-cal standard-of-rest velocity as determined from the line profile (see above), and I ⌫ the intensityat frequency ⌫ with corresponding velocity v ⌫ with respect to the local standard-of-rest velocity.For most targets, the SiO emission mainly traces the inner 1 region of the wind. For some tar-gets the line velocity map exhibits distinct red-shifted and blue-shifted components, which is theclassical signature of rotation (17) or bipolar outflow: U Del (Fig. S18B – Fig. S18C) and ⇡ Gru(Fig. S43B – Fig. S43C) are examples of this type of signature. Another example is the COmoment1-map of R Hya (Fig. S36). In the inner 0.5 region of R Hya there is some evidence ofpotential rotation/bipolar outflow in agreement with the offset between the red and blue shiftedemission in the stereogram. S Pav (Fig. S10B – Fig. S10C) and T Mic (Fig. S14B – Fig. S14C)have signs of complex rotation or bipolar dynamics, with their moment1-maps resembling thatof the red supergiant Betelgeuse (161) . Position-velocity diagrams:
Fig. S11, Fig. S15, Fig. S19, Fig. S23, Fig. S27, Fig. S31,Fig. S35, Fig. S40, Fig. S44, Fig. S49, Fig. S53, Fig. S57, Fig. S61, Fig. S65: Position-velocity(PV) diagrams of the CO J = 2 ! and SiO J = 5 ! emission are shown for each target.A PV diagram is obtained by taking a slice through the 3D ALMA data at an arbitrary angularaxis in the plane of the sky, and hence is a 2D plot of the emission along this chosen axis versusvelocity. In principle any slit width can be chosen. If the slit width is larger than one singulardata pixel, the emission is collapsed onto each other by summing up the emission with identicalPV coordinates. We here have chosen the slit width to encompasses the full emission in themoment0-maps. A set of two orthogonal pairs of PV diagrams are displayed with the positionangle (PA) chosen to produce the greatest difference between the two PV diagrams, maximizingthe asymmetry in the data. 33 Fig. S8:
ALMA medium resolution CO J = ! channel map of S Pav. North is up,East is left. Contours show the continuum emission at [10, 30, 60, 90, and 99]% of the peakcontinuum emission. The ordinate and co-ordinate axis give the offset of the right ascension anddeclination, respectively, with respect to the peak of the continuum emission. The velocity withrespect to the local standard of rest velocity ( v LSR as given in column 11 in Table S1) is given inthe upper right corner of each panel (in units of km s ). The ALMA beam is shown as a whiteellipse in the bottom left corner of the bottom left panel. The minimum and maximum velocityoffset are determined from the line profile, accounting for all channels in which the integratedflux is greater than ⇥ rms (see Table S6). The minimum intensity in the color scale is set to ( ⇥ rms , the maximum of the color scale is set to ⇠ Low resolution CO J = ! channel map of S Pav. Same as Fig. S8 but for thelow resolution observation. 35ig. S10:
SiO stereogram and moment1-map of S Pav.
North is up and east is to the left. Thecross indicates the position of the maximum continuum flux. The ordinate and co-ordinate axisgive the offset of the right ascension and declination, respectively, with respect to the peak ofthe continuum emission. Fig. S10A: SiO J = 5 ! stereogram. The blue and red contours areconstructed from 49% of the blue and red-shifted emission of the molecular lines, respectively.The black dotted contours are constructed with the remaining 2% of the velocity channels whichare centered on the local velocity of rest v LSR . The contours are evenly spaced between ⇥ rms in the (red, blue, black) velocity channel maps and the maximum of the moment0-map of therespective (red, blue, black) channels, where the moment0-map is the emission map integratedover all frequency channels with measurable emission in the line. Fig. S10B: SiO J = 5 ! moment1-map, Fig. S10C: SiO J = 6 ! moment1-map.36ig. S11: ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofS Pav. The position angle (PA) is indicated at the top of each panel. Contours (in white) areplotted at [0.1, 0.3, 0.5, 0.9] times the maximum value. Fig. S11A – Fig. S11B: PV diagramsof the SiO J = 5 ! medium resolution data. Fig. S11C – Fig. S11D: PV diagrams of theCO J = 2 ! medium resolution data. Fig. S11E – Fig. S11F: PV diagrams of the CO J = 2 ! low resolution data. 37 Fig. S12:
Medium resolution CO J = ! channel map of T Mic. Same as Fig. S8 butfor T Mic. 38ig. S13:
Low resolution CO J = ! channel map of T Mic. Same as Fig. S9, but forT Mic. 39ig. S14:
SiO stereogram and moment1-map of T Mic.
