Katholieke Universiteit Leuven · werden gebracht, al het over-en-weer gesleur met meubelen, elke...

Katholieke Universiteit LeuvenFaculteit Wetenschappen

Spitzer survey of dust grain processing in stablediscs around binary post-AGB stars

Clio Gielen

Promotor:Prof. Dr. H. Van Winckel

Proefschrift voorgelegd tothet behalen van het doctoraatin de Wetenschappen

LEUVEN 2009

PROMOTOR: COMMISSIELEDEN:Prof. Dr. H. Van Winckel Prof. Dr. C. Waelkens

Prof. Dr. L.B.F.M WatersProf. Dr. V. AfanasievDr. F. KemperDr. J. BlommaertDr. B. Acke

Acknowledgements:This PhD thesis was possible thanks to financial support from the Funds for Scientific Re-search of Flanders (FWO) under grant G.0178.02 and G.0470.07, and from the ResearchCouncil of the Catholic University of Leuven under FLOF grant 10379. This work is basedon observations made with the Spitzer Space Telescope, operated by the Jet Propulsion Lab-oratory, California Insitute of Technology under a contract with NASA.

Cover: Image of the sky with an overlay of the Hevilius constellations. Depicted the constellationsHercules and Lyra, home to two important sample stars studied in this work, namely AC Her and EP Lyr.Image made using Google SkyTM and the Hevelius overlay (credit: the U.S. Naval Observatory andthe Space Telescope Science Institute).

Copyright:c©2009 Faculteit Wetenschappen, Geel Huis, Kasteelpark Arenberg 11, 3001 Heverlee (Leuven)

Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenigvuldigd en/of openbaar gemaaktworden door middel van druk, fotokopie, microfilm, elektronisch of op welke andere wijze ook zondervoorafgaandelijke schriftelijke toestemming van de uitgever.All rights reserved. No part of the publication may be reproduced in any form by print, photoprint,microfilm, electronic or any other means without written permission from the publisher.ISBN 978-90-8649-263-3D/2009/10.705/41

Dankwoord

‘Sterrenkunde’, de fascinatie begon al vele jaren geleden, en het was dan ook geen wonderdat ik reeds tien jaar geleden wist: ‘dat ga ik nu eens studeren’. En zo begon ik aan eenopleiding wiskunde (wat ik gelukkig ook graag deed), om te eindigen in sterrenkunde. Datik uiteindelijk zelfs na men studie in de sterrenkunde kon blijven plakken, was de kers op detaart.

Het feit dat ik nu na 3,5 jaar dit thesiswerk kan voorleggen, is natuurlijk niet alleen maar mijnverdienste. Zoveel mensen hebben op verschillende manieren bijgedragen aan dit werk.

Als eerste verdient natuurlijk mijn promotor lofbetuigingen.Hans, jij bent diegene die dit project voor mij op poten heeft gezet, en me jaar na jaar dichterbij dit uiteindelijke proefschrift heeft gebracht. Jij was er altijd om me te helpen bij elkprobleem, groot of klein, kwam telkens met goede ideeen af die het werk nog interessanterzouden maken, en hield er mee de moed in toen er weer eens een proposal werd afgekeurd.

Maarten, het is dankzij jou dat ik heb ontdekt hoe leuk en interessant het was om een thesiste maken in sterrenkunde. Jouw begeleiding tijdens men master-thesisjaar, en ook nog weldaarna, hebben ervoor gezorgd dat ik met volle vertrouwen aan dit doctoraat kon beginnen.

I would also like to thank all the co-authors of my work, Rens, Michiel, Carsten, and manyothers, for all their input and suggestions. Without them I would never have gotten all thosechapters of this work published. Ik bedank ook graag de leden van de jury, waaronder Ciska,Valeri en Joris, voor al hun relevante vragen en opmerkingen.

The last years would not have been half as much fun without my colleagues and friends atthe IvS: Maarten, Bram, Tijl, Elvire, Pieter, Kristof, Els, Ben, ... All the social activities,conferences and holidays were memorable. And I’m sure all the coffee breaks did wondersfor my trivia knowledge. Ook de collegas die ondertussen al andere oorden hebben opge-zocht, het verre Amerika, het iets minder verre Brussel, de kerktoren van Duffel of gewoonden B, mogen hier natuurlijk niet ontbreken. And all the people I’ve met during conferences,meetings and observing runs.

Het IvS zou natuurlijk nooit geweest zijn wat het nu was zonder jou, Christoffel.Dankzij jou inspiratie, motivatie en harde werk is het instituut de geweldige onderzoeksgroepdie het nu is.

My office mates: Rachel and Djazia, for making sure the office had enough female influence.De Jonasses, for not complaining too much about all this girl stuff, and Jeroen, om de bloe-mekes terug tot leven te wekken en toe te geven dat het hier leuker is dan bij de burgies. I’msure that everyone on the 3th and 4th floor is jealous of our office!

Ook wil ik al mijn klasgenoten door de jaren heen bedanken, zowel van W1, W2 als W3. Mijnangst voor de wereldvreemde wiskundige vervloog meteeen toen ik jullie leerde kennen.Of het nu was door oxo te spelen tijdens lineaire, met de assistent te lachten tijdens ..., weltijdens vele lessen eigenlijk, of door me te helpen met de vele vragen die ik had over de cursus,

jullie zorgden ervoor dat ik altijd met plezier opstond om naar de lessen en oefenzittingengaan. En Annelies, om ervoor te zorgen dat ik ook met plezier eens een les ’s morgens mistena een avondje uitgaan.

Natuurlijk verdienen ook mijn ouders en familie een bedankje.Een dochter die in het verre Leuven op kot zat, de bergen was die in het weekend thuiswerden gebracht, al het over-en-weer gesleur met meubelen, elke keer ik weer eens van kotveranderde, er viel nooit een klacht of zucht te bespeuren. Bedankt voor de steun die jullieme altijd hebben gegeven, en om niet al te raar op te kijken toen ik lang geleden verkondigdedat ik wiskunde en sterrenkunde ging studeren.

Tijl, je staat hierboven al even vermeld als collega en vriend, maar natuurlijk ben je voormij zoveel meer. Nu deze thesis afgerond is, begin ik vol goede moed aan dat nieuwe groteproject: leren fietsen met klikpedalen. Ik ben er zeker van dat jij hiervoor de beste promotorzal zijn...

Clio,Heverlee, 13 Mei 2009

Contents

1 Introduction 1

1.1 AGB and post-AGB evolution . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 The asymptotic giant branch . . . . . . . . . . . . . . . . . . . . . . 1

1.1.2 The post-AGB phase . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Galactic Post-AGB stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Proto-planetary nebulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.4 Binary post-AGB stars and their circumbinary disc . . . . . . . . . . . . . . 5

1.5 Dust mineralogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.5.1 Interstellar and circumstellar dust . . . . . . . . . . . . . . . . . . . 9

1.5.2 The detection of dust . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.5.3 Dust characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.6 Astronomical dust species . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.6.1 Silicate dust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.6.1.1 Formation . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.6.1.2 Identification . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.6.2 Polycyclic aromatic hydrocarbons . . . . . . . . . . . . . . . . . . . 14

1.7 Discs in young stellar objects . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.7.1 Star formation and protoplanetary discs . . . . . . . . . . . . . . . . 16

1.7.2 Disc geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.7.3 Disc chemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.8 The self consistent disc model . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.9 The Spitzer sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.9.1 The Spitzer mission . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.9.2 The sample stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.9.3 Spitzer-IRS data reduction . . . . . . . . . . . . . . . . . . . . . . . 25

1.9.3.1 High-resolution . . . . . . . . . . . . . . . . . . . . . . . 25

1.9.3.2 Low-resolution . . . . . . . . . . . . . . . . . . . . . . . . 26

1.9.4 TIMMI2 data reduction . . . . . . . . . . . . . . . . . . . . . . . . . 27

1.10 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2 The RV Tauri spectral twins RU Cen and AC Her. 31

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.2 Observations and reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.2.1 Spitzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.2.2 TIMMI2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.2.3 CORALIE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.3 SED determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.4 Binary orbit of RU Cen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.5 Analysis infrared spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.5.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.5.2 Feature identification . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.5.3 Profile fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.6 SED fitting disc model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.6.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.6.1.1 2D disc model . . . . . . . . . . . . . . . . . . . . . . . . 44

2.6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

2.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3 Mineralogy of dust around evolved stars 51

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.2 Programme stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.3 Observations and data reduction . . . . . . . . . . . . . . . . . . . . . . . . 56

3.3.1 Spitzer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.3.2 TIMMI2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.4 Spectral energy distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.5 General overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

3.5.1 Identification of silicate dust features . . . . . . . . . . . . . . . . . 61

3.5.2 Mean spectra and complexes . . . . . . . . . . . . . . . . . . . . . . 61

3.5.2.1 The 10 µm complex (8− 13 µm) . . . . . . . . . . . . . . 62

3.5.2.2 The 14 µm complex (13− 15 µm) . . . . . . . . . . . . . 64

3.5.2.3 The 16 µm complex (15− 17 µm) . . . . . . . . . . . . . 64

3.5.2.4 The 19 µm complex (17− 21 µm) . . . . . . . . . . . . . 65

3.5.2.5 The 23 µm complex (20− 27 µm) . . . . . . . . . . . . . 66

3.5.2.6 The 33 µm complex (31− 37 µm) . . . . . . . . . . . . . 66

3.6 Analysis: profile fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.7 Full spectral fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3.7.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3.7.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.7.3 Correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

3.7.3.1 Mineralogy correlations . . . . . . . . . . . . . . . . . . . 79

3.7.3.2 Central star correlations . . . . . . . . . . . . . . . . . . . 79

3.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

3.8.1 Comparison with young stellar objects . . . . . . . . . . . . . . . . . 82

3.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

4 The peculiar post-AGB stars EP Lyr and HD 52961 85

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.2 Programme stars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.2.1 EP Lyr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.2.2 HD 52961 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.3 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4.4 Spectral analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

4.4.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

4.4.2 Silicate dust emission . . . . . . . . . . . . . . . . . . . . . . . . . . 92

4.4.3 CO2 emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.4.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.4.3.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

4.4.4 PAH features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

4.5 Spectral energy distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

4.5.1 2D disc modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

4.5.2 Comparison with interferometric data . . . . . . . . . . . . . . . . . 102

4.5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4.5.3.1 HD 52961 . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4.5.3.2 EP Lyr . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

5 Circumbinary discs in the LMC 111

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

5.2 Observations and data reduction . . . . . . . . . . . . . . . . . . . . . . . . 113

5.2.1 Optical high-resolution programme . . . . . . . . . . . . . . . . . . 113

5.2.2 IRS-spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

5.3 Photospheric abundance results . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.3.1 MACHO 79.5501.13 . . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.3.1.1 Photospheric model . . . . . . . . . . . . . . . . . . . . . 118

5.3.1.2 Chemical analysis . . . . . . . . . . . . . . . . . . . . . . 119

5.3.2 MACHO 81.9728.14 . . . . . . . . . . . . . . . . . . . . . . . . . . 119



5.3.3 MACHO 81.8520.15 . . . . . . . . . . . . . . . . . . . . . . . . . . 120



5.4 Infrared spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

5.4.1 MACHO 81.9728.14 . . . . . . . . . . . . . . . . . . . . . . . . . . 126

5.4.2 MACHO 79.5501.13 and MACHO 82.8405.15 . . . . . . . . . . . . 127

5.4.3 Comparison with infrared spectra of the Galactic sample . . . . . . . 128

5.5 Full spectral fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

5.6 2D disc modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

5.6.1 MACHO 79.5501.13 . . . . . . . . . . . . . . . . . . . . . . . . . . 132

5.6.2 MACHO 82.8405.15 . . . . . . . . . . . . . . . . . . . . . . . . . . 134

5.6.3 MACHO 81.9728.14 . . . . . . . . . . . . . . . . . . . . . . . . . . 134

5.6.4 MACHO 81.8520.15 . . . . . . . . . . . . . . . . . . . . . . . . . . 135

5.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

5.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

5.9 Appendix: Expanding the SAGE-Spec sample . . . . . . . . . . . . . . . . . 138

5.9.1 SAGE 054310 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

5.9.1.1 Dust approximation models . . . . . . . . . . . . . . . . . 140

5.9.1.2 Comparison with Galactic sources . . . . . . . . . . . . . 142

5.9.1.3 Full spectral fitting . . . . . . . . . . . . . . . . . . . . . . 143

5.9.1.4 2D disc modelling . . . . . . . . . . . . . . . . . . . . . . 144

5.9.1.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 145

6 Conclusions 149

7 Future work: Evolution of circumbinary discs around post-AGB binaries 153

7.1 Theoretical description of disc processes . . . . . . . . . . . . . . . . . . . . 154

7.2 Interferometry and far-IR spectroscopy of the disc . . . . . . . . . . . . . . . 154

7.3 Information on the dynamics and chemistry of the central star . . . . . . . . . 155

7.4 Global study of disc sources in the LMC . . . . . . . . . . . . . . . . . . . . 156

A Nederlandse samenvatting 159

A.1 Sterevolutie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

A.2 Schijven rond post-AGB dubbelsterren . . . . . . . . . . . . . . . . . . . . . 161

A.2.1 Schijfstructuur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

A.3 Kosmisch stof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

A.3.1 De detectie van stof . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

A.3.2 Astronomische stofsoorten . . . . . . . . . . . . . . . . . . . . . . . 165

A.4 Deze thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

B List of acronyms 169

C List of refereed publications 171

Bibliography 173

Chapter 1Introduction

In this work we present the results of a spectral survey of evolved stars, obtained with theIRS spectrograph aboard the Spitzer satellite. This satellite was the last mission of the “GreatObservatories Program” of NASA and is an integral part of the “Astronomical Search forOrigins Program”. We selected 21 Galactic and 4 extragalactic sources from a larger sampleof evolved stars, believed to be part of a binary system and surrounded by a stable dustydisc. By combining high-spectral resolution observations with detailed models describingdust properties and disc structure, we want to investigate if the infrared spectra confirm thesuspected disc hypothesis, and if they allow us to constrain the exact disc composition andstructure. With this study we hope to trace the evolution of these evolved binary stars andtheir disc, and determine the impact of the disc on the entire system. In the following chapterswe introduce this particular class of evolved stars.

1.1 AGB and post-AGB evolution

1.1.1 The asymptotic giant branch

Once the hydrogen and helium in the core of a low- and intermediate-mass stars (0.8 M¯ <M < 8 M¯) is exhausted, the core will contract until He-shell burning starts around thedegenerate CO core. During this phase the envelope expands and the luminosity of the starincreases. The star now ascends the Asymptotic Giant Branch (AGB). The star spends 90%of its time on the AGB on the so called early-AGB (E-AGB). Here helium is burned in ashell around the degenerate CO core. Once the He is exhausted, H will be ignited in a highershell. This burning H-shell produces He, replenishing the underlying He-shell. At a certaincritical mass, He can then be re-ignited, causing a shell instability: a thermal pulse. The starhas reached the thermally-pulsating AGB (TP-AGB). This process will repeat itself and thenumber of times is largely dependent on the initial mass, metallicity, and mass-loss of thestar.

2 Chapter 1. Introduction

Figure 1.1 — Hertzsprung-Russell diagram showing the theoretically predicted evolutionary track of asolar-mass star.

During a short relaxation phase after the thermal pulse, the convective envelope may reachdown to layers where nuclear reactions are occurring, dredging up elements from He-shellburning or s-process (slow neutron-capture process) elements to the stellar surface. Thisprocess is called ’the third dredge-up’. This enrichment can increase the carbon content inthe photosphere until C/O > 1. The star has now become a C-rich star instead of an O-rich Mstar. For a more detailed review on AGB evolution, we refer to Herwig (2005).

On the AGB, the star experiences strong mass loss, caused by a strong stellar wind. Herethe star reaches its largest radius (R ≥ 250 R¯) and lowest effective temperature (Teff ≤2500 K). This makes for a low gravity at the stellar surface, making it easier for the envelopepulsations to expel material from the star and allowing dust grains to condense in the stellarwind. AGB stars have strong stellar winds with mass-loss rates of M = 10−6−10−4 M¯yr−1

and characteristic wind speeds of 10− 30 km s−1. The mass loss probably peaks with everythermal pulse, because of the higher luminosity reached at this point, and increases along theAGB, with the strongest mass losses occurring at the end of the TP-AGB. When the stellarenvelope is reduced to a mass of about 10−2 M¯, the AGB-phase comes to an end. Figure 1.1shows a schematic evolutionary track of a solar-mass star.

1.1.2 The post-AGB phase

When the mass of the H-envelope is reduced, the pulsations decrease in amplitude and themass loss decreases. The star has reached the post-AGB phase: the effective temperature ofthe star will increase strongly, while the luminosity stays roughly constant (103 − 104 L¯).When the effective temperature reaches about 30 000 K, the star can ionise the circumstellarmaterial. The star is now visible as a Planetary Nebula (PNe). Not all AGB-stars will end upas PNe, however. If the circumstellar material is already too far away from the central star or

1.2. Galactic Post-AGB stars 3

Figure 1.2 — Models for post-AGB evolution for stars with initial masses of 7, 3 and 1 M¯ (top,middle, bottom). The marks are in units of 1000 years (Blocker 1995).

too dissipated, it will never become ionised.

If the post-AGB star crosses the Population II Cepheid instability strip (Wallerstein 2002), thestar can show strong radial pulsations. This particular class of evolved pulsators are calledRV Tauri stars (e.g. Jura 1986; Jura & Kahane 1999; De Ruyter et al. 2005).

Typical post-AGB timescales are very short, of the order of 1000−10 000 years (see Fig. 1.2),and strongly depend on the core mass and mass-loss history along the AGB (Blocker 1995).The former CO core eventually ends its life as a white dwarf, slowly cooling at constantradius.

1.2 Galactic Post-AGB stars

The observed post-AGB stars are evolved stars, with spectral types ranging from B to K, at thehigh-luminosity end of the HR- diagram. Their photospheres bear witness to the subsequentdredge-up processes during previous evolutionary phases. Depending on the initial mass andmetallicity of the star, strong 3th dredge-up signatures can be seen, such as helium-burningproduct (mainly 12C), and s-process enhancement, produced by the slow neutron-capture pro-cess. One of the most prominent observational characteristics is the strong observed infraredexcess. This excess comes from dusty circumstellar material, expelled during the strong AGBmass-loss phase. For a more detailed description of AGB and post-AGB evolution we referto Herwig (2005); Van Winckel (2003).

It is this strong infrared excess which made it possible to perform a systematic search fornew post-AGB sources. Most Galactic post-AGB candidates were discovered with the 1983IRAS mission, more specifically by looking for sources which occupied a specific region inthe ([12micron]-[25micron],[25micron]-[60micron]) colour-colour diagram. Ground-basedfollow-up programmes were then used to constrain between late-type AGB stars, super-giants or early-type PNe (e.g. Kwok et al. 1987; Hrivnak et al. 1989; Oudmaijer et al. 1992;Garcia-Lario et al. 1997; Suarez et al. 2006). In the recent catalogue of Szczerba et al. (2007),


about 300 possible post-AGB candidates are presented. Since the evolution in the post-AGBphase is very fast, few Galactic sources are indeed expected.

If we look at literature, most of the best-studied post-AGB stars are on individual, ratherspectacular, objects, such as AFGL 2688, AFGL 618, MZ-3 and OH 231.8+4.2. These objectsare probably all binaries, for which the strong binary processes are fundamental in producingthe observed characteristics.

1.3 Proto-planetary nebulae

One of the major questions in current research is whether the predominately symmetrical out-flows during the AGB phase can produce the observed wide variety in proto-PNe (PPNe) andPNe shapes and structures, including spherical, point symmetric, axisymmetric and bipolarnebulae. During this short transition time the star and its circumstellar envelope must be sub-ject to fundamental and rapid changes in structure, mass loss and geometry (Balick & Frank2002; Sahai et al. 2007).

One of the theories trying to explain the observed asymmetry, is known as the InteractiveStellar Wind (ISW) model (e.g. Kwok et al. 1978; Soker & Livio 1989; Garcıa-Segura et al.1999). Here the asymmetry is caused by interaction of the fast stellar wind, launched by thehot central star, and a slow, older AGB envelope, remnant of the mass loss on the AGB. Inrecent years it became clear, however, that major shaping processes are active, already at avery early stage after the AGB evolution. Examples of complex, multipolar structures can beseen in PPNe, e.g. Sahai et al. (2007, 2009), which have central stars too cold to drive a fastwind.

Additional processes, such as strongly-collimated outflows during the post-AGB phase, arenecessary to explain all the observed complex PPNe and PNe shapes (Sahai & Trauger 1998).The linear momenta deduced from CO gas rotational lines, are found to be much larger thanwhat could be driven by radiation pressure from single stars (Bujarrabal et al. 2001). Clearly,other models than the ISW model are necessary to explain the PNe shaping. Such mod-els, which include effects such as stellar rotation, magnetic fields and aspherical winds, arewidespread (Garcıa-Segura et al. 2005; Nordhaus et al. 2007), but probably still cannot ac-count for the large fraction in observed asymmetrical PPNe and PNe.

There is now growing evidence that the processes in the strong bipolar PPNe and PNe oc-cur because of strong interaction in a binary system (De Marco et al. 2008, and referencestherein). Many of these PNe have structures, such as bipolar lobes, jets, rings, dusty discsand tori, which are often explained by common envelope binary interactions, which is a badlyunderstood process by itself.

Direct evidence for the suspected high binary rate in PNe however is still lacking, but this isno longer the case for post-AGB stars (Van Winckel 2003, 2007). Optically-bright post-AGBstars are invaluable to study the transition region between the AGB and the PNe phase.

1.4. Binary post-AGB stars and their circumbinary disc 5

Figure 1.3 — Spectral energy distributions of a typical outflow source (HD 187885, left) and a discsource (TW Cam, right). The outflow source is characterised by a double-peaked SED, dominated bycool dust, which resides at larger distances from the central star. The disc source shows a very broadinfrared excess, showing the presence of both hot and cool dust.

1.4 Binary post-AGB stars and their circumbinary disc

With the discovery of the first binary post-AGB stars (e.g. Waelkens et al. 1991a; Waters et al.1993; Van Winckel et al. 1995), it became clear that these evolved stars have common obser-vational characteristics, such as a broad IR excess, coming from a hot dust component. Thepresence of hot circumstellar dust is either evidence of a very recent dusty mass loss, or of astable circumstellar environment close to the star (Sect. 1.4). Pilot studies on some interest-ing objects, such as HD 44179, the central star of the Red Rectangle (Cohen et al. 1975;Van Winckel et al. 1995; Waters et al. 1998; Bujarrabal et al. 2005; Witt et al. 2008), andHR 4049 (Johnson et al. 1999; Dominik et al. 2003; Hinkle et al. 2007), showed that theseobjects are indeed binaries with unusual orbits.

The observed broad IR excess in these binary post-AGB stars points to the presence of bothhot and cool dust around the star, which was first noted by Trams et al. (1991). De Ruyter et al.(2006) showed that, for all stars, the hot dust is close to sublimation temperature (∼ 1500 K).An outflow model is not adequate to explain the presence of such hot dust, since the near-infrared excess is expected to disappear within years after the stop of the dusty mass loss. Forthe sources we have observed data over these timescales, the observed IR excess remainedunchanged for over 30 years. Typical mass-loss velocities are around 10 km/s, which wouldplace the dust too far away from the central star to reach such high temperatures. Also, thesepost-AGB stars are currently too hot to produce a strong dusty outflow. The dust must thus re-side in a long-lived, stable reservoir close to the star. This lead us to suspect that these objectsare surrounded by a stable Keplerian disc. Some examples of spectral energy distributions oftypical outflow and disc sources are given in Figure 1.3.

These specific observational characteristics allowed for a more systematic search for thesevisually-bright objects with a broad IR excess and, so far, more than 60 Galactic objectshave been discovered (De Ruyter et al. 2006). For these stars our group started an extensivemulti-wavelength study, including radial velocity monitoring, high-spectral resolution opti-


Figure 1.4 — e− log P diagram of the orbits detected so far for our sample suspected binary post-AGBstars.

cal and infrared studies, submillimetre bolometric observations, and high-spatial resolutioninterferometric studies.

The formation of a disc probably only occurs in an interacting binary system. To detect thisbinary motion, we started an extensive radial velocity monitoring programme, which con-firmed the binary hypothesis and gave a typical semi-major axis for the orbit of around 1 AU(Van Winckel 2003, 2007). The detection of binarity is severely hampered by the pulsationof some objects (the RV Tauri subclass), but so far we have orbital parameters for about 30sources, with orbital periods ranging from 100 to 2000 days. Given the effective temperatureof the central star and its high luminosity, all the discs must be circumbinary, since the orbitsall lie well within the sublimation radius of the dust. Interaction of the binary system with thedisc can account for the observed high eccentricities. Suggested models include pumping-upof the eccentricity (Artymowicz et al. 1991; Waelkens et al. 1996a), or enhanced mass loss atperiastron transit (Soker 2000).

The mass functions calculated from the orbit indicate that the companion is likely a main-sequence star (Van Winckel 2007). The distribution of the orbital periods and eccentricities issurprising (Fig. 1.4), since theory predicts efficient circularisation by tidal forces on very shorttimescales (Pols et al. 2003). All orbits discovered so far are too small to have accommodateda full-grown AGB star. This implies that the system must have been subject to severe binaryinteraction, while the evolved star was at giant dimension.

The presence of the circumbinary disc has an impact on several observational characteristicsof the post-AGB star:

– These binary disc sources are characterised by a depletion pattern in their photospheres(Giridhar et al. 1994; Gonzalez et al. 1997b; Van Winckel et al. 1998; Maas et al. 2005).Generally, a lack of carbon or s-process enhancement in the stellar photospheres is found.These stars apparently did not undergo a third dredge-up process on the AGB, even though

1.4. Binary post-AGB stars and their circumbinary disc 7

Figure 1.5 — The abundance pattern in a typical post-AGB binary, AC Her, shown as [el/H] in functionof the condensation temperature of the chemical element. The condensation temperature of an elementis defined as the temperature at which 50% of the gas particles of that element are condensed, calculatedassuming equilibrium chemistry, solar initial composition and a constant pressure (Lodders 2003). Adescription of the abundance calculations is given in (Van Winckel et al. 1998). The photosphere of thisstar is strongly affected by the depletion process in which refractory elements are selectively under-abundant with respect to non-refractories.

they have initial masses which would theoretically make them evolve to carbon stars. Thisabundance pattern is the result of gas-dust separation, followed by a reaccretion of thegas component, which is now poor in refractory elements. Waters et al. (1992) alreadyproposed that the most likely circumstance for this process to occur is when the dust istrapped in a circumstellar disc. Recently, post-AGB objects with similar depletion patterns(Reyniers et al. 2007; Reyniers & Van Winckel 2007) and far-IR excesses (Meixner et al.2006), have been detected in the Large Magellanic Cloud (LMC), showing that here alsothe formation of a circumbinary disc is probably a common phenomenon in binary evolu-tion (see Fig. 1.5).

– The lack of a carbon-rich signature is also reflected in the circumstellar environment ofthese sources, which is highly oxygen-rich (Molster et al. 2002a). ISO (Infrared SpaceObservatory) and TIMMI2 infrared spectroscopic data of a limited sample show thatdust grain processing is strong, with oxygen-rich dust dominated by crystalline silicates(Molster et al. 2002a; De Ruyter 2006). Some mixed-chemistry sources, showing bothcarbon and oxygen dust species in their infrared spectra, are known. Examples includeHD 44179, the central star of the carbon-rich Red Rectangle nebula (Cohen et al. 1975)and HR 4049 (Johnson et al. 1999; Dominik et al. 2003; Antoniucci et al. 2005; Hinkle et al.2007). The spectra show a strong resemblance to the infrared observations of solar-systemcomets or protoplanetary discs around young stellar objects (YSOs) (Fig. 1.12), despitethe very different evolutionary history and the very different evolutionary time scales in-


Figure 1.6 — The circumbinary disc of the post-AGB star IRAS 08544-4431 in the K-band (left) andthe N-band (right) (Deroo et al. 2007a). The position of the star is indicated by the cross and thebrightest region is the inner wall of the disc on the far side.

volved. A description of the formation and characteristics of discs around young stellarobjects is given in Section 1.7.

– Recent interferometric studies indeed confirm the very compact nature of the circumstellarmaterial (Deroo et al. 2007a,b). The AMBER and MIDI instruments on the VLTI provedto be ideally suited to probe the disc structure. All the data indicate that the disc sizes arevery similar, with N-band sizes between 30 − 50 AU, although the geometrical locationof different dust components can differ significantly. Some objects show crystalline dustspecies close to the inner rim, other show strong processing throughout the disc. In Fig-ure 1.6 we show the results of a disc model, combined with interferometric measurements,for the post-AGB binary IRAS 08544-4431.

– Kinematical information of several objects detected in CO show typical rotation veloci-ties instead of outflows (Bujarrabal et al. 2001, 2007), again confirming the presence of aKeplerian disc.

– Submillimetre measurements (De Ruyter et al. 2005, 2006) indicate the presence of largegrains (cm-size and larger) in the circumstellar environment. A disc is an ideal environ-ment for grain growth to occur, since it is a long-lived stable reservoir, with relativelyhigh densities, close to the star. These large grains have relatively small dust-settlingtimes, causing the disc to be inhomogeneous, consisting of small hot grains in the surfacelayer of the disc and a cool midplane of mainly large grains.

During the evolution of these stars, there must have been a phase of strong interaction whenthe primary was at giant dimension, but the system avoided dramatic spiral in. It is during thisphase of strong interaction that non-conservative mass transfer may have occurred, creatinga circumbinary disc and allowing the system to avoid spiral-in. Another possible formationscenario include a wind-capture scenario (Mastrodemos & Morris 1999), where the AGBwind is captured by the companion. The exact disc formation, however, still remains unclear.

The strong bias found in literature towards binary post-AGB stars could be due to the stabilityof the circumbinary disc. Since the disc gives rise to a long-lived circumstellar environment,

1.5. Dust mineralogy 9

the binary stars will be detected more frequently in infrared studies, in contrast to the singlestars, where the circumstellar environment, coming from an outflow, will dissipate on muchshorter timescales. This selection bias towards atypical objects clouds our understanding ofthe late evolutionary phases of both single and binary stars. So far, a complete statisticalselection of post-AGB stars is not yet available, which prevents a detailed evolutionary studyof these stars.

With this thesis, however, we whish to answer some questions regarding the late phases ofbinary stellar evolution, the relevant physical processes and the impact of evolution on thecentral star and disc system, and how this information relates to the current knowledge of thetransition area between the AGB and PNe phase.

1.5 Dust mineralogy

A large part of this thesis is based on a systematic study of the disc mineralogy, using infraredspectra. We want to investigate which dust species are found in the disc and what the physicalparameters of the dust grains are, like grain size, structure, and dust temperatures.

For this we use high-resolution infrared Spitzer spectra in the 5− 35 µm region. This wave-length region is very sensitive to the thermal emission of dust, and thus rich in solid-stateemission and absorption bands. By comparing these observed features with laboratory data,we can identify the dust properties of these circumbinary discs around evolved objects.

The main goal of this mineralogy study is to determine whether the dust characteristics andgrain processing of these post-AGB binaries contribute to our understanding of the geometryof the circumstellar environment, and the chemical and physical processes which occur here.We will also see if the dust characteristics can be related to other parameters, such as theorbit, the stellar parameters, disc geometry, etc.

1.5.1 Interstellar and circumstellar dust

The interstellar medium (ISM) plays an important role in different stages of stellar evolutionand the life cycle of dust. Stellar evolution efficiently recycles and enriches the dust in theISM, with new stars being formed from dust ejected by dying stars. More than 50% of mate-rial in the ISM comes from mass loss from AGB stars and supernova ejecta (Boulanger et al.2000). Examples of typical ISM dust grains are shown in Figure 1.7.

A dust grain will typically spend about 5×108 years in the ISM (e.g. Jones et al. 1994), whereit may be chemically or physically altered, even destroyed by shocks and hard radiation, andundergo several different chemical processes before it may end up in a dense molecular cloudfrom which new stars are born (e.g. Tielens et al. 1994). If, somewhere in its lifecycle, theISM dusty environment evolves into a star-forming region, some ISM grains will be storedin a disc. The dust particles may start to grow, due to the high density prevailing in thesediscs. Small dust grains will stick together to form increasingly larger dust grains, possiblyeven comets, asteroids or planets. The formation of macroscopic objects is still a poorlyunderstood process, since collisions of dust agglomerates are thought to destroy rather than


Figure 1.7 — Example of an interplanetary dust grains, collected from the Earths stratosphere (e.g.Bradley et al. 2005). This grain is probably made up from smaller ISM-like grains.

enlarge the grains (Dullemond et al. 2008; Zsom & Dullemond 2008).

1.5.2 The detection of dust

The only way to detect the presence of dust in astronomical environments, is to study theinteraction of the dust particles with electromagnetic radiation.

Radiation can scatter (leaving the photon nearly unchanged, apart from its direction) on dustor it can be absorbed by the dust. This will heat up the dust particle, which in turn will startto emit radiation itself. This thermal emission is dependent on the temperature of the dustparticle and its specific characteristics, like size, shape and chemical composition. Typicalinterstellar and circumstellar dust grains have temperatures ranging from a few Kelvin up to∼1500 K, at higher temperatures they will evaporate. Temperatures between 100 K and 1000 Kcorrespond to emission features in the 3 − 30 µm region, thus at infrared wavelengths. So,to study the dust characteristics, we need instruments, sensitive in those wavelength regimes.The Spitzer-IRS spectrograph (see Sect. 1.9) proves to be the ideal tool to study the dust inthe circumbinary discs described in this work.

1.5.3 Dust characteristics

We can study different dust properties such as dust composition, grain size, grain shape andtemperature, by comparing calculated spectra with the solid state bands in the observed emis-sion spectrum. These bands come from stretching and vibration modes (phonons), caused byenergy transferred from an infalling photon to the lattice structure of the molecules.

Dust composition describes which atoms make up the dust, and how these atoms are struc-tured. If the lattice structure of the molecules is very structured, the dust is crystalline, other-wise, it is in amorphous (or glassy) form.

The size of a dust grain has an important effect on the observed emission features. Increasing

1.5. Dust mineralogy 11

the grain size generally decreases the strength of the resonant modes, and can change theprofiles of the infrared spectral features (e.g. Min et al. 2004). With increasing grain size,emission features become weaker and eventually disappear, leaving mainly a contribution tothe continuum, as a blackbody-radiating particle (Fig. 1.8). Features at short wavelengths areaffected most by a change in grain size, at longer wavelengths the features remain of similarstrength.

Models of grain shape are roughly divided into two categories: particles with a sphericalshape and particles with a more irregular structure. The spectrum produced by homoge-neous spherical particles is very different from that produced by more irregular particles (e.g.Fig. 1.8). This difference is much larger than the difference between synthetic spectra com-puted using approximations of different irregular particles (Min et al. 2003).

Synthetic spectra for all these shapes can be produced by modelling the interaction of radia-tion with the different dust particles. For this we need to solve the Maxwell equations for theinteraction of radiation with the dust particles, determining the refractive index values of thespecific dust species. With these refractive indices and the particle cross sections, the massabsorption coefficient is calculated. This calculation is not straightforward, so the shape ofthe dust particle is usually simplified.

If the particle is assumed to be a homogeneous spherical compact grain, then Mie theory(Aden & Kerker 1951; Toon & Ackerman 1981) can be used to calculate the absorption co-efficients, and thus the emission spectrum of the dust. Cosmic dust grains are, however, nothomogeneously spherical, but more irregular, as already can be seen from Figure 1.7. On theother hand, modelling the realistic dust structure, which needs complex computer algorithms,is very time consuming. Most models nowadays are based on the statistical approach, wherethe optical properties of irregularly shaped grains are modelled using an ensemble of simpleshapes with the same optical properties.

An approximation to model irregular grains is the GRF (Gaussian Random Field) approxima-tion (Grynko & Shkuratov 2003; Shkuratov & Grynko 2005). A GRF particle is constructedby first dividing space in small volume elements. Every volume element is then given a valuecorrelated with the values of surrounding volume elements following a lognormal distribu-tion. The resulting 3D-field is called a GRF. All volumes elements with values above a chosenthreshold value are inside the particle, all the other are considered vacuum. This creates anensemble of particles.

The CDE (Continuous Distribution of Ellipsoids) approximation of Bohren et al. (1983) isan example of the statistical approach. There a distribution of ellipsoidal dust shapes is usedto model characteristics of dust grains, smaller than the wavelength of the radiation (i.e. theRayleigh domain). Because of this Rayleigh domain constraint, this model cannot be used tostudy grain size effects on calculated emission spectra.

Another example of the statistical approach is the DHS (Distribution of Hollow Spheres)approximation (Min et al. 2003, 2005a). This model considers hollow spherical particles,averaging the optical properties uniformly over the fraction of the total volume occupied bythe central vacuum inclusion, f , over the range 0 < f < fmax, while keeping the materialvolume of the particles constant. The value of fmax reflects the degree of irregularity of theparticles.


Figure 1.8 — Mass absorption coefficients [cm2/g] of forsterite in different grain sizes and dust shapes.Top panel: GRF approximation with grain sizes of 0.1 µm (solid line), 2.0 µm (dashed line) and 4.0 µm(dotted line). Bottom panel: 0.1 µm forsterite in GRF (solid line), DHS (dashed line) and Mie (dottedline) dust shapes.

1.6 Astronomical dust species

Dust in interstellar and circumstellar matter mainly consists of silicate dust species or car-bonaceous species, such as amorphous carbon (C) or polycyclic aromatic hydrocarbons (PAHs).Examples of other dust species include metallic iron (Fe), different ices (H2O, CO, CO2...),silicon carbide (SiC) and iron sulfide (FeS).

Theoretical models describing the dust condensation in an oxygen-rich AGB outflow (e.g.Tielens et al. 1998a) show that the dust formation is dependent on the temperature, and thushas a specific formation order. At 1800 K, the first to form is corundum (Al2O3), followedat lower temperatures by species like melilite (Ca2Al2SiO7), diopside (CaMgSi2O6) andanorthite (CaAl2Si2O6). At lower pressures crystalline silicates are formed, such as forsterite(Mg2SiO4), enstatite (MgSi)3) and fayalite (Fe2SiO4). Some dust species condense directlyfrom the gas phase (corundum, forsterite), other form by gas-solid (e.g. melilite, enstatite,fayalite) and solid-solid (anorthite) reactions in the AGB outflow. Typical AGB outflowsshow strong features due to amorphous dust, and in lesser degree crystalline dust species.

In the following sections I will describe in short the two most important dust species for thisthesis, namely silicates and PAHs.

1.6.1 Silicate dust

The most abundant dust species in the ISM are amorphous silicates. The ISM mainly con-sists of small (< 0.1 µm) olivine (Mg2xFe2(1−x)SiO4), or pyroxene (MgxFe1−xSiO3) grains

1.6. Astronomical dust species 13

Figure 1.9 — Mass absorption coefficients [cm2/g] of amorphous and crystalline dust species. Toppanel: forsterite (solid line) and enstatite (dotted line). Bottom panel: olivine (solid line) and pyroxene(dotted line)

(where 0 ≤ x ≤ 1 denotes the magnesium content). A significant fraction of silicatedust is, however, also found in crystalline form, in objects such as comets (Crovisier et al.1997), circumstellar discs around young stars (Waelkens et al. 1996b; van Boekel et al. 2005;Bouwman et al. 2008), outflows of AGB stars and planetary nebula (Justtanont et al. 1996;Waters et al. 1996).

Surprisingly the crystalline dust is only found in the magnesium-rich component. Spectraobtained with the ISO satellite and more recently with the Spitzer satellite show strong fea-tures of forsterite (Mg2SiO4) and enstatite (MgSiO3) in stars at different evolutionary phases.To our knowledge, features of the Fe-rich crystalline species such as fayalite (Fe2SiO4) orferrosilite (FeSiO3) have never been observed.