Same as Fig. S10, but for T Mic.40ig. S15:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofT Mic. Same as Fig. S11, but for T Mic. 41
Fig. S16:
Medium resolution CO J = ! channel map of U Del. Same as Fig. S8, butfor U Del. 42ig. S17:
Low resolution CO J = ! channel map of U Del. Same as Fig. S9, but forU Del. 43ig. S18:
SiO stereogram and moment1-map of U Del.
Same as Fig. S10, but for U Del.44ig. S19:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofU Del. Same as Fig. S11, but for U Del. 45
Fig. S20:
Medium resolution CO J = ! channel map of RW Sco. Same as Fig. S8, butfor RW Sco. 46ig. S21:
Low resolution CO J = ! channel map of RW Sco. Same as Fig. S9, but forRW Sco. 47ig. S22:
SiO stereogram and moment1-map of RW Sco.
Same as Fig. S10, but for RW Sco.Owing to the low signal-to-noise ratio of the SiO emission, these images should be interpretedwith care (see also Sect. S3.2). 48ig. S23:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofRW Sco. Same as Fig. S11, but for RW Sco. 49
Fig. S24:
Medium resolution CO J = ! channel map of V PsA. Same as Fig. S8, butfor V PsA. 50ig. S25:
Low resolution CO J = ! channel map of V PsA. Same as Fig. S9, but forV PsA. 51ig. S26:
SiO stereogram and moment1-map of V PsA.
Same as Fig. S10, but for V PsA.52ig. S27:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofV PsA. Same as Fig. S11, but for V PsA. 53
Fig. S28:
Medium resolution CO J = ! channel map of SV Aqr. Same as Fig. S8, butfor SV Aqr. 54ig. S29:
Low resolution CO J = ! channel map of SV Aqr. Same as Fig. S9, but forSV Aqr. 55ig. S30:
SiO stereogram and moment1-map of SV Aqr.
Same as Fig. S10, but for SV Aqr.Owing to the low signal-to-noise ratio of the SiO emission, these images should be interpretedwith care (see also Sect. S3.2). 56ig. S31:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofSV Aqr. Same as Fig. S11, but for SV Aqr. 57
Fig. S32:
Medium resolution CO J = ! channel map of R Hya. Same as Fig. S8, butfor R Hya. 58ig. S33:
Low resolution CO J = ! channel map of R Hya. Same as Fig. S9, but forR Hya. 59ig. S34:
SiO stereogram and moment1-map of R Hya.
Same as Fig. S10, but for R Hya.60ig. S35:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofR Hya. Same as Fig. S11, but for R Hya. 61ig. S36:
ALMA CO J = ! moment1-maps of R Hya. Fig. S37:
Medium resolution CO J = ! channel map of U Her. Same as Fig. S8, butfor U Her. 63ig. S38:
Low resolution CO J = ! channel map of U Her. Same as Fig. S9, but forU Her. 64ig. S39:
SiO stereogram and moment1-map of U Her.
Same as Fig. S10, but for U Her.65ig. S40:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofU Her. Same as Fig. S11, but for U Her. 66 ⇡ Gru
Fig. S41:
Medium resolution CO J = ! channel map of ⇡ Gru.