1.6.1.1 Formation

Amorphous silicates can form by fast condensation, not allowing the particles to arrangethemselves in a very structured lattice. Forsterite condenses directly from the gas phase athigh temperatures (≈ 1200 − 1500 K) or it may form by thermal annealing of amorphoussilicates, diffusing the iron out of the lattice structure. Enstatite can form in the gas phasefrom a reaction between forsterite and silica (SiO2), or it may also form by a similar thermalannealing process as forsterite (Bradley et al. 1983; Tielens et al. 1998a). Laboratory exper-iments have indicated that silica (SiO2) can be formed when amorphous silicates anneal toforsterite (e.g. Fabian et al. 2000).

Tielens et al. (1998a) proposed three explanations for the observed low iron content in theobserved silicates. The first that reactions with iron require low enough temperatures. Highertemperatures favour the formation of Mg-rich silicates. Another explanation would be that,given high enough pressures, iron can condense out of the molecule as metallic iron. The


Figure 1.10 — Molecular structure of benzene (C6H6), phenanthrene (C14H10), tetracene (C18H12),C54H12 and C96H24.

third is that the reaction of iron with crystalline silicates leads to the formation of amorphousMg−Fe-rich silicates instead of Fe-rich crystalline silicates.

1.6.1.2 Identification

Amorphous silicates have very prominent broad spectral features around 10 µm and 18 µm(see Fig. 1.9). The 10 µm feature (also often called the 9.8 µm feature) is caused by theSi-O stretching in the molecule and the 18 µm feature is due to an O-Si-O bending mode.Amorphous pyroxene shows a 10 µm feature similar to that of amorphous olivine, but shiftedtowards shorter wavelengths. Also the shape of the 20 µm feature is slightly different.

The synthetic spectrum of crystalline silicates shows strong but narrow features (see Fig. 1.9),and is very dependent on the adopted Mg/(Mg+Fe) ratio: the peak position of the bands shiftsto longer wavelengths as this ratio decreases (e.g. Jaeger et al. 1998; Chihara et al. 2002;Koike et al. 2003). The Mg-rich end members of crystalline olivine and pyroxene, forsterite(Mg2SiO4) and enstatite (MgSiO3), show features at distinct wavelengths around 11.3 −16.2− 19.7− 23.7− 28− 33.6 µm (see Fig. 1.8). The peak positions not only depend on theadopted grain sizes but also on the dust shape approximation (Min et al. 2003, 2005a).

1.6.2 Polycyclic aromatic hydrocarbons

A PAH particle is made up from hydrocarbon molecules with the C atoms arranged in anaromatic benzene ring (like graphite) with H atoms bound at the edges. In its most simpleform it consists of just one benzene ring, with six H atoms. This is then actually a MonocyclicAromatic Hydrocarbon, the building block for the larger polycyclic forms. Examples ofdifferent PAHs are given in Figure 1.10.

On Earth, PAHs are most known as a by-product of fuel burning. They are found in e.g.oil, coal and cigarette smoke. In space they are found in a variety of astronomical envi-

1.7. Discs in young stellar objects 15

Figure 1.11 — PAH emission spectrum seen in the post-AGB star HD 44179, the central star of the RedRectangle.

ronments. They are abundantly found in the ISM, but also in circumstellar discs aroundyoung stars, evolved carbon-rich and mixed-chemistry stars, which are hot enough to ex-cite the PAH carriers, HII regions and galaxies (e.g. Hony et al. 2001; Peeters et al. 2002;Acke & van den Ancker 2004; Brandl et al. 2006; Armus et al. 2007; Geers et al. 2007).

The main source of excitation of the PAH carriers is UV radiation. When the UV field ispresent, emission features are seen in the infrared at typical wavelengths, 3.3 − 6.2 − 7.7 −8.6− 11.2 and 12.7 µm (Fig. 1.11). The feature at 3.3 µm is attributed to the C-H stretchingmode of neutral PAHs. The C-C modes produce features with typical central wavelengthsat 6.2 and 7.7 µm. The 8.6 µm feature is due to C-H in-plane bending modes and featureslongward of 10 µm can be attributed to C-H out-of-plane bending modes. For a detaileddescription of PAH chemistry and characteristics we refer to Tielens (2008).

1.7 Discs in young stellar objects

A natural environment to study grain processing is in the circumstellar environment of youngstellar objects. Molster et al. (2002a) and De Ruyter (2006) already pointed out that the in-frared spectra of the discs around evolved stars show emission due to the same dust speciesseen in discs around YSOs (Fig. 1.12). Discs around young stars have already been exten-sively studied, and a clear correlation is found between the evolutionary phase and parame-ters of the central star, and the disc characteristics (e.g. Meeus et al. 2001; Dullemond et al.2007). Below we will give a short description on the disc formation and characteristics ofprotoplanetary discs around young stars.


Figure 1.12 — Comparison between the binary post-AGB star AC Her, the young stellar objectHD 100546 and the solar-system comet Hale-Bopp.

1.7.1 Star formation and protoplanetary discs

Star formation occurs when molecular clouds, i.e. dense clouds of interstellar gas and dust,start to collapse due to self-gravitation. During the collapse the density in the core increasesand the core temperature rises until it is high enough for the protostar to initiate nuclear reac-tions. At this point the protostar becomes a stable star and moves from the pre-main sequence(PMS) to the main sequence (MS) in the Hertzsprung-Russell diagram (see Fig. 1.1).

Depending on the mass, young stars can be divided into three groups: Stars with M <1.5 M¯ are called T Tauri stars and are optically visible on the PMS; stars with M > 8 M¯remain optically invisible during the entire PMS. The last group are the intermediate Her-big Ae/Be stars, with 1.5 M¯ < M < 8 M¯: here the ’A/B’ denotes the spectral type andthe ’e’ the presence of emission lines. These stars are characterised by (sometimes strong)infrared flux coming from the surrounding circumstellar material (Hillenbrand et al. 1992;Mannings & Sargent 1997; Waters & Waelkens 1998).

Since the infalling cloud has intrinsic angular momentum, conservation of angular momen-tum will give rise to the formation of a circumstellar accretion disc, occasionnaly togetherwith the formation of a strong bipolar outflow of material from the star. The disc is in firstinstance very active. There will be strong accretion from the disc onto the star, causing thedisc to become very turbulent with efficient radial and vertical mixing of material in the disc.This active accreting phase will last about 0.1 million years with typical accretion rates of10−5M¯yr−1.

When the accretion has decreased significantly the disc enters a passive phase. It becomesa stable disc where processes like grain growth and dust settling can occur. Such a disc is

1.7. Discs in young stellar objects 17

also called a protoplanetary disc since it is the ideal environment for planet formation. Thepassive disc is dispersed on a timescale of a million years. The dust has grown to considerabledimension (including macrostructures like planetesimals and planets), and the bulk of the gasin the disc has been removed. The protoplanetary disc has become a debris disc. Collisionsin this debris disc can once again give rise to the formation of small grains.

1.7.2 Disc geometry

The disc surface is heated by irradiation from the central star, heating the dust grains in thedisc. If the densities in the disc are high enough, the gas and dust will be coupled, givingthem the same temperature distribution. Heating up the gas will increase the gas pressure inthe disc, which can balance the vertical component of the gravitational force, pulling materialto the midplane. In hydrostatic equilibrium, the vertical scale height of the disc depends onthe ratio of the gas pressure and gravity.

Looking at the spectral energy distribution of Herbig stars, we can divide them into twogroups (Meeus et al. 2001): Group I shows a strongly rising double-peaked IR excess with aclear 60 µm bump, Group II has only a moderate IR excess without a significant 60 µm bump(see Fig. 1.13). Both groups show a strong near-IR excess, pointing to the presence of hotdust in the inner parts of the disc. Within the sublimation radius no dust will be present, so aninner dust-free gap will be present around the star, allowing stellar radiation to hit the innerrim of the disc full-on, causing it to heat up and puff up (Natta et al. 2001; Dullemond et al.2001).

In recent years strong progress has been made in the construction of detailed disc models.Starting with the semi-analytical model for an irradiated flared disc of Chiang & Goldreich(1997), more elaborate and detailed 2D/3D disc models have been constructed, describingdifferent disc structures such as accretion discs, passive discs and flared discs with a puffed-upinner rim (D’Alessio et al. 1998; Dullemond & Dominik 2004; Baes et al. 2007a; Min et al.2009). A description of such a disc model is given in Section 1.8

1.7.3 Disc chemistry

The chemical composition of the circumstellar environment will reflect the chemical signa-ture of the molecular cloud from which they are born, showing strong features of moleculesand dust species abundant in the ISM, such as silicates, PAHs and different molecules. Dur-ing disc evolution both the gas and dust component will be strongly processed, allowing usto study different chemical and physical processes in the disc.

The discs are highly gas-rich, with emission and absorption features due to CO, H2, CN,HCN, C2H, etc (Thi et al. 2001; Herczeg et al. 2002; Thi et al. 2004). Also atomic linesof neutral hydrogen, oxygen and ionised calcium, silicon, magnesium and iron have beenobserved in emission.

The main dust species of the discs are amorphous silicates (Bouwman et al. 2001; Sargent et al.2006; Kessler-Silacci et al. 2006), but studies with ISO, TIMMI2 and Spitzer-IRS have shownthat dust processing, in the form of grain growth and crystallisation, is very efficient is


Figure 1.13 — The spectral energy distribution of a typical group I source (HD 97048), and a typicalgroup II object (HD 95881). The symbols represent photometric data and the grey line is the observedSpitzer spectrum. The black solid line gives the adopted Kurucz stellar atmosphere model.

these protoplanetary discs. In many stars strong features due to crystalline silicates are seen(Malfait et al. 1998; van Boekel et al. 2003, 2005). Other observed solid state bands includeFeO, H2O ices and FeS (Keller et al. 2002).

PAH molecules are abundant in the ISM and are therefore expected to be observed in youngstellar discs. Indeed, about 50% of all young passive discs show features due to PAHs(Van Kerckhoven et al. 2002; Acke & van den Ancker 2004; Sloan et al. 2005; Geers et al.2007) and the detection seems to be linked to the shape of the SED. Strongly flared discs willabsorb UV radiation more easily, which is necessary to excite the PAH carriers, and will thusshow stronger PAH features compared to shadowed sources.

1.8. The self consistent disc model 19

Figure 1.14 — Sketch of the geometrical disc model as described in Dullemond et al. (2001).

1.8 The self consistent disc model

To model the discs around evolved binaries we used the self-consistent method as describedby Dullemond et al. (2001), Dullemond (2002) and Dullemond & Dominik (2004). This codewas originally developed to study protoplanetary discs around young stars, but proved ide-ally suited to model our discs as well. The model performs 2D radiative transfer, under thecondition of radiative and vertical hydrostatic equilibrium. Also, the disc is assumed to bepassive: irradiation of the central star is the only source of heating in the disc and there is nodynamic coupling between the disc and its central star.

Calculations are done using axisymmetric spherical coordinates (r,θ) and consist of two ma-jor parts. The first being a 2D radiative transfer code (RADMC) which calculates the dustopacity and temperature. The gas component of the disc is assumed to be thermally coupledto the dust. The second part of the model is a 1D hydrostatic equilibrium code which calcu-lates the vertical density profile. This is done self-consistently by using an iteration process:an initial guess for the density distribution ρ(r, θ) is taken as input to the first radiative trans-fer step, which gives a temperature distribution T (r, θ) throughout the disc. A new densitydistribution is then calculated by integrating the equation of vertical hydrostatic equilibrium

∂P

∂z= −ρ

GM∗r3

z,

with z = r(π2 − θ), in the vertical direction. This process is repeated several times, until

convergence is reached. This gives a solution for the disc structure, with self-consistentvalues for ρ(r, θ) and T (r, θ).

Using these new values for the density and temperature distribution, the spectral energy dis-tribution of the disc is calculated. This is done with the RADICAL routine, a ray-tracingmethod which calculates the SED under a given inclination of the disc with respect to theline-of-sight.

The inner part of the disc is assumed to be dust free, and the gas which is present in this inner


gap is considered to be optically thin. This will cause the radiation of the star to hit the innerwall of the disc head-on. The gas/dust mixture will heat up which causes it to puff up. If theopacity in this puffed-up inner rim is high enough, it may cast a shadow over the subsequentparts of the disc, resulting in a decrease of local scale height. At larger distances, the disc canstill receive direct stellar radiation, which causes flaring of the outer disc parts. A sketch ofthe geometry of such a disc is shown in Figure 1.14.

Model input parameters are:

– Stellar parameters: We assume the evolved star to be the main source of irradiation,and thus neglect the contribution of the secondary. A Kurucz atmosphere model, with therelevant temperature Teff , luminosity L∗, gravity log g and metallicity, is taken to representthe primary. For the gravitational mass of the central star system we take the estimatedcombined mass of the primary and secondary.

– Composition of the disc: Input parameters include the different dust species, the ratiocrystalline/amorphous dust and the grain size distribution. This grain size distribution isassumed to be a power law distribution. We consider the disc to be homogeneous, so weuse the same grain-size distribution everywhere. For the gas-to-dust ratio a standard valueof 100 is taken.

– Total disc mass: Combined mass of the gas and dust components.

– Surface density distribution: The radial distribution of the mass throughout the disc isdescribed with the surface density distribution:

Σ(r) = Σ0

(r

r0

)p

,

with p the power-law index and Σ0 a normalisation factor at a reference distance r0.

– Disc size: The disc dimension is given by the inner radius Rin and outer radius Rout. Forcomputational reasons, the disc is not cut off after Rout, but instead the density power lawgoes down very steep, with a value of -12. In our modelling we do not allow the innerradius to be smaller than the dust sublimation radius.

1.9 The Spitzer sample

The infrared spectroscopic studies performed in this thesis are based on observations obtainedwith the spectrograph aboard the Spitzer Space Telescope. One of the major improvementsof the Spitzer telescope over the ISO mission, launched in 1995, is its high sensitivity (downto fluxes of a few mJy). Unfortunately, Spitzer cannot compete with ISO when it comes towavelength coverage, giving respectively infrared spectra from 5 − 40 µm and 2 − 240 µm.Spectroscopic information longwards of 40 µm will again become available with the launchof the Herschel satellite, and the release of the Herschel-PACS instrument, which is operativein the 60− 210 µm region.

1.9. The Spitzer sample 21

Figure 1.15 — The Spitzer Space Telescope. Taken from http://www.spitzer.caltech.edu/technology.

1.9.1 The Spitzer mission

The Spitzer Space Telescope (formerly SIRTF, the Space InfraRed Telescope Facility, Fig. 1.15)was launched into space by a Delta rocket from Cape Canaveral, Florida on 25 August 2003,in an earth-trailing heliocentric orbit. During its mission, Spitzer obtained images and spec-tra by detecting the infrared energy radiated by objects between 3− 180 µm. Consisting of a0.85 m telescope and three cryogenically-cooled science instruments, Spitzer was the largestinfrared telescope ever launched into space.

Spitzer was the final mission in NASA’s Great Observatories Program - a family of four or-biting observatories, each observing the Universe at different wavelengths. Other missionsin this program include the Hubble Space Telescope (HST), Compton Gamma-Ray Observa-tory (CGRO), and the Chandra X-Ray Observatory(CXO). Spitzer is also a part of NASA’sAstronomical Search for Origins Program, designed to provide information which will helpus understand our cosmic roots, and how galaxies, stars and planets develop and form.

On board of Spitzer are three infrared instruments: the Infrared Array Camera (IRAC), theInfrared Spectrograph (IRS) and the Multiband Imaging Photometer for Spitzer (MIPS).IRAC is a four-channel camera that provides simultaneous images at 3.6, 4.5, 5.8, and 8 µm.The IRS has four separate modules: a short-wavelength, low-resolution mode (SL) cov-ering the 5.3-14 µm interval; a short-wavelength, high-resolution (SH) mode covering 10-19.5 µm; a long-wavelength, low-resolution mode (LL) for observations at 14-40 µm and along-wavelength, high-resolution mode (LH) for 19-37 µm. The MIPS instrument consistsof three detectors at 24, 70 and 160 µm.

1.9.2 The sample stars

From the larger binary post-AGB sample described by De Ruyter et al. (2006), 35 sourcesfall within the dynamic range of the IRS high resolution mode. We selected 21 stars to be


observed with the Spitzer-IRS spectrograph, for which we have clear evidence that they areindeed binaries. The determination of the binary orbit was done using long-term radial ve-locity monitoring. An example is given in Figure 1.16. The coordinates of the program starswere well known since they are bright in the optical and near-IR, giving us exact coordinates.All sources were detected with IRAS and for a few of the brightest ones, ISO-SWS data areavailable. A list of targets and coordinates is given in Tables 1.1 and 1.2. Also given are thestellar parameters and the orbital characteristics (if available).

Exposure times were chosen to achieve a S/N ratio of around 150 for the high-resolutionmodes, which we complemented with short exposures in low-resolution mode with a S/Nratio around 50, using the first generation of the exposure time calculator of the call forproposals. Proper motions were taken from the UCAC2 and/or Tycho reference catalogue.The science objects were too bright for peak-up, so 2MASS or PSC-RS stars were taken.The total requested integration time was 7.7 hours and all sources were observed betweenSeptember 2004 and March 2005.

For stars where we lack the Spitzer-IRS SH observations we obtained additional ground-based N-band infrared spectra with the Thermal Infrared Multi Mode Instrument 2 (TIMMI2,Reimann et al. 2000; Kaufl et al. 2003), mounted on the 3.6 m telescope at the ESO La SillaObservatory. These data were taken in June 2006.

The low-resolution (R ∼ 160) N band grism was used in combination with a 1.2 arcsecslit, the pixel scale in the spectroscopic mode of TIMMI2 is 0.45 arcsec. For the reduction ofthe spectra we used the method described in van Boekel et al. (2005). We scaled the TIMMI2spectra to the Spitzer-IRS spectra and found a very good agreement in spectral shape betweenthe two data sets.


Figure 1.16 — The spectral energy distribution and orbit of one of our sample stars, IRAS 19125+0343.The plus-signs in the SED represent the dereddened photometry, the full line gives the adopted Kuruczmodel atmosphere, with the appropriate temperature, metallicity, and gravity. The orbital period iscalculated to be Porb = 508 days, with an eccentricity of e = 0.22.


Tabl

e1.

1—

The

equa

tori

alco

ordi

nate

sα

and

δ(J2

000)

,eff

ectiv

ete

mpe

ratu

reT

eff

,sur

face

grav

itylo

gg

and

met

allic

ity[F

e/H

]of

our

sam

ple

star

s.Fo

rth

em

odel

para

met

ers

we

refe

rto

De

Ruy

tere

tal.

(200

6).

Als

ogi

ven

isth

eor

bita

lper

iod,

the

ecce

ntri

city

and

the

sem

i-m

ajor

axis

ofth

eor

bita

lmot

ion

asi

ni

(see

refe

renc

esin

De

Ruy

tere

tal.

2006

).

Nam

eα

(J20

00)

δ(J

2000

)T

eff

log

g[F

e/H

]P

orb

ite

asi

ni

(hm

s)(

’”)

(K)

(cgs

)(d

ays)

(AU

)E

PLy

r19

1817

.5+

2750

3870

002.

0-1

.5H

D13

1356

1457

00.7

−68

5023

6000

1.0

-0.5

1490

0.36

2.07

HD

2139

8522

3527

.5−1

715

2782

501.

5-1

.025

90.

000.

78H

D52

961

0703

39.6

+10

4613

6000

0.5

-4.8

1310

0.21

1.60

IRA

S05

208−

2035

0522

59.4

−20

3253

4000

0.5

0.0

236

0.00

0.41

IRA

S09

060−

2807

0908

10.1

−28

1910

6500

1.5

-0.5

371

0.14

0.16

IRA

S09

144−

4933

0916

09.1

−49

4606

5750

0.5

-0.5

1770

0.25

2.53

IRA

S10

174−

5704

1019

18.1

−57

1936

323

0.00

0.17

IRA

S16

230−

3410

1626

20.3

−34

1712

6250

1.0

-0.5

IRA

S17

038−

4815

1707

36.3

−48

1908

4750

0.5

-1.5

1381

0.56

1.69

IRA

S17

243−

4348

1727

56.1

−43

5048

6250

0.5

0.0

484

0.00

0.18

IRA

S19

125+

0343

1915

00.8

+03

4841

7750

1.0

-0.5

517

0.22

0.55

IRA

S19

157−

0247

1918

22.5

−02

4209

7750

1.0

0.0

120.

50.

330.

07IR

AS

2005

6+18

3420

0754

.8+

1842

5758

500.

7-0

.4R

UC

en12

0923

.7−4

525

3560

001.

5-2

.0SA

O17

3329

0716

08.3

−23

2702

7000

1.5

-0.8

115.

90.

000.

14ST

Pup

0648

56.4

−37

1633

5750

0.5

-1.5

410

0.03

0.65

SUG

em06

1400

.8+

2742

1257

501.

125

-0.7

SXC

en12

2112

.6−4

912

4160

001.

0-1

.060

00.

001.

12T

WC

am04

2048

.1+

5726

2648

000.

0-0

.5U

YC

Ma

0618

16.4

−17

0235

5500

1.0

0.0


Table 1.2 — The name, equatorial coordinates α and δ(J2000), effective temperature Teff , surface grav-ity log g and metallicity [Fe/H] of the MACHO RV Tauri pulsators described in Chapter 5.

Name α (J2000) δ (J2000) Teff log g [Fe/H](h m s) ( ’ ”) (K) (cgs)

MACHO 79.5501.13 05 14 18.1 -69 12 34.9 5750 0.5 -2.0MACHO 81.8520.15 05 32 54.5 -69 35 13.2 6250 1.0 -1.5MACHO 81.9728.14 05 40 00.5 -69 42 14.6 5750 1.5 -1.0MACHO 82.8405.15 05 31 51.0 -69 11 46.4 6000 0.5 -2.5

Figure 1.17 — Example of bcd.fits image of one of our sample stars, taken in SH mode. The differentechelle orders are clearly visible, as are the abundant bad pixels.

1.9.3 Spitzer-IRS data reduction

1.9.3.1 High-resolution

For the data reduction of the high-resolution Spitzer-IRS spectra the ’Cores2Discs (c2d)Interactive Analysis’ reduction environment was used, developed by the c2d legacy teamfor the legacy program “From molecular cores to planet forming disks”. Here we give inshort the different steps in the reduction process, for a more detailed description we refer toLahuis et al. (2006).

For wavelength calibration and IRS aperture definition, tools from the SMART softwarepackage (Higdon et al. 2004) were used. Extraction of the spectra was done using interme-diate calibrated data (RSC), pipeline version S13.2.0. These two dimensional spectra havebeen processed through the Spitzer Science Center (SSC) pipeline, including saturation flag-ging, dark-current subtraction, linearity correction, cosmic ray correction, ramp integration,droop (a slope proportional to the photocurrent summed over all pixels) correction, stray light(falling on the short-low modules from the peak-up arrays) removal and crosstalk correction


(light leakage between the high-resolution modules).

1D spectra are extracted from the long-slit (SL,LL) and echelle (SH,LH) images (Figs. 1.17-1.18) using two extraction methods: a full aperture extraction and a point spread function(PSF) extraction. As a first step the spectra have to be corrected for hot/bad pixels. SSCprovided mask files for permanently bad pixels but in the c2d extraction method additionalbad pixels are identified for the PSF extraction and corrected for all known bad pixels in thefull aperture extraction.

The full aperture extraction consists of a fixed-width aperture extraction from the RSC prod-ucts (intermediate products with stray light and cross-talk removed) followed by a correctionfor the spectral response function (SRF). For the low-resolution mode the width is taken sothat 99% of a point-source flux falls within the window, for the high-resolution mode the fullslit width is used. The spectra are extracted separately for each dither position and then aver-aged into one spectrum. For the low-resolution mode the different nod positions are used forsky correction, for high-resolution mode a sky estimate is derived using the PSF extractionmethod.

The optimal PSF extraction method is done with combined RSC data after correcting for thecross dispersion offsets of the separate dither position images. The PSF in the cross dipsersiondirection is defined using sky corrected high signal-to-noise calibrator data. 1D spectra areextracted for all apertures and flux calibrated using the SRF.

For both methods the calibration is done using standard stars observed within the SpitzerCalibration Program. Cohen templates and MARCS model atmospheres (Decin et al. 2004)provided by the SSC are used to derive the SRF.

During the extraction, errors are assigned to each spectral data point. The errors are propa-gated along the pipeline and include the relative spectral response uncertainty, the absoluteflux calibration uncertainty, and, for the full aperture extraction, the deviation between ditherpositions.

After extraction and flux calibration, the high-resolution (SH, LH) and the low-resolution LLorder spectra are defringed using the IRSFRINGE package (Lahuis & Boogert 2003), wherea fringe model based on robust sine-wave fitting removes fringe-residuals from the spectra.These fringes originate from interferences on plane-parallel surfaces in the light-path of theinstrument.

As a last step, order matching is performed for the SH and LL modules. Orders within eachmodule are scaled in flux such that consecutive overlapping orders are matched. For SL andLL this is not done as order mismatches can be useful for assessing pointing errors.

1.9.3.2 Low-resolution

For low-resolution spectra we used the FEPS pipeline, developed for the Spitzer legacy pro-gram “Formation and evolution of planetary systems”. A detailed description can be foundin Hines et al. (2005).

For the extraction, intermediate droop data products, as delivered by the SSC, are used, to-gether with the SMART reduction package tools (Higdon et al. 2004) and reduction tools


Figure 1.18 — Example of bcd.fits images of one of our sample stars, taken in SL (left) and LL (right)modes. The bright squared regions in the SL mode are the peak-up arrays.

developed by the FEPS team. As a first step, the spectra are background-corrected by sub-tracting images of the two slit positions for each module and order. Using the mask filesprovided by the SSC, bad/hot pixels are identified and replaced with an average pixel value.

The spectra are extracted using a fixed width aperture, where the width is determined so that99% of the source flux falls within the aperture. After the extraction of the spectrum for eachnod position and cycle, a mean spectrum over all slit positions and cycles is computed foreach individual order. After that the orders are combined. In regions where there is orderoverlap, the fluxes are replaced by the mean flux value at each wavelength point. Here alsothe IRSFRINGE package was used for the defringing of the spectra.

For the calibration, spectral response functions were calculated, derived from standard starsand corresponding stellar models. The estimated average error on the SRF and error on theextracted spectra are incorporated in the final estimated error.

Finally, for each target a one-dimensional spectrum is calculated by stitching together over-lapping orders. This is done by taking the mean flux at overlapping wavelength points, nowavelength shifting or spectral tilting is performed.

1.9.4 TIMMI2 data reduction

For some Spitzer sample stars we lack spectra in the 10 µm region. For these stars weobtained additional spectra with the Thermal Infrared Multi Mode Instrument 2 (TIMMI2,Reimann et al. 2000; Kaufl et al. 2003), mounted on the 3.6 m telescope at the ESO La SillaObservatory. The low-resolution (R ∼ 160) N band grism was used in combination with a1.2 arcsec slit, the pixel scale in the spectroscopic mode of TIMMI2 is 0.45 arcsec.

For the reduction we used the procedure as described in van Boekel et al. (2005). The pipelinereduced data, which computes the mean of all subtracted images, consist of a FITS image of


10 15 20 25 30 35Wavelength [µm]

2.0

2.5

3.0

3.5

4.0

4.5

Fν

[Jy]

Figure 1.19 — Reduced Spitzer-IRS of one of our sample stars, IRAS 16230, together with the pre-dicted errors.

the source. Due to chopping and nodding, this FITS image has a positive spectrum containinghalf the signal, and two negative spectra, each containing a quarter of the signal. The recoverthe signal, the spectra are extracted and combined.

To correct for the strong observed ozone atmospheric absoprtion around 9.7 µm, we deter-mine, for each science observation, the atmospheric extinction per unit airmass Aλ and theinstrumental response Rλ, from two calibration measurements, using:

sλ,1 = Iλ,1e−τλ,1Rλ,1 τλ,1 = AλmA,1

sλ,2 = Iλ,2e−τλ,2Rλ,2 τλ,2 = AλmA,2.

Here, sλ,i is the measured calibrator spectrum, Iλ,i the intrinsic model calibrator spectrum,τλ,i the optical depth of the earth atmosphere during the calibration, and mA,i the airmass atwhich the calibrators were observed.

Solving these equations gives:

Aλ =ln(Iλ,2/Iλ,1) + ln(sλ,2/sλ,1)

mA,2 −mA,1

Rλ =sλ,1

Iλ,1e−Aλmλ,1.

Since the atmospheric extinction per unit airmass Aλ is relatively constant with wavelengthover the spectrum, we consider a mean Aλ.

We then obtain the intrinsic spectrum of the science target by dividing the observation of thescience target by the calibration measurement, and then multiply the result with the knownintrinsic calibrator spectrum. The instrinsic spectrum of a science target observed at airmassmA is thus calculated from its measured spectrum sλ as

Iλ =sλ

RλeAλmA .

1.10. Outline of the thesis 29

1.10 Outline of the thesis

The work presented in this thesis focuses on a detailed study of the composition and ge-ometry of the circumbinary structure around evolved binary stars, both for Galactic and forextragalactic targets. High-resolution infrared spectroscopic observations are used to study indetail the dust composition whereas photometric and interferometric high-spatial resolutiondata give us more information on the characteristics of the circumstellar environment, suchas size, density, temperature and dust mass. By comparing results in a large-sample study, wecan study observed trends and individual deviations from general characteristics. This way,we can investigate possible relations between the circumstellar environment and the charac-teristics of the central star, the binary orbit and/or evolutionary phase. We also hope to shedsome light on some more fundamental questions relating late phases of stellar evolution, and,more specifically, binary evolution and the impact on the central system. Below we give ashort overview of the different chapters.

In Chapter 2 we perform a pilot study on one of our Spitzer sample stars, namely the RV Tauripulsating star RU Cen. By combining a wide range in observational data and techniques, westudy the binary nature of this star and compare it with another well-known RV Tauri star,AC Her.

In Chapter 3 we extend our study and investige the mineralogy and dust processing in thecircumbinary discs of 21 binary post-AGB stars using high-resolution Spitzer and TIMMI2infrared spectra. We perform a full spectral fitting, comparing the observed spectral withsynthetic calculated spectra. This allows us to constrain the different dust species and graincharacteristics.

In Chapter 4 we study in more detail two of our sample stars, namely EP Lyr and HD 52961.As found in Chapter 3, both stars show unique solid-state and gas features in their spectraand have a very small infrared excess, in comparison with the other sample stars.

In Chapter 5 we expand our study by investigating the circumstellar environment of 4 extra-galactic RV Tauri stars in the LMC. Central star parameters and chemistry are derived fromhigh-resolution UVES optical spectra. This, combined with the luminosity derived from theknown LMC distance, confirms the post-AGB evolutionary status of these stars. The dustcomposition of the circumstellar environment is traced with low-resolution Spitzer infraredspectra, and we compare the observed features with those seen in the Galactic sample.

In Chapter 6 we give the overall conclusions and remaining questions of this thesis. We alsotake a preview of future research in Chapter 7.

Chapter 2The RV Tauri spectral twins RU Cenand AC Her.

This chapter is published as

Dust-grain processing in circumbinary discs around evolved binaries. The RV Taurispectral twins RU Cen and AC Her.

C. Gielen, H. Van Winckel, L.B.F.M. Waters, M. Min and C. Dominik, A&A 2007, 475,629

Abstract:CONTEXT: We study the structure and evolution of the circumstellar discs aroundevolved binaries and their impact on the evolution of the central system.AIMS: By combining a wide range of observational data and techniques, we aim tostudy in detail the binary nature of RU Cen and AC Her, as well as the structure andmineralogy of the circumstellar environment.METHODS: We combine a multi-wavelength observational program with a detailed 2Dradiative transfer study. Our radial velocity program is instrumental in the study of thenature of the central stars, while our Spitzer spectra complimented with the broad-bandspectral energy distribution (SED) are used to constrain mineralogy, grain sizes andphysical structure of the circumstellar environment.RESULTS: We determine the orbital elements of RU Cen, showing that the orbit ishighly eccentric with a large velocity amplitude despite the rather long period of 1500days. The infrared spectra of both objects are very similar and the spectral dustfeatures are dominated by magnesium-rich crystalline silicates. The small peak-to-continuum ratios are interpreted as being due to large grains. Our model contains twocomponents with a cold midplain dominated by large grains, and the near- and mid-IRwhich is dominated by the emission of smaller silicates. The infrared excess is well

32 Chapter 2. The RV Tauri spectral twins RU Cen and AC Her.

modelled assuming a hydrostatic passive irradiated disc. The profile-fitting of the dustresonances shows that the grains must be very irregular.CONCLUSIONS: These two prototypical RV Tauri pulsators with circumstellar dustare binaries where the dust is trapped in a stable disc. The mineralogy and grainsizes show that the dust is highly processed, both in crystallinity and grain size. Thecool crystals show that either radial mixing is very efficient and/or that the thermalhistory at grain formation has been very different from that in outflows. The physicalprocesses governing the structure of these discs are very similar to those observed inprotoplanetary discs around young stellar objects.

2.1 Introduction

The RV Tauri class of objects contains highly luminous stars showing large-amplitude pho-tometric variations with alternating deep and shallow minima. The members are located inthe high-luminosity end of the population II instability strip and the photometric variationsare interpreted as being due to radial pulsations. There are two different photometric classes:the RVa stars are objects with a constant mean magnitude while the RVb objects displaya long-term variation in their mean magnitude. Preston et al. (1963) introduced a spectro-scopic classification of the RV Tauri stars, using unfortunately the same alphabetic letters forthe naming: RVA objects show strong absorption lines, RVB objects are of a somewhat hotterspectral type but are weak lined with enhanced CN and CH bands. The RVC objects are alsoweak lined but show no enhanced CN and CH molecular band strength. A significant fractionof the RV Tauri stars show a large IR excess due to circumstellar dust and Jura (1986) identi-fied them as post-AGB objects on the basis of this IR excess, their luminosities and mass-losshistory.

The photospheric content of RV Tauri stars is, however, very different from what could beexpected in post-3rd dredge-up objects: they do not show high carbon abundances or s-process overabundances but instead often show a depletion pattern in their photospheres(Giridhar et al. 1994, 1998, 2000; Gonzalez et al. 1997a,b; Van Winckel et al. 1998; Maas et al.2005). This abundance pattern is the result of gas-dust separation followed by reaccretion ofthe gas, which is poor in refractory elements. Waters et al. (1992) proposed that the mostlikely circumstance for this process to occur is when the dust is trapped in a circumstel-lar disc. This depletion phenomenon is also observed in binary post-AGB stars with a disc(Van Winckel et al. 1995). This led Van Winckel et al. (1999) to suggest that the depletedRV Tauri stars must also be binaries with a disc. The likely presence of a Keplerian circum-stellar disc was further shown by the systematic study by De Ruyter et al. (2005, 2006) of alarge sample of binary post-AGB and RV Tauri stars.

Blocker (1995) list the post-AGB evolutionary tracks of single stars of different initial mass.Evolutionary tracks for binary post-AGB stars have not yet been determined, however, butsincce we find evidence that these stars have been shortcut on their AGB evolution (Sect.2.8), we expect longer lifetime scales because of the expected lower core masses. Typicalpost-AGB lifetimes of both stars are estimated to be of the order of ∼ 104 yr.

To investigate the special evolutionary status of RV Tauri stars and to research in detail the

2.1. Introduction 33

Table 2.1 — The name, equatorial coordinates α and δ (J2000), the effective temperature Teff , thesurface gravity log g and the metallicity [Fe/H] of the programme stars RU Cen and AC Her. Modelparameters are taken from Maas et al. (2002) and Van Winckel et al. (1998). Typical errors are ∆ T=250 K, ∆log g = 0.5.

name α (J2000) δ (J2000) Teff log g [Fe/H](h m s) ( ’ ”) (K) (cgs)

RU Cen 12 09 23.7 -45 25 35 6000 1.5 -2.0AC Her 18 30 16.2 +21 52 00 5500 0.5 -1.5

interplay between the photospheric and circumstellar environment in these systems, we focusin this paper on two well known RV Tauri stars: AC Her and RU Cen.

AC Her has been extensively studied in the literature. It is a binary RV Tauri star of photomet-ric class a, with an orbital period of 1200 days (Van Winckel et al. 1998). It is a very regularpulsator with a formal pulsation period (timespan between two successive deep photometricminima) of 75.5 days (Zsoldos 1993). The mean magnitude of AC Her is mV = 7.69 magand the amplitude4mV = 2.31 mag (Lloyd Evans 1985). The presence of two strong shockwaves in every formal pulsation cycle, causing line-profile deformations in the spectra ofAC Her and R Scuti has been discussed by Gillet et al. (1989, 1990). AC Her shows a chem-ical depletion pattern (Van Winckel et al. 1998; Giridhar et al. 2000) that is attributed to thepresence of a stable Keplerian dusty disc. The presence of such a disc has also been pro-posed by Jura & Kahane (1999) who interpret the detection of weak CO rotational emissionlines with a small velocity width (Bujarrabal et al. 1988; Jura et al. 1995) as a signature ofsuch a long-lived reservoir. The presence of highly crystalline silicates in the infrared spec-trum (Molster et al. 1999) and the strong millimeter continuum flux from large dust grains(Shenton et al. 1995; Jura & Kahane 1999) further corroborate this conclusion. Jura et al.(2000) claimed to have resolved the circumstellar material around AC Her using N and Q-band imaging. Close et al. (2003) however detect no significant extended structure aroundAC Her, using adaptive optics at mid-infrared wavelengths with a higher spatial resolution.

RU Cen also is an RV Tauri star of spectroscopic class B and photometric class a. It is aregular pulsator with a pulsation period of 64.6 days, a mean magnitude of mV = 9.05 magand an amplitude of 4mV = 1.28 mag (Pollard et al. 1996). During every formal pulsationcycle two shock waves propagate through the atmosphere (Maas et al. 2002). The same pa-per reports the detection of a significant radial velocity variation on a longer time scale andattributes this to be due to orbital motion. No orbit could be determined, however.

Both stars are very regular RV Tauri pulsators with very similar chemical depletion patterns,atmospheric parameters and pulsational stability. In this paper we report our detailed com-parative study of both objects based on our optical monitoring and Spitzer Space Telescopespectra. The outline of the paper is as follows: in Section 2.2 we give an overview of thedifferent observations and reduction strategies. Section 2.3 contains the construction of thespectral energy distributions and colour excess determination. In Section 2.4 we discuss thespectral monitoring and deduce the binary model of RU Cen. The analysis of the infraredspectra and the spectral fitting is done in Section 2.5. In Section 2.6 we model the observed


SEDs using a passive disc model. The discussion of our different results and our conclusionsare presented in Section 2.7 and Section 2.8.

2.2 Observations and reduction

2.2.1 Spitzer

We used the Infrared Spectrograph (IRS) aboard the Spitzer Space Telescope in February2005 to obtain high- and low-resolution spectra of RU Cen. The spectra were observedwith combinations of the short-low (SL), short-high (SH), long-low (LL) and long-high (LH)modules. SL (λ=5.3-14.5 µm) and LL (λ=14.2-40.0 µm) spectra have a resolving power ofR=λ/ 4 λ ∼ 100 and the SH (λ=10.0-19.5 µm) and LH (λ=19.3-37.0 µm) spectra have aresolving power of ∼ 600.

The spectra were extracted from the SSC raw data pipeline version S13.2.0 products, us-ing the c2d reduction software package (Lahuis et al. 2006; Kessler-Silacci et al. 2006). Thisdata processing includes bad-pixel correction, extraction, defringing and order matching. In-dividual orders are corrected for offsets, if necessary, by small scaling corrections to matchthe bluer order.

2.2.2 TIMMI2

We completed our infrared spectra of RU Cen with spectra in the 10 µm region, taken withthe TIMMI2 instrument mounted on the 3.6 m telescope at the ESO La Silla Observatory inMarch 2004. The low-resolution (R ∼ 160) N band grism was used in combination with a1.2 arcsec slit, the pixel scale in the spectroscopic mode of TIMMI2 is 0.45 arcsec. For thereduction of the spectra we used the method described in van Boekel et al. (2005).

2.2.3 CORALIE

We extended the radial velocity monitoring reported by Maas et al. (2002) with new dataobtained with the same spectrograph CORALIE attached to the same 1.2 m Swiss Euler tele-scope. In total we accumulated 151 raw spectra between June 2000 and July 2006 with atypical sampling of 3 runs of 10 days spread over every semester. The radial velocity wasdetermined by off-line cross-correlation, using a spectral mask tuned to the spectral proper-ties of RU Cen (Maas et al. 2002). The internal error for every measurement was quantifiedby the standard deviation of a 50-point bisector through the cross-correlation profile. The bi-sector was determined on an equidistant sampling, starting from 2 times the width (σ) of theGaussian fit through the cross-correlation profile down to the minimum. From January 2005onwards, we used the HARPS-software release to determine the cross-correlation profile. Weextended the baseline by including the radial velocity data of Pollard et al. (1997), which arebased on high-quality high-resolution spectra.

2.3. SED determination 35

2.3 SED determination

We collected photometric data so as to construct the spectral energy distributions (SEDs)homogeneously (De Ruyter et al. 2005, 2006). The main problem in constructing such SEDsof pulsating stars with large amplitudes and often strong cycle-to-cycle variability is thatequally-phased data are necessary. Since we do not have these data available, photometricmaxima were used for the SED construction.

Our SED construction gives E(B−V ) = 0.4±0.3 for RU Cen and E(B−V ) = 0.2+0.3−0.2 for

AC Her. The colour excess E(B−V ) determination method was adopted from De Ruyter et al.(2006). We estimate total extinction by dereddening observed photometry, using the averageextinction law of Savage & Mathis (1979). Minimising the difference between the dered-dened observed fluxes and the adopted Kurucz model gives total colour excess E(B − V ).Model parameters for our programme stars (Table 2.1) are taken from Maas et al. (2002) andVan Winckel et al. (1998).

The SEDs of RU Cen and AC Her are also discussed in Maas et al. (2002) and De Ruyter et al.(2006), who find a total reddening for RU Cen of respectively E(B − V ) = 0.6 ± 0.1 andE(B − V ) = 0.3± 0.3 and a total reddening for AC Her of E(B − V ) = 0.1+0.3

−0.2.

For both stars we find a broad infrared excess, starting around L-M. We find a value for theenergy ratio LIR/L∗ ≈ 0.15 for RU Cen and LIR/L∗ ≈ 0.35 for AC Her.

2.4 Binary orbit of RU Cen

Together with AC Her, RU Cen is known to be one of the most regular RV Tauri star pulsators.It has a stable formal period of 64.60 days and a total peak-to-peak amplitude of 1.3 magni-tudes in V (Pollard et al. 1997). There is no long-term photometric modulation detected.

The pulsational modulation of the radial velocity is very significant (see Fig. 2.1) makingorbital detection far from straightforward. Any cycle-to-cycle variability will make that asystematic cleaning of the pulsation from the raw radial velocity data yields a residual. More-over, strong atmospheric shocks associated with the RV Tauri pulsations passing through theline-forming region (e.g. Gillet et al. 1990), have a strong effect on the line-profiles. Thismakes the determination of the stellar radial velocity at those pulsational phases problematic.In Figure 2.2 we show a few cross-correlation profiles at different phases in the pulsation cy-cle. The propagation of the shock is well illustrated and in the case of RU Cen the non-linearbehaviour leads to a very significant drop in velocity of more than 20 km s−1 over a smallphase interval. The shocks in RU Cen are so energetic that during these phases, He lines areobserved in emission (Maas et al. 2002).

Despite the strong pulsational modulation in the radial velocity data, variability in the radialvelocity is detected with a much longer time scale. We interpret this as being due to orbitalmotion and modelled this with a Keplerian model of a binary star.

To obtain the orbital elements, we only retained the high-quality data outside the pulsationphases where the strong shock is visible in the cross-correlation profile. To do so, we requiredthat in all data used for the orbital detection, the 50-point bisector has a variance of less


Figure 2.1 — The pulsational modulation of the radial velocity of RU Cen after cleaning of the orbit.The full line is the harmonic fit (one overtone included) with the pulsation period of 32.36 days. Zerophase was taken arbitrarily on the first datapoint obtained (JD2448388.93). The number of pulsationcycles covered is 170. The fractional variance reduction of the fit is 77%. Note the very high pulsationalamplitude.

Figure 2.2 — The cross-correlation profiles obtained at different pulsational phases. The upper leftpanel gives the cross correlation profile at phase 0.33 with the 50-point bisector which is used to quantifythe radial velocity. In the other panels, we witness the propagation of a strong shock in the line-formingregion of the atmosphere.

than 1 km s−1. We performed an iterative process on the raw velocity data in which wecleaned the orbital solution from the raw radial velocity data to obtain a good model of thepulsation cycle itself. We used PDM (Phase Dispersion Minimalisation method developedby Stellingwerf 1978) to quantify the fundamental pulsation period (time between successive

2.4. Binary orbit of RU Cen 37

Table 2.2 — The orbital elements of the binary RU Cen. P is the orbital period, JD the periastronpassage, K gives the semi-amplitude, e the eccentricity, ω is the longitude of periastron, γ is the systemradial velocity, a1 gives the semi-major axis of the primary and i is the inclination.

Element Value Formal Error (1σ) UnitP 1489 4 days

JDperiastron 2451378 10 daysK 21.9 0.7 km s−1

e 0.60 0.03ω 133 4 degreesγ -28.4 0.4 km s−1

rms 4.5 km s−1

Mass Function 0.83 M¯a1 sin i 2.4 AU

Figure 2.3 — The radial velocity data of RU Cen, cleaned for the pulsational modulation and folded onthe orbital period of 1489 days. The epoch of zero phase is the periastron passage (JD2451378). Thefull line represents the least square orbital solution discussed in the text. Note that for clarity the figuresamples all data twice.

deep and shallow photometric minima) and determined a harmonic fit with one overtone asa model description of the pulsation. We then cleaned the raw radial velocity data by thepulsation model and performed the next least-square fit of the orbit. We stopped the iterationwhen the changes in the orbital parameters became less than the error.

The final result is that we indeed found an orbital solution with a period of 1489±4 days. Theerrors given in the table are the formal errors obtained using the covariance matrix (Hadrava2004). The mass function is 0.83 M¯ and the semi-major axis is a1 sin i = 2.4 AU.


0 10 20 30 40Wavelength [µm]

0.0

0.5

1.0

1.5

2.0

Fν+

0.90

[Jy]

Hale-Bopp

ACHer

RUCen

Figure 2.4 — ISO-SWS spectra from the solar-system comet Hale-Bopp (Crovisier et al. 1997), theRV Tauri star AC Her (Molster et al. 1999) and combined TIMMI2 and Spitzer-IRS spectrum ofRU Cen. The spectra are normalised and offset for comparison. The dust-emission features on topof the continuum are identified as being due to enstatite and forsterite grains.

2.5 Analysis infrared spectra

2.5.1 General

A first look at the infrared spectra of AC Her and RU Cen shows their striking similarity (seeFig. 2.4), both in the global shape and in the dust emission features. In both spectra there isa lack of a strong 10 µm amorphous silicate feature, while the 20 µm amorphous feature isprominent. AC Her and RU Cen show strong emission features around 11.3− 16.2− 19.7−23.7−28−33.6 µm which we can identify as features of forsterite and enstatite, two abundantsilicate crystals. In none of the spectra is there evidence for a carbon-rich component. Notonly are the infrared spectra of AC Her and RU Cen very similar to each other, they alsoshow a strong resemblance to the infrared spectrum of the solar-system comet Hale-Bopp(Bouwman et al. 2003; Min et al. 2005b).

Min et al. (2005b) model the thermal emission and degree of linear polarisation of radiationscattered by grains in the coma of the comet Hale-Bopp. The largest contribution in dust,about 75% of total dust mass, is made up of amorphous silicate grains, with dust sizes from0.01 µm up to 93 µm, and large amorphous carbon grains (∼ 10 µm). Small crystallinesilicates make up only 5% of the total dust mass of Hale-Bopp but this is sufficient to havethis strong spectral signature in the IR spectrum.

2.5.2 Feature identification

The amorphous and crystalline features seen in the spectra of AC Her and RU Cen are iden-tified as caused by the most commonly found dust species in the circumstellar environment,namely glassy and crystalline silicates (Molster et al. 2002a,b,c; Min et al. 2007). The most

2.5. Analysis infrared spectra 39

commonly used amorphous silicates have an olivine or pyroxene stoichiometry. We furthertake the commonly used term “amorphous and crystalline olivine and pyroxene” to describethese dust species. Amorphous olivine (Mg2xFe2(1−x)SiO4, where 0 ≤ x ≤ 1 denotes themagnesium content) has very prominent broad features around 9.8 µm and 18 µm. These fea-tures (also called the 10 µm and 20 µm features) arise respectively from the Si-O stretchingmode and the O-Si-O bending mode. For large grains the 9.8 µm feature gets broader andshifts to redder wavelengths. Amorphous pyroxene (MgxFe1−xSiO3) shows a 10 µm featuresimilar to that of amorphous olivine, but shifted towards shorter wavelengths. Also the shapeof the 20 µm feature is slightly different. We cannot exclude the presence of other amorphoussilicate stoichiometries, for instance, Na-, Ca- or Al-based silicates. All these amorphous sil-icates have very similar emission profiles, making it hard to distinguish between them. In thisthesis however, we will only focus on the typical olivine and pyroxene descriptions.

Crystalline olivine and pyroxene have very distinct emission features and comparing withthe features seen in the spectra of AC Her and RU Cen, we conclude that the Mg-rich endmembers, forsterite (Mg2SiO4) and enstatite (MgSiO3), dominate our spectra. Forsteritecondenses directly from the gas phase at high temperatures (≈ 1500 K), or it may formby thermal annealing of amorphous silicates, diffusing the iron out of the lattice structure.Enstatite can form in the gas phase from a reaction between forsterite and silica, or it may alsoform by a similar thermal-annealing process as forsterite (Bradley et al. 1983; Tielens et al.1998b).

The observed spectra of AC Her and RU Cen show a shift from the amorphous 18 µm featuretowards 20 µm when comparing with synthetic spectra of amorphous olivine and pyroxene.This could point to the dominance of Mg-rich amorphous dust, which also shows this shift toredder wavelengths. Photospheric depletion in iron, which we detect in RU Cen and AC Her(Maas et al. 2002; Van Winckel et al. 1998), can be understood when the iron is locked upin the circumstellar dust (Waters et al. 1992). The lack of iron in the detected silicates istherefore surprising. If both the crystalline and amorphous silicates are devoid of iron, thiscould mean that iron is stored in metallic iron or iron oxide (Sofia et al. 2006). Metallic ironhas no distinct features but still a significant contribution in opacity, especially at shorterwavelengths, making it very hard to detect directly.

2.5.3 Profile fitting

Our aim is to fit the observed crystalline emission features of AC Her and RU Cen with syn-thetic spectra of forsterite and enstatite. The conversion from laboratory-measured opticalconstants of dust to mass absorption coefficients is not straightforward and is largely depen-dent on the adopted size, shape, structure and chemical composition of the dust (Min et al.2003, 2005a). These different dust approximations result in very different emission fea-ture profiles. The spectrum produced by homogeneous spherical particles is very differentfrom that produced by more irregular particles. This difference is much larger than the differ-ence between synthetic spectra computed using approximations of different irregular particles(Min et al. 2003). We have access to a large sample of mass absorption coefficients of variousdust shapes and sizes. The sample consists of forsterite and enstatite in Mie approximation(Aden & Kerker 1951; Toon & Ackerman 1981), CDE (continuous distribution of ellipsoids,


Figure 2.5 — Normalised and continuum-subtracted 33.6 µm emission feature of RU Cen together withmass absorption coefficients for different forsterite shape distributions. The DHS forsterite grains bothhave grain sizes of 0.1 µm. From the figure it is clear that the grains in DHS approximation must havefmax = 1.0. For this emission feature the difference between DHS grains of 0.1 µm and 1.5 µm isminimal.

Bohren et al. 1983), GRF (Gaussian random field particles, e.g. Grynko & Shkuratov 2003;Shkuratov & Grynko 2005) and DHS (distribution of hollow spheres, Min et al. 2003, 2005a)grain shapes. The DHS shaped particles are further characterised by the fraction of the totalvolume occupied by the central vacuum inclusion, f , over the range 0 < f < fmax. Thevalue of fmax reflects the degree of irregularity of the particles (Min et al. 2003, 2005a).Mie theory is used to model homogeneous spherical compact grains, while CDE, GRF andDHS particles are more irregular. Cross sections in CDE and GRF are computed under theassumption that the grains are in the Rayleigh limit (that the grains are much smaller thanthe wavelength of radiation, thus smaller than 0.1 µm). The different grain sizes for Mie andDHS dust particles range from 0.1 µm till 10 µm. These different grain sizes produce emis-sion features at very different central wavelengths (see Figs. 2.5 and 2.6) and larger grainsmainly contribute to the dust continuum.

To keep the number of free parameters in our χ2 minimisation reasonable, we first deter-mined and then fixed the best forsterite dust-opacity description. We carefully compared theobserved dust-emission profiles of RU Cen and AC Her with emission profiles of differentadopted forsterite shapes. From Figs. 2.5 and 2.6 it is clear that different dust shapes resultin very different emission profiles. We calculated that the best fit is obtained when usingsmall (< 0.1 µm) forsterite particles in CDE approximation and big (1.5 µm) forsterite par-ticles in DHS approximation with fmax = 1.0. The identification of the best enstatite dustspecies is not as straightforward since the enstatite emission features are often blended withmore prominent forsterite features. We have opted to keep the same grain-size distributionfor forsterite as for enstatite since this is physically more plausible.

Identifying the underlying dust continuum distribution is by no means straightforward. We


Figure 2.6 — Normalised and continuum subtracted 23.7 µm emission feature of RU Cen together withmass absorption coefficients for different forsterite shape distributions. The DHS forsterite grains bothhave fmax = 1.0. Grains of 1.5 µm are necessary to fit the observed bump at 25 µm.

opted to extract the dust continuum by connecting the local minima of the crystalline fea-tures. This method was used both for the synthetic spectra of forsterite and enstatite andthe observed spectra of RU Cen and AC Her to allow for a quantative comparison. The bestmodel fit was then determined using a χ2 minimalisation. Free parameters are dust species,the fraction of the given dust species, dust temperatures and the dust fractions at a given dusttemperature. This allows us to study the contribution of different dust temperatures and thedifferent dust species to the infrared spectral features.

The model emission profiles are then given by

Fλ ∼ (∑

i

αiκi) ∗ (∑

j

βjBλ(Tj))

where κi is the mass absorption coefficient of dust component i and αi gives the fractionof that dust component, Bλ(Tj) denotes the Planck function at temperature Tj and βj thefraction of dust in that given temperature. In this approach we assume that different grainspecies and grain sizes can have equal temperatures and that the observed flux originatesfrom an optically thin region.

We find that the best fit is obtained using very irregular grains, which was also found in recentstudies of dust in comets and protoplanetary discs (e.g. Crovisier et al. 1997; Bouwman et al.2001, 2003). As a next step, the dust species are kept fixed and the free parameters in the χ2

minimalisation are dust temperatures, temperature fractions and dust species fractions.

It is clearly not possible to fit the observed crystalline dust spectral features of RU Cen andAC Her with only one dust temperature (Fig. 2.7), since this fails to model all the observedemission features. The range of temperatures needed is, however, limited and good fits canbe achieved by allowing only two different dust temperatures between 50 K and 1000 K for


Table 2.3 — Dust parameters deduced from our spectral fitting for RU Cen and AC Her. Listed are thetwo dust temperatures, the fraction of dust with that given temperature and the different fractions of dustin the given dust species. Small forsterite/enstatite dust consists of CDE dust shapes with grain sizes< 0.1 µm, large forsterite/enstatite dust consists of DHS dust shapes of 1.5 µm sized dust particles withfmax = 1.0.

RU Cen AC HerT1=dust temperature 1 (K) 150 100T2=dust temperature 2 (K) 600 800β1=fraction dust in T1 0.40 0.60β2=fraction dust in T2 0.60 0.40α1=fraction small forsterite 0.40 0.50α2=fraction large forsterite 0.30 0.20α3=fraction small enstatite 0.00 0.10α4=fraction large enstatite 0.30 0.20

Figure 2.7 — Continuum-subtracted and normalised spectrum of RU Cen. Overplotted two models ofa forsterite-enstatite mixture at different temperatures. The dashed line represents a model at 150 K andthe dotted line a model at 600 K. It is clear that a single-temperature model is insufficient to reproducethe spectrum of RU Cen.

the adopted dust features (small/big forsterite/enstatite). An increase in the number of dusttemperatures yielded only a minor improvement in the χ2 minimisation.

Results of our model fit are given in Table 2.3 and Figs. 2.8 and 2.9. We find that both hot andcool dust are necessary to reproduce the observed spectra and that AC Her and RU Cen havea similar temperature distribution. The hot dust temperature is less constrained and similarfits could be derived with a temperature a few hundreds of Kelvin higher. All our best modelsgive a fraction of big enstatite grains that is higher than the fraction of small enstatite grains,


Figure 2.8 — Best model fit (dashed line) to the continuum-subtracted and normalised spectrum ofRU Cen. Fit parameters are given in Table 2.3.

Figure 2.9 — Best model fit (dashed line) to the continuum-subtracted and normalised spectrum ofAC Her. Fit parameters are given in Table 2.3.

sometimes the fraction of small grains is even zero. One explanation for the need of a highfraction of large grains is the rather broad 11.3 µm feature, which can only be fitted using aconsiderable fraction of large grains. Including a contribution of amorphous silicates in thisbroad feature makes the crystalline emission feature much narrower and more peaked around11.3 µm, which is distinctive for small forsterite grains. Of all the different observed emissionfeatures, the 11.3 µm feature is most sensitive to adopted grain size. Molster et al. (2002b)already found for AC Her that the 11.3 µm feature is well fitted by only crystalline silicates butthat an amorphous contribution could not be excluded. Including an amorphous component


would reduce the 11.3 µm feature and result in a larger fraction of small forsterite grains anda reduction in big enstatite grains. The maximum dust temperature of the forsterite-enstatitemixture decreases from temperatures around 700 K to around 500 K.

Our best models do not always match the observed features: in our modelled RU Cen spec-trum the 19 µm feature is stronger than the 23 µm feature, while this is not the case in theobserved spectrum of RU Cen. This problem could arise from a problem in the data reduction,since at around 20 µm, getting a good order overlap of the different echelle orders proved tobe problematic. Also the shape of the 27 µm feature is quite different in observed and mod-elled spectra. Another puzzling fact is that the 16.2 µm forsterite feature seems to be shiftedto the left and it is surprisingly strong in AC Her. This is not a temperature effect and it maycall for the inclusion of another mineral since in RU Cen this feature is not well reproducedeither.

2.6 SED fitting disc model

2.6.1 Method

Broad-band SEDs are notoriously degenerate but together with the Spitzer infrared spectralinformation, the circumstellar physical characteristics are much better constrained. A firmfirst conclusion is that all spherical models failed to fit both the SED and the infrared spectraldata. Moreover, spherical outflow models which fit the SED have evolutionary timescaleswhich are much too short compared to any evolutionary track in which a post-AGB star ofspectral type F is involved. We therefore concentrate on constructing detailed 2D disc models.

2.6.1.1 2D disc model

As a next step we performed an SED-fitting using a Monte-Carlo code, assuming 2D-radiativetransfer in a passive-disc model (Dullemond et al. 2001; Dullemond & Dominik 2004). Thiscode computes the temperature and density structure of the disc. The vertical scale height ofthe disc is computed by an iteration process, demanding vertical hydrostatic equilibrium. Thedust grain property distribution is fully homogeneous, and although this model can reproducethe SED, dust-settling timescales indicate that settling of large grains to the midplane occursand thus that an inhomogeneous disc model is necessary. Large grains are necessary to ac-count for the 850 µm flux in AC Her. As we do not possess a 850 µm fluxpoint for RU Cen,we estimated the 850 µm flux by assuming a blackbody slope from the IRAS 60 µm fluxredwards, as is observed in AC Her. This submillimeter data is invaluable to constrain grainsizes in the disc. In all other similar sources sampled (De Ruyter et al. 2006) the 850 µm fluxshows that the flux is at the Rayleigh-Jeans slope connecting the 60 µm IRAS flux point.

Dust settling time for a grain to migrate from height z0 to z is calculated using

tset =π

2Σ0

ρda

1Ωk

lnz

z0

with Σ0 the surface density, ρd the particle density, a the grain size and Ωk =√

GM∗r3 the

2.6. SED fitting disc model 45

Keplerian rotation rate (Miyake & Nakagawa 1995). From our SED modelling (Sect. 2.6.2)we find that the surface density of these discs approximately is Σ = Σ0( R

1 AU )−0.5, with Σ0 ≈0.2 gcm−2. This estimation means that grains larger than 100 µm can descend a distance of50 AU in less than 4× 105 years. Grains of 850 µm will even travel this distance in less than6× 104 years. This is similar to the estimated lifetime (Sect. 2.1) of the disc and means thatthere will be a vertical distribution in grain size, where the largest grains settle in the discmidplane, if no strong turbulence or stirring is present.

The disc structure and mid-IR flux is almost fully determined by the small grains while thecool midplane, consisting of large grains, mainly contributes to the long-wavelength part ofthe SED and the total mass of the disc. We construct an inhomogeneous 2-component modelwhere the near- and mid-IR flux comes from the small grains to which we add a blackbodyflux to represent the midplane.

Stellar input parameters of the model are luminosity, mass (which we take fixed at L =3000 L¯ and M = 1 M¯) and Teff . Input disc parameters are Rout and Rin, where the dustsublimation radius is used as a zero-order approximation of Rin, the total disc mass and thepower law for the density distribution. For the dust sublimation temperature for silicates weuse the typical value Tsub = 1500 K and we assume blackbody radiation. For RU Cen andAC Her this gives Rsub ≈ 2 AU. In the more detailed modelling we take the inner radius ofthe disc so that the start of the IR-excess fits the photometric data. Since we are not dealingwith outflow sources a power law > −2 is used. Using this disc model, the SED can now becalculated, given a specific inclination angle of the system.

2.6.2 Results

The SED-fitting gives an estimate of the distance to the systems, d = 2.4 kpc for RU Cenand d = 1.4 kpc for AC Her, using a luminosity L = 3000 L¯. This distance estimation islargely dependent on luminosity of the star and adopted inclination of the system. Indepen-dent distances estimated from the P-L relation of Alcock et al. (1998) are d = 1.6 ± 0.6 kpcfor RU Cen and d = 1.3±0.4 kpc for AC Her (De Ruyter et al. 2006). For RU Cen this couldmean that the adopted luminosity of L = 3000 L¯ is too high.

When modelling the near- and mid-IR part of the SED the feature-to-continuum ratio of thesilicate features is too strong in comparison with the ratio observed in the infrared spectra.Including an extra continuum opacity source is needed to reduce the strength of the features.Since the silicates are devoid of iron (Sect. 2.5.2), metallic iron is a potential opacity source:while its near-IR opacity is large, the absorption coefficient is unfortunately featureless sodirect detection is difficult. Inclusion of metallic iron has a strong impact on the modellingbecause the near-IR excess increases significantly with a given inner radius. The opacity ofmetallic iron alone would require an inner radius > 70 AU to maintain a reasonable fit tothe SED, but this is inconsistent with the constraints from high-spatial-resolution imaging(Close et al. 2003). So relatively large grains (up to 20 µm) were included in the modellingas well.

Our final model of the near- and mid-infrared part of the SED with a homogeneous discmodel consists of grains with sizes between 0.1 µm and 20 µm. In a disc, grain evolution canproduce larger grains than what is typically seen for ISM grains. A grain-size distribution


Figure 2.10 — SED disc modelling of RU Cen. The dashed line represents the homogeneous disc modelconsisting of grains between 0.1 µm and 20 µm. The solid line gives the disc model with an addedblackbody to represent the cool midplane. Crosses represent photometric data and we also overplottedthe observed Spitzer spectrum. Note that the 850 µm photometric point (asterisk) is an estimation.

Figure 2.11 — SED disc modelling of AC Her. The dashed line represents the homogeneous disc modelconsisting of grains between 0.1 µm and 20 µm. The solid line gives the disc model with an addedblackbody to represent the cool midplane. Crosses represent photometric data and we also overplottedthe observed ISO-SWS spectrum.

∼ ap (a is the grain size) with a power law −3.5 < p < −2.5 is expected (D’Alessio et al.2001). Bouwman et al. (2003) for example find for the Herbig Be star HD 100546, a grain-size distribution p ∼ −2 and p ∼ −2.8 for the comet Hale-Bopp. We take p = −3.0 forthe full range of grain sizes between 0.1 and 20 µm to still have a significant component of

2.7. Discussion 47

small grains. Further disc input parameters are adopted dust species, for which we take amixture of 6% metallic iron and 94% amorphous and crystalline silicates. Best models forthe SED-fit of RU Cen and AC Her are presented in Figures 2.10 and 2.11. For RU Cen wefind a model with Rin = 35 AU, Rout = 400 AU and a total disc mass (in small grains andgas) of 3 × 10−4 M¯. AC Her has a disc model with Rin = 35 AU, Rout = 300 AU and atotal disc mass (in small grains and gas) of 5× 10−4 M¯. For both stars we find that the bestfit is obtained when using a rather flat surface-density distribution (Σ ∼ Rα, with α > −1.0).To reduce the number of free parameters we kept the value of this power law fixed at −0.5.The inclusion of both metallic iron and larger grains cause a degeneracy in the value for theinner radius. If one would increase the fraction of large grains, as main contributors to thecontinuum opacity, the inner radius could have values between 15 AU and 30 AU. This is stillsignificantly larger than the sublimation radius for these stars (Sect. 2.6.1.1). The larger innerradius could be an evolutionary effect of the disc. Interferometric measurements are clearlyneeded to further constrain the disc radii. The added blackbody to represent the midplaneand explain the far-IR part of the SED has a temperature of 120 K for RU Cen and 170 K forAC Her.

These temperatures and the 850 µm fluxes can be used to estimate the dust mass in theselarge grains. In the optical thin approach (at 850 µm) the disc mass can be estimated by using(Hildebrand 1983)

Md =F850 D2

κ850 B850(T ).

Assuming a cross section of large spherical grains, the mass absorption coefficient κ =πa2

43 πa3ρ

, with a the grain size and ρ typically 3.3 g cm−3 for astronomical silicate, of 850 µm

grains in blackbody approximation is about 2.4 cm2 g−1. This results in dust mass estimatesof 5× 10−4 M¯ for RU Cen and 2× 10−4 M¯ for AC Her.

The resulting discs for RU Cen and AC Her are flared discs, with the scale height H ≈ 0.15×R1.2 (Fig. 2.12). This enables the discs to reprocess the large fraction, 15% for RU Cen and35% for AC Her (Sect. 2.3), of light emitted by the central star.

The models for RU Cen and AC Her have comparable discs, with similar geometries, scaleheights, temperature distributions and grain sizes. Temperature distributions in the discs arein agreement with dust temperatures derived from the spectral fitting. Both discs are opticallythick in the equator direction between 0.1 µm and 1 µm. After 1 µm the optical thicknessdecreases rapidly, making the disc optically thin for infrared emission.

2.7 Discussion

A major problem in the SED model fitting of the near- and mid-infrared flux is that the modelsare degenerate, especially the outer radius and the total disc mass are poorly constrained.Almost equally-fitting models for RU Cen and AC Her can be found with an outer radius afew hundred AU larger which would result in a larger total disc mass. It is also difficult todiscriminate between the two opacity sources we added to reduce the feature-to-continuumratio of the silicate features: metallic iron and larger grains. If grains up to 50 µm are includedin the disc, the amount of metallic iron needed to fit the spectrum will be reduced significantly.


0 100 200 300 400R(AU)

0

50

100

150

200

250

H(A

U)

Figure 2.12 — The scale height - radius relation H ≈ 0.15×R1.2, found for the discs of RU Cen andAC Her.

To compare the modelled SED with observational photometric data, these data need to becorrected for the interstellar extinction. A value for the interstellar extinction would constrainthe inclination of the system. Since no such estimate is available we put a minimum on theinterstellar extinction of E(B−V )is = 0.1. As a maximum value we use the total extinctionwhich we have deduced in Sect. 2.3. This gives us possible inclination values smaller than70 for RU Cen and smaller than 50 for AC Her.

With these values for the inclination a minimal mass for the companion star can be esti-mated, using the mass function and a typical value for the primary of M1 = 0.5 − 0.6 M¯.The mass functions of f(M) = 0.83 M¯ for RU Cen and f(M) = 0.25 M¯ for AC Her(Van Winckel et al. 1998) yield a minimal mass for the companion star of M2 = 1.7 M¯ forRU Cen and M2 = 1.1 M¯ for AC Her. Even the minimal mass of the companion of RU Cenis therefore not compliant with a possible white-dwarf mass, and the companions are in bothcases likely to be unevolved main-sequence stars.

2.8 Conclusion

AC Her and RU Cen are known to be prototypical RV Tauri pulsators which are normally seenas transition objects in their evolution from the AGB to the PNe evolutionary phase. Further-more RU Cen turned out to be an evolved binary with an orbital period similar to AC Her in-dicating that it must have been subject to severe binary interaction when at giant dimensions.Neither of the two currently fills its Roche Lobe. In this paper we focused on modelling thecircumstellar environment as constrained by our high quality Spitzer spectra and the broad-band SED. Our analysis showed that both stars are surrounded by a circumbinary dusty discin hydrostatic equilibrium. Since the disc modelling is a degenerate problem, interferometricmeasurements are needed to constrain further the disc geometry (e.g. Deroo et al. 2007a,b).

The observed Spitzer spectra clearly show that the circumstellar grains are extremely pro-

2.8. Conclusion 49

cessed. The IR spectrum of both objects is dominated by crystalline dust features. Themineralogy is magnesium-rich and large grains and/or metallic iron is necessary to explainthe low feature-to-continuum ratio of the silicate features. The temperature estimate of thecrystalline silicates shows that a significant fraction must be rather cool, and certainly wellbelow the annealing temperature. This shows that either radial mixing from the hot innerboundary (where annealing can take place) to far out in the disc must have occurred, or thatthe formation process and thermal history of the grains is quite different in discs than inoutflows. The profile fitting shows that the grains must be very irregular.

Despite the very different evolutionary history and the very different evolutionary timescalesinvolved, it is remarkable that the mineralogy of the small hot silicates around the evolvedobjects RU Cen and AC Her is extremely similar to what is found in some young stellarobjects (YSO) such as HD100546 (Malfait et al. 1998; Lisse et al. 2007) and that of primitivecomets such as Hale-Bopp (Bouwman et al. 2003; Lisse et al. 2007; Min et al. 2005b). In allthose systems, the thermal history of the grains has been such as to promote the dominanceof forsterite, the Mg-rich end member of the crystalline olivine family. In YSO as well asin comets, the dust processing is thought to be the tracer for the process of disc clearing andplanet building in the protoplanetary disc.

In evolved objects the circumstellar material is coming from the star themselves, and is muchmore chemically homogeneous than the ISM composition around YSO. Moreover the evo-lutionary timescales are likely to be orders of magnitude smaller than the disc evolution inYSO. Detailed dust mineralogy studies around evolved stars can therefore yield importantinformation on dust formation and dust processing in discs, and this under very differentchemical, dynamical and evolutionary environments than the processing in protoplanetarydiscs. Our study appears to indicate that the chemico-physical processes of dust grains inthe hydrostatic discs of the evolved binaries RU Cen and AC Her, are very similar to thosegoverning in the protoplanetary discs around YSO. We are in the process of expanding ourstudy to a wider sample of evolved binaries. This will enable us to build up a broader viewof the chemico-physical dust processes of the grains around rapidly evolving stars.

The inclination and mass function give a minimal mass for the companion star of 1.1 M¯for AC Her and 1.7 M¯ for RU Cen. The unevolved companions have masses such that theywould normally (on single-star evolutionary tracks) evolve to carbon stars on the AGB. Giventhe silicate dominated circumstellar dust, it is however, clear that even the primaries, whowere more massive, did not evolve to become carbon stars. The actual orbits combined withthe observed chemical evolution of the stars show that the binary interaction processes dom-inate the final evolution of these objects. The internal chemical evolution seems to have beencut short by binary interaction. As such the stars do not evolve on single-star evolutionarytracks and both objects should be seen as remarkably evolved binaries. It is likely that theformation of the circumstellar dust in these objects is closely related to binary interaction.

Both objects show that the formation, structure and evolution of the circumstellar discsplay a leading role in their final evolution and illustrate once again the intimate relationbetween the depleted photospheres and the presence of a circumbinary disc. Since post-AGB stars with similar SEDs and/or chemistry are abundant (e.g. De Ruyter et al. 2006;Reyniers & Van Winckel 2007) it is clear that disc formation is a process relevant in the lateevolution of a considerable fraction of binary stars.


AcknowledgementsThe authors want to acknowledge: the Geneva Observatory and its staff for the generous time alloca-tion on the Swiss Euler telescope; the 1.2 m Mercator staff as well as the observers from the Instituutvoor Sterrenkunde who contributed to the monitoring observations using both the Euler and Mercatortelescopes. CG acknowledges support of the Fund for Scientific Research of Flanders (FWO) under thegrant G.0178.02. and G.0470.07. We also thank Fred Lahuis for his assistance with the Spitzer datareduction.

Chapter 3Mineralogy of dust around evolvedstars

This chapter is published as

Spitzer survey of dust grain processing in stable discs around binary post-AGB stars.

C. Gielen, H. Van Winckel, M. Min, L.B.F.M. Waters and T. Lloyd Evans, A&A, 2008,490, 725

Abstract:AIMS: We investigate the mineralogy and dust processing in the circumbinary discs ofbinary post-AGB stars using high-resolution TIMMI2 and Spitzer infrared spectra.METHODS: We perform a full spectral fitting to the infrared spectra using the mostrecent opacities of amorphous and crystalline dust species. This allows for theidentification of the carriers of the different emission bands. Our fits also constrainthe physical properties of different dust species and grain sizes responsible for theobserved emission features.RESULTS: In all stars the dust is oxygen-rich: amorphous and crystalline silicatedust species prevail and no features of a carbon-rich component can be found, theexception being EP Lyr, where a mixed chemistry of both oxygen- and carbon-richspecies is found. Our full spectral fitting indicates a high degree of dust grainprocessing. The mineralogy of our sample stars shows that the dust is constituted ofirregularly shaped and relatively large grains, with typical grain sizes larger than2 µm. The spectra of nearly all stars show a high degree of crystallinity, wheremagnesium-rich end members of olivine and pyroxene silicates dominate. Other dustfeatures of e.g. silica or alumina are not present at detectable levels. Temperatureestimates from our fitting routine show that a significant fraction of grains must becool, significantly cooler than the glass temperature. This shows that radial mixing

52 Chapter 3. Mineralogy of dust around evolved stars

is very efficient is these discs and/or indicates different thermal conditions at grainformation. Our results show that strong grain processing is not limited to youngstellar objects and that the physical processes occurring in the discs are very similarto those in protoplanetary discs.

3.1 Introduction

Post-AGB stars are low- and intermediate-mass stars that evolve rapidly from the AsymptoticGiant Branch (AGB) towards the Planetery Nebula (PNe) phase, before cooling down aswhite dwarves. During the preceding AGB phase, the star has undergone severe mass loss,leaving behind a slowly expanding circumstellar dust shell. Depending on the dilution inthe line-of-sight, the expected spectral energy distribution (SED) is double-peaked, with thefirst UV-visible peak indicative of the central star and the second infrared peak coming fromthermal emission of the cool dust in the circumstellar environment (CE). The release of theIRAS All-Sky Survey mission made a large-scale identification of post-AGB stars on the basisof their infrared colours possible (Lloyd Evans 1985; Hrivnak et al. 1989; Oudmaijer et al.1992). More recent surveys on post-AGB stars include Szczerba et al. (2007), who presentthe latest catalogue of Galactic post-AGB stars with almost 400 objects.

IRAS colour-colour diagrams revealed a large sample of stars that did not show the ex-pected double-peaked SED (Lloyd Evans 1985; De Ruyter et al. 2006). Instead, they showa strong near-IR excess, pointing to the presence of hot dust in the system, while the cen-tral stars are currently too hot to have an ongoing dusty mass loss. Correlations with opticaldatabases of variable stars point to the presence of pulsating stars that also show this specificinfrared excess. These RV Tauri stars are luminous evolved stars that cross the Population IICepheid instability strip and populate a well-defined part of the IRAS colour-colour diagram(Lloyd Evans 1985; Raveendran 1989; Lloyd Evans 1999). Recent interferometric studies(Deroo et al. 2007a,b) prove that the circumstellar emission originates from a very compactregion and that the SEDs of these stars are well modelled with a passive 2D disc model(Dullemond & Dominik 2004; Deroo et al. 2007a, and Chap. 2). The spectral slope of thesubmillimetre SED points to the presence of large, up to µm and cm sized, grains in thesediscs (De Ruyter et al. 2006, and Chap. 2), which have settled to form a cool disc midplane.The inner radius of these discs is determined by the dust sublimation radius. The hot innerrim is puffed up by gas pressure from the central star and radiates mainly in the near-IR, whilethe outer parts of the disc can be strongly flared and are responsible for the strong absorptionand re-radiation of the stellar light. In a few cases the direct confirmation of the Kepleriankinematics of the circumstellar discs comes from resolved interferometric CO measurements(Bujarrabal et al. 1988, 2005, 2007). Recent studies have shown that these post-AGB starsare very likely all binaries (Maas et al. 2002, 2003; Van Winckel et al. 1999; Van Winckel2007, and Chap. 2).

Famous examples of these objects include HD 44179, the central star of the Red Rectangle.It is a carbon-rich post-AGB object for which the binarity was proven by Van Winckel et al.(1995). The star is surrounded by an oxygen-rich disc (Waters et al. 1998) and is resolvedin ground-based high-spatial resolution imaging (Roddier et al. 1995; Men’shchikov et al.

3.1. Introduction 53

2002) as well as in HST optical images (Cohen et al. 2004). The disc is also resolved ininterferometric CO measurements, where the Keplerian velocity of the disc was detected(Bujarrabal et al. 2005).

Infrared spectroscopy is the ideal tool to study the physico-chemical characteristics of thecircumstellar material, since it samples resonances of the most dominant dust species andimportant ro-vibrational bands of dominant molecules. Spectroscopic data obtained withthe Infrared Space Observatory (ISO) allowed for the first time to study the mineralogyof circumstellar environments of objects in different evolutionary stages (e.g. Waters et al.1996; Waelkens et al. 1996b). More advanced studies of the CE of young stellar objects (e.g.van Boekel et al. 2005; Lisse et al. 2007) and mass-losing objects (Molster et al. 2002a,b,c)further identified the dominant dust species and grain sizes of the circumstellar dust.

Infrared spectral studies of post-AGB objects (Van Winckel 2003, and references therein)allowed for the detection of different CE chemistries. The more typical C-rich post-AGBstars are characterised by a strong amorphous SiC feature at 11.3 µm. In these stars, a strongunidentified 21 µm feature can sometimes be found (Kwok et al. 1989). In addition to thedetection of O-rich post-AGB stars, dominated by silicate species, surprisingly, also mixedchemistries were found, showing features of both O-rich and C-rich dust species.

Young stellar objects are the natural environment to study circumstellar disc physics, andprevious studies of the mineralogy of discs around the few brightest infrared evolved objectswith ISO (Molster et al. 1999, 2002a,b,c) show strong dust processing, with oxygen-rich andhighly-crystalline dust. Molster et al. (2002a,b,c) found that a high degree of crystallisationis indicative for the presence of a stable disc, and not a dusty outflow.

In our pilot study (Chap. 2), based on Spitzer-IRS and ISO data, we investigate the mineralogyand spectral energy distribution of two post-AGB stars, RU Cen and AC Her. The spectra ofthese stars are extremely similar and show a strong resemblance to the infrared spectrumof the solar-system comet Hale-Bopp and young stellar objects, such as HD 100546. Theobserved crystalline emission features are well modelled using magnesium-rich crystallinesilicates. The grains are irregularly shaped, with grain sizes of 0.1 µm and 1.5 µm. Bothhot and cool grains are necessary to reproduce the observed spectrum. The spectral energydistributions of both stars were fitted using a 2D passive disc model (Dullemond & Dominik2004). The discs surrounding these objects start at 35 AU from the central star, which is wellbeyond the dust sublimation radius. The outer radius is less well constrained, but extends tillabout 300 AU. The discs are strongly flared.

In this work we study the mineralogy of the discs in a large sample of post-AGB binariesand perform a detailed fitting of the observed emission features. We have observed 21 binarypost-AGB objects with the Spitzer-IRS spectrograph and look for relations between the dustparameters and stellar characteristics.