Same as Fig. S8, butfor ⇡ Gru. Only 1/4 of the channel map is shown, the velocity resolution is ⇠ .67ig. S42: Low resolution CO J = ! channel map of ⇡ Gru.
Same as Fig. S9, but for ⇡ Gru. Only 1/4 of the channel map is shown, the velocity resolution is ⇠ .68ig. S43: SiO stereogram and moment1-map of ⇡ Gru.
Same as Fig. S10, but for ⇡ Gru.69ig. S44:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram of ⇡ Gru.
Same as Fig. S11, but for ⇡ Gru. 70ig. S45:
ALMA CO J = ! moment1-maps of ⇡ Gru. Fig. S46:
Medium resolution CO J = ! channel map of R Aql. Same as Fig. S8, butfor R Aql. 72ig. S47:
Low resolution CO J = ! channel map of R Aql. Same as Fig. S9, but forR Aql. 73ig. S48:
SiO stereogram and moment1-map of R Aql.
Same as Fig. S10, but for R Aql.74ig. S49:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofR Aql. Same as Fig. S11, but for R Aql. 75
Fig. S50:
Medium resolution CO J = ! channel map of W Aql. Same as Fig. S8, butfor W Aql. 76ig. S51:
Low resolution CO J = ! channel map of W Aql. Same as Fig. S9, but forW Aql. 77ig. S52:
SiO stereogram and moment1-map of W Aql.
Same as Fig. S10, but for W Aql.78ig. S53:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofW Aql. Same as Fig. S11, but for W Aql. 79
Fig. S54:
Medium resolution CO J = ! channel map of GY Aql. Same as Fig. S8, butfor GY Aql. 80ig. S55:
Low resolution CO J = ! channel map of GY Aql. Same as Fig. S9, but forGY Aql. 81ig. S56:
SiO stereogram and moment1-map of GY Aql.
Same as Fig. S10, but for GY Aql.82ig. S57:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofGY Aql. Same as Fig. S11, but for GY Aql. 83 Fig. S58:
Medium resolution CO J = ! channel map of IRC Same as Fig. S8,but for IRC Low resolution CO J = ! channel map of IRC Same as Fig. S9, butfor IRC SiO stereogram and moment1-map of IRC Same as Fig. S10, but forIRC ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofIRC Same as Fig. S11, but for IRC Fig. S62:
Medium resolution CO J = ! channel map of IRC + Same as Fig. S8,but for IRC +10011. 88ig. S63:
Low resolution CO J = ! channel map of IRC + Same as Fig. S9, butfor IRC +10011. 89ig. S64:
SiO stereogram and moment1-map of IRC +10011.
Same as Fig. S10, but forIRC +10011. 90ig. S65:
ALMA CO J = ! and SiO J = ! position-velocity (PV) diagram ofIRC +10011. Same as Fig. S11, but for IRC +10011.91 upplementary Text
S8 Hydrodynamical simulations of binary systems
Hydrodynamical simulations of the influence of a companion on a wide orbit around a mass-losing AGB star ( ˙ M ⇠ ⇥ M yr , orbital separation between 2 – 50 au) have demon-strated that spiral patterns emerge (92) . Depending on the binary separation, wind velocity, andsecondary mass the global large scale wind geometry can be described as bipolar, elliptical, orquasi-spherical (92) . The spiral structure in the stellar wind is caused by the orbital motion ofthe mass-losing AGB star around the common center-of-mass, or by the accretion wake of thecompanion owing to BHL flow (19) . The latter case gives rise to a spiral which is much morefocussed toward the orbital plane. For binary systems with an eccentric orbit, a bifurcationof the spiral pattern might occur, including an asymmetry in the interarm density cavity and aspiral/ring appearance (162, 163) .Simulations for companions with a mass of 0.1–0.5 M and separation of 3–10 au orbitingaround an AGB star with mass-loss rate around ⇥ M yr show the ejected AGB masscan form a circumbinary disk, or contribute to an accretion disk around the secondary (20) in addition to a spiral structure caused by BHL accretion. As exemplified by L Puppis (17) ,a low mass secondary orbiting an AGB star can shape the AGB wind into a bipolar structure.When the effects of the AGB pulsations are included in simulations, the bipolar outflow