The outline of this paper is as follows: in Sections 3.2 and 3.3 we introduce our programmestars and the different observation and reduction strategies used. Section 3.4 contains theconstruction of the spectral energy distributions and the total extinction determination. InSect. 3.5 we give a general overview of the spectra and the observed emission features. Theprofile and full spectral fitting is presented in Sections 3.6 and 3.7. The discussion and con-clusion of our analysis are given in Sections 3.8 and 3.9.


3.2 Programme stars

From the total sample of 51 published binary post-AGB stars likely surrounded by a stabledisc (De Ruyter et al. 2006), we selected the 21 stars with fluxes below the saturation limit ofthe Spitzer-IRS spectrograph. Spectral types range from A to M.

Our radial velocity monitoring programme, using the CORALIE spectrograph attached to the1.2 m Swiss Euler Telescope at the ESO La Silla Observatory, is still ongoing but we havealready found orbital parameters for 15 of our 21 program stars. Orbital periods range from100 up to more than 2000 days (Van Winckel 2007). The objects for which we do not yethave orbital parameters, have pulsational amplitudes which are too large for a straightforwarddetection of the binary motion. The results of the radial velocity monitoring program will besubject of a forthcoming paper.

The orbital periods indicate that strong binary interaction must have occurred during thelate stages on the AGB. The binaries are now not in contact but the orbits are too small toaccommodate a full-grown AGB star. The discs are all circumbinary since all orbits lie wellwithin the dust sublimation radius. The distribution of the mass functions gives a range inminimal mass for the companion between 1 M¯ and 2 M¯. The companion stars are likelyunevolved main-sequence stars (e.g. Van Winckel 2007, and Chap. 2).

These binary post-AGB stars are characterised by a depletion pattern in their photospheres(Giridhar et al. 1994, 1998, 2000; Gonzalez et al. 1997a,b; Van Winckel et al. 1998; Maas et al.2005). This abundance pattern is the result of gas-dust separation followed by reaccretion ofthe gas, which is poor in refractory elements. Waters et al. (1992) proposed that the mostlikely circumstance for this process to occur is when the dust is trapped in a circumstellardisc. Photospheric chemical studies of post-AGB candidates in the LMC have revealed thatthere also, depletion patterns are common (Reyniers & Van Winckel 2007). The observationsin the recent release of the SAGE database (Meixner et al. 2006) showed that these depletedLMC sources have infrared excesses similar to the Galactic binary post-AGB stars, and aretherefore thought to be disc sources as well.

The disc formation itself is badly understood. Possible formation scenarios of the discs in-clude a wind-capture scenario (e.g. Mastrodemos & Morris 1999), or a formation scenariothrough non-conservative mass transfer in an interacting binary. In the first scenario, theAGB wind is captured by the companion. In the second scenario, which is still not very wellexplored theoretically, the disc formation precedes the dust-grain formation and it is likelythat the thermal history of the grains was very different from that in normal AGB winds. Thiscould lead to very different chemico-physical properties during formation. The sizes of theorbits suggest this scenario to be more likely.

We may witness the formation of a circumbinary disc by Roche-Lobe overflow in the evolvedbinary system SS Lep. The optical/IR interferometric observables can be best understood,assuming a Roche-lobe filling M star in a system with a circumsystem disc (Verhoelst et al.2007).

Our programme stars are therefore all likely binaries which are surrounded by a dusty stabledisc. This disc seems to have an important impact on the objects and in this paper we focuson the analysis of the IR spectra as probes for the disc physics.

3.2. Programme stars 55Ta

ble

3.1

—T

hena

me,

equa

tori

alco

ordi

nate

sα

and

δ(J2

000)

,eff

ectiv

ete

mpe

ratu

reT

eff

,sur

face

grav

itylo

gg

and

met

allic

ity[F

e/H

]ofo

ursa

mpl

est

ars.

For

the

mod

elpa

ram

eter

sw

ere

fer

toD

eR

uyte

reta

l.(2

006)

.A

lso

give

nis

the

orbi

talp

erio

d(s

eere

fere

nces

inD

eR

uyte

reta

l.20

06,a

ndC

hap.

2).

The

tota

lred

deni

ngE

(B−

V) t

ot,t

heen

ergy

ratio

LIR

/L∗

and

the

calc

ulat

eddi

stan

ce,a

ssum

ing

alu

min

osity

ofL∗

=5000±

2000

L¯

.D

ista

nces

mar

ked

with

*ar

eup

perl

imits

due

toth

eas

pect

angl

eof

the

disc

.

N

Nam

eα

(J20

00)

δ(J

2000

)T

eff

log

g[F

e/H

]P

orb

itE

(B−

V) t

ot

LIR

/L∗

d(h

ms)

(’”

)(K

)(c

gs)

(day

s)(%

)(k

pc)

1E

PLy

r19

1817

.5+

2750

3870

002.

0-1

.50.

52±0

.01

3±0

4.1±

0.8

2H

D13

1356

1457

00.7

−68

5023

6000

1.0

-0.5

1490

0.20±0

.01

50±2

3.0±

0.6

3H

D21

3985

2235

27.5

−17

1527

8250

1.5

-1.0

259

0.27±0

.01

24±1

3.1±

0.6

4H

D52

961

0703

39.6

+10

4613

6000

0.5

-4.8

1310

0.06±0

.01

12±1

2.1±

0.4

5IR

AS

0520

8−20

3505

2259

.4−2

032

5340

000.

50.

023

60.

00±0

.00

38±2

3.9±

0.8

6IR

AS

0906

0−28

0709

0810

.1−2

819

1065

001.

5-0

.537

10.

57±0

.02

63±3

5.4±

1.1

7IR

AS

0914

4−49

3309

1609

.1−4

946

0657

500.

5-0

.517

701.

99±0

.05

53±5

2.7±

0.6

8IR

AS

1017

4−57

0410

1918

.1−5

719

3632

39

IRA

S16

230−

3410

1626

20.3

−34

1712

6250

1.0

-0.5

0.56±0

.02

60±3

6.1±

1.2

10IR

AS

1703

8−48

1517

0736

.3−4

819

0847

500.

5-1

.513

810.

22±0

.02

69±5

4.5±

1.0

11IR

AS

1724

3−43

4817

2756

.1−4

350

4862

500.

50.

048

40.

59±0

.02

68±4

3.8±

0.8

12IR

AS

1912

5+03

4319

1500

.8+

0348

4177

501.

0-0

.551

71.

08±0

.02

52±3

1.8±

0.4

13IR

AS

1915

7−02

4719

1822

.5−0

242

0977

501.

00.

012

0.5

0.68±0

.01

63±2

4.2±

0.9

14IR

AS

2005

6+18

3420

0754

.8+

1842

5758

500.

7-0

.40.

51±0

.02

905±

4210

.9±2

.315

RU

Cen

1209

23.7

−45

2535

6000

1.5

-2.0

1489

0.55±0

.01

13±1

2.3±

0.5

16SA

O17

3329

0716

08.3

−23

2702

7000

1.5

-0.8

115.

90.

39±0

.01

36±1

6.5±

1.3

17ST

Pup

0648

56.4

−37

1633

5750

0.5

-1.5

410

0.00±0

.00

55±1

5.7±

1.2

18SU

Gem

0614

00.8

+27

4212

5750

1.12

5-0

.70.

58±0

.02

111±

74.

8±1.

0*19

SXC

en12

2112

.6−4

912

4160

001.

0-1

.060

00.

32±0

.02

34±2

3.8±

0.7

20T

WC

am04

2048

.1+

5726

2648

000.

0-0

.50.

40±0

.02

42±3

3.2±

0.6

21U

YC

Ma

0618

16.4

−17

0235

5500

1.0

0.0

0.00±0

.00

89±3

9.6±

2.0*


3.3 Observations and data reduction

3.3.1 Spitzer

High- and low-resolution spectra for 21 post-AGB stars were obtained using the InfraredSpectrograph (IRS; Houck et al. 2004) aboard the Spitzer Space Telescope (Werner et al.2004) in February 2005. The spectra were observed using combinations of the short-low(SL), short-high (SH) and long-high (LH) modules. SL (λ=5.3-14.5 µm) spectra have a re-solving power of R=λ/4λ ∼ 100 , SH (λ=10.0-19.5 µm) and LH (λ=19.3-37.0 µm) spectrahave a resolving power of ∼ 600. Exposure times were chosen to achieve a S/N ratio ofaround 400 for the high-resolution modes, which we complemented with short exposures inlow-resolution mode with a S/N ratio around 100, using the first generation of the exposuretime calculator of the call for proposals.

The spectra were extracted from the SSC raw data pipeline version S13.2.0 products, us-ing the c2d reduction software package (Kessler-Silacci et al. 2006; Lahuis et al. 2006). Thisdata processing includes bad-pixel correction, extraction, defringing and order matching. In-dividual orders are corrected for offsets, if necessary, by small scaling corrections to matchthe bluer order.

3.3.2 TIMMI2

For stars where we lack the Spitzer IRS-SH observations we obtained additional ground-based N-band infrared spectra with the Thermal Infrared Multi Mode Instrument 2 (TIMMI2,Reimann et al. 2000; Kaufl et al. 2003), mounted on the 3.6 m telescope at the ESO La SillaObservatory. The low-resolution (R ∼ 160) N band grism was used in combination with a1.2 arcsec slit, the pixel scale in the spectroscopic mode of TIMMI2 is 0.45 arcsec. For thereduction of the spectra we used the method described in van Boekel et al. (2005). We scaledthe TIMMI2 spectra to the Spitzer spectra and found a very good agreement in spectral shapebetween the two data sets.

3.4 Spectral energy distribution

For all sample stars spectral energy distributions were updated from the photometric data asgiven by De Ruyter et al. (2006). The resulting SEDs are presented in Figs. 3.2 and 3.3. Thetotal extinction E(B−V )tot was determined by dereddening the observed photometry, usingthe average extinction law of Savage & Mathis (1979). The relation between E(B−V )tot andAV is given by AV = RV ×E(B−V ), with a typical value for RV = 3.1 (Savage & Mathis1979). Minimising the difference between the dereddened observed optical fluxes and the ap-propriate Kurucz model (Kurucz 1979) gives the total colour excess E(B−V )tot (Table 3.1).Model parameters for our sample stars are given in Table 3.1 and are based on the analysis ofhigh-resolution spectra as given by the literature (see De Ruyter et al. (2006) for references).A considerable fraction of our sample stars show a photometric variability due to pulsation,so we use only photometric maxima for the SED construction.

3.4. Spectral energy distribution 57

Figure 3.1 — Comparison between the dereddened (solid line) and reddened (dotted line) ofIRAS 19125. Dereddening mainly influences the spectral signature in the 10 µm region.

The errors on the value for E(B − V )tot are calculated using a Monte-Carlo simulationon the photometric data. We use an error of 0.05 mag for the photometric measurementsin a Gaussian distribution. Since we do not know the distances to the sources, we adopt alikely luminosity for evolved low-gravity objects of L∗ = 5000 ± 2000 L¯. The distance toIRAS 20056 is likely overestimated since visible light from this source probably reaches usonly by scattering through a nearly edge-on optically-thick disc (Menzies & Whitelock 1988;Gledhill et al. 2001).

We also corrected the infrared spectra for reddening by extending the average extinction lawof Savage & Mathis (1979) with the theoretical extinction law of Steenman & The (1989,1991), and by using the derived total extinction values. This is done under the assumptionthat the extinction is fully due to interstellar extinction. Since the total extinction probablyconsists of both an interstellar and a circumstellar component, the applied dereddening is thusa maximal correction.

Dereddening the infrared spectra mainly influences the shape of the silicate feature around9.8 µm (see Fig. 3.1). Since our total extinction values are on average quite small, the effectof deredding the infrared spectra is negligible for most stars. For stars with a higher value forE(B−V )tot, like IRAS 09144 and IRAS 19125, one has to be careful interpreting the shapeof the silicate feature, since the partitioning between interstellar and circumstellar reddeningis not known at this point.

We compute the energy ratio LIR/L∗ to determine the amount of energy reprocessed by theCE (Table 3.1). 80% of our stars have a ratio LIR/L∗ of 30% and higher, showing that theabsorption and re-radiation of stellar light by the CE is extremely efficient.


Figure 3.2 — The SEDs of our sample stars. The dereddened fluxes (diamonds) are given together withthe scaled photospheric Kurucz model (solid line). In gray we plot the corresponding TIMMI2 andSpitzer spectra.


Figure 3.3 — See previous caption.



3.5 General overview

Figures 3.15 to 3.17 show the wide range in observed spectra. All spectra are characterised bymany distinct spectral structures in emission over a smooth continuum. There is quite somevariety over the sample, with different continuum slopes, but for nearly all stars the spectrumis dominated by emission bands that can be attributed to silicate dust species.

In a few stars only, gas-phase emission is detected in the form of bandhead emission of 12CO2

and 13CO2. These stars are HD 52961, EP Lyr and IRAS 10174.

Unique to our sample is the spectrum of EP Lyr in the 7-20 µm region. This spectral rangeis dominated by a strong and narrow 11.3 µm emission peak and a complex plateau withnarrow emission features in the 14-18 µm region. The strongly asymmetric band between7 µm and 10 µm is real as well. These features are observed in C-rich evolved stars andin the C-rich component of the ISM, and are associated with PAH emission and the CH-out-of-plane and C-C-C bending modes (e.g. Van Kerckhoven et al. 2000; Hony et al. 2001;Peeters et al. 2002). Longward of 18 µm, the spectrum is dominated by silicate emission.EP Lyr is the only object in our sample displaying a mixed chemistry. It is remarkable that inour considerable sample of post-AGB stars, only one possible post-carbon star is found!

3.5. General overview 61

3.5.1 Identification of silicate dust features

Amorphous and crystalline silicates are the most commonly found dust species in the inter-stellar and circumstellar environment (Molster et al. 2002a,b,c; Kemper et al. 2004; Min et al.2007). To describe the glassy and crystalline silicates with an olivine and pyroxene sto-ichiometry we will use the commonly used term “amorphous and crystalline olivine andpyroxene”. Amorphous olivine (Mg2xFe2(1−x)SiO4, where 0≤x≤1 denotes the magnesiumcontent) has very prominent broad features around 9.8 µm and 18 µm (see Fig. 3.5), whichare easily detected in our spectra. These features (also called the 10 µm and 20 µm features)arise respectively from the Si-O stretching mode and the O-Si-O bending mode. For largegrains the 9.8 µm feature gets broader and shifts to redder wavelengths. Amorphous pyrox-ene (MgxFe1−xSiO3) shows a 10 µm feature similar to that of amorphous olivine, but shiftedtowards shorter wavelengths. Also the shape of the 20 µm feature is slightly different.

The Mg-rich end members of crystalline olivine and pyroxene, forsterite (Mg2SiO4) andenstatite (MgSiO3), show strong but narrow features at distinct wavelengths around 11.3 −16.2−19.7−23.7−28−33.6 µm (see Fig. 3.5), making them easily identifiable in our spectra.Forsterite condenses directly from the gas phase at high temperatures (≈ 1500 K) or it mayform by thermal annealing of amorphous silicates, diffusing the iron out of the lattice struc-ture. Enstatite can form in the gas phase from a reaction between forsterite and silica (SiO2),or it may also form by a similar thermal annealing process as forsterite (Bradley et al. 1983;Tielens et al. 1998b). Laboratory experiments have indicated that silica can be formed whenamorphous silicates anneal to forsterite (e.g. Fabian et al. 2000). No evidence for the pres-ence of silica, with strong features around 9 and 21 µm, can be found in our spectra. There isalso no evidence for the presence of Fe-rich crystalline silicates, like fayalite (Fe2SiO4).

Nearly all stars show strong crystalline dust features, both at short and long wavelengths,showing that both hot and cool crystallines seem to be abundant.

In our spectra, the amorphous silicate dust seems to peak at 20 µm, rather than 18 µm, aswould be expected from synthetic spectra of amorphous silicates. This could point to thepresence of Mg-rich amorphous dust, which also shows this shift to redder wavelengths (seeFig. 3.5).

3.5.2 Mean spectra and complexes

Different complexes at 10, 14, 16, 19, 23 and 33 µm can be identified in our spectra. In orderto study the systematics between these complexes in our spectra we compare them to a meanspectrum in that region. The mean-complex spectra were obtained by adding continuum-subtracted spectra of sample sources with clear spectral structures, using a weighing factorproportional to the S/N in that spectral region. The continuum was determined by linearlyinterpolating between the beginning and end of the studied region. The mean complex spectraare shown in Figures 3.9 to 3.13. We also plot the mass absorption coefficients of forsteriteand enstatite in every complex.


Figure 3.5 — The mass absorption coefficients [cm2/g] of forsterite and enstatite in GRF approxima-tion. Grain sizes of 0.1 µm (full line), 2.0 µm (dashed line) and 4.0 µm (dotted line) are plotted. Thethird box shows olivine in 0.1 µm GRF approximation. The Mg-rich member (Mg2SiO4) is plotted infull line, the standard Mg-Fe member (MgFeSiO4) is plotted in dot-dashed line. As an example we alsoplotted the Spitzer spectrum of one of our sample stars, ST Pup.

3.5.2.1 The 10 µm complex (8− 13 µm)

In Fig. 3.6 we plot the continuum-divided spectra in the 10 µm region, ordered by peak value.We plot 1+Fν,cs/ < Fν,c >, where Fν,cs is the continuum-subtracted spectrum (Fν −Fν,c)and < Fν,c > is the mean of the continuum. The continuum was determined by linearlyinterpolating between 7.5 and 13 µm. This method preserves the shape of the emission bandand allows a good comparison between the profiles themselves. The figure clearly illustratesthe very wide variety of the spectral appearance of the warm silicates.

In Fig. 3.7 we show the continuum-subtracted flux at 11.3 and 9.8 µm versus the peak-to-continuum ratio of the 10 µm silicate feature. The 11.3/9.8 µm ratio is a measure for theamount of processing that the dust has undergone, where the peak/continuum ratio is a mea-sure for the typical grain size, iron content or other opacity sources. We find a high degree ofcrystallisation (11.3/9.8 µm ratio) for all sources, but rather small peak-to-continuum values,compared to the typical ISM profile. EP Lyr is a clear outlier with a very strong 11.3/9.8 µmratio, due to a very sharp peak at 11.3 µm which is due to PAH emission. No clear correlation,and thus no evolutionary trend, is observed between the shape and the strength of the 10 µm


Figure 3.6 — Continuum-divided spectra in the 10 µm region, ordered by peak value. The peak valueincreases from left to right and from bottom to top. The numbering corresponds with the star numbersas given in Table 3.1. For SU Gem and TW Cam we lack flux points in the 7.5 µm region to perform agood continuum determination.

feature in our sample stars, however. This is in contrast with the clear correlation that is seenin young stellar objects (e.g. van Boekel et al. 2003, 2005), where stars with a strong featurehave low 11.3/9.8 µm ratios and stars with a weaker silicate feature show higher 11.3/9.8 µmratios. Sources that display a strong emission feature show rather unprocessed silicate band,similar to the ISM profile, with little evidence for crystalline grains. Sources with a weak10 µm profile show a broader and flatter silicate feature, dominated by larger grains.

In Fig. 3.8 we compare the 10 µm complexes of our different sample stars with the cal-culated mean 10 µm complex. On overall the stars show good agreement with the meanspectrum, although some stars show a slightly stronger amorphous 9.8 µm feature over thecrystalline 11.3 µm feature. RU Cen however shows a very strong 11.3 µm feature. Clearoutliers are EP Lyr, HD 52961, IRAS 10174 and IRAS 20056. EP Lyr shows no silicate fea-


Figure 3.7 — continuum-subtracted flux at 11.3 and 9.8 µm versus the peak-to-continuum ratio of the10 µm silicate feature. SU Gem and TW Cam are not plotted since for these stars we lack a goodcontinuum determination.

tures but strong PAH emission instead. HD 52961 does show an amorphous and crystallinefeature but different in shape and central wavelength than the other stars. These two starswill be discussed in detail in Chapter 4. IRAS 10174 only shows an amorphous feature, butagain of different in shape and central wavelength. IRAS 20056 shows two strongly peakedfeatures at 9.8 and 11.3 µm, but it is not clear whether these can be explained with standardamorphous and crystalline silicates.

3.5.2.2 The 14 µm complex (13− 15 µm)

In Fig. 3.9 the 14 µm complexes are plotted. The mean spectrum has two strong emissionpeaks, around 13.7 and 14.7 µm. The profile at 13.7 µm seems to blend of two peaks, ofwhich one can be identified as enstatite. The small feature in the mean spectrum at 14.3 µmis also due to enstatite. The feature around 14.7 µm remains unidentified. It seems to be madeup of a broader feature from 14.3 µm till 15 µm, with a strong narrow peak at 14.7 µm. Inthe sample stars, there is some variation in the ratio between the two strong peaks at 13.7 and14.7 µm, meaning they are probably due to different dust species.

3.5.2.3 The 16 µm complex (15− 17 µm)

In Fig. 3.10 the 16 µm complexes are plotted. The mean spectrum shows a clear broad featurearound 16 µm. The shape and strength is very similar to the forsterite feature at 16.2 µm,with a contribution of enstatite at 15.3 µm. On top of this broad band more narrow features at15.9 µm and 16.2 µm can be distinguished. The small peaks at 15.4 µm and 16.2 µm are dueto CO2 gas emission, which can only be clearly seen in EP Lyr, HD 52961 and IRAS 10174.


Figure 3.8 — The 10 µm complex, continuum-subtracted and normalised. Overplotted in gray the meanspectrum. The mass absorption coefficients of forsterite and amorphous olivine are plotted in dotted anddashed lines respectively. TW Cam and SU Gem are not plotted since their spectrum is not complete inthis wavelength region.

Most sample sources show a similar profile as the mean spectrum, although the ratio betweenthe two main contributors sometimes differs, due to a different enstatite/forsterite ratio. Starswith a weak 16 µm feature (like TW Cam) show a very clear separation between the threefeatures, with widths of about 0.3 µm. This shows that the mean spectrum probably consistof a very broad forsterite feature topped with more narrow features of another dust (or gas?)species. HD 52961 has a strong profile that is shifted bluewards in comparison to the meanspectrum, with clear CO2 emission lines.

3.5.2.4 The 19 µm complex (17− 21 µm)

In Fig. 3.11 the 19 µm complexes are plotted. The very broad profile is again a blend offorsterite and enstatite. The sometimes very sharp feature around 19.7 µm is a data reduc-tion artefact. Most stars have the same profile shape, except for HD 52961, EP Lyr andIRAS 10174. The first two are clearly outliers in our sample stars, since they have verydistinct spectra, very different from other sample stars. Looking at the Spitzer spectrum ofIRAS 10174, this star has almost no crystalline features and the observed complex is actuallydue to poor continuum subtraction and normalisation.


Figure 3.9 — The 14 µm complex, continuum-subtracted and normalised. Overplotted in gray themean spectrum. The mass absorption coefficients of forsterite and enstatite (in CDE approximation)are plotted in dotted and dashed lines respectively.

3.5.2.5 The 23 µm complex (20− 27 µm)

In Fig. 3.12 the 23 µm complexes are plotted. The mean profile is dominated by forsteriteand there is an extremely good agreement between the mean spectrum and sample stars.Remarkable little variation is detected in the whole sample.

3.5.2.6 The 33 µm complex (31− 37 µm)

In Figure. 3.13 the 33 µm complexes are plotted. The observed profile is mainly due toforsterite at 33.6 µm. At 32.5 µm another feature can be observed. The ratio 32.5/33.6 µmvaries slightly from source to source, indicating that a different dust species is responsible forthe feature at 32.5 µm. In ST Pup, the feature is clearly broader than in other sample sources.The bump at 35.7 µm is not a reliable result since it sits at the end of the spectrum.

3.6. Analysis: profile fitting 67


3.6 Analysis: profile fitting

To fit the observed emission features with synthetic spectra, mass absorption coefficients fordifferent dust species need to be calculated. The conversion from laboratory-measured opticalconstants of dust, to mass absorption coefficients is not straightforward and is largely depen-dent on the adopted size, shape, structure and chemical composition of the dust (Min et al.2003, 2005a). These different dust approximations result in very different predicted emis-sion feature profiles, with a clear division between the synthetic spectra of homogeneous andirregular grains. If one would assume homogeneous spherical particles, one could use Mietheory (Aden & Kerker 1951; Toon & Ackerman 1981). However, cosmic dust grains are notperfect spheres, so we have to reside to other methods. Examples of irregular dust approxi-mations are CDE (continuous distribution of ellipsoids, Bohren et al. 1983), GRF (Gaussianrandom field particles, e.g. Grynko & Shkuratov 2003; Shkuratov & Grynko 2005) and DHS(distribution of hollow spheres, Min et al. 2003, 2005a) grain shapes. The DHS shaped par-ticles are further characterised by the fraction of the total volume occupied by the centralvacuum inclusion, f , over the range 0 < f < fmax. The value of fmax reflects the degree ofirregularity of the particles (Min et al. 2003, 2005a). Cross sections in CDE are computed un-der the assumption that the grains are in the Rayleigh limit (that the grains are much smaller



than the wavelength of radiation, thus smaller than 0.1 µm), and do not allow the study ofgrain growth. The spectra for the GRF particles were computed using the Discrete DipoleApproximation (Draine 1988). More details on the particle shapes and the computationsof the spectra can be found in Min et al. (2007, 2008). Refractory indices for the materialsused are taken from Servoin & Pirou (1973); Dorschner et al. (1995); Henning & Stognienko(1996); Jaeger et al. (1998).

Not only grain shape, but also grain size, has a profound influence on the mass absorption co-efficients. Since a circumstellar disc is the perfect environment for grain growth to occur, wealso study the effect of different grain sizes on the observed emission profiles. Different grainsizes produce emission features at different central wavelengths and with different profiles,as shown in Figure 3.5. With increasing grain size, emission features will become weakerand eventually disappear, leaving mainly a contribution to the thermal continuum. This effectis already considerable for grain radii a > 2 µm. Previous studies (Bouwman et al. 2001;Honda et al. 2004) in the 10 µm region have shown that the variety of emission features canbe described using two typical grain sizes: generally 0.1 µm to describe the grains with radii< 1 µm, and 1.5-2 µm for larger grains.

Comparing the observed crystalline emission features with calculated synthetic spectra, we

3.6. Analysis: profile fitting 69


find that irregular grains are needed to explain the dust profiles (Fig. 3.14). When we look atthe prominent forsterite features at 16− 19− 23.7− 33.6 µm, we find a good fit is achievedby using grains in CDE, GRF or DHS (with f = 1.0). None of the tested dust shapesproves an accurate fit to the, in some stars very strong, forsterite 16 µm feature, which seemsto be shifted to bluer wavelengths in the observed spectra. This was also seen in the discspectra of Molster et al. (2002a). The best fit to this feature is achieved using dust in GRFapproximation. The fit of the 23.7 µm feature is strongly improved when using large grainsin DHS. The GRF approximation does not result in a good fit to the 33.6 µm feature, which isthe purest forsterite feature, since at other wavelengths the features may be blended with e.g.enstatite emission features. Both CDE and DHS reproduce the shape of this feature, but sincethe CDE approximation does not allow us to study the effect of large grains we use GRF andDHS approximations in our modelling routine.

A similar approach to determine the best enstatite or amorphous silicate dust approximationsis not straightforward. Enstatite emission features are mostly blended with forsterite features,but in the 14 µm region there are some small unblended features. These are ideal features totrace the enstatite fraction of the grains. The central wavelengths of these peaks are at 13.8,14.4 and 15.4 µm, with the 15.4 µm feature being the most prominent. The sample sourcesthat show clear emission in this region all show the 13.8 µm feature and in lesser degree



the 15.4 µm feature. The 14.4 µm feature is absent but a rather strong unidentified featureappears around 14.8 µm. In order to keep the fitting homogeneous we will also use GRF andDHS dust approximations to describe the amorphous and crystalline pyroxene dust species.

3.7 Full spectral fitting

3.7.1 Method

For an exact spectral modelling, full 2D radiative transfer in a realistic disc model should bestudied. Such models are not yet available so as a first approximation we assume the emissionfeatures to be formed in the thin surface layer of the disc. Assuming the flux originates froman optically-thin region, which is motivated by the fact we see the dust features in emission,we can make linear combinations of the absorption profiles to calculate the model spectrum.

This approach is a first approximation but allows for a general study of trends in dust shapes,grain sizes and processing in these discs. Disc models like the one of Dullemond & Dominik(2004), which we used in our pilot study to fit the SED of RU Cen and AC Her, do not allow

3.7. Full spectral fitting 71

Figure 3.14 — Normalised and continuum-subtracted emission features of the mean of our samplestars, together with mass absorption coefficients for different forsterite shape distributions. CDE grainsare plotted in red, GRF in blue and DHS in green. 0.1 µm grains are plotted in full lines, 2 µm grains indashed.

for a detailed spectral modelling. The code computes the temperature structure and density ofthe disc. The vertical scale height of the disc is computed by an iteration process, demandingvertical hydrostatic equilibrium. Other important processes, like dust settling and turbulentmixing, probably also occur in these discs but are not included. The code also does notinclude an independent dust species and grain-size distribution throughout the disc, makinga detailed mineralogy study impossible.

Our model emission profiles are then given by

Fλ ∼ (∑

i

αiκi)× (∑

j

βjBλ(Tj))

where κi is the mass absorption coefficient of dust component i and αi gives the fractionof that dust component, Bλ(Tj) denotes the Planck function at temperature Tj and βj thefraction of dust in that given temperature. The temperature of dust grains will depend ongrain size and grain shape, as well as the distance to the central star and the integrated line-of-sight opacity from the stellar surface to the grains. This can only be calculated using fullradiative transfer, so we assume all grains to have the same dust temperatures, irrespective ofsize and shape.

In this modelling the continuum is given as a sum of Planck functions and we take it as another


dust component with a constant mass absorption coefficient. To keep the number of freeparameters in our fitting routine reasonable, we allow only two different dust temperaturesand continuum temperatures (between 100 and 1000 K), with different ratios. A fit withthree temperatures was also tested and proved only a minor improvement. The total numberof fit parameters in the model is thus 15, with contributions of four silicate species, wherewe use two grain sizes, the continuum mass absorption coefficient, two dust and continuumtemperatures which each their relative fractions.

To fit the spectra we minimise the reduced χ2 of the full Spitzer spectrum, extended with theIRAS 60 µm flux point to constrain the continuum at larger wavelengths, given by

χ2 =1

N −M

N∑

i=1

∣∣∣∣Fmodel(λi)− F observed(λi)

σi

∣∣∣∣2

,

with N the number of wavelength points λi, M the number of fit parameters and σi theabsolute error on the observed flux at wavelength λi. The errors σi represent the statisticalnoise on the spectra. They are generated so that the errors are proportional to the square rootof the flux and they are scaled in such a way that the best fit has a reduced χ2 of approximately1.

The errors on the fit parameters are calculated using a Monte-Carlo simulation. We randomlyadd Gaussian noise with a distribution of width σi at each wavelength point. This generates100 synthetic spectra, all consistent with our data, on which we perform the same fitting pro-cedure. This results in slightly different fit parameters for which we calculate the mean, ourbest-fit value, and the standard deviation. The uncertainty on the mass absorption coefficientsare not taken into account in the χ2-minimalisation.

We find that the mass absorption coefficient of small and large amorphous grains are quitesimilar, this degeneracy could introduce a large error on the derived fractions for small-largeamorphous grains.

3.7.2 Results

We tested both the DHS and the GRF dust approximations in the fitting routine, where theGRF approximation proved a far stronger match. In order to test for the presence of Fe-pooramorphous dust, as postulated in Section 3.5.1, we perform the fitting both with pure Mg-richamorphous silicates (x = 1, see Sect. 3.5.1) and with the more standard Mg-Fe amorphoussilicate dust (x = 0.5).

When using 0.1 µm and 2.0 µm grain sizes, we find that nearly 80% of the dust resides in the2.0 µm grains. The amount of grain growth in crystalline material versus amorphous materialis plotted in Figure 3.18. Almost all crystalline grains seem to be large, with a fraction oflarge grains in crystalline component higher than 0.7. For the amorphous grains most sourceshave fractions between 0.3 and 0.9. No correlation between grain growth in crystalline andamorphous grains is seen. An efficient removal of the smallest grains must have occurred inthese discs or grain growth was the dominant factor to reduce the number of small grains atgrain formation.


Figure 3.15 — Best model fits for our sample stars, showing the contribution of the different dustspecies. The observed spectrum (black curve) is plotted together with the best-model fit (red curve) andthe continuum (black solid line). Forsterite is plotted in dash-dot lines (green) and enstatite in dash-dot-dotted lines (blue). Small amorphous grains (2.0 µm) are plotted as dotted lines (magenta) and largeamorphous grains (4.0 µm) as dashed lines (magenta).

Because of this lack of small grains, as a next step we ram models using grain sizes of 2.0 µmand 4.0 µm. This resulted in a better fit for 17/21 stars, giving slightly better values for χ2. ForEP Lyr and HD 52961 the quality of the fit decreased considerably. This was to be expected,


Figure 3.18 — The mass fraction of large grains in the amorphous component versus the mass fractionof large grains in the crystalline component, using the fitting with grain sizes of 0.1 µm and 2.0 µm.

since the strong narrow crystalline features in these stars already indicated the presence ofsmall grains.

In general the fit with the x = 1.0 Mg-rich amorphous silicates proved the best, be it only aminor improvement. The observed trends in the fit parameters and stellar characteristics donot change significantly between the fits when using x = 1.0 or x = 0.5 amorphous dust. Ofour sample stars, 6 stars show a better fit using the x = 0.5 amorphous silicate dust.

The best model spectra, overplotted with the observed spectra are shown in Figures 3.15-3.17. The resulting values of the best fit parameters are given in Table 3.2 and Table 3.3. Thequality of our fitting is generally very good, 70% of stars have a χ2 < 5.

Some trends can be observed (e.g. Figs. 3.15-3.17). As already explained in Section 3.6, theshape of the 33.6 µm feature is not well reproduced, although the strength is well modelled.Also the 29 µm feature seems to be more prominent in the observed spectra, than in the modelfits. The 19 µm feature is slightly overestimated in the model spectra, while the neighbouring23 µm feature is slightly underestimated. This discrepancy could be due to a data reductioneffect, since in this region there can be a bad overlap between the short and long Spitzer-IRShigh-resolution bands.

The clear outliers in our modelling are EP Lyr and HD 52961 with very high values of χ2.These stars have unusual features in their observed spectra, including strong narrow featuresand CO2 emission lines. A detailed study of these outliers will be given in Chapter 4.

Nearly all sources show the presence of both hot and cool dust, both in dust temperature asin continuum temperature. As seen in Table 3.2, dust temperatures can differ strongly fromcontinuum temperatures. The modelling now assumes the amorphous and crystalline dust tohave the same temperature. Dropping this constraint could increase the crystallinity fractioneven more. Since iron is more heated more efficiently than magnesium, this would give theamorphous dust a higher temperature. This implies we need to increase the crystalline dustto get a similar fit to the observed features.

3.7. Full spectral fitting 77Ta

ble

3.2

—B

estfi

tpar

amet

ers

dedu

ced

from

our

full

spec

tral

fittin

g.L

iste

dth

eχ

2,d

usta

ndco

ntin

uum

tem

pera

ture

san

dth

eir

rela

tive

frac

tions

(βj

inth

efo

rmul

agi

ven

inSe

ct.3

.7).

N

Nam

eχ

2T

du

st1

Tdu

st2

Frac

tion

Tcon

t1T

con

t2Fr

actio

n(K

)(K

)T

du

st1

-Tdu

st2

(K)

(K)

Tcon

t1-T

con

t2

1E

PLy

r56

.111

697

16

2372

77

37

0.80

0.1

00.3

0−

0.20

0.3

00.1

019

545

100

9218

0342

0.98

0.0

10.0

2−

0.02

0.0

20.0

1

2H

D13

1356

3.5

2262

69

46

9287

3187

0.80

0.1

00.5

0−

0.20

0.5

00.1

019

512

104

5349

550

0.92

0.0

30.0

2−

0.08

0.0

20.0

3

3H

D21

3985

4.1

1854

587

9267

4333

0.80

0.1

00.4

0−

0.20

0.4

00.1

018

713

91

8387

3162

0.98

0.0

10.0

0−

0.02

0.0

00.0

1

4H

D52

961

72.2

1991

0140

8451

21

106

0.90

0.0

00.7

0−

0.10

0.7

00.0

010

9497

999

82 139

0.99

0.0

00.0

3−

0.01

0.0

30.0

0

5IR

AS

0520

84.

529

1176

101

9049

7124

0.80

0.1

00.3

0−

0.20

0.3

00.1

017

730

79

3742

676

0.84

0.0

20.0

3−

0.16

0.0

30.0

2

6IR

AS

0906

03.

620

6176

675

5140

163

0.90

0.0

00.3

0−

0.10

0.3

00.0

022

586

122

8051

62

198

0.93

0.0

20.0

2−

0.07

0.0

20.0

2

7IR

AS

0914

46.

121

291

24

5351

37

78

0.90

0.0

00.2

0−

0.10

0.2

00.0

019

616

105

7678

989

0.94

0.0

20.0

2−

0.06

0.0

20.0

2

8IR

AS

1017

413

.923

472

125

3784

779

0.70

0.2

00.4

0−

0.30

0.4

00.2

012

576

25

3229

522

0.96

0.0

20.0

4−

0.04

0.0

40.0

2

9IR

AS

1623

04.

921

0143

22

5071

82

74

0.90

0.0

00.2

0−

0.10

0.2

00.0

015

1154

51

5922

86

93

0.94

0.0

10.0

3−

0.06

0.0

30.0

1

10IR

AS

1703

82.

925

0140

69

8541

20

192

0.80

0.1

00.2

0−

0.20

0.2

00.1

019

82 139

5638

485

0.96

0.0

10.0

2−

0.04

0.0

20.0

1

11IR

AS

1724

32.

321

0110

38

4418

979

0.80

0.1

00.5

0−

0.20

0.5

00.1

020

2113

15

6331

68

39

0.91

0.0

30.0

1−

0.09

0.0

10.0

3

12IR

AS

1912

53.

910

2139

220

4277

40.

900.0

00.1

0−

0.10

0.1

00.0

032

1169

138

7021

26

84

0.84

0.0

40.0

4−

0.16

0.0

40.0

4

13IR

AS

1915

75.

519

926

111

6091

76

159

0.90

0.0

00.4

0−

0.10

0.4

00.0

018

614

89

6898

5137

0.96

0.0

10.0

2−

0.04

0.0

20.0

1

14IR

AS

2005

63.

811

884

18

2198

319

0.60

0.2

00.6

0−

0.40

0.6

00.2

023

887

51

6581

19

72

0.88

0.0

10.0

4−

0.12

0.0

40.0

1

15R

UC

en3.

423

387

67

5172

00

115

0.80

0.1

00.4

0−

0.20

0.4

00.1

019

91 054

577

125

0.98

0.0

10.0

2−

0.02

0.0

20.0

1

16SA

O17

3329

3.1

2051

60

36

7641

79

133

0.80

0.1

00.5

0−

0.20

0.5

00.1

019

55 106

5812

3100

0.93

0.0

10.0

2−

0.07

0.0

20.0

1

17ST

Pup

8.4

2079

913

4808

686

0.80

0.1

00.2

0−

0.20

0.2

00.1

020

799

13

5241

89

57

0.95

0.0

20.0

3−

0.05

0.0

30.0

2

18SU

Gem

1.8

2413

36

98

6292

84

184

0.80

0.1

00.3

0−

0.20

0.3

00.1

017

723

79

7631

37

192

0.95

0.0

20.0

2−

0.05

0.0

20.0

2

19SX

Cen

4.3

2252

02

44

9188

3183

0.80

0.1

00.3

0−

0.20

0.3

00.1

019

19 96

6599

580

0.94

0.0

20.0

2−

0.06

0.0

20.0

2

20T

WC

am2.

321

6123

82

3901

07

171

0.70

0.2

00.4

0−

0.30

0.4

00.2

013

1321

31

5462

05

46

0.95

0.0

10.0

2−

0.05

0.0

20.0

1

21U

YC

Ma

2.9

2051

04

12

7611

31

113

0.90

0.0

00.0

0−

0.10

0.0

00.0

020

3119

351

1126

11

0.84

0.0

30.0

2−

0.16

0.0

20.0

3


Tabl

e3.