displaysripple-like structures with lobes extending to 10 – 20 au (20) .Focussing on the effect of the mass ratio between secondary and primary AGB star, it wasshown that the mass-accretion efficiency by the secondary increases with the mass ratio ofthe system (164) . Those simulations were for a system with a primary star of 3 M , orbitalseparation of 3 au, mass-loss rate of M yr , and a wind velocity of 25 km s . Thehigher the mass ratio, the more complex the structure of the outflow. The wind morphology wasdescribed as being ‘spider-like’ for q = 0 . and as ‘rose-like’ for q = 1 . (164) .Simulations have been performed for a variety of binary mass ratios, initial wind velocities,orbital separations, and rotation rates of the donor star, and where the primary star loses massat a rate of . ⇥ M yr and has a mass of 1.2 M (135) . That study showed that thestrength of interaction depends on the binary mass ratio and the ratio of the orbital velocity towind velocity. Smaller values for the orbital separation and larger companion masses inducestronger interaction. The authors also find that the corotation of the primary star modifies theoutflow morphology if the initial wind velocity is below 5 km s .Comparing these simulations to our ALMA observations, the morphologies can be ex-plained by gravitational interaction with the companion causing the initially isotropic windto be heavily disturbed. Other mechanisms, such as the stellar pulsations, magnetic fields, androtation might contribute in an initial wind anisotropy thus adding complexities to the binary-induced morphological phenomena. 92 In addition to our ATOMIUM survey, there are previously-published ALMA observations of12 AGB stellar winds. Some have been described as being perturbed by a spiral-like signature.For carbon-rich AGB stars, examples include R Scl ( ˙ M ⇠ ⇥ M yr ) (108, 109) , CIT 6( ˙ M ⇠ ⇥ M yr ) (105) , and IRC +10216 ( ˙ M ⇠ . ⇥ M yr ) (106, 107) , whereall the three targets have terminal wind velocities of approximately 14–18 km s . Two oxygen-rich high mass-loss rate targets, OH 26.5+06 and OH 30.1 . ⇥ M yr and terminal wind speed of 15–18 km s (55) . The spiral patternsin these stars have been interpreted as being caused either by the orbital motion of the AGBstar around the center-of-mass of a binary system, if the companion is relatively massive; orbecause the companion’s gravity focusses a fraction of the wind material toward the equatorialplane, producing a (more flattened) spiral structure associated with the companion owing to theBHL flow it induces. Position-velocity diagrams can differentiate between both causes (54) .Bifurcated spiral-like structures in the circumstellar envelope of CIT 6, have been shown to beproduced by companions on a wide, eccentric orbit (105, 162) .The oxygen-rich AGB star L Puppis ( ˙ M ⇠ ⇥ M yr ) is surrounded by a circum-stellar disk in which the rotating gas has a Keplerian velocity profile within the central cavityof the dust disk (r < < r <
20 au) beyond theinner dust rim (17) . The continuum map revealed a secondary source with an estimated massof ⇠
10 Jupiter masses. Two more oxygen-rich examples are R Dor ( ˙ M ⇠ ⇥ M yr )and EP Aqr ( ˙ M ⇠ . ⇥ ), although the velocity field in the EDE/disk-like environmentsis more complex than Keplerian (79, 101) . In the case of EP Aqr, the CO emission exhibitsthe characteristic features of a nearly face-on spiral with a biconical outflow. The narrow widthof the spiral signature in velocity space has been interpreted as an indication that the spiral iscaused by a hydrodynamical perturbation in a face-on differentially rotating disk. For thesethree cases (L Puppis, R Dor, and EP Aqr), the widths of the rotational SiO lines in the groundvibrational state are considerably larger than the