3—

Bes

tfitp

aram

eter

sde

duce

dfr

omou

rfu

llsp

ectr

alfit

ting.

The

abun

danc

esof

smal

l(2.

0µ

m)

and

larg

e(4

.0µ

m)

grai

nsof

the

vari

ous

dust

spec

ies

are

give

nas

frac

tions

ofth

eto

tal

mas

s,ex

clud

ing

the

dust

resp

onsi

ble

for

the

cont

inuu

mem

issi

on.

The

last

colu

mn

give

sth

eco

ntin

uum

flux

cont

ribu

tion,

liste

das

ape

rcen

tage

ofth

eto

tali

nteg

rate

dflu

xov

erth

efu

llw

avel

engt

hra

nge.

N

Nam

eO

livin

ePy

roxe

neFo

rste

rite

Ens

tatit

eC

ontin

uum

Smal

l-L

arge

Smal

l-L

arge

Smal

l-L

arge

Smal

l-L

arge

1E

PLy

r0.

177.5

90.1

7−

5.61

46.8

25.6

119

.223

4.2

317.1

0−

26.9

840.9

624.4

220

.423

2.7

814.2

5−

7.80

27.6

37.4

63.

1726.2

13.1

3−

16.6

329.1

614.5

858

.268

.69

8.6

3

2H

D13

1356

1.58

25.2

91.5

7−

29.7

515.7

721.0

912

.991

8.6

111.3

4−

30.4

721.4

725.3

513

.931

0.9

17.2

9−

3.63

18.7

33.5

70.

495.8

50.4

9−

7.17

17.4

96.8

277

.783

.10

2.7

2

3H

D21

3985

1.51

21.4

91.5

2−

28.8

114.6

221.1

614

.121

5.9

412.0

8−

31.3

727.4

022.9

07.

6410.2

05.5

8−

8.50

17.3

37.6

00.

556.5

30.5

6−

7.50

10.0

76.3

175

.053

.51

5.5

3

4H

D52

961

0.21

15.7

50.2

1−

0.59

45.9

90.5

954

.401

6.4

230.8

8−

7.00

55.0

57.0

417

.331

5.2

212.9

9−

11.5

322.9

710.6

42.

2621.8

72.2

6−

6.69

32.3

16.4

964

.914

.75

6.8

3

5IR

AS

0520

81.

9116.8

81.9

0−

7.39

12.8

16.5

929

.091

1.1

913.8

6−

9.42

25.2

18.9

723

.048

.08

10.0

5−

2.70

17.4

72.6

95.

7010.3

55.2

5−

20.7

513.5

815.6

067

.642

.74

4.2

7

6IR

AS

0906

07.

2024.6

67.0

4−

19.6

716.6

817.2

034

.081

3.7

318.5

0−

9.12

32.1

19.0

214

.037

.88

6.3

3−

5.10

15.9

94.9

71.

2510.0

11.2

4−

9.55

10.3

07.1

771

.992

.60

3.4

6

7IR

AS

0914

41.

0523.7

51.0

6−

32.8

615.7

019.8

120

.982

1.4

514.7

6−

30.2

527.1

023.3

88.

398.0

55.4

2−

2.96

10.2

52.8

90.

266.4

50.2

6−

3.23

10.7

43.1

872

.993

.48

3.2

5

8IR

AS

1017

421

.582

7.4

818.7

2−

28.9

420.1

024.3

127

.361

5.2

215.1

6−

17.8

121.5

215.5

50.

856.8

80.8

4−

1.70

9.5

31.6

60.

688.7

20.6

8−

1.09

7.0

91.0

935

.706

.39

5.7

2

9IR

AS

1623

03.

1330.3

73.1

3−

43.7

519.0

830.3

44.

1512.0

43.9

6−

16.5

618.2

615.2

418

.021

1.4

58.8

9−

9.20

22.7

47.5

50.

000.0

00.0

0−

5.19

20.9

14.9

577

.274

.17

2.6

6

10IR

AS

1703

80.

7956.9

90.8

0−

22.9

819.8

917.8

64.

0213.0

53.9

5−

24.5

618.2

017.7

819

.331

1.0

711.2

4−

10.2

523.9

39.8

01.

318.9

71.3

1−

16.7

618.1

412.8

881

.622

.00

2.7

8

11IR

AS

1724

38.

6231.1

38.4

1−

31.9

917.5

124.7

122

.071

6.6

913.7

0−

13.1

428.7

712.0

516

.369

.05

8.4

4−

1.27

15.3

21.2

50.

124.4

10.1

2−

6.43

13.4

55.3

683

.091

.90

2.5

3

12IR

AS

1912

510

.682

4.6

39.7

9−

9.13

19.9

48.3

815

.452

1.1

111.3

2−

27.5

725.1

821.8

59.

046.5

25.2

3−

10.3

512.0

26.7

41.

5411.4

41.5

3−

16.2

411.0

49.7

472

.703

.73

3.2

3

13IR

AS

1915

79.

1440.2

59.1

2−

48.9

920.7

436.6

09.

6713.4

77.7

9−

12.1

725.8

711.6

910

.719

.97

7.7

0−

5.13

16.7

05.0

00.

188.9

90.1

8−

4.02

14.1

63.7

880

.172

.85

4.8

0

14IR

AS

2005

66.

1227.2

46.0

8−

20.8

423.7

317.1

827

.052

0.0

219.6

1−

14.0

135.3

413.2

511

.599

.43

7.2

5−

5.04

15.7

04.8

61.

6710.6

61.6

7−

13.6

915.2

911.0

581

.362

.73

3.3

2

15R

UC

en0.

9816.0

00.9

8−

22.3

716.2

317.5

25.

7411.4

95.3

2−

19.0

317.4

716.4

927

.791

4.3

011.4

6−

10.3

121.4

48.7

21.

0712.3

41.0

7−

12.7

017.0

910.1

080

.232

.89

3.3

4

16SA

O17

3329

1.07

23.2

51.0

7−

41.5

218.1

028.7

14.

1920.4

54.1

5−

20.9

318.3

616.7

79.

1011.3

67.0

3−

7.62

24.1

27.2

72.

0625.1

22.0

4−

13.5

221.7

810.4

083

.302

.81

2.2

9

17ST

Pup

2.28

32.7

02.2

9−

26.0

715.6

317.2

79.

3318.1

98.5

3−

36.9

118.5

927.3

210

.696

.40

6.9

4−

9.14

19.1

98.2

80.

4311.6

00.4

3−

5.15

14.6

54.4

956

.074

.32

4.1

4

18SU

Gem

3.52

43.0

33.5

1−

45.5

621.9

732.8

44.

6820.7

44.4

6−

12.3

921.5

611.2

321

.231

6.8

411.6

0−

5.50

19.7

85.2

20.

519.9

20.5

1−

6.61

17.5

75.9

786

.852

.62

6.3

8

19SX

Cen

3.69

31.5

63.7

0−

40.5

316.9

423.3

012

.071

5.1

510.1

1−

23.0

722.3

920.9

39.

007.3

55.2

4−

4.15

16.2

24.0

80.

847.8

00.8

4−

6.64

12.7

25.7

775

.853

.10

3.5

1

20T

WC

am2.

2644.4

52.2

7−

56.5

226.8

144.4

60.

020.0

00.0

2−

4.65

38.7

64.6

317

.293

3.3

811.4

5−

8.64

40.9

18.5

00.

8213.0

00.8

2−

9.79

35.3

28.7

990

.612

.12

2.1

4

21U

YC

Ma

0.94

30.9

50.9

4−

14.7

514.5

811.3

18.

6318.6

97.6

4−

45.3

617.5

025.1

215

.639

.36

8.5

3−

8.54

16.8

87.8

10.

335.6

60.3

3−

5.82

15.9

45.0

679

.112

.10

1.9

7


3.7.3 Correlations

To gain insight in the dust formation process and evolution in the circumstellar environmentof our sample stars we look for trends in the derived fit parameters and correlations with dustand stellar parameters.

3.7.3.1 Mineralogy correlations

In Figure 3.19 the mass fraction in large grains is plotted against the mass fraction in crys-talline grains. Mass fractions are calculated as fractions of the total dust mass, excluding thedust responsible for the continuum. Note that small grains now indicate the 2.0 µm grainsizes and large grains the 4.0 µm grains.

85% of our sources show a mass fraction in large grains above 0.5. None of the sourceshave a mass fraction in large grains below 0.25. Most sources have a mass fraction in crys-talline grains between 0.1 and 0.6, strongly centred around 0.3. There is one clear outlier:IRAS 10174 with a degree of crystallinity of 0.05. This source shows the more standard ISMprofile, dominated by amorphous grains. No clear correlation can be found.

The fraction of large grains in the crystalline component has values between 0.25 and 0.7,while the fraction of large grains in the amorphous component has values ranging from 0.3till 0.95. Grain growth appears to be more efficient in the amorphous dust component. Nocorrelation between grain growth in crystalline and amorphous grains is seen however.

The fraction of enstatite grains in the crystalline component lies between 0.1 and 0.5, showingforsterite to be the dominant crystalline species. No correlation between the enstatite fractionand crystallinity can be found.

In Figure 3.20 we plot the continuum-to-dust luminosity ratio of our observed spectra againstthe mass fraction in large grains. We assume here that the dust component is only responsiblefor the features and not the underlying continuum. A correlation can be found between thecontinuum/dust ratio and the mass fraction in large grains. Sources with a high fractionof large grains show a high value for the continuum/dust ratio. This could be expected sincelarger grains show less prominent emission features and have a larger continuum contribution.A high value for the continuum/dust ratio also indicates the presence of even larger grains inthe disc. This indicates that the abundance of these larger grains correlates with the abundanceof the 4 µm grains, and thus that the size distribution continues beyond 4 µm grains.

3.7.3.2 Central star correlations

By comparing the derived fit parameters with stellar characteristics, like Teff , LIR/L∗ andthe orbital period, we can investigate possible evolutionary trends in our sample. However,no such correlation between fit parameters and stellar parameters are found. This was alsoseen in the SEDs (Sect. 3.4), where nearly all stars show a similar SED, irrespective of thecentral star. Our results do not show an obvious evolutionary trend in the mineralogy of oursample stars.


Figure 3.19 — The mass fraction in large grains plotted against the mass fraction in crystalline grains,as derived from our best fit parameters. A high degree of crystallinity is found.

Figure 3.20 — The continuum to dust ratio of the observed spectra plotted against the mass fraction onlarge grains.

3.8 Discussion

The mineralogy of our sample stars show that the dust is purely O-rich. Amorphous andcrystalline silicate dust species prevail and no features of a C-rich component are found,except the PAH emission feature seen in EP Lyr.

This is remarkable since some of our sample stars are thought to have initial masses whichwould make them evolve to carbon stars on the AGB on single-star evolutionary tracks. Thelack of third dredge-up is also seen in the two objects with clear CO2 gas emission lines.

3.8. Discussion 81

EP Lyr has a 12C/13C ratio of ≈ 9 (Gonzalez et al. 1997b), illustrating that 12C is not en-riched during the preceding AGB evolution. The internal chemical evolution of our sourcesseems to have been cut short by binary interaction processes.

EP Lyr is the only star in our sample which also has a mixed chemistry, and strongly resem-bles HD 44179. Like HD 44179, EP Lyr also is strongly depleted (Gonzalez et al. 1997b).PAH emission features dominate the spectrum untill 20 µm, at larger wavelengths crystallinesilicates start to dominate. The observed broad and asymmetric emission feature at 8.2 µm isvery similar to the “class C” objects as discussed in Peeters et al. (2002). The famous “EggNebula” also shows this “class C” 8.2 µm emission feature. EP Lyr, together with HD 52961,will be studied in detail in Chapter 4.

Some binary post-AGB stars with oxygen-rich discs are known that did evolve to carbon stars,like HD 44179, which has a large carbon-rich resolved nebula. These stars show infraredspectra which are indicative for mixed chemistries, with features of both oxygen-rich andcarbon-rich species. The most likely scenerio for this is that the formation of the O-rich discantedated the C-rich transition of the central star. Whether other of our sample stars will alsoundergo this evolution is still unknown.

Our full spectral fitting indicates a high degree of dust grain processing. The dust seems toconsist of considerably large grains, with grain sizes larger than 2 µm. An efficient removal ofsmall grains must have occurred in the discs. The dust shape is highly irregular, showing thatMie theory is not applicable for the dust in these discs. The spectra of nearly all stars showa high degree of crystallinity, where Mg-rich end members of olivine and pyroxene silicatesdominate. The dust condensation sequence of dust in winds of oxygen-rich AGB stars pre-dicts the formation of aluminium- or calcium-rich dust grains, like corundum (Al2O3), spinel(MgAl2O4) or anorthite (CaAl2Si2O8) (Tielens et al. 1998a; Cami 2002). There is howeverno evidence for the presence of these dust species.

Most features are well reproduced and only the 14.7 and 32.5 µm features remain unidenti-fied (Sect. 3.5.2). Diopside (CaMgSiO3) could be a possible candidate (which has a featurearound 14.7 µm and a weak feature around 32.1 µm). However, for diopside only mass ab-sorption coefficients for very small grains are available, so we cannot include it in the fittingprocedure in a homogeneous way.

The dust is highly magnesium-rich, leaving a large fraction of iron unaccounted for. Pre-vious studies (e.g. Chapter. 2) suggest the iron may be locked in the form of metalliciron. Photospheric depletion in iron, which we detect in our sample stars (Maas et al. 2002;Van Winckel et al. 1998), can be understood when the iron is locked up in the circumstellardust (Waters et al. 1992). The lack of iron in the detected silicates is therefore surprising. Ifboth the crystalline and amorphous silicates are devoid of iron, this could mean that iron isstored in metallic iron or iron-oxide (Sofia et al. 2006). Metallic iron has no distinct featuresbut still makes a significant contribution in opacity, especially at shorter wavelengths, makingit very hard to detect directly.

In our fitting method, for both dust temperatures equal silicate fractions are used. This seemsto fit features indicative of both hot and cool crystalline dust, meaning we do not have a strongradial gradient in crystallinity throughout the disc. The inner regions of the discs, with tem-peratures above the annealing temperature, are expected to be fully crystalline. The presence


of a considerable amount of cool crystalline grains implies thus that we have strong turbu-lent mixing in the disc or that the crystalline grains were already abundant at disc formation.It is interesting to note that the very young circumbinary disc in the evolved binary SS Lep(Verhoelst et al. 2007), which may be formed in a similar process as the discs around ourpost-AGB sources, is dominated by small and large amorphous grains. The crystalline com-ponent is very small (Schutz et al. 2005). This would indicate that the crystallisation processhappens during disc evolution and is not already present at formation.

The evolution of these discs is still unknown. Our analysis shows no clear correlation betweendust parameters and any fundamental parameter of the central star, such as Teff or the orbitalperiod. In a recent study of Chesneau et al. (2007), a compact dusty disc was discoveredin “the Ant”, a well studied bipolar planetary nebula. Interferometric MIDI observationsprovided evidence for a flat, nearly edge-on disc, primarily composed of amorphous silicates.This is in contrast with the high crystallinity observed in the discs around binary post-AGBstars, suggesting that the disc in the Ant is relatively young. Whether there is a link betweenthis disc and the post-AGB discs remains unclear.

Some degeneracy is present in our spectral fits. We performed the spectral fitting both withpure Mg-rich amorphous silicates (x = 1) and with the more standard Mg-Fe amorphoussilicate dust (x = 0.5). Both models often have similar χ2 values and the best model stronglyvaries from star to star. Some stars also show equally well-fitting models, when using 0.1-2.0 µm or 2.0-4.0 µm grain sizes.

3.8.1 Comparison with young stellar objects

The mineralogy of the observed spectra shows a striking resemblance to the infrared spectraof young stellar objects, like Herbig Ae/Be stars (Lisse et al. 2007; van Boekel et al. 2005)or T Tauri stars (Watson et al. 2007), and primitive comets such as Hale-Bopp (Lisse et al.2007; Bouwman et al. 2003; Min et al. 2005b). There also, amorphous silicates and Mg-rich crystalline features prevail. The dusty disc is the relic of the star formation process soboth silicates together with a carbon-rich component in the form of PAH emission are oftendetected, depending on the disc geometry (Acke & van den Ancker 2004).

We compare our findings with the ones discussed in van Boekel et al. (2005). They presentspectroscopic observations of a large sample of Herbig Ae/Be stars in the 10 µm region.Similar studies on young stellar objects have also been done by e.g. Bouwman et al. (2001).One has to be careful comparing studies, since our study includes a far larger wavelengthrange, and an exact comparison is thus not possible.

The degree of crystallinity found in the Herbig Ae/Be stars is clearly smaller than for oursample of post-AGB binaries. The Herbig stars all show a degree of crystallinity below 0.35,whereas our stars have significantly higher values. van Boekel et al. (2005) use both small(0.1 µm) and large (1.5 µm) grains to fit the observed emission features and find that mostsources have a mass fraction in large grains of more than 80%.

For both young and evolved objects a substantial removal of the smallest grains has occurred,but seems to be more efficient in the discs around the evolved stars. A similar physical processmight be responsible for the observed grain-size distribution in both cases, like aggregation

3.9. Conclusions 83

of small grains or the removal of small grains by radiation pressure. Since several processesoccur in the formation of the disc it could also be that silicate grains in the post-AGB disc areformed at large grain sizes. These large grains could be broken up by grain interactions in thedisc, producing a small fraction of small grains.

van Boekel et al. (2005) found a clear correlation between the crystallinity and the dominanceof enstatite or forsterite in these grains. For sources with a high degree of crystallinity mostcrystalline grains appear to be in the form of enstatite, while for sources with a low crys-tallinity, forsterite seems to be the dominant species. For all our sample stars forsterite is thedominant species, despite the high crystallinity factors. This difference could relate to the ini-tial dust species when the disc is formed. For discs around YSO, which are formed from theISM, the abundant dust species is likely amorphous olivine. Forsterite is expected to be thedominant species formed by thermal annealing, while enstatite is expected to be the dominantspecies formed by chemical equilibrium processes in most of the inner disc. Our discs seemto prefer the formation of forsterite, but the initial dust species is not known. The innermostdisc regions are hot enough to crystallise dust, but the very high degree of crystallinity seemsto point to the presence of another crystallisation process, possibly at disc formation.

3.9 Conclusions

We present high-resolution TIMMI2 and Spitzer infrared spectra of 21 binary post-AGB starssurrounded by a stable Keplerian disc. We summarise our main conclusions:

– Almost all discs display only O-rich spectral signatures. The noticeable exception isEP Lyr, which shows a very similar spectrum as the central star of the Red Rectangle.PAH emission features dominate the spectrum untill 20 µm, at larger wavelengths crys-talline silicates start to dominate.

– Our mineralogy study indicates the dominance of Mg-rich amorphous and crystalline sil-icate dust in the disc. The high crystallinity and the large fraction of large grains, asdeduced from our full spectral fitting, show strong dust grain processing in the discs.

– The temperature estimates from our fitting routine show that a significant fraction of crys-talline grains must be cool. This shows that radial mixing is efficient is these discs orindicate a different thermal history at grain formation.

– Trend analysis of our fitting parameters show no clear correlation with stellar characteris-tics. For the moment it is not clear if and how the observed diversity in observed spectrarelates to specific structural elements of the disc, the star and/or the orbits or whether wewitness directly an evolutionary change between different sources.

To further improve our understanding of these circumbinary discs, as a next step, we willcombine our photometric and spectroscopic data with interferometric measurements. Com-paring spatial information from the MIDI and AMBER interferometric instruments with arealistic disc model (Dullemond & Dominik 2004), constrained by photometric and spectro-scopic data, will allow us, not only to study the mineralogy, but also the structure of the


discs. So far, interferometric measurements for five of our sample stars have been obtained(Deroo et al. 2006; Deroo 2007).

Probing the dust processing in the discs around evolved objects proves to be an excellentcomplement to study physics in planet-forming young discs.

AcknowledgementsThe authors want to acknowledge: the 1.2 m Mercator staff as well as the observers from the Instituutvoor Sterrenkunde who contributed to the monitoring observations using the Mercator telescope. CG ac-knowledges support of the Fund for Scientific Research of Flanders (FWO) under the grant G.0178.02.and G.0470.07. We also thank Fred Lahuis for his assistance with the Spitzer data reduction.

Chapter 4The peculiar post-AGB stars EP Lyrand HD 52961

This chapter is accepted and to be published by A&A as

Spectral analysis of the infrared spectra of the peculiar post-AGB stars EP Lyr andHD 52961.

C. Gielen, H. Van Winckel, M. Matsuura, M. Min, P. Deroo, L.B.F.M. Waters and C.Dominik, A&A, 2009

Abstract:AIMS: We aim to study in detail the peculiar mineralogy and structure of the circum-stellar environment of two binary post-AGB stars, namely EP Lyr and HD 52961. Bothstars were selected from a larger sample of evolved disc sources observed with Spitzerand show unique solid-state and gas features in their infrared spectra. Moreover, theyshow a very small infrared excess in comparison with the other sample stars.METHODS: On the basis of high resolution Spitzer-IRS spectra, the different dustand gas species are identified. We fit the full spectrum to constrain grain sizes andtemperature distributions in the discs. This, combined with our broad-band spectralenergy distribution and interferometric measurements allows us to study the physicalstructure of the disc, using a self-consistent 2D radiative-transfer disc model.RESULTS: We find that both stars have strong emission features due to CO2 gas, dom-inated by 12C16O2 but with clear 13C16O2 and even 16O12C18O isotopic signatures.Crystalline silicates are apparent in both sources but proved very hard to model.EP Lyr also shows evidence for a mixed chemistry, with emission features of the rareclass-C PAHs. Whether these PAHs reside in the oxygen-rich disc or in a carbon-richoutflow is still unclear. With the strongly processed silicates, the mixed chemistryand the low 12C/13C ratio, EP Lyr resembles some silicate J-type stars. We find that

86 Chapter 4. The peculiar post-AGB stars EP Lyr and HD 52961

the disc environment of both sources is, to first approximation, well modelled with apassive disc, but additional physics such as grain settling, radial dust distributionsand an outflow component must be included to explain the details of the observedspectral energy distributions in both stars.

4.1 Introduction

The infrared spectra of post-AGB stars are often characterised by strong spectral signatures.These are formed in the gas and dust rich circumstellar environment (CE), which is a remnantof the strong mass loss that occurred during the previous asymptotic giant branch (AGB)evolutionary phase. The chemistry in this circumstellar environment is found to be oxygen-rich or carbon-rich, depending on whether oxygen or carbon is more abundant. The lessabundant of the two, will be locked in the very stable CO molecule which forms in the stellarphotosphere.

Typical post-AGB outflow sources that have O-rich CE not only show the well-known 9.7and 18 µm features of amorphous silicates but also narrower features, arising from crys-talline silicates. (e.g. Waters et al. 1996; Molster et al. 2002a). The condensates in C-richoutflows show features of carbon-species such as SiC, MgS or polycyclic aromatic hydrocar-bons (PAHs) (e.g. Hony et al. 2001, 2002; Peeters et al. 2002). They are also characterised byan often very strong feature at 21 µm (Kwok et al. 1989; Volk et al. 1999; Hony et al. 2003).The photospheres of these 21 µm sources show strong enhancements of s-process elements(e.g. Reyniers et al. 2004, 2007) and the 21 µm stars are recognised to be post-AGB carbonstars (e.g. Van Winckel & Reyniers 2000).

Some objects show, however, features of both O-rich and C-rich dust species in their spectra.They are called mixed chemistry sources. This mixed chemistry is detected in a number ofsources in a wide range of different evolutionary stages. Some examples include Herbig Aestars or AGB stars, such as J-type carbon stars with silicate dust emission (Little-Marenin1986; Lloyd Evans 1990). Others are red giants, for example HD 233517, an evolved O-rich red giant with PAHs in a circumstellar disc (Jura et al. 2006). Other examples areplanetary nebulae (PNe) with evidence for silicates as well as PAHs (Kemper et al. 2002;Gutenkunst et al. 2008), or the hydrogen-poor [WC] central stars of PNe (Waters et al. 1998;Cohen et al. 1999). Also some M supergiants are associated with emission due to PAHs(Sylvester et al. 1998; Sloan et al. 2008).

Post-AGB stars with evidence for mixed chemistry include HD 44179, the central star of thecarbon-rich Red Rectangle nebula (Cohen et al. 1975). The central star is a binary surroundedby a Keplerian O-rich circumbinary disc (e.g. Van Winckel et al. 1995; Waters et al. 1998;Men’shchikov et al. 2002; Bujarrabal et al. 2005). Here the formation of the disc is believedto have antedated the C-rich transition of the central star (e.g. Cohen et al. 2004; Witt et al.2008).

Studies have shown that these evolved binaries with circumbinary discs are much more abun-dant than anticipated (De Ruyter et al. 2006; Van Winckel 2007). Interferometric studies(Deroo et al. 2006, 2007a) prove that the discs are indeed very compact, with radii around

4.2. Programme stars 87

50 AU in the N-band. The discs are also the natural environment of the observed photosphericchemical depletion pattern in these stars (Van Winckel et al. 1998; Giridhar et al. 2000), dueto chemical fractionation by dust formation in the circumstellar environment (Waters et al.1992) and subsequent accretion of the gas component. The presence of a long-lived stablereservoir of dust grains also could allow for the observed strong processing of the silicatedust grains, both in size as well as in crystallinity (Molster et al. 2002a, and Chap. 3).

Dusty RV Tauri stars are a distinct class in these binary post-AGB stars. They cross the insta-bility strip, and are therefore pulsating stars (Jura 1986; Jura & Kahane 1999; De Ruyter et al.2005). RV Tauri stars show large-amplitude photometric variations with alternating deep andshallow minima. The members are located in the high-luminosity end of the population IIinstability strip, and the photometric variations are interpreted as being due to radial pulsa-tions. Circumstellar dust emission was observed in many of them (Jura 1986), and this wasgenerally acknowledged to be a decisive character to place these stars in the post-AGB phaseof evolution. The grains in almost all dusty RV Tauri stars are, however, not freely expandingbut likely also trapped in a disc (Van Winckel et al. 1999; De Ruyter et al. 2005, 2006).

In this paper we focus on two peculiar post-AGB stars with RV Tauri pulsational character-istics: EP Lyr and HD 52961. These stars show unique spectral features and have very smallinfrared excesses in comparison to the larger sample.

The outline of the paper is as follows: We start with a short description of the programmestars in Sect. 4.2. In Sect. 4.3 we give an overview of the different observations and reductionstrategies. The analysis based on the Spitzer spectra is given in Sect. 4.4 and subdivided indifferent subsections. Sect. 4.4.2 contains a description of the silicate dust features and themodelling of the Spitzer-IRS spectra. The CO2 gas features are discussed in Sect. 4.4.3 andthe observed PAH features in EP Lyr in Sect. 4.4.4. In Sect. 4.5 we model the observed SEDsusing a passive disc model, also constrained with MIDI interferometric measurements. Thediscussion of our different results and our conclusions are presented in Sect. 4.6.

4.2 Programme stars

In our previous study we described and modelled the Spitzer-IRS spectra of 21 sources andfound that the dust around these stars is all O-rich and on average highly crystalline (Chap. 3).The two stars discussed here have the lowest LIR/L∗, respectively 12% and 3%, in the largerSpitzer sample, where an average of about 50% was found. The large infrared luminositycan be explained with a passive disc model, provided that the inner rim is close to the starand the scale height of the disc is significant (e.g. Deroo et al. 2007a). The low observedLIR/L∗ values of both stars point to a small disc scale height and/or a much larger inner gap,as it is unlikely that a disc is optically thin in the radial direction. Not only do they have thelowest LIR/L∗ values, both stars show unique spectral signatures in comparison to the largersample. We therefore selected these objects for a more detailed analysis.


Tabl

e4.

1—

The

nam

e,eq

uato

rial

coor

dina

tesα

and

δ(J2

000)

,eff

ectiv

ete

mpe

ratu

reT

eff

,sur

face

grav

itylo

gg

,and

met

allic

ity[F

e/H

]ofo

ursa

mpl

est

ars.

Fort

hem

odel

para

met

ers

we

refe

rto

De

Ruy

tere

tal.

(200

6).A

lso

give

nar

eth

eor

bita

lper

iod

and

the

ecce

ntri

city

(see

refe

renc

esin

De

Ruy

tere

tal.

2006

,an

dC

hap.

2).T

heto

talr

edde

ning

E(B

−V

) tot,t

heen

ergy

ratio

LIR

/L∗

and

the

calc

ulat

eddi

stan

ce,a

ssum

ing

alu

min

osity

ofL∗

=3000±

2000

L¯

.

Nam

eα

(J20

00)

δ(J

2000

)T

eff

log

g[F

e/H

]P o

rbe

E(B

−V

) tot

LIR

/L∗

d(h

ms)

(’”

)(K

)(c

gs)

(day

s)(%

)(k

pc)

EP

Lyr

1918

17.5

+27

5038

7000

2.0

-1.5

0.51±

0.01

3±

03.

2±

1.0

HD

5296

107

0339

.6+

1046

1360

000.

5-4

.813

100.

210.

06±

0.02

12±

11.

6±

0.5

4.3. Observations 89

4.2.1 EP Lyr

Schneller (1931) discovered the variability of EP Lyr and classified it as an RVb star. RVbstars are objects with a variable mean magnitude, in the General Catalogue of Variable Stars(Kholopov et al. 1999). Other studies (Zsoldos 1995; Gonzalez et al. 1997b) classify thelight curve as an RVa photometric variable, having a constant mean magnitude, with a pe-riod of P = 83.46 days. Preston et al. (1963) classify it as an RVB spectroscopic variable.Gonzalez et al. (1997b) performed an abundance analysis on EP Lyr where they deduced stel-lar parameters (see Table 4.1) and found the star to be metal-poor and severely depleted.Using the molecular lines found in the spectra, they also quantified the 12C/13C ratio to be9 ± 1. In the radial velocity data there is also evidence that EP Lyr must have a stellar masscompanion, but additional observations are necessary to determine the orbit.

4.2.2 HD 52961

HD 52961 is an RV Tauri like object, similar to class RVb objects (Waelkens et al. 1991b),with a photometric variability of 72 days due to clear radial pulsations (Waelkens et al. 1991b).The binarity of HD 52961 was first reported by Van Winckel et al. (1995) and further re-fined in Van Winckel et al. (1999) and Deroo et al. (2006), where an orbital period of Porb =1297 ± 7 days and an eccentricity of e = 0.22 ± 0.05 was found. On top of the stablephotometric variation due to the pulsation, another long-term photometric variation was de-tected, correlated with the orbital period. Van Winckel et al. (1999) conclude that this can beunderstood as caused by variable circumstellar extinction during the orbital motion.

The star is a highly metal-poor object with [Fe/H] =−4.8 (Waelkens et al. 1991a) and has anextremely high zinc to iron ratio of [Zn/Fe] =+3.1 (Van Winckel et al. 1992).

HD 52961 has been studied with mid-IR long-baseline interferometry using the VLTI/MIDIinstrument (Deroo et al. 2006). They find that the dust emission originates from a very smallbut resolved region, estimated to be ∼ 35 mas at 8 µm and ∼ 55 mas at 13 µm, likely trappedin a stable disc. The dust distribution through the disc is not homogeneous: the crystallinityis higher in the hotter inner region.

4.3 Observations

High- and low-resolution spectra of 21 post-AGB stars were obtained using the Infrared Spec-trograph (IRS; Houck et al. 2004) aboard the Spitzer Space Telescope (Werner et al. 2004) inFebruary 2005. The spectra were observed using combinations of the short-low (SL), short-high (SH) and long-high (LH) modules. SL (λ=5.3 − 14.5 µm) spectra have a resolvingpower of R=λ/ 4 λ ∼ 100, SH (λ=10.0 − 19.5 µm) and LH (λ=19.3 − 37.0 µm) spectrahave a resolving power of ∼ 600. Exposure times were chosen to achieve a S/N ratio ofaround 400 for the high-resolution modes, which we complemented with short exposures inlow-resolution mode with a S/N ratio around 100, using the first generation of the exposuretime calculator of the call for proposals.

The spectra were extracted from the SSC data pipeline version S13.2.0 products, using the


Figure 4.1 — The combined Spitzer-IRS high- and low-resolution spectrum of EP Lyr (top) andHD 52961 (bottom).

c2d Interactive Analysis reduction software package (Kessler-Silacci et al. 2006; Lahuis et al.2006). This data processing includes bad-pixel correction, extraction, defringing and ordermatching. To match the different orders, we applied small scaling corrections.

4.4. Spectral analysis 91

4.4 Spectral analysis

4.4.1 General

A look at the spectra of EP Lyr and HD 52961 (Fig. 4.1) shows that the continuum slopes ofboth stars are very different. EP Lyr shows strong emission features at longer wavelengths,with peak emission in the 20 µm region, whereas HD 52961 is characterised by a strong10 µm emission feature on top of a much steeper continuum.

Common dust species found in oxygen-rich post-AGB stars are amorphous silicates, namelyolivine and pyroxene. Amorphous olivine (Mg2xFe2(1−x)SiO4, where 0 ≤ x ≤ 1 de-notes the magnesium content) has very prominent broad features around 9.8 µm and 18 µm.Amorphous pyroxene (MgxFe1−xSiO3) shows a 10 µm feature similar to that of amorphousolivine, but shifted towards shorter wavelengths. Also the shape of the 18 µm feature isslightly different. For EP Lyr it is unclear whether there is a significant contribution of amor-phous silicates. Small amorphous silicates could contribute to the observed strong emissionbump at 20 µm in EP Lyr, but as there does not seem to be a 10 µm amorphous feature, the20 µm bump could be purely continuum dominated. HD 52961 has clear strong emission ofamorphous silicates at 10 µm, but the profile shows complex narrow subfeatures. Very littlecontribution at 20 µm is seen.

Both stars show strong narrow emission features which can be identified as being due tocrystalline silicates. The Mg-rich end members of crystalline olivine and pyroxene, forsterite(Mg2SiO4) and enstatite (MgSiO3), show strong but narrow features at distinct wavelengthsaround 11.3 − 16.2 − 19.7 − 23.7 − 28 and 33.6 µm. For EP Lyr the silicate emission onlyclearly starts longward of 18 µm, where strong emission features around 19 − 23 − 27 and33 µm can be seen. HD 52961 has strong narrow features at 9.8− 11.3 µm and a remarkablystrong 16 µm feature. If this strong 16 µm band is only due to forsterite it has shifted con-siderably to shorter wavelengths. A significant 16 µm feature is seen in several evolved discsources but it is never as strong as in HD 52961 (Chap. 3).

EP Lyr shows evidence for the presence of carbon-rich dust species with probable PAH iden-tifications at 8.1 and 11.3 µm. The detection of PAH emission together with silicates is sur-prising and only observed in a few other post-AGB sources. The analysis of the PAH featuresis given in Sect. 4.4.4.

The spectrum of EP Lyr shows a strong resemblance to that of IRAS 09245-6040 (Fig. 4.3), asilicate J-type carbon AGB star (Molster et al. 2001; Garcıa-Hernandez et al. 2006a). SilicateJ-type carbon stars have surprisingly low 12C/13C ratios and do not show the typical s-processoverabundances seen in N-type carbon stars (Abia & Isern 2000). The infrared spectrum ofthese stars shows features of both carbon- and oxygen-rich dust species.

Of the silicate J-type carbon stars, only 10% show emission bands due to crystalline mate-rial (Lloyd Evans 1991; Ohnaka & Tsuji 1999). The formation history of these stars is stillunclear, but the most promising scenario for the presence of silicates in these stars, is thatthey are binaries with an undetected companion (Lloyd Evans 1990; Yamamura et al. 2000).A disc is supposed to be formed when the primary was still an O-rich giant. After that thestar underwent thermal pulses and evolved into a carbon star. The silicate disc could be either


captured from the wind (Mastrodemos & Morris 1999) or the result of a phase of strong bi-nary interaction in a narrow system. Systems with strong crystalline features in their spectra,such as IRAS 09425-6040 or IRAS 18006-3213 (Deroo et al. 2007b), would then be a resultof mass-transfer into a circumbinary system, whereas sources dominated by amorphous sili-cates, such as V778 Cygni (Yamamura et al. 2000) or BH Gem (Ohnaka et al. 2008), consistof a wide binary with a circumcompanion disc. To date, no orbits are known however, anddirect evidence for binarity is found in a few objects only (Izumiura et al. 2008).

For IRAS 09245-6040, the 12C/13C ratio is calculated to be 15 ± 6 (Garcıa-Hernandez et al.2006a). In the ISO-SWS spectrum features of C2H2, HCN, CO, C3 and SiC are seen short-ward of 15 µm; after 15 µm the spectrum is dominated by strong emission features of Mg-richcrystalline silicates (Molster et al. 2001). As in EP Lyr, there is no evidence for a strong con-tribution of amorphous silicates.

Finally in both EP Lyr and HD 52961, clear CO2 gas emission features are detected the 13−18 µm region. This is discussed in Sect. 4.4.3.

4.4.2 Silicate dust emission

We optimised the fitting procedure as discussed in Chapter 3 for these two outliers, where wemodelled the full Spitzer sample, consisting of 21 stars. In short we assume the flux to beoriginating from an optically thin region, so we can make linear combinations of the absorp-tion profiles to calculate the model spectrum. In our previous modelling we found that, onaverage for the full sample, the best fit was obtained using relatively large grains (≥ 2 µm) inan irregular Gaussian Random Fields (GRF) dust model. For EP Lyr and HD 52961 however,we already found that using smaller grain sizes (≤ 2 µm) improved the fit considerably.

So we repeated the analysis for EP Lyr and HD 52961, allowing for different dust shapes,grain sizes and Mg/Fe content in the amorphous grains. We tested Mie, GRF and DHS(Distribution of Hollow Spheres) dust models in grain sizes ranging from 0.1 to 4.0 µm. Inorder to test for the presence of Fe-poor amorphous dust, we perform the modelling both withpure Mg-rich amorphous silicates (x = 1) and with the more standard Mg-Fe amorphoussilicate dust (x = 0.5). For a detailed description of the fitting routine we refer to Sect. 3.7.The results of the fitting can be found in Table 4.2. As for EP Lyr the silicate signatures onlyappear after 18 µm; we only fit this part of the Spitzer spectrum.

The χ2 values of our fitting (Table 4.2) are still quite high for HD 52961 but, confirmingthe result of Chapter 3, we can already tell that for both stars Mie theory is not a good dustapproximation. For EP Lyr the GRF grains prove the best match, but the difference in χ2 withthe small DHS grain approximation is only minimal. The best fit to EP Lyr is obtained usingboth small (0.1 µm) and larger (2.0 µm) silicate grains. The small difference in calculated χ2

values for EP Lyr is due to the low signal-to-noise ratio, making it hard to distinguish betweendifferent synthetic emission profiles. For HD 52961 small grains in Mg-rich silicates give thebest χ2. Plots of our best fitting models can be found in Fig. 4.2. Table 4.3 gives the resultingparameters.

The large χ2 value of HD 52961 quantifies that this star has a very peculiar, unique chemistry,and we did not succeed in explaining all of the observed features. The strong forsterite


Figure 4.2 — Best fits for EP Lyr and HD 52961. The observed spectrum (black curve) is plottedtogether with the best model fit (red curve) and the continuum (black solid line). Forsterite is plotted indash-dot lines (green) and enstatite in dash-dot-dotted lines (blue). Small amorphous grains are plottedas dotted lines (magenta) and large amorphous grains as dashed lines (magenta).

11.3 µm feature in the GRF dust approximation is clearly too broad. DHS grains fit thefeature better, but other feature profiles are fitted less well with this approximation. Therealso appears to be a short wavelength shoulder on the amorphous 9.8 µm feature, which isnot explained in the modelling. The strong 16.5 µm feature is not reproduced in centralwavelength by any of the different models. We already observed this trend in our full samplefitting (Chap. 3), where the feature seemed to be shifted bluewards in comparison with themean spectrum of the full sample. The two narrow features around 19 µm could be an artifactof the data reduction, since in this region there can be a bad overlap between the SH and LHSpitzer-IRS high-resolution bands.


Figure 4.3 — The Spitzer-IRS spectrum of EP Lyr compared to the ISO-SWS spectrum of IRAS 09425-6040. The spectrum of IRAS 09425-6040 is normalised and offset for comparison.