widths of the ground-state CO lines obtainedfrom the same observational setup (as is also the case for five of our sources, see Table S2). TheSiO lines are radiatively excited, and hence are more diagnostic of the inner wind dynamics.The Mira AB system is another well-known binary containing an oxygen-rich AGB star (99, 100) . When imaged at a spatial resolution of ⇠ , the map of CO J = 3 ! showsa lot of complexity. Spiral arcs oriented in the orbital plane are found around Mira A ( ˙ M ⇠ ⇥ M yr ). The accretion wake behind the companion (Mira B), which resides at anorbital separation of 0 . ⇠ ) opposing arcs were alsofound. Some morphological resemblance between the CO J = 3 ! emission of the Mira ABsystem and the CO channel map of S Pav can be discerned [compare Fig. S8 with figure A1of (99) ].R Aqr is a symbiotic stellar system, consisting of interacting AGB star and a white dwarf (103, 104) . In these systems, the strong binary interaction results in high mass transfer betweenboth stars, equatorial flows, and ejection of fast bipolar jets. The two-arcminute-wide nebula of93 Aqr is composed of an equatorial structure, and a precessing jet powered by the accretion onthe white dwarf (103) . The orbital period is ⇠
44 yr with binary separation of ⇠
45 mas (or 10 auat 218 pc). CO data show that the CO structure traces mass ejection focused in the orbitalplane, with two plumes in opposite directions probably corresponding to (the start of) a doublespiral (103) . The mass-loss rate is ⇥ M yr (104) .Three of the fourteen ATOMIUM AGB sources are known to be part of a binary system:the two S-type AGB stars W Aql and ⇡ Gru (whose C/O ratio is slighter lower than 1), andR Hya. W Aql ( ˙ M ⇠ ⇥ M yr ) has a known companion at a separation of 0.46 (or ⇠
150 au), which is classified as an F8 to G0 main sequence star (165) . The CO J = 3 ! linein W Aql in the inner 10 of the circumstellar envelope is asymmetric with arc-like structures atseparations of 2–3 (50) . Farther out, weaker spiral structures are present at greater separations.The larger separations can be explained by the interaction with the known companion in anorbit with low eccentricity, but not the smaller separation pattern. Potential physical causes thathave been suggested include a second, closer companion residing at an eccentric orbit; a recentchange in the wind velocity; or variations in the mass-loss-rate (50) . ⇡ Gru has a companion ofspectral type G0V at a separation of 2 . ⇠
400 au) (166) . Observations of the CO J = 3 ! and CO J = 3 ! at a spatial resolution of ⇠ confirmed the envelope structure includes aradially expanding equatorial torus (with a velocity of 8–13 km s ), and a fast bipolar outflowwith a linear velocity increase of up to 100 km s (49) . Because wide binary systems arenot expected to produce this kind of envelope structure, a second, closer companion (at 10–30au) has been suggested (167) . R Hya is thought to be a wide binary system with an angularseparation of 21 and a very long orbital period (168) . S10 ATOMIUM observation properties
The ATOMIUM observation properties are listed in Table S5.94able S5:
Observing properties of the ATOMIUM project.
Data for GY Aql only are shown as an example; full data forall targets are given in Data S1. Given are the Science Goal scheduling block (SG; in which a, b, c and d are observations inthe medium resolution configuration, and e and f in the low resolution configuration), the observing date, the ALMA archivalname (ASDM), the target name, the right ascension (R.A.) and declination (Dec.) of that specific observation, the time onsource (ToS), the precipitable water vapour at the date of observations (PWV), the phase-reference source (Ph. Ref.), and thebandpass and flux density scale calibrator (Bandpass).