Table 4.2 — χ2 values for different models used in our full spectral fitting. For each model we give theused dust approximation, grain size and Mg-Fe content in the amorphous grains. x = 1.0 denotes pureMg-rich amorphous dust, x = 0.5 the more standard Mg-Fe amorphous silicates.

EP Lyr HD 52961 model descriptionχ2 χ2

model1 21.7 129.4 Mie - 0.1− 2.0 µm - x = 0.5model2 6.2 67.5 DHS - 0.1− 1.5 µm - x = 0.5model3 6.2 63.8 DHS - 0.1− 1.5 µm - x = 1.0model4 8.4 101.4 DHS - 1.5− 3.0 µm - x = 0.5model5 8.6 140.8 DHS - 1.5− 3.0 µm - x = 1.0model6 5.9 64.2 GRF - 0.1− 2.0 µm - x = 0.5model7 6.3 50.0 GRF - 0.1− 2.0 µm - x = 1.0model8 5.8 96.5 GRF - 2.0− 4.0 µm - x = 0.5model9 5.4 72.2 GRF - 2.0− 4.0 µm - x = 1.0

For EP Lyr we fit the spectrum longwards of 18 µm, where the silicate features are seen.This gives dust temperatures between 100 and 230 K. This model, however, does not fit thespectrum before 18 µm, since the continuum does not follow the observed strong downwardslope before 20 µm. If we try to fit the full Spitzer wavelength range we find we can get abetter fit to the underlying continuum but then the features at 27 and 33 µm are much strongerin the observed spectrum than in our best model. Unlike in other sources (Chap. 3), a twotemperature approach fails to model both the observed continuum and the coolest featuresfor the full Spitzer wavelength spectrum of EP Lyr. Irrespective of the derived continuumtemperature, all the tested models give estimates of the dust temperatures between 100 −300 K, which agrees with the temperatures derived in the SED modelling (Sect. 4.5). Clearly,the crystalline dust particles must be quite cold.


Tabl

e4.

3—

Lis

ted

are

the

best

fitpa

ram

eter

sde

duce

dfr

omou

rfu

llsp

ectr

alfit

ting:

the

χ2,d

usta

ndco

ntin

uum

tem

pera

ture

san

dth

eir

rela

tive

frac

tions

(βj

inth

efo

rmul

agi

ven

inSe

ct.3

.7).

The

abun

danc

esof

smal

l(0.

1µ

m)a

ndla

rge

(2.0

µm

)gra

ins

ofth

eva

riou

sdu

stsp

ecie

sar

egi

ven

asfr

actio

nsof

the

tota

lmas

s,ex

clud

ing

the

dust

resp

onsi

ble

fort

heco

ntin

uum

emis

sion

.The

last

colu

mn

give

sth

eco

ntin

uum

flux

cont

ribu

tion,

liste

das

ape

rcen

tage

ofth

eto

tali

nteg

rate

dflu

xov

erth

efu

llw

avel

engt

hra

nge.

The

erro

rsw

ere

obta

ined

usin

ga

Mon

te-C

arlo

sim

ulat

ion

base

don

100

equi

vale

ntsp

ectr

a.D

etai

lsof

the

mod

ellin

gm

etho

dar

eex

plai

ned

inSe

ctio

n3.

7.

Nam

eχ

2T

du

st1

Tdu

st2

Frac

tion

Tcon

t1T

con

t2Fr

actio

n(K

)(K

)T

du

st1

-Tdu

st2

(K)

(K)

Tcon

t1-T

con

t2

EP

Lyr

5.4

1141

22

14

2284

88

89

0.90

.10.6−

0.10

.60.1

2057

02

103

6413

31

301

0.96

0.0

20.0

6−

0.04

0.0

60.0

2

HD

5296

150

.020

010

072

4186

96

0.90

.00.1−

0.10

.10.0

1113

56

11

9964 1

11

0.99

0.0

00.0

3−

0.01

0.0

30.0

0

Nam

eO

livin

ePy

roxe

neFo

rste

rite

Ens

tatit

eC

ontin

uum

Smal

l-L

arge

Smal

l-L

arge

Smal

l-L

arge

Smal

l-L

arge

EP

Lyr

641

6−

525

583

47−

623

634

19

15−

933

852

05−

2817

21

5311

20

HD

5296

101

30−

141

155

12

18−

243

261

85−

3313

18

18 1−

326

369

4 4


Figure 4.4 — Comparison between the continuum-subtracted spectrum of EP Lyr and HR 4049, anothermixed chemistry source. Both stars show clear emission due to CO2 gas around 15 µm and strong PAHfeatures before 13 µm. The CO2 features at 14.9 − 16.2µm, 15.3 µm and 15.1 µm are respectivelydue to 12C16O2, 13C16O2 and 16O12C18O. The PAH emission is clearly very different in HR 4049,where class B PAHs are found, whereas in EP Lyr the PAH features can be attributed to class C (seeSect. 4.4.4).

4.4.3 CO2 emission

4.4.3.1 Introduction

CO2 emission has been found in approximately 30% of all O-rich AGB stars (Justtanont et al.1998; Ryde et al. 1999; Sloan et al. 2003), but CO2 detections in post-AGB stars are rare. Toour knowledge CO2 gas has been found in only two post-AGB stars, the Red Rectangle andHR 4049 (Waters et al. 1998; Cami & Yamamura 2001), which are also binaries surroundedby a stable circumstellar disc. HR 4049 is the only example of a post-AGB star showing CO2

in emission in the 13 − 16 µm region. Cami & Yamamura (2001) argued that the isotopicdistribution of oxygen in HR 4049 is abnormal, based on the isotope ratio analysis of the CO2

emission features. This was not confirmed by Hinkle et al. (2007), who use high-resolutionspectra of the fundamental and first overtone CO vibro-rotational transition in the near-IR.

Both EP Lyr and HD 52961 show clear gas phase emission lines of 12CO2 and 13CO2. Theseemission lines were also seen in only one other source in our Spitzer sample (Chap. 3),namely in IRAS 10174-5704.

The CO2 emission of EP Lyr seems to be lying on top of a “plateau” that extends from 13 to17 µm. A similar plateau in this region is observed in PAH-rich sources (Peeters et al. 2004),but this plateau is much broader and ranges from 15 to 20 µm and is often characterised bystrong emission features at 16.4 µm (and less prominent at 15.8, 17.4 and 19 µm), and thusquite different to the one seen in EP Lyr.


Table 4.4 — Resulting parameters for the model calculations. The isotope ratio is as follows:12C16O2:13C16O2:16O12C18O.

Tex N r isotope ratio(K) (cm−2) (R∗)

EP Lyr 900 8× 1018 4.7 0.7 : 0.2 : 0.1HD 52961 800 5× 1018 4.7 0.93 : 0.05 : 0.02

4.4.3.2 Analysis

To retrieve the very rich spectral information of the CO2 emission bands, we calculate spec-tra of CO2, using HITRAN line lists (Rothman et al. 2005) and a circular slab model for theradiative-transfer (Matsuura et al. 2002). The model has four parameters: the excitation tem-perature (Tex), the total CO2 column density (N ), the radius of CO2 gas (r) and the isotoperatio. The radius of the CO2 layer is given relative to the radius of the background continuumsource at 13 µm. The dependence of the CO2 model spectra on these parameters are describedby Cami & Yamamura (2001). We estimate a pseudo-continuum by using a spline fit and alinear fit for HD 52961 and EP Lyr, respectively. For HD 52961, we interpolate the spectrumat the spectral range where CO2 bands have little influence on the observed spectrum. Thecontinuum was also chosen so that the forsterite feature at 16 µm would be removed. A splinefit was tested for EP Lyr but failed because of the richness of CO2 features in the mid-infraredrange, so we simply use a linear interpolation between 13.4 and 17.8 µm. Estimated continuaare displayed as dotted lines in the top panels of Figs. 4.5 and 4.6. The resulted parametersfor the model calculations are summarised in Table 4.4, and the resulted spectra show theidentifications of the different CO2 bands (bottom panels of Figs. 4.5 and 4.6).

Many small features in EP Lyr in the 13.5 − 17 µm region are due to CO2: features at 13.5,13.9, 14.7, 14.9, 16.2 µm are attributed to the CO2 main isotope 12C16O2. The main isotopic12C16O2 bands are probably optically thick, surpressing the line intensities. 13C16O2 and16O12C18O bands are found at 15.3 µm and 15.1 µm, respectively. The prominent 16O12C18Ofeature is surprising. This feature was also found in the other binary post-AGB star HR 4049(Cami & Yamamura 2001).

The model uses a high fraction of isotopes, but actual abundance ratios remain largely un-certain, mainly because of the uncertainty of the interpolated continuum spectrum and theoptical thickness of the main isotope. Nevertheless, these two features are particularly promi-nent in the spectrum of EP Lyr, more than in HD 52961, suggesting different isotope ratiosfor EP Lyr. We see that the observed “plateau” in EP Lyr can be explained by the richnessof the 12C16O2 features, but the 12C/ 13C ratio is confirmed to be low. The strength of theisotopes, including the very rare 18O (in the Sun 16O/18O is ∼ 500), is an exclusive featureof post-AGB stars.


Figure 4.5 — EP Lyr: top panel shows the observed spectrum (solid line) and pseudo-continuum spec-tra (dotted line). The bottom panel shows the continuum divided spectrum, CO2 model spectra (com-bining all of the isotopes) and individual CO2 isotope spectra (from top to bottom 12C16O2, 13C16O2,16O12C18O).

4.4.4 PAH features

Polycyclic aromatic hydrocarbons are found in a large variety of objects, including the diffuseISM, HII regions, young stellar objects, post-AGB stars and planetary nebulae. They havestrong emission features in the 3-13 µm region (e.g. Tielens 2008). The feature at 3.3 µmis arises from the C-H stretching mode of neutral PAHs. The C-C modes produce featureswith typical central wavelengths at 6.2 and 7.7 µm. The 8.6 µm feature is due to C-H in-plane bending modes and features longward of 10 µm can be attributed to C-H out-of-planebending modes.

Peeters et al. (2002) defined three groups of PAH spectra based on their emission profilesand peak positions. The “class A” sources have features at 6.22, 7.6 and 8.6 µm. “Class B”sources show the same features but shifted to the red, peaking at 6.27, 7.8 and > 8.6 µm. Theyalso identified two “class C” sources, the Egg Nebula (AFGL 2688) and IRAS 13416-6243,both post-AGB objects. These rare “class C” sources show emission features at 6.3 µm, noemission near 7.6 µm, and a broad feature centered around 8.2 µm, extending beyond 9 µm.


Figure 4.6 — Same as Fig 4.5 but for HD 52961.

With the release of the IRS aboard the Spitzer Space Telescope, a limited number of addi-tional class-C sources were discovered (Fig. 4.7). MSX SMC 029, a class-C post-AGB star inthe SMC, was detected by Kraemer et al. (2006). Sloan et al. (2007) report on the detection ofclass-C PAH features in HD 100764, a carbon-rich red giant with evidence for a circumstellardisc. Jura et al. (2006) also report on the detection of class-C PAH features in a circum-stellar disc around the oxygen-rich K-giant HD 233517. Two young objects, the T Tauri starSU Aur (Furlan et al. 2006) and the Herbig Ae/Be source HD 135344 (Sloan et al. 2005), alsoshow PAH spectra of class C, although in HD 135344 the PAH features seem to be somewhatmore in between B and C. A comparison of the PAH features in all these sources is givenby Sloan et al. (2007). They find that all the known class-C spectra are excited by relativelycool stars of spectral type F or later and argue that the hydrocarbons in these sources have notbeen exposed to much ultraviolet radiation. The class-C PAHs are then relatively protectedand unprocessed, while class A and B PAHs have been exposed to more energetic photonsand are hence more processed.

EP Lyr shows PAH emission bands at 8.1, 11.3 and 12.6 µm. Another features can be seenat 13.25 µm, whether this can be attributed to PAHs remains uncertain. As Figure 4.8 shows,the spectrum of EP Lyr has a striking resemblance to that of HD 233517, shortward of 13 µm


Figure 4.7 — continuum-subtracted spectra (based on a spline fit) of class-C sources as described inSloan et al. (2007). The vertical dashed lines are at 8.1, 11.3 and 12.6 µm. For comparison we also plotHD 44179, which has class B PAH emission.

Figure 4.8 — Comparison between the Spitzer-IRS spectrum of HD 233517 and EP Lyr in the 6-14 µmregion.

(Jura et al. 2006), and we classify the PAH spectrum of EP Lyr as class C. EP Lyr is a highamplitude-variable with an effective temperature around 7000 K, which is on the hot end ofthe other class-C emitters (see Sloan et al. 2007).

Jura (2003) hypothesises that HD 233517 was a short-period binary on the main sequence.A circumstellar disc was then formed when the companion star was engulfed by the moremassive star when it entered giant evolution, followed by a phase of strong mass ejection in


the equatorial plane. Since HD 233517 is an oxygen-rich star, it is remarkable that the discshows features of carbon-rich chemistry. Jura et al. (2006) propose a scenario in which thePAHs could be formed inside the disc due to Fischer-Tropsch (FT) catalysis on the surfaceof solid iron grains. These FT reactions can convert CO and H2 into water and hydrocarbons(Willacy 2004), these hydrocarbons could then be converted in into PAHs.

Another explanation could be that the PAH carriers do not reside in the disc but in bipolarlobes created by a more recent mass-loss event, as observed in HR 4049 (Johnson et al. 1999;Dominik et al. 2003; Antoniucci et al. 2005; Hinkle et al. 2007) and in the Red Rectangle(Men’shchikov et al. 2002; Cohen et al. 2004). For HR 4049, Johnson et al. (1999) foundthat the optical polarisation seems to vary with orbital phase, suggesting that the polarisationin the optical is due to scattering in the circumbinary disc. In the UV, the polarisation iscaused by scattering in the bipolar lobes, which should contain a population of small grains,including the PAH carriers. HR 4049 and HD 44179, as well as EP Lyr and HD 52961, arestrongly depleted, and molecules or the formation of dust in these very depleted environmentsis likely very different from solar condenstation. As the CO molecule is abundant in thecircumstellar environment, accretion of circumstellar gas will result in a C∼O photosphere.HR 4049 and HD 44179 are stars with PAH emission belonging to the more standard class B.So far it remains unclear whether the shape of the observed emission features can detect ifthe PAH carriers reside in the disc or an outflow.

4.5 Spectral energy distribution

4.5.1 2D disc modelling

For both stars extensive photometric data are available. This, together with the Spitzer in-frared spectral information, allows us to constrain some of the physical characteristics of thecircumbinary disc.

We fit the SED using a Monte Carlo code, assuming 2D-radiative-transfer in a passive discmodel (Dullemond et al. 2001; Dullemond & Dominik 2004). This code computes the tem-perature structure and density of the disc. The vertical scale height of the disc is computedby an iteration process, demanding vertical hydrostatic equilibrium. The distribution of dustgrain properties is fully homogeneous and, although this model can reproduce the SED, dustsettling timescales indicate that settling of large grains to the midplane occurs. Using the dustsettling timescale

tset =π

2Σ0

ρda

1Ωk

lnz

z0

with Σ0 the surface density, ρd the particle density, a the grain size and Ωk =√

GM∗r3 the

Keplerian rotation rate (Miyake & Nakagawa 1995), we find that grains with sizes between500 µm and 0.1 cm can descend 50 AU on timescales similar to the estimated lifetime of thedisc. An inhomogeneous disc model with a vertical gradient in grain-size distribution is thusnecessary (Chap. 2). These large and cooler grains in the disc midplane mainly contribute tothe far-IR part of the SED and constitute the main fraction of the total dust mass. The discstructure and near- and mid-IR flux are almost fully determined by small grains. So we use


a homogeneous 2D disc model to fit the near- and mid-IR part of the SED and add a singleblackbody temperature to represent the cooler midplane made up of large grains. The lack ofobservational constraints on the temperature structure of this component does not allow us toconstrain this extra parameter.

Stellar input parameters of the model are the luminosity, the total mass (we assume the totalgravitational potential to be spherically symmetric with a total mass of M = 1 M¯), andTeff . The luminosity (and thus the distance) for these sources is not well constrained sowe use values between L = 1000 − 5000 L¯, typical values for post-AGB sources. Inputdisc parameters are Rin and Rout, the different dust components, the total disc mass and thepower law for the surface-density distribution. Since we are not dealing with outflow sourcesa power law α > −2 is used. The gas-to-dust ratio is kept fixed at 100.

The modelling is still degenerate, especially in parameters like the outer radius and the totaldisc mass which can be easily interchanged, without strongly influencing the SED. We usea dust mixture of amorphous and crystalline silicates in grain sizes ranging of 0.1 − 20 µm,with a power law distribution of −3.0, for the homogeneous disc. For HD 52961 the 850 µmsubmillimetre data points to the presence of extremely large grains in the disc, but these largegrains are assumed not to influence the near- and mid-IR part of the SED, and will only beimportant for the blackbody component. We use a value of 300 AU for Rout. For the innerrim we use the radius at which the temperature of the inner rim equals about 1500 K. Thisis a typical value for the dust sublimation temperature of silicates, although values as lowas 1200 K are sometimes also used. The total SED energetics are then calculated, given aspecific inclination angle of the system.

We do not aim to reproduce the observed narrow features in the Spitzer spectrum, since thesefeatures originate from an optically thin upper layer of small grains at the disc surface. Theycan only be fitted well using an inhomogeneous disc model with grain settling. Instead,we want to model the observed general energetics of the SED spectrum, thus the observedcontinuum and amorphous dust features.

4.5.2 Comparison with interferometric data

The circumstellar environment of HD 52961 has been resolved using the VLTI/MIDI instru-ment, with angular sizes in the N-band between 35 mas and 55 mas in a uniform disc approx-imation (Deroo et al. 2006). EP Lyr is too faint for current interferometric capabilities.

To compare our disc model of HD 52961 with the MIDI data we made model images fromwhich we extracted visibilities, using the same projected baseline lengths (40 m/46 m) andangles (45/46) as the observations. The only free parameters here are the inclination of thedisc and the orientation angle of the system on the sky. A range of inclinations which still fitthe observed SED was tested, in steps of 15. The orientation angle is varied 1 at a time.The result of this comparison can be found in Sect. 4.5.3.1.


Figure 4.9 — Some SED disc models showing the influence of the inclusion of metallic iron and differ-ent grain sizes in the fitting. All models have the same physical parameters, the only difference beingthe amount of iron in the disc and the grain sizes. The solid line depicts a pure silicate disc with grainsizes between 0.1-20 µm. The dashed line represents a model with 5% metallic iron and 95% silicatewith grain sizes between 0.1-20 µm. The dotted line represents a silicate disc model with grain sizesbetween 0.1-50 µm and the dot-dashed line one with grain sizes between 0.1-100 µm.

4.5.3 Results

When modelling the near- and mid-IR part of the SEDs, the feature-to-continuum ratio of thesilicate features is too strong in comparison with the observed spectra for both stars. More-over, the near-IR flux is often underestimated. This was also observed in Chapter 2, wherewe fitted the SED similarly to two post-AGB stars, RU Cen and AC Her. Including an extracontinuum opacity source is needed to reduce the strength of the features (see Fig. 4.9) andto increase the near-IR contribution. As the silicates are devoid of iron (Sect. 4.4.2), we usemetallic iron as a potential opacity source: its near-IR opacity is large, but the absorption co-efficient is unfortunately featureless so direct detection is difficult. Inclusion of free metalliciron has a strong impact on the modelling because the near-IR excess increases significantlywith a given inner radius. Another possible opacity source is the inclusion of hot large grainsin the homogeneous disc model.

4.5.3.1 HD 52961

We can use the interferometric data for HD 52961 to constrain the distance to the system(Fig. 4.11). When we compare the modelled visibilities with the observations for HD 52961we find that the visibilities of our model are too high, when using standard models of L =5000 L¯ which fit the SED and impose the inner radius to be at sublimation radius. Thismeans we need to increase the angular size of the N-band emission region, either by increas-ing the physical size or by decreasing the distance. Increasing the disc size to outer radii


Table 4.5 — Results of our SED disc modelling. The inner and outer radius (Rin-Rout), the total discmass m for the homogeneous disc model, the surface-density distribution power law α, the percentageof iron in the homogeneous disc model, the blackbody temperature and the inclination of the system.

Rin −Rout m α iron Tbb iAU 10−5M¯ % K

HD 52961: A 10− 500 1.7 -1.5 0 160 0− 90HD 52961: B 10− 500 0.7 -1.5 10 160 < 65EP Lyr: A 40− 300 6 -1.0 0 - 0− 90

> 500 AU does not influence the N-band emission so we need to increase the inner radius,flatten the density distribution power law and/or decrease the luminosity in the disc model.Changing the surface-density distribution power law to values > −1.5 proved incompatiblewith the observed SED. We therefore use an average luminosity of 3000 L¯ and move theinner radius to larger distances. A good fit to the SED was obtained using an inner radius of10 AU. This assumed luminosity gives a distance to the system of about 1700 pc. At 10 AUthe temperature of the inner rim is around 1100 K, which is slightly below the canonical dustsublimation temperature for silicates.

For HD 52961 we tested models with and without the inclusion of metallic iron. The resultingfits can be seen in Figure 4.10, parameters can be found in Table 4.5. Since this star showsa rather strong 10 µm feature, pointing to relatively large amounts of hot dust in the disc, wefind we need a rather steep surface-density distribution, α < −1. We add a 160 K blackbodyto fit the observed submillimetre emission.

When no iron is included (model A) we see that the flux around from 2 to 8 µm is strongly un-derestimated. Including about 10% metallic iron (model B) increased this flux significantly.The inclusion of metallic iron has only a minor influence on the N-band interferometric mea-surements. It decreases the modelled visibilities by about 10%, which is still consistent withthe observed visibilities.

The modelled visibilities (Fig. 4.11) lie within the observed visibility range but a remark-able detection is that the variation in visibilities between the two observations is quite large,despite the very limited difference in lengths (41.3 − 46.3 m), as well as in projected angle(45.6 − 46.3). This is illustrated when we plot the visibility versus the spatial frequencyfor a given wavelength, as seen in Figure 4.12. In this figure we illustrate that when using auniform disc for the intensity distribution, the steep visibility drop can be accounted for. Thephysical disc model is, however, much smoother and does not contain the very sharp edgecharacteristic of the uniform disc. The model also do not reproduce the observed ‘bump’ invisibility between 9 and 12 µm. Deroo et al. (2006) explain this observed increase in visibil-ity as being due to a non-homogeneous distribution of the silicates, which contribute most tothe inner regions of the disc. The current disc model does not include the physics to be ableto reproduce this radial distribution of dust species.

The submillimetre 850 µm flux for HD 52961 and the derived blackbody temperature of160 K can be used to estimate the dust mass of large grains in the disc. In the optically


Figure 4.10 — SED disc modelling of HD 52961 (top: model A without metallic iron, bottom: modelB with metallic iron). The dashed line represents the homogeneous disc model consisting of grainsbetween 0.1 µm and 20 µm. The solid line gives the disc model with an added blackbody to representthe cool midplane. Crosses represent photometric data and in the infrared we overplot the observedSpitzer-IRS spectrum. The dotted line represents the adopted Kurucz stellar model.

thin approach the disc mass can be estimated using (Hildebrand 1983)

Md =F850 D2

κ850 B850(T ).

Assuming a cross section of large spherical grains, the mass absorption coefficient of 850 µmgrains in blackbody approximation is about 2.4 cm2 g−1. The mass absorption coefficientis given by κ = πa2

43 πa3ρ

, with a the grain size, and ρ typically 3.3 g cm−3 for astronomical

silicates. This results in dust-mass estimates in large grains of 3.2× 10−6 M¯ for HD 52961.


Figure 4.11 — The observed MIDI measurements of HD 52961 (gray data with error bars) and thevisibilities deduced from disc model A (black solid line). The model has an inclination of 45 and aposition angle on the sky of 227 East of North.

Figure 4.12 — Comparison between modelled and observed visibilities at 12.6 µm, at the two differentbaseline lengths. The dashed line represents a uniform disc model with an angular size of 50 mas. Thesolid line gives the visibilities as calculated from the 2D disc modelling (model B).

4.5.3.2 EP Lyr

The SED-fitting gives an estimate of the distance to the system of d = 3.4 kpc, assuming aluminosity L = 3000 L¯. This estimation is not only dependent on the assumed luminosityof the star, but also on the adopted inclination of the system. Other derived disc parameterscan be found in Table 4.5.

4.6. Conclusions 107

Figure 4.13 — The SED disc model of EP Lyr. The model parameters are given in Table 4.5.

For EP Lyr it proved very problematic to get a good fit to the observed strong 20 µm feature,without introducing a strong 10 µm silicate feature. The mixed chemistry adds to the com-plexity of the object as the PAH emission and underlying continuum in the near-IR may verywell come from a distinct structural component, in for example the polar direction. This isalready seen in HD 44179, were the observed PAH emission comes from a bipolar outflow(Men’shchikov et al. 2002; Cohen et al. 2004). The lack of data shortwards of 7 µm makes ithard to get a good continuum estimate of the near- and mid-IR energetics. Cool dust clearlydominates the SED, but without additional information from interferometry, we cannot con-strain parameters like the inner radius or the surface-density distribution. We opted to keepa rather flat density distribution of α = −1.0 and do not force the model to fit the spectrumshortward of 15 µm.

To get a good fit to the SED a very large inner radius of about 40 AU is necessary. At thisdistance the inner rim reaches temperatures of ∼ 300 K. This temperature agrees with thetemperatures found in the spectral modelling (Sect. 4.4.2). If we include metallic iron in themodel we find we need inner radii even larger than 200 AU, which seems physically implau-sible. Unfortunately we do not posses submillimetre data for this star so we cannot determinethe blackbody temperature associated with the midplane. The small LIR/L∗ as well as thelack of a near-IR dust excess shows that the inner rim is quite far from the sublimation radius.

4.6 Conclusions

HD 52961 and EP Lyr both have rich infrared spectra, and the assembled multi-wavelengthdata show that these evolved objects are surrounded by a stable circumstellar disc. While thebinary nature of HD 52961 is well established, the binarity of EP Lyr is suspected but not yetfirmly proven. The discs must be circumbinary as the sublimation radius of the dust is larger


than the determined (HD 52961) and suspected (EP Lyr) binary orbit.

Recent studies have shown that many of these binary post-AGB systems are already detected(Van Winckel 2003; De Ruyter et al. 2006; Deroo et al. 2006, 2007a, and Chap. 3), but EP Lyrand HD 52961 both show quite distinct characteristics in dust and gas chemistry as well as inphysical properties of their discs.

EP Lyr and HD 52961 are the only stars from the larger Spitzer sample that have clear CO2

gas emission lines in the mid-IR. Our modelling shows that the emission in both stars canbe well fitted and is dominated by 12C16O2 features, but clear detections of other isotopesare present as well. Similar excitation temperatures and column densities are found in bothobjects, but with different ratios for 12C16O2 and 13C16O2. Why these two stars are the onlyones from the larger sample showing strong CO2 features, and if there is any relation with thelow observed infrared flux remains unclear. Similar column densities observed in the otherstars would have been easily detected. One possibility is that the low dust emission in thetwo sample stars, reflected in the low LIR/Lstar, makes it easier for the CO2 gas to becomevisible. This effect is also seen in AGB stars, where CO2 gas emission is strongest in sourceswith the lowest mass-loss rates.

One of the most remarkable features is the clear detection of 18O isotopes of CO2 in bothobjects. Together with HR 4049 (Cami & Yamamura 2001), the strong 16O12C18O band is asystematic feature of the gas emission in the discs of post-AGB binaries when CO2 emissionis detected. The high 18O abundance of HR 4049 derived in an optically thin approximationof a putative nucleosynthetic origin (Lugaro et al. 2005) was not confirmed by the analysis ofCO first overtone absorption (Hinkle et al. 2007) in the same object. It is likely that the CO2

gas is strongly optically thick, also in EP Lyr and HD 52961, so that very rare isotopes aredetectable.

The high-resolution Spitzer spectra also reveal unique solid state features. As in the bulk ofthe disc sources (Chap. 3), crystalline silicate features prevail in both stars, but unlike whatwe found for the larger sample, they proved very hard to model. In HD 52961 we observesome unique strong crystalline features at 11.3 and 16 µm, which could not be reproducedin the modelling, irrespective of the grain size used in the models, shape or assumed grainmodel. In the 2D disc modelling we could not fit the steep rise around 10 µm, without theinclusion of metallic iron. Combining our physical model constrained by the SED, togetherwith our interferometric data, we concluded that the inner dust rim is slightly beyond the dustsublimation radius. This is in contrast to similar binary objects like IRAS 08544-4431 wherethe interferometric data shows that the dusty disc has to start very near to the dust sublimationradius (Deroo et al. 2007a). Assuming a luminosity of 3000 L¯, we find that the inner discradius of HD 52961 is rather large, around 10 AU. The strong 850 µm flux shows that thisobject has a component of very large grains. This contribution was added to the SED fittingby an additional colder Planck curve.

EP Lyr has only a very small infrared excess, but the Spitzer spectrum is very rich in spectraldetails. The most remarkable characteristic is the clear PAH emission, in combination withthe strong crystalline features at longer wavelengths. There is no evidence that the central starevolved into a carbon star when on the AGB, yet unprocessed class-C PAH features are clearlydetected. Whether these PAH species are formed in the circumbinary disc or in a recent, likelypolar outflow of the depleted central star, remains unclear. An extra component of cold dust


is necessary in this object as well, to fit the entire Spitzer spectrum. Unfortunately, EP Lyr istoo faint for the current interferometric possibilities.

The mixed chemistry, the strongly processed cold crystalline silicates and low 12C/13C ratioare in common with the subgroup of silicate J-type carbon stars, which can also displaystrong crystalline material. This corroborates the conclusion that in the latter, the disc iscircumbinary. The abundance studies of J-type carbon stars are not complete enough to probewhether photospheric depletion affected these objects as well.

Both objects are extreme examples of post-AGB binary stars, with characteristics dominatedby the presence of a stable circumbinary disc. This disc environment is, to first order, wellmodelled by assuming a passive, irradiated stable disc. In this paper we corroborate that thisgeometry is ideal to induce strong grain processing and a rich, even mixed chemistry. Weconclude also that a homogeneous disc model is too primitive to model the spectral detailsas evidence for grain settling is strong. The route to PAH formation (and excitation) in theO-rich EP Lyr remains to be studied in detail as PAH emission is only observed in a verylimited number of such sources.

This detailed study of HD 52961 and EP Lyr shows that many questions still remain in ourcurrent understanding of the evolution of a significant number of post-AGB binary stars, andthe impact of the circumbinary discs on the entire system.

AcknowledgementsCG acknowledges support of the Fund for Scientific Research of Flanders (FWO) under the grantG.0178.02. and G.0470.07.

Chapter 5Circumbinary discs in the LMC

This chapter is to be published as

Chemical depletion in the Large Magellanic Cloud: RV Tauri stars andthe photospheric feedback from their dusty discs.

C. Gielen, H. Van Winckel, M. Reyniers, T. Lloyd Evans and the SAGE-Spec team. A&A,2009

5.1 Introduction

With a distance of ∼ 50 kpc (Feast 1999), the Large Magellanic Cloud (LMC) is one of thebest galaxies to study stellar evolution in connection to the lifecycle of dust. The knowndistance allows us to calculate the luminosity for stellar sources and it is the ideal laboratoryto study physical and chemical processes in an environment with a sub-solar metallicity ofZ ∼ 0.3− 0.5 Z¯ (Westerlund 1997). Moreover, it is close enough to allow detailed studiesof individual sources using large-aperture ground based telescopes.

Here we focus on the post-AGB sample in the LMC. The first extragalactic RV Tauri stars inthe LMC were discovered by the MACHO experiment (Alcock et al. 1998). RV Tauri starsare pulsating evolved stars with a characteristic light curve showing alternating deep andshallow minima. The post-AGB status of RV Tauri stars was, at that time, still not generallyacknowledged but the detection of extragalactic pulsators and their large derived absoluteluminosities were in line with the expected evolutionary tracks of post-AGB stars.

Reyniers & Van Winckel (2007) and Reyniers et al. (2007) performed a chemical study ontwo RV Tauri stars, selected from those reported by Alcock et al. (1998), and found that,like in the Galaxy, post-AGB stars are chemically much more diverse than previously antic-ipated. One of the stars, MACHO 47.2496.8, proved to be strongly enhanced in s-process

112 Chapter 5. Circumbinary discs in the LMC

elements, in combination with a very high carbon abundance (C/O > 2 and 12C/13C≈ 200).The metallicity of [Fe/H] =−1.4 is surprisingly low for a field LMC star. The s-processenrichment is large: the light s-process elements of the Sr-peak are enhanced by 1.2 dexcompared to iron ([ls/Fe] = +1.2), while for the heavy s-process (Ba-peak) elements, an evenstronger enrichment is measured: [hs/Fe] =+2.1. Lead was not found to be strongly en-hanced. The patterns can only be understood assuming a very low efficiency of the 13Cpocket (Bonacic Marinovic et al. 2007) which is created during the dredge-up phenomenonand the associated partial mixing of protons into the intershell. It was the first detailed studyof the s-process of a post-AGB star in an external galaxy.

Another object, MACHO 82.8405.15, turned out to be chemically altered by the depletionphenomenon (Reyniers & Van Winckel 2007). Depletion of refractory elements in the pho-tosphere is a chemical process in which chemical elements with a high dust condensationtemperature are systematically underabundant (e.g. Maas et al. 2005; Giridhar et al. 2005).The special photospheric abundance patterns are the result of gas-dust separation in the cir-cumstellar environment, followed by re-accretion of only the gas, which is poor in refractoryelements. The photospheres become deficient in refractories (as Fe and Ca and the s-processelements), while the non-refractories are not affected. The best abundance tracers of the de-pletion phenomenon are the Zn/Fe and S/Ti ratios because the elements involved in eitherratio have a similar nucleosynthetic formation channel, but have very different condensationtemperatures. Intrinsically Fe-poor elements have [Zn/Fe] and [S/Ti] close to solar which isnot the case for depleted objects. With [Fe/H] =−2.6, in combination with [Zn/Fe] =+2.3and [S/Ti] =+2.5, in MACHO 82.8405.15, there is no doubt that the depletion affected thephotosphere of this LMC star very strongly. The very low abundances, as well as the clearcorrelation with condensation temperature, are shown in Figure 5.8, which is a reproductionof the figure in Reyniers & Van Winckel (2007).

Photospheric depletion is surprisingly common in Galactic post-AGB stars (e.g. Giridhar et al.2005; Maas et al. 2005, and references therein). In almost all depleted post-AGB objects,there is observational evidence that a stable circumbinary disc is present (Van Winckel 2007).

Thanks to the efficient infrared detectors of the Spitzer satellite, very sensitive infrared ob-servations allow us to probe for circumstellar dust, even around individual objects in externalgalaxies. The SAGE (Surveying the Agents of a Galaxy’s Evolution) Spitzer LMC survey(Meixner et al. 2006) mapped the LMC, using all photometric bands of the Spitzer IRAC(InfraRed Array Camera, Fazio et al. 2004) and MIPS (Multiband Imaging Photometer forSpitzer, Rieke et al. 2004) instruments. This survey resulted in the detection of over 4 millionsources. Thanks to the release of this database, we found that the LMC RV Tauri stars asdiscovered with the MACHO experiment in the visible, indeed have infrared excesses withvery similar SED shapes as many Galactic post-AGB binaries (see Fig. 5.1).

In this contribution we focus on the MACHO RV Tauri objects in the LMC, and connect thechemical studies based on high-resolution optical spectroscopy to the SED energetics and theinfrared spectra obtained by Spitzer. This research was possible thanks to the SAGE-Specinternational program (http://sage.stsci.edu/ ). The goal of this large spectroscopic infraredprogram is to complement the wealth of photometric data from the SAGE photometric survey,with an extensive spectroscopic follow-up programme using the infrared IRS spectrographaboard of Spitzer (Kemper et al.,in prep.). The main goal of the survey is to determine

5.2. Observations and data reduction 113

1 0 1 2 3 4 5 6[ 3 . 6 ] [ 8 . 0 ] 10 12 34 567J K

Figure 5.1 — Colour-colour diagram indicating the Galactic post-AGB sources with discs (grey circles)and extragalactic RV Tauri stars as presented by Alcock et al. (1998) (black circles). The grey dotsrepresent the LMC objects as found by the SAGE-LMC survey (Meixner et al. 2006).

the composition, origin and evolution, and observational characteristics of interstellar andcircumstellar dust and its role in the LMC. A total of 193 sources, in all stages of stellarevolution, were observed in Spitzer-IRS and MIPS-SED mode.

5.2 Observations and data reduction

5.2.1 Optical high-resolution programme

The optical high-spectral-resolution data were obtained with the UVES (Dekker et al. 2000)spectrograph mounted on the Nasmith focus of UT2 of the VLT, in February 2005 in visitormode under mediocre sky conditions.

To cover a wide spectral domain, two wavelength settings of the spectrograph were used:one with the dichroic, to use both arms of the spectrograph simultaneously (Dic2 860+437nm setting) and one centered on 580 nm using only the red arm. The detailed log of theobservations is given in Table 5.1. The run was plagued with bad weather conditions, so fornone of the three stars discussed here, we could obtain the full optical spectral window.

The data were reduced using the dedicated UVES context in the MIDAS software packageand include the standard reduction steps of echelle data reduction. Spectral normalisationwas performed by fitting polynomial functions through interactively determined continuumwindows. For the lower S/N data, optimal extraction was used. Sample spectra are shown in


Table 5.1 — Log of the UVES observations and the obtained final S/N at a given spectral band. TheS/N is measured in the middle of the spectral window covered.

Star Date UT Exp. Time Wavelength S/N(start) (sec.) (nm)

MACHO 79.5501.13 09/02/2005 4:43 3600 375.8 – 498.3 70670.5 – 1008.4 100

MACHO 81.8520.15 08/02/2005 4:56 3600 478.0 – 680.8 70MACHO 81.9728.14 08/02/2005 00:37 3600 478.0 – 680.8 65

Wavelength (nm)

Figure 5.2 — Sample UVES spectra of the MACHO LMC stars. The spectra were normalised andoffset for illustrative purposes. AC Her is a Galactic source which is a spectral analogue of the LMCsources.

Figures 5.2 and 5.3.


Wavelength (nm)

Figure 5.3 — Sample UVES spectra of the MACHO LMC stars. The spectra were normalised andoffset for illustrative purposes. AC Her is a Galactic source which is a spectral analogue of the LMCsources. The upper spectrum is from HR1017, a F5Iab star with solar composition.


Tabl

e5.

2—

The

nam

e,eq

uato

rial

coor

dina

tesα

and

δ(J2

000)

,eff

ectiv

ete

mpe

ratu

reT

eff

,sur

face

grav

itylo

gg

and

met

allic

ity[F

e/H

]ofo

ursa

mpl

est

ars.

Als

ogi

ven

are

the

tota

lred

deni

ngE

(B−

V) t

ot,t

heen

ergy

ratio

LIR

/L∗,

the

calc

ulat

edlu

min

osity

(com

pute

dby

inte

grat

ing

the

dere

dden

edph

otos

pher

e),

assu

min

ga

dist

ance

ofd

=50

kpc,

and

the

peri

odP

asgi

ven

in(A

lcoc

ket

al.1

998)

.

Nam

eα

(J20

00)

δ(J

2000

)T

eff

log

g[F

e/H

]E

(B−

V) t

ot

LIR

/L∗

L∗

P(h

ms)

(’”

)(K

)(c

gs)

(%)

L¯

(day

s)M

AC

HO

79.5

501.

1305

1418

.1-6

912

34.9

5750

0.5

-2.0

0.14±

0.01

60±

350

00±

500

48.5

MA

CH

O81

.852

0.15

0532

54.5

-69

3513

.262

501.

0-1

.50.

27±

0.03

0±

142

00±

500

42.1

MA

CH

O81

.972

8.14

0540

00.5

-69

4214

.657

501.

5-1

.00.

05±

0.02

53±

342

00±

500

47.1

MA

CH

O82

.840

5.15

0531

51.0

-69

1146

.460

000.

5-2

.50.