SG Date Obs. ASDM Target R.A. Dec. ToS PWV Ph. Ref. ToS Bandpass ToS(h:min:sec) (deg:min:sec) (sec) (mm) (sec) (sec)GY Aql a 06 TM2 2018-11-13 22:46:33 uid A002 Xd51939 X6300 GY Aql 19:50:06.31 –07:36:52.3 302 0.79 J1951-0509 60 J2000-1748 302GY Aql b 06 TM2 2018-11-11 22:31:26 uid A002 Xd50463 X9b77 GY Aql 19:50:06.31 –07:36:52.3 302 2.68 J1951-0509 60 J1924-2914 302GY Aql b 06 TM2 2018-11-12 23:36:25 uid A002 Xd51939 X2e4 GY Aql 19:50:06.31 –07:36:52.3 302 1.71 J1951-0509 60 J2000-1748 302GY Aql b 06 TM2 2018-11-13 22:28:25 uid A002 Xd51939 X6277 GY Aql 19:50:06.31 –07:36:52.3 302 0.96 J1951-0509 60 J2000-1748 302GY Aql c 06 TM2 2018-11-14 23:37:44 uid A002 Xd52fc8 X876 GY Aql 19:50:06.31 –07:36:52.3 302 0.75 J1951-0509 60 J2000-1748 302GY Aql d 06 TM2 2018-11-13 23:04:44 uid A002 Xd51939 X638d GY Aql 19:50:06.31 –07:36:52.3 90 0.64 J1951-0509 60 J2000-1748 302GY Aql e 06 TM1 2019-01-06 16:19:01 uid A002 Xd7aa27 X7ceb GY Aql 19:50:06.31 –07:36:52.3 302 1.84 J1951-0509 60 J1751+0939 302GY Aql e 06 TM1 2019-03-03 11:43:28 uid A002 Xd90607 X3948 GY Aql 19:50:06.31 –07:36:52.3 302 1.65 J1951-0509 60 J2000-1748 302GY Aql f 06 TM1 2019-01-13 14:34:05 uid A002 Xd80784 X62d5 GY Aql 19:50:06.31 –07:36:52.3 302 1.12 J1951-0509 60 J1751+0939 302
11 ATOMIUM image cube properties
The properties of each cube and continuum image are listed in Table S6 and Table S7, respec-tively.Table S6:
Image cube properties.
Data for GY Aql only are shown as an example; full data forall targets are given in Data S2. Given are the source name, the configuration (low or mediumspatial resolution), the cube number (see Fig. S1), the start and end frequency of each cube (inGHz), the size of the major and minor axis of the synthesized beam (bmaj and bmin, in arcsec),the angle of the synthesized beam (bpa, in degrees), the size of the image (imsize, in arcsec)and the noise rms (in mJy) of the cube as measured in an emission-free channel of the cube.target config cubeNo start end bmaj bmin bpa imsize rms (GHz) (GHz) (arcsec) (arcsec) (deg) (arcsec) (mJy)GY Aql low 00 213.838 215.711 1.343 1.045 64 24.0 2.8GY Aql low 01 216.038 217.910 1.360 1.047 66 24.0 3.3GY Aql low 04 227.234 229.107 1.286 1.013 69 24.0 2.9GY Aql low 05 229.589 231.462 1.265 0.986 67 24.0 3.1GY Aql low 08 244.051 244.987 1.294 0.929 65 24.0 3.0GY Aql low 09 245.350 247.223 1.278 0.925 66 24.0 2.6GY Aql low 12 258.623 260.496 1.223 0.885 66 24.0 2.9GY Aql low 13 262.104 263.039 1.207 0.876 66 24.0 3.5GY Aql medium 00 213.838 215.711 0.382 0.319
74 24.0 2.2GY Aql medium 01 216.037 217.911 0.375 0.318