05±

0.01

84±

340

00±

500

46.5


Figure 5.4 — Spitzer-IRS spectra of MACHO 79.5501.13, MACHO 81.9728.14 and MA-CHO 82.8405.15.

5.2.2 IRS-spectroscopy

In the follow-up programme SAGE-Spec to the SAGE LMC survey several post-AGB sources,showing RV Tauri-like behaviour, were observed with the IRS instrument aboard the SpitzerSpace Telescope (Houck et al. 2004; Werner et al. 2004). IRS was used in low-resolution SLand LL spectroscopic staring modes. SL (λ=5.3-14.5 µm) and LL (λ=14-44 µm) spectra havea resolving power of R=λ/4 λ ∼ 100. Exposure times were chosen to achieve a S/N ratioof around 60 for the SL modes, LL modes have a S/N of 30. For a detailed description of theobservations we refer to Kemper et al. (in prep.). Spitzer-IRS spectra for three of our samplestars are shown in Figure 5.4.

The extraction was done from intermediate droop data products, version S17.0.4 up to S18.0.2,as delivered by the SSC. SMART reduction package tools (Higdon et al. 2004) and reductiontools developed by the FEPS (the Spitzer legacy program “Formation and evolution of plan-etary systems”) team. For a detailed description of the different reduction steps we refer toHines et al. (2005). The spectra are background-corrected and bad/hot-pixel-corrected anda fixed-width aperture is used for the extraction. The spectra are then combined and ordermatched. For the defringing of the spectra the IRSFRINGE package was used. Calibrationspectral response functions were calculated, derived from standard stars and correspondingstellar models.

In this contribution we limit the IRS spectral analyses to those objects for which we also haveUVES data.


5.3 Photospheric abundance results

One of the features of the spectra displayed in Fig. 5.2 and 5.3 is the general weakness ofmetallic lines. The LMC RV Tauri stars are proven to be spectral analogues of the GalacticRV Tauri star AC Her also shown in the figures. This Galactic RV Tauri star is considered tobe one of the prototypes of its pulsation class, but it is also known to be a strongly depletedobject (Giridhar & Ferro 1989; Van Winckel et al. 1998).

Our analysis of the final product spectra is similar to those used in Van Winckel & Reyniers(2000) and Reyniers et al. (2007). We performed a line identification using the solar linelists of Thevenin (1989, 1990). Weak lines, on the linear part of the curve-of-growth arethe best tracers of the chemical conditions of the photospheres, as they are not saturatedand are formed deeper in the photosphere, where possible non-LTE effects are known tobe weaker. For the chemical analysis, we only used lines with strength below 150 mA andaccurate oscillator strengths (Van Winckel & Reyniers 2000).

As the stars seem chemically peculiar, we performed a relative abundance analysis using theGalactic RV Tauri star AC Her. The quantified abundance analysis was performed using thelatest Kurucz model atmospheres of the appropriate metallicity (http://kurucz.harvard.edu/ ),and the LTE abundance analysis program MOOG (April 2002 version) written by Prof. C.Sneden.

5.3.1 MACHO 79.5501.13

5.3.1.1 Photospheric model

To determine the most appropriate photospheric model, we combined the classical spectro-scopic methods with Balmer line-profile fitting and the constraints from the star being amember of the LMC. We refer to Reyniers et al. (2007) for a more detailed description ofthe methods.

The spectroscopic analysis assumes that the excitation of the Fe I lines follow a Boltzmandistribution and by requiring that the abundances of lines from low and high excitation levelsmust be the same, a spectroscopic temperature is deduced. The gravity is determined byimposing that the abundance of different ions must be the same (e.g. Van Winckel & Reyniers2000). For pulsating stars with large amplitude pulsations and associated shocks, this methodturns out to be too sensitive to non-LTE effects. We therefore refined the effective temperatureand gravity determination by fitting the Balmer δ and γ line profiles with synthetic models.As the object is known to be a member of the LMC, we assumed the distance to be 50 kpc,and determined the absolute luminosity after dereddening of the colours. A surface gravityestimate is then deduced using a standard mass for a post-AGB star of 0.6 M¯. With thisprocess, we find best values of Teff = 5750 K, log g = 0.5 and a [Fe/H] abundance of −1.7.Typical errors are ±250 K for the Teff , and 0.5 for the log g.

5.3. Photospheric abundance results 119

5.3.1.2 Chemical analysis

The results of our quantitative analyses are given in Table 5.3 and displayed in Figure 5.5.The spectral coverage of this object is not optimal as the intermediate spectral window ofUVES with central wavelength of 580 nm was not observed. The results of our chemicalstudy clearly confirm that the object is a spectral analogue to AC Her. Its Fe abundance is low([Fe/H] =−1.8), even for the LMC, but this does not reflect the initial abundance. A clearcorrelation is found when depicting the abundance of a given element to its condensationtemperature (Fig. 5.5): refractory elements are strongly depleted. Zn and S may be tracingthe original metallicity which was likely to be around −0.6. This indicates that the objectwas originally slightly metal deficient compared to the overall metallicity of the LMC.

Also in this LMC source, the depletion is found to be affecting the chemical content of thephotosphere very strongly. For Sc, an element with one of the highest condensation tempera-tures, the abundance is about 250 times lower than what we would expect from a star of thisinitial metallicity ([Fe/H]original =−0.6). Also Y and Ce, the only two s-process elements forwhich we found lines suitable for abundance determination, the underabundance is a factor80 and 20 respectively.

5.3.2 MACHO 81.9728.14


We obtained a good spectroscopic solution for the model atmosphere parameters, showingthat this star is a slightly cooler analogue of the previous object. The only Balmer linescovered are Hβ and Hα and these are affected by emission cores which make them unsuitablefor model parameter determination.


The chemical pattern displayed by this star is depicted in Figure 5.6. This object has thepoorest quality spectrum, and the line-to-line scatter in the abundances remains quite high.

The abundance distribution is different in this object, since we find the Zn abundance tobe very low ([Zn/H] =−1.22). Interpreting this as the initial chemical condition, this staris intrinsically of low metallicity, even for the LMC. The abundance pattern is flat exceptfor the refractory elements with the highest condensation temperature. These refractoriesshow abundances which are significantly lower, down by up to −0.8 relative to the [Zn/H]abundance. Therefore, we interpret these abundances as affected by depletion. The effect hasbeen marginal and only detectable for elements with the highest condensation temperature.

The metallicity dependence of the depletion phenomenon has been found in the Galacticsample as well (Giridhar et al. 2005; Maas et al. 2005), and may be confirmed here: at lowinitial metallicity, the effect is strongly reduced and often only visible for elements with thehighest condensation temperature.


5.3.3 MACHO 81.8520.15


The spectrum of MACHO 81.8520.15 is of high quality which is reflected in the number oflines used in our chemical analyses and in a smaller line-to-line scatter.

The star is found to be somewhat hotter than previous objects. This is also seen directly in thespectra, more specifically in the ratio of line strengths between lines of ionised versus neutralions of the same species. Our best spectroscopic estimate is a Teff = 6250 K with a gravityof log g = 1.0.


Also in MACHO 81.8520.15, the abundances are significantly lower than the mean metallic-ity of the LMC. Assuming the Zn abundance reflects the initial condition, a [Zn/H] = −1.4makes MACHO 81.8520.15 too a star of rather low initial metallicity.

The abundance pattern shown in Figure 5.7 does show a general trend with condensationtemperature. The main difference is that the C abundance turned out to be very high. Thecarbon abundance is determined on the basis of 8 lines which show little internal scatter, sothe high C abundance is certainly significant.

The low Zn abundance can be interpreted either as the initial value, which means that theobject increased its C abundance by a factor of 10, or it is the C abundance which reflects theinitial conditions. In this case the depletion has been efficient to such low temperatures thateven the Zn abundance was affected.


Figure 5.5 — Top: Spectral energy distribution of MACHO 79.5501.13. Triangles represent photomet-ric data and the solid line gives the corresponding Kurucz model. Overplotted in gray the Spitzer-IRSspectrum. Bottom: The abundance pattern in MACHO 79.5501.13 shown as [el/H] as a function of thecondensation temperature.


Figure 5.6 — Top: Spectral energy distribution of MACHO 81.9728.14. Triangles represent photomet-ric data and the solid line gives the corresponding Kurucz model. Overplotted in gray the Spitzer-IRSspectrum. Bottom: The abundance pattern in MACHO 81.9728.14 shown as [el/H] as a function of thecondensation temperature.


Figure 5.7 — Top: Spectral energy distribution of MACHO 81.8520.15. Triangles represent photomet-ric data and the solid line gives the corresponding Kurucz model. Bottom: The abundance pattern inMACHO 81.8520.15 shown as [el/H] as a function of the condensation temperature.


Figure 5.8 — Top: Spectral energy distribution of MACHO 82.8405.15. For a description of the abun-dance determination we refer to (Reyniers & Van Winckel 2007). Triangles represent photometric dataand the solid line gives the corresponding Kurucz model. The newly obtained Spitzer-IRS spectrumis overplotted in gray. Bottom: The abundance pattern in MACHO 82.8405.15 shown as [el/H] as afunction of the condensation temperature.


Tabl

e5.

3—

Qua

ntifi

edab

unda

nce

resu

ltsfo

rM

AC

HO

79.5

501.

13,

MA

CH

O81

.972

8.14

and

MA

CH

O81

.852

0.15

(the

abun

danc

eta

bles

ofM

A-

CH

O47

.249

6.8

and

MA

CH

O82

.840

5.15

can

befo

und

inre

sp.

Rey

nier

s&

Van

Win

ckel

(200

7)an

dR

eyni

ers

etal

.(2

007)

).T

heex

plan

atio

nof

the

colu

mns

isas

follo

ws:

Ngi

ves

the

num

bero

flin

esus

ed;l

og

εis

the

abso

lute

abun

danc

ede

rived

log

ε=

log

(N(X

)/N

(H))

+12;

σlt

lis

the

line-

to-l

ine

scat

ter;

and

[el/H

]ist

heab

unda

nce

rela

tive

toth

eSu

n.Fo

rthe

refe

renc

esof

the

sola

rabu

ndan

cesn

eede

dto

calc

ulat

eth

e[e

l/H]v

alue

s:se

eR

eyni

ers

&V

anW

inck

el(2

007)

.T

hedu

stco

nden

satio

nte

mpe

ratu

res

are

take

nfr

omL

odde

rs(2

003)

,and

refe

renc

esth

erei

n.T

hey

are

com

pute

dus

ing

aso

lar

abun

danc

em

ixat

apr

essu

reof

10−

4at

m.

MA

CH

O79

.550

1.13

MA

CH

O81

.972

8.14

MA

CH

O81

.852

0.15

Teff

=5750

Klo

gg

=0.5

(cgs)

ξ t=

3.0

km

s−1

Teff

=5500

Klo

gg

=1.0

(cgs)

ξ t=

3.0

km

s−1

Teff

=6250

Klo

gg

=1.0

(cgs)

ξ t=

3.0

km

s−1

ion

Nlo

gε

σlt

l[e

l/H]

Nlo

gε

σlt

l[e

l/H]

Nlo

gε

σlt

l[e

l/H]

Sun

Tcond

CI

67.

650.

12-0

.92

27.

520.

04-1

.05

88.

080.

08-0

.49

8.57

40.

Na

I2

5.07

0.21

-1.2

66.

3395

8.M

gI

26.

090.

33-1

.45

16.

78-0

.76

26.

340.

05-1

.20

7.54

1336

.Si

I2

6.48

0.20

-1.0

64

6.69

0.12

-0.8

57.

5413

10.

SiII

26.

180.

06-1

.36

7.54

1310

.S

I5

6.71

0.06

-0.6

27.

3366

4.C

aI

24.

670.

13-1

.69

95.

250.

14-1

.11

124.

860.

19-1

.50

6.36

1517

.Sc

II1

0.19

-2.9

82

1.15

0.17

-2.0

25

1.23

0.20

-1.9

43.

1716

59.

TiI

13.

13-1

.89

5.02

1582

.Ti

II6

2.20

0.25

-2.8

24

3.26

0.29

-1.7

616

3.03

0.18

-1.9

95.

0215

82.

CrI

33.

850.

02-1

.82

34.

390.

30-1

.28

43.

910.

16-1

.76

5.67

1296

.C

rII

103.

760.

16-1

.91

114.

700.

18-0

.97

124.

050.

19-1

.62

5.67

1296

.M

nI

64.

340.

21-1

.05

24.

320.

31-1

.07

13.

58-1

.81

5.39

1158

.Fe

I41

5.85

0.15

-1.6

673

6.43

0.16

-1.0

812

25.

920.

17-1

.59

7.51

1334

.Fe

II14

5.56

0.22

-1.9

514

6.47

0.15

-1.0

423

5.95

0.15

-1.5

67.

5113

34.

NiI

24.

270.

22-1

.98

104.

990.

11-1

.26

64.

470.

12-1

.78

6.25

1353

.Z

nI

33.

980.

04-0

.62

13.

38-1

.22

13.

21-1

.39

4.60

726.

YII

1-0

.26

-2.5

01

0.49

-1.7

52

0.11

0.01

-2.1

32.

2416

59.

ZrI

I1

0.74

-1.8

62.

6017

41.

Ba

II4

0.23

0.06

-1.9

02.

1314

55.

Ce

II1

-0.2

8-1

.86

1.58

1478

.E

uII

20.

120.

08-0

.40

0.52

1356

.


5.4 Infrared spectroscopy

Depletion was found to be very efficient in the RV Tauri stars for which we have UVESspectra available. In the Galaxy, depletion patterns are abundant but only around stars wherestable dusty discs are present. We therefore used the infrared catalogue released by the SAGEteam to investigate the spectral energy distributions. We used the photospheric parametersfrom our UVES study and dereddened the data until the match between the photometry andthe Kurucz photospheric model was optimal. The SED figures are shown in Figures 5.5 to5.8. It is clear that also around these objects, which were selected on the basis of their lightcurve, circumstellar dust is present with specific colours typical of discs. In the next sectionswe discuss our analysis of the spectrophotometric data we obtained for these objects.

5.4.1 MACHO 81.9728.14

For this star we unfortunately only possess spectral information in the 5− 14 µm region. Thespectrum is unique in our sample, and is dominated by strong emission features at 6.2−7.6−8.6 − 11.3 − 12.8 and 14 µm, which can be identified as PAH emission (Tielens 2008, andreferences therein).

The C-C stretching and bending modes produce features with typical central wavelengthsat 6.2 and 7.7 µm. The 8.6 µm feature is due to C-H in-plane bending modes and featureslongward of 10 µm can be attributed to C-H out-of-plane bending modes.

Following the classification as described in Peeters et al. (2002) the PAH emission bands arecharacterised as being of class A. This class A is usually linked to very processed PAHsresiding in interstellar material, directly illuminated by a star.

PAH bands are seen in the infrared spectra of several post-AGB sources, but surprisingly al-most always together with features of silicate dust species. Such mixed chemistry objects in-clude HR 4049 (Johnson et al. 1999; Dominik et al. 2003; Antoniucci et al. 2005; Hinkle et al.2007), and HD 44179, the central star of the ‘Red Rectangle’ (Men’shchikov et al. 2002;Cohen et al. 2004). In these stars the PAH emission is classified as of class B. Class B sourcesare mainly associated with circumstellar material. HR 4049 and HD 44179 are both binarypost-AGB stars surrounded by an O-rich circumstellar disc. The PAH carriers however do notreside in the disc but in a more recent C-rich bipolar outflow. In Chapter 4 we describe somepost-AGB sources where the less frequent class C PAHs are seen. Class A PAH features areseen in other post-AGB sources, both post-AGB stars with evidence for a carbon-rich chem-istry, e.g. IRAS 16594-4656 (Garcıa-Hernandez et al. 2006b; van de Steene et al. 2008) as inmixed-chemistry sources, e.g. IRAS 16279-4757 (Molster et al. 1999; Matsuura et al. 2004).In these two examples there is evidence for a dusty disc/torus and bipolar outflow in whichthe PAH carriers reside.

If the PAH carriers reside in the disc then there has to be a formation process where carbon-rich grains can be formed in an oxygen-rich environment. Such a scenario was proposed byJura et al. (2006) to explain PAH features observed in the oxygen-rich giant HD 233517. Thespectrum of MACHO 81.9728.14 does not show any other features due to C-rich or O-richdust species, such as SiC or silicates, although it cannot be ruled out that these features are

5.4. Infrared spectroscopy 127

Figure 5.9 — Continuum-subtracted spectrum of MACHO 81.9728.14 (black solid line). Below weplot the class A PAH emission features as described in Peeters et al. (2002).

Figure 5.10 — The 10−14−16−19−23−33µm complexes of MACHO 79.5501.13 (left), the Galacticmean spectrum (middle) and MACHO 82.8405.15 (right), continuum-subtracted and normalised. Over-plotted in gray is the mean spectrum of the Galactic sources. For the mean spectrum of the Galacticsources we also plot the comparison with mass absorption coefficients of forsterite (dotted), enstatite(dashed). In the 10 µm complex we plot amorphous olivine instead of enstatite, in dashed lines.

present at longer wavelengths, for which no Spitzer-IRS spectrum were taken. Our chemicalanalysis of the photosphere shows that although the photosphere is affected by depletion,there is no evidence for carbon enhancement (this, contrary to MACHO 81.8520.15).

5.4.2 MACHO 79.5501.13 and MACHO 82.8405.15

MACHO 79.5501.13 and MACHO 82.8405.15 both show clear emission around 10 µm. Thisemission feature is mainly due to amorphous silicates, namely olivine (Mg2xFe2(1−x)SiO4)


and pyroxene (MgxFe1−xSiO3), which peak near 9.8 µm and 18 µm. The observed shouldernear 11.3 µm is caused by crystalline silicate dust, namely forsterite (Mg2SiO4), the Mg-rich end member of the olivine family. We cannot exclude the presence of other amorphoussilicate species, but will focus here only on the olivine and pyroxene members.

MACHO 79.5501.13 shows a strong slope in the continuum, and no strong features of amor-phous or crystalline silicates at wavelengths longer than 10 µm are detected. Some emissionmight be seen around 15 − 19 and 23 µm. This shows that there is little cool silicate dust inthe circumstellar environment. The peak at 36 µm is probably a reduction residual at the endof the spectrum.

In MACHO 82.8405.15 some small features around 15− 19− 23 and 33 µm are seen. Thesefeatures can also be attributed to the crystalline silicates forsterite and enstatite (MgSiO3).The 18 µm amorphous silicate bump is detected, albeit not very strongly.

5.4.3 Comparison with infrared spectra of the Galactic sample

Both MACHO 79.5501.13 and MACHO 82.8405.15 show a strong similarity to our sampleof Galactic post-AGB sources, with similar spectral shapes and emission of amorphous andcrystalline silicates. On average the Galactic sources do show a slightly higher fraction ofcrystallinisation.

In our previous study of the Galactic sample stars (Chap. 3) we defined 6 different complexes(10 − 14 − 16 − 19 − 23 − 33 µm complexes) and compared these to a calculated meancomplex for the 21 Galactic sources. We will use this calculated mean Galactic spectrum tocompare the observed features of these extragalactic sources. For this we subtracted a linearcontinuum from the different complexes and then normalised them. The comparison can beseen in Figure 5.10.

The low signal-to-noise for these stars makes it hard to make a detailed comparison, but wecan already see that several features appear in both Galactic and extragalactic sources. Forboth stars the 10 µm complex is very similar to the Galactic mean complex, although it seemsslightly narrower in MACHO 82.8405.15. The double-peaked structure due to the presenceof both amorphous silicates and forsterite is apparent in both stars.

Around 14 µm we do find the enstatite 13.8 µm, and possibly the 14.4 µm, feature, but notthe strong unidentified feature near 14.8 µm which is seen in the Galactic stars.

For the Galactic sample we found the forsterite 16 µm to be shifted to shorter wavelengths.In MACHO 79.5501.13 we do not find strong evidence for this feature. The peak at 16 µmcould be due to forsterite but note that the noise is large and enhanced by the continuumsubtraction and normalisation. A strong feature at 15.5 µm is seen in MACHO 82.8405.15.If this is due to forsterite the feature is shifted to even shorter wavelengths in this star. Inboth stars there appear features around 19 and 23 µm, but only MACHO 82.8405.15 shows astrong resemblance to the Galactic mean feature. The 23 µm complex in MACHO 79.5501.13seems to be tilted to shorter wavelengths, instead of longer wavelengths. Whether this pointsto the presence of a large enstatite fraction is unclear, since the other enstatite features seemto be only minimal.


The 33.6 µm forsterite feature is clearly present in MACHO 82.8405.15, albeit rather noisy,but can not be detected in MACHO 79.5501.13.

5.5 Full spectral fitting

For MACHO 79.5501.13 and MACHO 82.8405.15 we perform a full spectral fitting as al-ready discussed in Chapter 3. We allow for different dust species, grain sizes (from 0.1 to4.0 µm) and dust approximations (DHS, GRF, Mie) but the lack of strong emission featuresand the low signal-to-noise makes it hard to distinguish between different models, and thecalculated χ2 values are thus quite similar. Similar to what was seen in the Galactic sources(Chap. 3), we find that the best fit is obtained using irregular grain compositions, and not aspherical description, such as Mie theory. In Figure 5.11 and Table 5.4, we give the result ofour calculated best fit, consisting of small grains (0.1− 2.0 µm) in GRF dust approximation.

For MACHO 79.5501.13 we find a good fit to the observed spectrum. The model continuumdoes not follow the strong downward trend, meaning that the modelled continuum temper-ature is too high. The 10 µm feature is nicely reproduced, as are the small bumps near 15,19 and 23 µm. The small feature around 13.5 µm is not explained by the model but could bedue to enstatite, and was also seen in some Galactic sources. We find the model produces asignificant 33.6 µm forsterite feature, which is not observed and thus shows that the amountof cool dust is overestimated.

A good fit is also obtained for MACHO 82.8405.15. Again the 10 µm feature and featuresnear 20 µm are very well fitted. There appears to be a rather strong feature at 15.5 µm whichis not reproduced in the model. Forsterite has an emission feature at 16 µm and for the Galac-tic sources we already found that this features seemed to be shifted to shorter wavelengths(Chap. 3), but the strength was generally well reproduced. The feature observed here howeverseems to be too strong to be explained by this shifted forsterite feature alone.


Figure 5.11 — Best fits for MACHO 79.5501.13 and MACHO 82.8405.15. The observed spectrum(black curve) is plotted together with the best model fit (red curve) and the continuum (black solidline). Forsterite is plotted in dash-dot lines (green) and enstatite in dash-dot-dotted lines (blue). Smallamorphous grains (0.1 µm) are plotted as dotted lines (magenta) and large amorphous grains (2.0 µm)as dashed lines (magenta).


Tabl

e5.

4—

Bes

tfitp

aram

eter

sde

duce

dfr

omou

rful

lspe

ctra

lfitti

ng.L

iste

dar

eth

eχ

2,d

usta

ndco

ntin

uum

tem

pera

ture

san

dth

eirr

elat

ive

frac

tions

(βj

inth

efo

rmul

agi

ven

inSe

ct.3

.7).

The

abun

danc

esof

smal

l(0.

1µ

m)a

ndla

rge

(2.0

µm

)gra

ins

ofth

eva

riou

sdu

stsp

ecie

sar

egi

ven

asfr

actio

nsof

the

tota

lm

ass,

excl

udin

gth

edu

stre

spon

sibl

efo

rthe

cont

inuu

mem

issi

on.T

hela

stco

lum

ngi

ves

the

cont

inuu

mflu

xco

ntri

butio

n,lis

ted

asa

perc

enta

geof

the

tota

lin

tegr

ated

flux

over

the

full

wav

elen

gth

rang

e.T

heer

rors

wer

eob

tain

edus

ing

am

onte

-car

losi

mul

atio

nba

sed

on10

0eq

uiva

lent

spec

tra.

Det

ails

ofth

em

odel

ling

met

hod

are

expl

aine

din

Cha

pter

3.

Nam

eχ

2T

du

st1

Tdu

st2

Frac

tion

Tcon

t1T

con

t2Fr

actio

n(K

)(K

)T

du

st1

-Tdu

st2

(K)

(K)

Tcon

t1-T

con

t2

MA

CH

O79

.550

1.13

5.1

1807

281

8431

45

444

0.75

0.2

20.7

5−

0.25

0.7

50.2

245

8119

284

7432

20

122

0.62

0.2

80.4

5−

0.38

0.4

50.2

8

MA

CH

O82

.840

5.15

3.9

4901

06

91

7372

07

339

0.78

0.2

10.6

9−

0.22

0.6

90.2

127

959

145

5632

58

105

0.86

0.1

0.2

4−

0.14

0.2

40.1

0

Nam

eO

livin

ePy

roxe

neFo

rste

rite

Ens

tatit

eC

ontin

uum

Smal

l-L

arge

Smal

l-L

arge

Smal

l-L

arge

Smal

l-L

arge

MA

CH

O79

.550

1.13

337

3−

448

428

33

27−

1249

12

843

8−

2924

25

838

8−

835

887

8 10

MA

CH

O82

.840

5.15

934

9−

660

630

32

27−

1548

14

922

8−

1631

14

732

7−

928

884

10

11


5.6 2D disc modelling

To model the energetics of the observed RV Tauri stars, we use the disc model as describedin Chapter 2, namely the 2D radiative transfer passive disc model given by Dullemond et al.(2001) and Dullemond & Dominik (2004). Stellar input parameters are the mass, luminosityand effective temperature of the star. For the total mass of the system we use a value of1 M¯. We use the estimated luminosity as given in Table 5.2, which corresponds to a distanced = 50 kpc. The disc parameters consist of the disc size, total mass, the different dustcomponents and grain sizes and the surface-density distribution α. The disc is made upfrom an astronomical silicate dust mixture with a gas-to-dust ratio of 100. We use a value−2.0 < α < −1 for the surface-density distribution, as expected in a disc environment. Theinclination is kept fixed at 45.

The observed feature-to-continuum ratio can be used to determine the grain sizes in the disc,as small grains will produce much stronger feature-to-continuum ratios than larger grains.For the Galactic sources we already modelled (Chaps. 2 and 4), we opted to include anextra opacity source such as metallic iron and/or larger grains (up to 20 µm), but for theseextragalactic stars we need to use considerably smaller grain sizes, ranging from 0.1 µm to5 − 10 µm. In the Galactic sources, grain growth is very efficient in a disc environment andwill produce grains with sizes larger than 10 µm, which will settle towards the midplane. Thiswill cause the disc to be inhomogeneous, consisting of a disc made up from small grains anda cool midplane of larger grains.

For a few stars in the Galactic sample we possessed submillimetre data showing a blackbodyslope from 60 µm to 850 µm, pointing to the presence of extremely large grains, up to cen-timetre size or larger. The presence of such large grains in the disc can not be confirmed forthese stars since we lack data longwards of 40 µm.

The models are calculated using the standard gas-to-dust ratio of 100. Since this value forthe LMC is likely higher (Gordon et al. 2003), we tested how a higher ratio influences themodelling. Increasing the gas-to-dust ratio means that dust formation and grain growth wereless effective, and will thus give less pronounced dust signatures. We find that using a gas-to-dust ratio of 400, we need to increase the total disc mass by about a factor 3 − 4 to get asimilar dust mass, and thus to get a similar fit to observed infrared dust features. Changingthe gas-to-dust ratio does not strongly influence the disc structure.

As noted in Chapters 2 and 4 the models are quite degenerate and equally fitting models withslightly different sizes, total masses and surface-density distributions can be found. Inter-ferometric measurements, combined with the disc model, are invaluable to constrain furtherthe disc geometry (Chap. 4). Unfortunately, the stars are too distant and too faint for currentinterferometric facilities, like the MIDI and AMBER instruments on the VLTI.

5.6.1 MACHO 79.5501.13

For MACHO 79.5501.13 we find we need a very small outer radius (< 100 AU) to fit theobserved SED. This could already have been deduced from the spectrum, where the strongdownward slope of the continuum pointed to the presence of little cool dust in the disc. A

5.6. 2D disc modelling 133

Table 5.5 — Results of our SED disc modelling. Given are the used grain-size distribution, the inner andouter radius (Rin-Rout), the total disc mass m for the homogeneous disc model and the surface-densitydistribution power law α.

grain size Rin −Rout m αµm AU 10−4M¯

MACHO 79.5501.13 0.1− 10 10− 30 1 -1.5MACHO 82.8405.15 0.1− 10 17− 200 6 -1.3MACHO 81.9728.14 (sil) 5 7− 100 2 -1.5MACHO 81.9728.14 (carb) 0.1− 10 10− 100 0.6 -1.0

Figure 5.12 — 2D disc model of MACHO 79.5501.13. Crosses represent photometric data and the grayline the observed Spitzer spectrum. The black solid line give the homogeneous disc model. The dottedline gives the Kurucz stellar model.

good fit was obtained by using an outer radius of 30 AU and an inner radius of 10 AU. Withthese radii the inner rim has a temperature of 1200 K, close to the canonical dust sublimationtemperature for silicates (∼ 1200− 1500 K).

We find that at longer wavelengths the observed Spitzer spectrum falls more quickly thanthe model (even within the error bars), which still overestimates the amount of cool dust.However, this could be a reduction effect, since the observed infrared spectrum at the longestwavelengths is typically less reliable.


Figure 5.13 — 2D disc model of MACHO 82.8405.15. Crosses represent photometric data, and theobserved Spitzer spectrum is overplotted in gray. The solid line give the homogeneous disc model. Thedotted line gives the Kurucz stellar model.

5.6.2 MACHO 82.8405.15

The 2D disc model also provides a good fit to the observed SED of MACHO 82.8405.15.For this star we find an inner radius around 17 AU. At this distance the inner rim reaches atemperature of 1000 K, slightly below the dust sublimation temperature. The outer radius isless well constrained, but we find that it cannot be as small as in MACHO 79.5501.13. In thisstar we see a clear 20 µm silicate feature, which can only be reproduced using an outer radius> 150 AU, with a value of ∼ −1.5 for the surface-density distribution.

5.6.3 MACHO 81.9728.14

Modelling the SED of MACHO 81.9728.14 using the 2D disc model is less straightforward.The main problem lies in the fact that we only possess the PAH dominated Spitzer spectrumfrom 5 − 15 µm, and no information at longer wavelengths. In other post-AGB sources theobserved PAH carriers do not reside in the disc but in a more recent outflow, and so wecannot use the disc model to reproduce this part of the spectrum. Moreover, PAHs are smalland subject to single photon heating effects, so they are not in equilibrium.

If we do try to fit the SED with a standard silicate disc, we find that the model gives a strong10 µm silicate feature, which should have been detected in the Spitzer spectrum. If the PAHfeatures do come from the disc, then the silicate grains must be either very large, to avoidproducing a strong silicate feature, or very optically thick. Some small grains, however, haveto be present to explain the observed near-IR excess. Another explanation would be that thedisc is made up from another, probably carbon-rich, dust species. The chemical composition

5.6. 2D disc modelling 135

Figure 5.14 — 2D disc model of MACHO 81.9728.14. Crosses represent photometric data, and theobserved Spitzer spectrum is overplotted in gray. The solid line give the homogeneous silicate discmodel. The dotted line gives the Kurucz stellar model.

of the photosphere does not show evidence for a possible strong carbon enhancement.

A disc consisting of 5 µm grains, and with inner and outer radii of 7 and 100 AU, proves agood fit to the observed spectrum. A similar fit however can be found using a carbon-richdisc, with grain sizes of 0.1− 10 µm and a disc size of 10− 100 AU.

5.6.4 MACHO 81.8520.15

The light curve of MACHO 81.8520.15 was studied by Alcock et al. (1998). The classifica-tion of the light curve is uncertain but the period of 42.1 days makes it, together with thecalculated stellar parameters, a possible RV Tauri candidate. This star has a very low ampli-tude light curve with slight variability in the depth of its minima, but the period seems to berelatively stable. Alcock et al. (1998) suggest that it also might be classified as a long-period,low-amplitude W Vir pulsator.

Looking at the spectral energy distribution of MACHO 81.8520.15 (Fig. 5.7), we see thatthe star has a very small infrared excess, starting near 8 µm. This is very different from thetypical strong infrared excesses seen in RV Tauri stars. The SED resembles that of debris discstars. These are stars surrounded by a circumstellar disc of dust and planetesimals, clearedof the gas content. These discs are the last phases of disc evolution and can even possessplanetary systems. Debris discs have been found around main-sequence stars of all spectraltypes, including brown dwarves.

The small infrared excess could also point to the presence of cool material at larger distancesfrom the star, a remnant from an outflow during the previous AGB phase. Such a detached


shell of dust is seen around single-star post-AGB sources, but usually gives rise to a stronginfrared excess at longer wavelengths.

But since the stellar photosphere is depleted, a process which likely only occurs by interactionwith a disc, a disc must be present, or was present in the past. We therefore interpret the dustexcess as coming from a very processed disc.

Unfortunately the lack of photometric or spectroscopic data at long wavelengths makes itimpossible to model this star with a disc or outflow model.

5.7 Discussion

The stellar parameters derived from high-resolution optical spectra, combined with amplephotometric data, allow for a good estimate of the total reddening of the object. The highluminosities and the effective temperature prove the suspected post-AGB nature of these stars.All sample stars have effective temperatures ranging from 5750 to 6250 K and luminositiesbetween 4000 and 5000 L¯, assuming a typical LMC distance of 50 kpc.

Our chemical study shows that, also in the LMC, the photospheres of RV Tauri stars are com-monly affected by the depletion process. The analysis of the Spitzer observations shows that,in the LMC as well as in the Galaxy, this depletion process is very closely related to thepresence of a stable, dusty circumstellar disc. The postulated long life-time of the dust nearthe luminous star is a favourable circumstance for the gas-dust separation to occur and forsubsequent gas accretion (Waters et al. 1992). For three sample stars we possess Spitzer low-resolution infrared spectra. Two sample stars have spectra strongly resembling our previouslystudied Galactic post-AGB disc sources, with clear emission features due to amorphous and,in lesser degree, crystalline silicates. The atmospheric abundances, combined with the in-frared spectral characteristics and the evidence for dust processing, show that RV Tauri starsin the LMC have, on average, very similar observational characteristics to the dusty RV Tauristars in our own Galaxy.

Based on our very limited sample here, the depletion patterns as observed in the LMC differsignificantly from star to star, as was also observed in the Galactic sample.

The most remarkable object in the infrared is MACHO 81.9728.14, which is dominated byemission peaks due to PAHs and does not show evidence for the presence of oxygen-richdust species. In the photosphere, there is no evidence for an enhanced carbon abundance.The object shows some evidence of depletion, but only marginally. The abundance pattern isflat with abundance around−1.0 and−1.2, with depletion which affects only the elements ofthe highest condensation temperature. With a [C/H] =−1.0 (based on two lines), the carbonfollows this trend, and we interpret this total distribution as coming from an object with lowinitial metallicity. Despite this low initial metallicity, the dust excess is significant and theLIR/L∗ of 53% shows that the inner rim of the dusty disc covers a wide solid angle as seenfrom the star. The PAH carriers can be categorised as being of class A, which are the stronglyprocessed interstellar PAHs. So far, it remains unclear whether the PAH carriers reside in thedisc, or more likely, in an outflow from the central star.

Also the abundance pattern of MACHO 81.8520.15 is remarkable. The star shows a very high


[C/Zn] ratio and a very clear correlation of the underabundances with the condensation tem-perature. If we interpret the high [C/Zn] value as due to depletion, the Zn abundance does notreflect the initial condition, but must also be affected by gas-dust separation and subsequentaccretion. This is remarkable as Zn is an element with one of the lower dust condensationtemperatures. The alternative interpretation, in which MACHO 81.8520.15 is an intrinsicallylow metallicity object and carbon is enhanced by the 3rd dredge-up, is problematic: in sucha scenario the s-process elements should be strongly enhanced as well, which is clearly notobserved.

Interestingly MACHO 81.8520.15 has a very small infrared excess, which seems to contradictthe presence of a stable dusty disc. However, as the star is depleted we postulate that a discmust have been present at some time, and is now strongly evolved. The SED does showa strong resemblance to the debris discs seen around young stellar objects. These are olddiscs, in which grain growth has already formed dust particles of considerable size (rocksand possibly even planetesimals), and the gas component has been removed. If so, this wouldbe the first time this late stage of disc evolution has been seen around a post-AGB star. Thelow Zn abundance would then indicate that the active gas-dust separation and accretion of thecleaned gas also takes place in colder regions of the evolving disc.

MACHO 79.5501.13 and MACHO 82.8405.15 are very similar in photospheric chemical com-position, SED, and in infrared spectral appearance to the strongly depleted Galactic objects.The underabundance in some elements is a factor of 1000 smaller than the solar value. Theinitial metallicity of both objects is harder to recover, but the flat abundance distribution ofC, Zn, S and Na suggests that both stars have an intrinsic metallicity of about −0.5, which isonly slightly lower than the average value of the LMC.

The spectral energy distributions of the sample stars are well modelled using a passive discmodel. The inner radius is close to the dust sublimation radius for all stars, while the outerradius is poorly constrained, but a good fit is obtained with outer radii around 100− 200 AU.We find, in comparison to most Galactic sources, that there is no need for an extra opacitysource, such as metallic iron or larger grains, to reproduce the observed feature-to-continuumratio of the dust emission. This could be an effect of the lower metallicity of the LMC sam-ple which reduces the formation of metallic iron. Dust formation itself, and the subsequentprocessing, has been efficient enough to produce the large scale height needed for explainingthe infrared luminosity. Unfortunately, we lack submillimetre data to probe the large graincomponent. Individual crystalline silicate bands do show some differences from the Galacticmean spectrum. To deduce whether this is a general trend in the LMC, and thus points toa different dust composition, grain size or dust model, we need to considerably enlarge theLMC sample. This will be done in a future study (see Chap. 7).

5.8 Conclusions

Clearly the formation of stable dusty discs and the significant feedback of this disc on thecentral star is not exclusive to our Galaxy alone. Four out of five LMC RV Tauri objects, forwhich we have high resolution data, are found to be strongly affected by the depletion processin which the atmospheres became poor in refractory elements. Moreover, the infrared colours


and spectral data show that three sources are surrounded by a highly processed, stable discin which dust processing has been efficient. In one source PAH particles are formed, whilethere is no other evidence for intrinsic carbon enhancement in the photosphere. This objectis intrinsically the most metal poor object of the sample. For MACHO 81.8520.15 we onlyfound a small dust excess at 8 µm and we interpret this as being the most evolved disc. In thisstar, even the Zn abundance is affected by depletion which means that the gas-dust separationoccurred at low temperatures.

Overall the RV Tauri stars in the LMC display many characteristics of their Galactic peers.Whether these extragalactic stars also reside in a binary system remains unclear with theavailable data at hand. The low magnitude of these stars does not allow for a long-term radialvelocity monitoring programme but, given the strong similarities to the Galactic binaries, abinary evolutionary channel is very likely needed to understand the RV Tauri stars in the LMCas well.

5.9 Appendix: Expanding the SAGE-Spec sample

In a recent follow-up to the SAGE-LMC survey, 13 possible binary post-AGB sources wereobserved with Spitzer-IRS. This follow-up program is called “SAGE-Spectroscopy: the lifecycle of dust and gas in the Large Magellanic Cloud” (http://sage.stsci.edu/index.php). Itexploits the SAGE-LMC program by conducting a comprehensive spectroscopic study, withthe goal to determine the composition, origin, evolution, and observational characteristics ofinterstellar dust and its role in the LMC (Kemper et al., in prep.). Of the 13 sample stars, 6 areconfirmed to be of RV Tauri nature and periods are calculated on the basis of their light curve(Alcock et al. 1998). Three are classified as variable stars in the SIMBAD AstronomicalDatabase, but periods are not yet derived. Stellar parameters will be determined using UVEShigh-resolution or SSO/SSAO low-resolution optical spectra (see Chap. 7.4).

Observations were made in low-resolution SL and LL spectroscopic staring mode. SL (λ=5.3-14.5 µm) and LL (λ=14-44 µm) spectra have a resolving power of R=λ/4λ ∼ 100. Exposuretimes were chosen to achieve a S/N ratio of around 60 for the SL modes, LL modes have aS/N of 30. Ten of these stars have overlap with the sample stars for which we have opticalspectra, and thus known stellar parameters. Stellar parameters for these LMC Spitzer sourcesare given in Table 5.6.