76 24.0 2.3GY Aql medium 02 220.237 222.110 0.364 0.295
73 24.0 2.2GY Aql medium 03 223.631 225.504 0.358 0.290
76 24.0 1.9GY Aql medium 04 227.235 229.108 0.357 0.304
79 24.0 2.4GY Aql medium 05 229.589 231.462 0.358 0.298
78 24.0 2.4GY Aql medium 06 235.438 237.311 0.340 0.272
76 24.0 2.3GY Aql medium 07 239.157 240.092 0.340 0.275
74 24.0 2.4GY Aql medium 08 244.051 244.987 0.397 0.278
75 24.0 3.0GY Aql medium 09 245.350 247.223 0.395 0.273
76 24.0 2.8GY Aql medium 10 251.583 253.456 0.351 0.273
71 24.0 4.9GY Aql medium 11 253.947 255.820 0.349 0.271
71 24.0 4.7GY Aql medium 12 258.624 260.497 0.377 0.260
75 24.0 3.0GY Aql medium 13 262.104 263.039 0.371 0.262
75 24.0 3.4GY Aql medium 14 265.533 267.406 0.334 0.257
70 24.0 6.1GY Aql medium 15 267.783 269.657 0.333 0.257
73 24.0 5.896able S7:
Continuum image properties.
Given are the source name, the configuration (lowor medium resolution), the size of the major and minor axis of the synthesized beam (bmaj andbmin, in arcsec), the angle of the synthesized beam (bpa, in degrees), the maximum recoverablescale (MRS, in arcsec), the size of the image (imsize, in arcsec), the noise contrms (in mJy), andin the last column the total line-free bandwidth (BW, in GHz), spread over the 56 GHz span ofobservations.target config bmaj bmin bpa MRS imsize contrms BW(arcsec) (arcsec) (deg) (arcsec) (arcsec) (mJy) (GHz)GY Aql low 1.220 0.897 64 9.5 24.0 0.040 8.69GY Aql medium 0.324 0.247
70 4.0 24.0 0.026 18.03IRC +10011 low 0.722 0.686
59 7.4 24.0 0.051 8.53IRC +10011 medium 0.112 0.100 38 1.6 6.0 0.033 18.49IRC
63 2.0 4.0 0.027 15.33 ⇡ Gru low 0.866 0.774
86 9.3 24.0 0.036 10.36 ⇡ Gru medium 0.248 0.235 30 3.9 8.0 0.034 20.49RW Sco low 0.928 0.701 86 9.0 24.0 0.034 10.30RW Sco medium 0.147 0.120
86 1.9 4.0 0.040 6.75R Aql low 0.764 0.648 83 7.7 24.0 0.042 10.25R Aql medium 0.306 0.238
54 3.8 8.0 0.030 20.31R Hya low 0.830 0.600 79 8.7 24.0 0.051 10.09R Hya medium 0.256 0.223 70 3.5 8.0 0.028 19.02SV Aqr low 0.886 0.747 74 9.8 24.0 0.038 10.92SV Aqr medium 0.124 0.104
75 1.6 8.0 0.023 27.55S Pav low 1.026 0.983
56 8.7 24.0 0.051 10.22S Pav medium 0.304 0.234 56 3.3 8.0 0.022 20.20T Mic low 1.047 0.730
79 9.3 24.0 0.059 10.99T Mic medium 0.268 0.225
89 4.0 8.0 0.025 19.36U Del low 1.167 1.043 42 9.6 24.0 0.036 6.47U Del medium 0.316 0.235
33 3.3 8.0 0.028 14.84U Her low 0.997 0.843 26 9.7 24.0 0.054 9.75U Her medium 0.352 0.249
33 2.2 8.0 0.048 16.83V PsA low 0.995 0.753 87 9.0 24.0 0.030 11.28V PsA medium 0.283 0.229 85 4.0 8.0 0.020 20.99W Aql low 0.920 0.667 76 8.9 24.0 0.056 6.61W Aql medium 0.351 0.223