Our preliminary study on these LMC sources shows they have a wide range in spectral typesof central stars (between B and K), as well as in luminosity (2000 − 12 000 L¯), indicat-ing that discs are detected over a wide range of evolutionary phases along the HR-diagram.This wide range of spectral types of irradiating sources, as well as their accurate evolutionaryposition, allows for the first time a systematic large-scale correlation study of the disc prop-erties in relation to the fundamental characteristics of the central stars. This will help us findevolutionary trends within the sample and describe different distinct phases in the life cycleof discs around evolved objects, as is already done for discs around young stellar objects.Another advantage of this LMC sample is that we can study the influence of the low LMCmetallicity and the impact of a metal poor environment on the evolution of the binary systemand the surrounding disc.

5.9. Appendix: Expanding the SAGE-Spec sample 139

Figure 5.15 — Spitzer-IRS low-resolution spectra of our LMC post-AGB binaries. The spectra arenormalised and offset for comparison.

A quick look at the spectra (Fig. 5.15) shows their strong resemblance to our Galactic post-AGB binaries. Nearly all stars show features of silicate dust species, with a clear 10 µmamorphous silicate feature. A few sources (HV 12631, SAGE 050830 and SAGE 054310)also show the high degree of crystallinity seen in some of the Galactic sources, such asIRAS 16230, IRAS 17038 or RU Cen (see Chap. 3).

Two clear outliers are visible: MACHO 81.9728.14 shows no silicate features but is domi-nated by PAH emission bands. This star was described in detail in Chapter 5. MSX LMC 949


Table 5.6 — The name, equatorial coordinates α and δ (J2000), effective temperature, luminosity, andperiod of our LMC sample stars. First order determinations of the effective temperature and luminosityare given, as described in Section 7.4. Periods are taken from Alcock et al. (1998). “?” means the star isclassified as variable but no value for the period is known. For stars with a “-” no information is knownon the variability.

Name α (J2000) δ (J2000) Teff L∗ P(h m s) ( ’ ”) (K) L¯ (days)

HV 2281 05 03 05.0 −68 40 24.9 6000± 40 2410± 60 31.7SAGE 050830 05 08 30.6 −69 22 37.4 10000± 300 11400± 800 -HV 915 05 14 18.1 −69 12 34.9 5750 5000± 500 48.5HV 2444 05 18 45.5 −69 03 21.7 6000± 160 2810± 60 ?HV 5829 05 25 19.5 −70 54 09.8 6500± 500 2730± 60 ?HV 2522 05 26 27.2 −66 42 58.7 5750± 10 3010± 30 ?SAGE 052707 05 27 07.1 −70 20 02.1 − − -MACHO 82.8405.15 05 31 51.0 −69 11 46.4 6000 4000± 500 46.5HV 12631 05 39 33.2 −71 21 55.5 − − 31.1MACHO 81.9728.14 05 40 00.5 −69 42 14.6 5750 4200± 500 47.1MSX LMC 949 05 40 14.8 −69 28 49.3 − 5886± 160 -SAGE 054310 05 43 10.9 −67 27 28.0 3750± 130 7300± 140 -HV 2862 05 51 22.3 −69 53 51.1 − 2568± 100 ?

shows some features due to silicate dust species but the bands are very weak and broad incomparison to the other sample stars, especially the 10 µm amorphous silicate feature, whichis barely visible.

5.9.1 SAGE 054310

For the 13 possible LMC post-AGB binaries with a disc, we performed an already more indepth pilot study on one star which shows a strong resemblance to our Galactic post-AGB discsources, both in the SED as in the infrared spectrum. SAGE 054310 shows a strong infraredexcess with LIR/L∗ = 27% (Fig. 5.16). This strong excess is indicative of a disc source andis comparable to the Galactic sample, where an average of LIR/L∗ = 50% was found. Usingthe method described in Section 7.4, SAGE 054310 was estimated to be of spectral type K,with an effective temperature Teff = 3750 ± 130 K and luminosity L∗ = 7300 ± 140 L¯.Stellar parameters can be found in Table 5.7

5.9.1.1 Dust approximation models

Looking at the spectrum of SAGE 054310 we define six different regions with strong emis-sion due to amorphous and crystalline silicate dust, which we call the 10-14-16-19-23 and33 µm complexes. The computed shape and central wavelength of crystalline silicates is


Figure 5.16 — Left: Spectral energy distribution of SAGE 054310. In black the dereddened Kuruczmodel, the gray line gives the observed Spitzer-IRS spectrum. Right: Spitzer-IRS low-resolution spec-trum of SAGE 054310 compared with Spitzer spectra of two Galactic post-AGB disc sources, ST Pupand EP Lyr. The spectra of ST Pup and EP Lyr are scaled and offset for comparison.

Table 5.7 — The name, effective temperature Teff , surface gravity log g and metallicity [Fe/H]. Alsogiven are the total reddening E(B − V )tot, the energy ratio LIR/L∗ and the calculated luminosity L∗,assuming a distance of d = 50 kpc.

Name Teff log g [Fe/H] E(B − V )tot LIR/L∗ L∗(K) (cgs) (%) L¯

SAGE 054310 3750± 130 1.5 -0.3 0.17± 0.03 27± 4 7300± 140

highly dependent on the adopted dust approximation which makes these features easy toidentify in the spectra. We therefore compare the strongest observed features with forsteritein CDE, GRF and DHS dust approximation and different grain sizes (0.1 − 2.0 µm) (seeFig. 5.17). The features at 23 and 33 µm allow for the best comparison since here the con-tamination with emission of enstatite is only minimal. A similar comparison with enstatitegrains is not possible since all prominent enstatite features are blended with forsterite.

We find that the 19 µm feature is clearly best fitted with CDE or GRF grains. The 23 µmfeature shows that the best fit is obtained when using small DHS or GRF grains and the 33 µmfeature excludes large GRF grains. At 28 µm it is not that clear which dust approximationexplains best the observed feature. A good fit is also obtained using CDE grains. Thisapproximation however, is calculated in the Rayleigh limit (for grains < 0.1 µm) and thusdoes not allow to study grain growth.

Since we cannot say which dust approximation proves the best match, we opted to perform afull spectral fitting, were we test both GRF and DHS dust grains. This full spectral modellingis described in Section 5.9.1.3.


Figure 5.17 — Normalised and continuum-subtracted emission features of SAGE 054310, together withmass absorption coefficients for different forsterite shape distributions. CDE grains are plotted in red,GRF in blue and DHS in green. 0.1 µm grains are plotted in full lines, 2 µm grains in dashed.

5.9.1.2 Comparison with Galactic sources

The infrared spectrum of SAGE 054310 strongly resembles some of previously studied Galac-tic post-AGB disc sources (Chaps. 3 and 4). The overall spectral shape of SAGE 054310follows very well that of ST Pup, while the observed strong emission features are similar tothose seen in EP Lyr, longwards of 18 µm (see Fig. 5.16).

In our previous study of the Galactic sample stars (Chap. 3), we compared the different com-plexes described in Sect. 5.9.1.1 to a calculated mean complex for the 21 Galactic sources.We will use this calculated mean Galactic spectrum to compare the observed features of thisextragalactic source. For this we subtracted a linear continuum from the different complexesand normalised the resulting spectra. The comparison can be seen in Figure 5.18.

The 10 µm complex of SAGE 054310 is very similar to the Galactic mean spectrum, withstrong features of amorphous and crystalline olivine. In this star the contribution of forsteriteseems to be stronger than the mean spectrum, where the amorphous olivine feature is slightlystronger. For the 14 µm complex we do see some emission around the first peak around13.7 µm, but we cannot say this feature is clearly present. Emission is seen around the secondpeak at 14.7 µm, but clearly has a different shape. This second peak remains unidentified, butis probably due to a different dust species. The Galactic sample showed a prominent feature at16 µm, probably due to forsterite, but, looking at the spectrum of SAGE 054310, this featureis not detected here. The peak at 15.5 µm seen in Figure 5.18 could be due to enstatite, butis more likely noise which is enhanced by the continuum subtraction and normalisation. The19 µm complex shows some differences to the mean spectrum. It seems to be slightly shiftedtowards longer wavelengths but this could just be due to the continuum subtraction. In themean spectrum the complex shows a maximum on the short-wavelength side whereas this starhas a peak towards longer wavelengths. Unfortunately, the noise level at this point is quitehigh, making it hard to attribute this to a different forsterite-enstatite ratio or a difference in


Figure 5.18 — The first and third column give the 10 − 14 − 16 − 19 − 23 − 33µm complexes ofSAGE 054310, continuum-subtracted and normalised (black). Overplotted in gray the mean spectrumof the Galactic sources. In the second and fourth column we give the mean spectrum of the Galacticsources, compared with mass absorption coefficients of forsterite (dotted) and enstatite (dashed) in CDEapproximation. For the 10 µm complex we compare with amorphous olivine (dashed) and forsterite(dotted).

GRF-DHS dust approximation. The overall shape of the 23 µm complex resembles the meanspectrum, but seems to be less broad and slightly blueshifted. This was also seen in someGalactic stars, such as IRAS 19125+0343. The noise level in the 33 µm regime is substantialbut we can still see that this complex follows the mean Galactic spectrum, with a strongforsterite feature.

5.9.1.3 Full spectral fitting

We use the fitting procedure as described in Chapter 3 to model the observed spectrum ofSAGE 054310. Using this procedure we found that for the Galactic sample the best fit, onaverage, was obtained using relatively large grains (> 0.1 µm) in an irregular GRF dustapproximation.

For SAGE 054310 we find the best fit is obtained with small silicate grains (0.1 − 1.5 µm)in DHS approximation. The difference in χ2 with models using GRF grains is howeververy minimal. For this fit we find very low average dust and continuum temperatures ofrespectively ∼ 150 K and ∼ 200 K. About 65% of the dust is in amorphous form, 35% isin crystalline form. Of the crystalline fraction, 20% consists of small grains (0.1 µm). Thebest fits are obtained using the standard Mg-Fe amorphous silicates over the purely Mg-richamorphous silicates.

From Figure 5.19 it is clear that the model strongly underestimates the features at longerwavelengths, and thus underestimates the amount of cool grains in the disc. This was also


Figure 5.19 — Best model fit for SAGE 054310. The observed spectrum (black curve) is plotted to-gether with the best model fit (red curve) and the continuum (black solid line). Forsterite is plotted indash-dot lines (green) and enstatite in dash-dot-dotted lines (blue). Small amorphous grains (0.1 µm)are plotted as dotted lines (magenta) and large amorphous grains (1.5 µm) as dashed lines (magenta).

seen when modelling the spectrum of EP Lyr (see Chap. 4). The central wavelengths of thefeatures are relatively well reproduced, showing that the correct dust species are used in themodelling. Also the strong 10 µm complex is nicely fitted.

5.9.1.4 2D disc modelling

We use the disc model as described in Chapter 3. Stellar input parameters are the mass,luminosity and effective temperature of the central star. For the mass of the system we usea value of 1 M¯ and we use the calculated value Teff = 3750 K. We use the integratedluminosity L = 7300 L¯, assuming a distance d = 50 kpc. We keep the value of the outerradius fixed at 500 AU and use a value of -1.0 for the surface-density distribution. The disc ismade up from an amorphous silicate-forsterite dust mixture with a gas-to-dust ratio of 100.

Looking at the SED we find that the feature-to-continuum ratio is very high, in comparison towhat was seen for most of the studied Galactic sources. For the Galactic sources we opted toinclude an extra opacity source such as metallic iron and/or larger grains (up to 20 µm), buthere we need to use considerably smaller grain sizes. We tested both grain-size distributionsof 0.1− 5 µm and 0.1− 10 µm, and found that the first reproduces best the observed featuresbut not the far-IR slope of the spectrum (Fig. 5.20: top panel). Grain growth in the disc willproduce grains with sizes larger than 5 µm, which will settle towards the midplane. Addinga 120 K blackbody to represent these cool larger grains gives a good fit to the observed SED.The bottom panel of Figure 5.20 shows a disc model with a homogeneous distribution ofgrain with sizes of 0.1 − 10 µm, so without a vertical distribution in grain size. This model


Table 5.8 — Results of our SED disc modelling. Given are the used grain-size distribution, the innerand outer radius (Rin-Rout), the total disc mass m for the homogeneous disc model, the surface-densitydistribution power law α, the blackbody temperature and the inclination of the system.

grain size Rin −Rout m α Tbb iµm AU 10−3M¯ K

model A 0.1-5 22− 500 1.0 -1.0 120 0− 55model B 0.1-10 27− 500 3.0 -1.0 - 0− 50

does fit the far-IR part of the spectrum nicely, but cannot account for the high feature-to-continuum ratio. For a few stars in the Galactic sample, we possessed submillimetre datashowing a blackbody slope from 60 µm to 850 µm, pointing to the presence of extremelylarge grains, up to centimetre size or larger. The presence of such large grains in the disc cannot be confirmed for SAGE 05430 since we lack data longwards of 40 µm.

Both models have an inner radius of∼ 25 AU, which corresponds to an inner rim temperatureof ∼ 800 K, well below the dust sublimation temperature for silicates (∼ 1200 K). Along themidplane the temperature decreases from 800 K to about 130 K. The discs are in both casesoptically thick (1 < τ < 10) before 40 µm in the orbital plain, after that τ decreases fast(Fig. 5.21).

5.9.1.5 Discussion

SAGE 054310 shows a remarkable similarity with some of our highly crystalline Galacticpost-AGB disc sources, showing that, in the LMC also, dust processing is very effective.The SED is nicely fitted using a passive disc model, with disc parameters similar to thoseobserved in our Galaxy.

Both in the full spectral fitting and in the SED disc modelling, there is no need to includemetallic iron. This is in contrast with the Galactic sources where we generally found a betterspectral fit including this dust species. Also, small grains are likely quite abundant in thissource.

To make a detailed comparison between the observed disc characteristics of these extragalac-tic LMC stars and the previously studies Galactic post-AGB binaries, we need to significantlyincrease the LMC sample.


Figure 5.20 — 2D disc model of SAGE 054310. Crosses represent photometric data, and the observedSpitzer spectrum is overplotted in gray. For the top panel (model A) the dashed line represents thehomogeneous disc model consisting of grains of 0.1 − 5µm. The solid line give the total disc modelwith the added blackbody of 120 K to represent the large cool grains. In the bottom panel (model B)the solid line give the homogeneous disc model consisting of grains of 0.1− 10µm.


Figure 5.21 — Radial optical depth versus wavelength for disc model A (full line) and model B (dashedline).

Chapter 6Conclusions

In this work we addressed some of the key questions on the late evolutionary phases of aparticular class of evolved stars. At the beginning of this thesis research, there was strongevidence that these stars probably all reside in a binary system and are surrounded by a stabledusty circumbinary disc. This work further corroborates this disc hypothesis. The origin andevolution of this disc is still unknown, but appears to play a lead role in the further evolutionof the entire binary system.

Our Galactic sample consists of 21 possible evolved sources, selected from the larger samplepresented in De Ruyter et al. (2006), with infrared colours indicative of a stable dusty disc.The broad-band SEDs of our sample stars have very similar characteristics, with on averagevery high LIR/L∗ values: 60% of our sample stars have LIR/L∗ ≥ 50%. Several sourceshave been studied using high-resolution spatial interferometric data, and they all show N-band disc diameters of ∼ 40 AU (Deroo et al. 2006; Deroo 2007).

The Spitzer high- and low-resolution spectral scans, however, show an amazing diversityin spectral-continuum shapes, but with common observed strong and specific dust-emissioncharacteristics. In Chapters 2 and 3 we showed that all sample stars show signatures ofoxygen-rich dust species in their infrared spectra, more specifically amorphous and crys-talline silicates. Only one star displays evidence for carbon-rich dust species, namely EP Lyr(Chap. 4). The magnesium-rich end members of crystalline silicates seem to prevail in thespectra, with strong emission due to forsterite, and to lesser extent enstatite. From our fullspectral fitting we also find that the amorphous dust might be Mg-rich, instead of the morecommon magnesium-iron amorphous silicates. Detailed chemical studies of these stars (e.g.Maas et al. 2005) have shown that the stellar photospheres are devoid of iron, with the ironinstead being locked up in the dust of the disc. If no iron is detected in the observed dustspecies, this could imply that iron is stored in the form of metallic iron or iron oxide, whichhave no distinct observable dust signatures.

We find that the discs are the ideal environment for strong dust grain processing to occur,both in grain size and in crystallinity. For our sample stars the mass fraction in crystalline

150 Chapter 6. Conclusions

grains ranges from 10 to 60%, with typical values around 35%, which is much higher thanwhat is observed in outflow sources. The observed dust emission features can be explainedby very irregular dust grains. The observed peak-to-continuum ratio shows that, on aver-age, grain sizes above 0.1 µm prevail. Since the dust features seen in the infrared spectraprobably originate from an optically-thin upper layer of the disc, we are only sensitive to themicron-size grains in our Spitzer study, but submillimetre data, presented in Chapter 2 and 4show that larger grains (up to cm-sizes and possibly larger) must be present as well. Severalsources show strong crystalline emission features at longer wavelengths, coming from rathercool crystalline dust grains in the disc. Crystalline dust may be formed close to the inner rin,where temperatures are high enough for crystallisation to occur. The presence of cool crys-talline grains shows that either radial mixing must be very efficient or that the grains have acrystallinisation process which is already efficient at low temperatures.

EP Lyr, discussed in detail in Chapter 4, is the only star in our sample that shows emissiondue to both oxygen- and carbon-rich dust species. The observed bands at 8 and 11 µm canbe identified as being due to the elusive class C PAHs. Longwards of 11 µm the spectrumis dominated by features of CO2 gas and crystalline silicates. Such mixed chemistry is alsoobserved in a few other post-AGB binaries, such as HD 4417 and HR 4049, showing, how-ever, other PAH families. For these sources, the carbon-rich species are located in an outflow.The oxygen-rich disc is believed to have antedated the carbon-rich transition of the centralstar. For EP Lyr, there is no evidence that the star evolved into a carbon star on the AGB.Whether the PAH carriers in EP Lyr reside in the disc or an outflow remains unclear, andfurther data, preferably polarimetry, is needed to solve this question. If the carriers do residein the disc, an unknown formation mechanism has to occur to produce carbon-rich carriers inan oxygen-rich environment.

CO2 gas emission is only clearly seen in two of our sample stars, EP Lyr and HD 52961,and possible weakly in a third (IRAS 10174-5704). This CO2 traces the gas component ofthe circumstellar disc. From our detailed fitting in Chapter 4, we find that the emission isdominated by features of 12C16O2, 13C16O2 and 16O12C18O. The gas is optically thick, sothe strength cannot easily be converted to isotope ratios. In the Sun 16O/18O is ∼ 500, soit is remarkable we can detect these 18O signatures. CO2 gas emission is seen in only twoother post-AGB sources, again HD 4417 and HR 4049. In our modelling, the temperaturesneeded to excite the gas are relatively high, showing that the gas must reside close to the star.Whether this means the gas is located in the inner disc gap, the hot parts of the disc such asthe inner rim, or an outflow is not yet certain.

Our radial velocity monitoring program is still ongoing, but the high binarity rate found sofar is consistent with all these post-AGB sources indeed residing in a binary system. Forone of our sample stars, RU Cen, we derived new orbital parameters in Chapter 2. We findsurprisingly high eccentricities, and orbits which are too short to have accommodated a full-grown AGB star. This shortcut of the AGB evolution is further strengthened by the lack ofthird dredge-up elements, such as the low observed 12C/13C ratio we derive for some stars(Chap. 4). From the mass functions and the estimated inclinations of the system, as derivedin Chapter 2, we find that the companion is probably an unevolved main-sequence star. Theprimary stars would normally evolve to become carbon stars, but given the oxygen-rich cir-cumstellar environment, it is clear that this has not happened. The only exception beingEP Lyr, where the observed carbon-rich dust species could indicate a carbon-rich outflow, as

151

seen in HD 44179, the central star of the Red Rectangle (Cohen et al. 2004; Witt et al. 2008).It is clear that the binary interaction has an immense impact on the late evolutionary stagesof these particular stars, and is responsible for the observed shaping of the circumstellar en-vironment.

To determine the structure of the circumbinary discs, we performed a 2D disc modelling onthe observed photometry and spectroscopy of a few sample stars (Chaps. 2, 4 and 5). Wefound that a good fit is obtained with this passively irradiated flared disc with a puffed-upinner rim. Submillimetre data indicates the presence of extremely large grains (centimetre-size and larger) in these discs. These large grains will settle quickly to the midplane, creatingan inhomogeneous disc where the disc structure is determined by the smaller grains (0.1 −20 µm), and the total mass by the large cool grains in the midplane. The inner radius of thedisc is found slightly below dust sublimation radius, but the outer radius is unfortunately lesswell constrained. Additional high-spatial information from interferometric measurements isneeded to remove some of the model degeneracies and determine the exact disc structure. InChapter 4 we compare the interferometric observations of HD 52961 with a 2D disc model,and find that the disc model is, in first order, consistent with the observed visibilities, but failsto reproduce the non-homogeneous distribution of the dust species and the observed intensityprofile. Clearly, additional physics needs to be included in the disc modelling to get a goodfit to each individual source. This was also the case for EP Lyr, where the inclusion of anoutflow might improve the fit to the SED considerably.

Both from our mineralogy study and from our disc modelling, we found that the chemicaland physical properties of these discs around evolved binaries are very similar to those seenin protoplanetary discs around young stellar objects, which is remarkable since these discshave very different formation histories. In YSOs the disc is a relic of star formation, andmainly consists of ISM dust species such as silicates and carbon-rich species, mainly PAHs.Since the exact formation history of the circumbinary discs around evolved stars it not known,we have no clear view on the initial dust species. This makes it hard to determine whetherthe observed dust species and grain-size distribution is already present at disc formation orindicative of disc evolution. In both disc types, dust processing is very effective: largergrain sizes (> 0.1 µm) prevail and the crystallinity fraction is quite high, although slightlylower values are seen in the protoplanetary discs. The observed correlation between theevolutionary phase and the disc parameters found for YSOs, is not yet observed in our limitedsample of evolved binary stars.

Recently, similar objects with evidence for stable discs have been found in the LMC. Forthese extragalactic stars, the known distance to the LMC allows us to calculate the luminos-ity, which confirms the evolved post-AGB phase of the objects. Chemical studies of our fourextragalactic sample stars, with the newly obtained high-resolution optical UVES spectra,show that, here also, the observed photospheric depletion is frequent and very strong. Thecalculated stellar parameters, together with the derived luminosities, allow for the first timeto compare these objects based on their evolutionary status. This is discussed in Chapter 7,where we present a larger sample of LMC post-AGB sources. The low-resolution infraredSpitzer spectra, presented in Chapter 5, show a striking similarity to the observed Galacticsources. The spectra are almost all dominated by amorphous and crystalline silicates, exceptfor one star, which shows clear emission due to PAHs. Unfortunately, we are not able to de-termine whether these stars also reside in a binary system, since they are too faint for current

152 Chapter 6. Conclusions

observational monitoring techniques, at telescopes which are accessible to us on a long-termregular time basis. However, given the strong resemblance they show to the Galactic sourcesin all other characteristics, we can assume that these stars too are binaries. Overall, we cansafely say that, as in the Galaxy, circumstellar discs around evolved stars in the LMC areclearly present.

Clearly, many open questions still remain on the formation, structure and evolution of thesecircumbinary discs around evolved stars, and on the impact the disc has on the evolution ofthe central binary system. The interaction in the binary system is fundamental in trappingpart of the dust in a stable circumbinary environment, and determines the further evolutionof the binary star and the disc. The observed shortcut of the evolution of these stars on theAGB, implies that they are actually not even “post-AGB” stars in the exact definition of thisterm.

In the next chapter we present some of our upcoming work, in which we aim to tackle someof the open issues and remaining questions.

Chapter 7Future work: Evolution ofcircumbinary discs aroundpost-AGB binaries

The formation and evolution of circumbinary discs around evolved stars is clearly a veryintriguing phenomenon, playing an important role in the life of evolved binary stars. Re-cently, some new light has been shed on a few possible formation scenarios (Soker & Livio1989; Mastrodemos & Morris 1999; Verhoelst et al. 2007; Bonacic Marinovic et al. 2008;Jorissen et al. 2009), but so far, no detailed study has been performed on the subsequentevolution of these discs. However, major questions remain: what is the impact of stellarevolution on the disc? How does the central-star system evolve? Are the timescales andprocesses in the disc similar to the ones seen in discs around young stars? Does the disc’sfeedback prolong the stellar post-AGB evolution? Is planet formation possible in the discaround evolved stars, etc.?

With the small sample of 21 Galactic and four extragalactic post-AGB disc sources describedin this work, we were not able to relate the observed diversity in the spectra to specific struc-tural elements of the disc, the central star, the orbit and/or whether we witness directly anevolutionary change between different sources.

To answer these questions and study the evolution of these discs, we plan to combine a theo-retical description with a multiwavelength observational approach. This to investigate quan-titatively the major physical processes in these discs, with the goal to determine accuratelythe actual disc structure, probe its evolution, and study its relation with the central star.

154 Chapter 7. Future work: Evolution of circumbinary discs around post-AGB binaries

7.1 Theoretical description of disc processes

We plan to expand existing theoretical passive-disc models to simulate disc evolution and thedifferent processes occurring in the disc.

From our pilot study on a few stars, described in the previous chapters, we found that ad-ditional physical and chemical processes in the disc models are necessary to explain all theobserved features. For this we will work with more advanced theoretical disc models, whichwill be able to describe grain settling towards the midplane, scattering on dust, turbulenceand viscosity effects, outflows and circumsystem dust/gas shells, etc. This work will be donein cooperation with groups in Ghent (Baes et al. 2007b), Amsterdam (Min et al. 2009) andHeidelberg (Dullemond & Dominik 2004; Klahr & Kley 2006).

Numerous models, with different complexities, for discs around young stars are currentlyin development (Dullemond et al. 2006, 2007; Klahr & Kley 2006; Lyra et al. 2008) and wewill tune them to the characteristics of the discs around evolved stars. Opposed to youngstellar objects, the evolutionary timescale of an evolved central star versus the timescales ofits disc is short, which has to be taken into account when modelling disc evolution. Alsothe disc composition at formation could possibly be different. Discs around young stars areformed outside-in, from ISM material in the molecular cloud from which they originate. Thismaterial mainly consists of small amorphous silicates and carbonaceous dust particles, suchas PAHs. Discs around evolved stars are formed inside-out, and it is, so far, not clear whatthe exact dust species and grain sizes are at disc formation. If the discs are formed by non-conservative mass transfer, the densities and temperatures might be high, making it easier forgrain growth and crystallinisation to occur. We therefore plan to focus on these models forYSO but will need to adopt them to the relevant physical environment.

7.2 Interferometry and far-IR spectroscopy of the disc

In Chapters 2, 4 and 5 we already presented the SED disc modelling of a few of our binarypost-AGB stars. For these stars we found that in general the SED is well reproduced usinga passive circumbinary disc with a puffed-up inner rim, a shadowed region and a large scaleheight.

All the discs (with the possible exception of the disc around MACHO 81.9728.14, see Chap. 5)consist of amorphous and crystalline silicates, with grain sizes from 0.1 µm to about 10 −20 µm. Larger grains, up to centimetre size and possibly larger, are present in the disc aswell, but have likely settled towards the midplane. In the disc modelling described in thiswork, we use a simple blackbody to model this cold midplane of large grains. We find thatthe disc inner rim is usually close to the sublimation radius of the dust, or slightly larger.The outer radius is unfortunately less well constrained and often degenerate with the densitydistribution and total disc mass.

We now plan to extend this study to model in detail all 21 Galactic and 14 extragalactic (seeSect. 5.9) Spitzer sample stars. With this larger study we can investigate possible relation-ships between the observed SED, disc characteristics, stellar parameters and infrared spectralinformation.

7.3. Information on the dynamics and chemistry of the central star 155

For about 6 stars we also possess interferometric measurements taken with the MIDI andAMBER instruments on the VLTI. Combining interferometric measurements with the discmodel already proved to be a big improvement in constraining some of the disc parameters,such as the disc size or the surface density distribution (see Chap. 4 and Deroo et al. 2007a).The AMBER instrument is ideally suited to study the hot puffed-up inner rim, whereas theMIDI instrument is sensitive to a more extended disc region, albeit not sensitive to the struc-ture of the outer disc.

Whereas interferometric data sheds light on the angular size of the disc and the dust distri-bution, infrared spectroscopic data traces the emitting upper dust layers of the disc, givinginformation on the dust composition in the disc. The Spitzer spectroscopic data will be ex-tended to longer wavelengths by using the PACS instrument aboard the Herschel satellite,which operates in the 60 − 120 µm region. This extension to longer wavelengths will allowus to detect new dust signatures and species.

Additional information on the total disc mass and amount of large grains will be obtainedusing submillimetre data. Recently APEX-LABOCA 870 µm observations have been per-formed on several of our sample stars and are ready for reduction and first inspection.

7.3 Information on the dynamics and chemistry of the cen-tral star

To study the impact of the disc on the evolution of the central star and vice versa, we need tobe able to derive the stellar parameters of the central binary system.

A long term radial velocity monitoring programme has already resulted in orbital parametersfor about 50 Galactic post-AGB binaries, and with the new and very efficient HERMESspectrograph on the Flemish Mercator telescope we will extend this sample significantly. Bychecking the binary nature of possible progenitors and descendants of our sample stars, suchas AGB disc sources, silicate J-type carbon stars and binary PNe, we can gain more insight inthe late phases of binary evolution. Such a study along the HR-diagram also allows to studythe impact of a changing irradiation field on the disc.

For Galactic post-AGB stars extensive chemical studies have been performed, giving stel-lar parameters for most known Galactic post-AGB objects (e.g. Van Winckel et al. 1995;Gonzalez et al. 1997b,a; Giridhar et al. 2000; Maas et al. 2005). With the high sensitivity ofthe UVES spectrograph mounted on the VLT, it recently has been possible to deduce stellarparameters for extragalactic sources. Six extragalactic LMC post-AGB sources have alreadybeen studied (see Chap. 5 and Reyniers & Van Winckel 2007; Reyniers et al. 2007), but thesample will be extended to include 14 more LMC sources, for which some will be observedwith UVES (Proposal ID: 082.D0941(A)).


−1 0 1 2 3 4 5 6 7 8K-[5.8]

0

2

4

6

8

[8]

- [2

4]

Figure 7.1 — Colour-colour diagram showing the LMC disc sources that fall within our selection crite-ria, and have visual fluxes (dark gray squares). The light gray circles depict the known RV Tauri objectsin the LMC, the black triangles give the Galactic disc sources and the gray dots are the SAGE-LMCobjects with K magnitudes larger than 14.5 mag.

7.4 Global study of disc sources in the LMC

To increase the LMC sample, we searched the SAGE Spitzer Catalogue (Meixner et al. 2006)for the presence of other likely post-AGB candidates in the LMC. All objects with 24 µmfluxes 2mJy < F24 < 1 Jy, to exclude young stellar objects and supergiants, were selected.Other selection criteria were chosen to distinguish between post-AGB stars with an expandingshell (F24 > F8) and binary post-AGB sources with a circumbinary disc (F24 > 0.5 F8 andJ−K < 1). For a detailed description of the selection criteria we refer to van Aarle et al.(in prep.). With these criteria 430 sources remained. They were correlated with Massey’scatalogue of the UBVR CCD survey of the Magellanic clouds (Massey, 2002), the GuideStar Catalog 2.3.2 (Space Telescope Science Institute & Osservatorio Astronomico di Torino2007), and the LMC Stellar Catalogue of Zaritsky et al. (2004). Searching the SIMBADAstronomical Database allowed to remove the stars that are known not to be of post-AGBnature. Stars without known V-magnitude were removed, which resulted in a final sample of327 stars for which we collected all photometric data available in literature. In Figure 7.1 weshow the location, in a colour-colour diagram, of our final sample in comparison to all LMCsources within the SAGE Catalogue.

For 82 of the 327 stars in our final sample, low-resolution optical spectra were obtainedat Siding Spring Observatory (Australia) or at the South African Astronomical Observatory(SAAO). Optical high-resolution, high-signal-to-noise spectra obtained with UVES have al-ready been studied for 4 LMC post-AGB sources (see Chap. 5), and 15 more are scheduled

7.4. Global study of disc sources in the LMC 157

Figure 7.2 — Spectral energy distribution of some of our newly discovered disc sources in the LMC.The black solid line gives the reddened Kurucz model, overplotted in grey the low-resolution opticalground-based spectra. Photometric observations are represented by crosses.

to be observed in near future. These observations are part of the PhD thesis of Els van Aarle,and a paper describing the observations and data analysis is currently in preparation (vanAarle et al., in prep.). From the optical spectra, spectral types could be deduced. From thephotometric data, the spectral type and the known LMC-distance, the SED, the luminosityand effective temperature for the different objects were calculated. Some examples of thesenewly discovered disc sources in the LMC are given in Figure 7.2.

The results show a wide range in spectral types and luminosities, indicating a wide range inevolutionary phases along the HR-diagram. With this systematic study of objects with knowndistances, we hope to constrain, for the first time, distinct evolutionary phases in the life ofthese particular evolved disc sources, and also study the impact of these binaries on our globalview of the final evolution of stars.

Appendix ANederlandse samenvatting

In dit proefschrift bestuderen we de circumstellaire omgeving van een bijzondere klassevergeevolueerde sterren. Deze sterren maken deel van een dubbelster-systeem, twee sterrendie mekaars aantrekking voelen en daardoor rond elkaar draaien. Ergens tijdens de evolutievan deze dubbelster is een grote fractie materiaal uitgestoten door de oude ster, en ingevan-gen in een stabiele schijf rond de dubbelster. In deze thesis bestuderen we in detail de eigen-schappen van deze schijf. Zo willen we onder andere weten uit welke gas- en stofsoorten zeis gemaakt, en welke geometrische structuur ze heeft. We doen dit door gebruik te makenvan hoge-spectrale-resolutie infrarood observaties, gemaakt met de Spitzer Ruimtesatelliet.

A.1 Sterevolutie

Sterren worden gevormd wanneer massieve wolken van interstellair stof en gas in mekaarstorten onder hun eigen gewicht. De zwaartekracht zorgt ervoor dat deze wolken steedsmeer worden samengeperst, tot de temperatuur en druk zo hoog zijn opgelopen dat in hetcentrum van de wolk kernreacties kunnen optreden. Een nieuwe ster is geboren, maar blijftnog een tijdje voor het oog verborgen door de hoeveelheid gas en stof er rond. De jongester zet waterstof om in helium, en we noemen deze fase de Hoofdreeksfase. De ster zalhet grootste gedeelte van haar leven, zo’n 90%, doorbrengen in deze fase. Ook onze Zonbevindt zich momenteel in deze fase, waar ze al zo’n 4.5 miljard jaar heeft doorgebracht enhoogstwaarschijnlijk nog eens zo lang zal blijven.

Een bijproduct van deze stervorming is de vorming van een schijf rond de jonge ster. Dit komtdoordat de ineenstortende wolk een bepaald intrinsiek angulair moment heeft. Hierdoor kanhet invallende materiaal niet recht op de ster invallen, maar eerder spiraalsgewijs toestromen,en zal zich ophopen in een schijf rond de ster. Tijdens de eerste fases van de sterevolutiezal het stof in deze schijf beginnen groeien, eerst tot kleine korreltjes en brokstukken, latereventueel tot planeten. Daarom dat we deze schijven ook proto-planetaire schijven noemen(Figuur A.1). Ons eigen Zonnestelsel is op deze manier gegroeid uit de proto-planetaire schijf

160 Chapter A. Nederlandse samenvatting

Figure A.1 — Artistieke schets van de proto-planetaire schijf rond een jonge ster (Figuur: NASA/JPL-CalTech).

rond de jonge Zon.

Wanneer het waterstof in de kern van de ster is opgebruikt, fuseert de ster enkel nog waterstofin een schil rond de kern. Hierdoor zal de ster opzwellen en het steroppervlak afkoelen: dester is een Rode Reus geworden. Tijdens deze Rode Reuzen Fase trekt de kern verder samen,tot de temperatuur hoog genoeg is om nu helium te verbranden tot koolstof en zuurstof.Wanneer de Zon binnen 5 miljard jaar deze fase bereikt zal ze ongeveer 250 keer in straaltoenemen, en zo misschien wel tot de huidige baan van de Aarde komen.

Wanneer vervolgens ook al het helium in de kern is opgebrand, zal de ster enkel nog he-lium verbranden in een schil rond de ontaarde koolstof/zuurstof kern. De ster zal opnieuwuitzetten en zijn lichtkracht vergroten. De ster heeft de Asymptotische Reuzentak bereikt, enwe noemen de ster nu een AGB (Asymptotic Giant Branch) ster. Wanneer nu ook het heliumin de schil is uitgeput, zal opnieuw waterstof verbrand worden in een bovenliggende schil.Deze laatste verbranding vult dan weer de uitgeputte heliumvoorraad aan, zodat helium-schil-verbranding opnieuw kan beginnen. Dit proces noemen we een thermische puls, en kan ver-scheidene malen voorkomen tijdens de AGB.

AGB sterren worden ook gekenmerkt door een groot massaverlies, veroorzaakt door eensterke sterrenwind. Tijdens het periodiek uitzetten van de ster, worden de buitenste lagen opzo’n grote afstand gebracht dat ze niet meer sterk gebonden worden door de zwaartekrachtvan de ster. Er stroomt dus telkens een beetje gas en stof weg van het steroppervlak, dat danverder weggeblazen wordt door straling van de ster.

Wanneer de meeste materie van de buitenlagen is afgeworpen, begint de ster aan de snellepost-AGB fase: de temperatuur van de ster stijgt fel, terwijl de lichtkracht nagenoeg dezelfdeblijft. Een specifieke subklasse van post-AGB sterren zijn de RV Tauri sterren. Dit zijn ster-ren die de Populatie II Cepheide instabiliteitsstrook passeren, en dus sterke radiele pulsatiesvertonen.

Wanneer de temperatuur hoog genoeg is, kan de ster het omringende uitgeworpen materiaalioniseren en doen oplichten. De post-AGB ster is een Planetaire Nevel geworden. Niet alle

A.2. Schijven rond post-AGB dubbelsterren 161

Figure A.2 — Twee sterren met een duidelijk sferische circumstellaire omgeving. Links: de AGB sterCW Leo, waarbij het periodiek sferisch massaverlies aanleiding geeft tot ringvormige structuur rond dester. Rechts: Abell 39, een uitzonderlijk sferische planetaire nevel.

post-AGB sterren zullen in staat zijn om hun circumstellaire materiaal zo te doen oplichten,soms is het afgestoten materiaal al te ver weggeblazen. De koolstof/zuurstof kern zal zijnleven eindigen als Witte Dwerg, waarbij hij langzaam afkoelt en uitdooft.

De bovenstaande beschrijving geldt enkel voor sterren met een massa kleiner dan acht keerde zonsmassa. Sterren met een grotere massa zullen nog extra fases van kernverbrandingdoorgaan vooraleer ze ontploffen als supernova. De ster zal dan haar leven eindigen alsneutronenster of zwart gat, afhankelijk van haar massa.

A.2 Schijven rond post-AGB dubbelsterren

Een van de grote vragen omtrent de late fasen van sterevolutie blijft hoe de sferisch sym-metrische sterwinden tijdens de AGB fase (Fig. A.2) verantwoordelijk kunnen zijn voor degeobserveerde sterke asymmetrie van de planetaire nevels (Fig. A.3). Verschillende verk-laringen werden reeds naar voor gebracht, zoals botsende sterwinden, invloed van magnetis-che velden, rotatie, etc.

Er is nu echter groeiend bewijs dat de grote fractie aan asymmetrische planetaire nevels enkelkan verklaard worden door de interactie van het materiaal met een centraal dubbelstersys-teem. Bij planetaire nevels kan deze binariteit spijtiggenoeg nog niet met zekerheid bevestigdworden. Dit in tegenstelling tot de post-AGB sterren, waar recent onderzoek de grote fractieaan binaire sterren heeft gedetecteerd. Door te kijken naar de periodieke verschuiving vanbepaalde absorptielijnen of de verandering in lichtcurve van het object kunnen we de baanvan de dubbelster bepalen. Zo vinden we dat deze binaire post-AGB sterren rond mekaardraaien met baanperiodes van 100 tot meer dan 2000 dagen. Om zulke lange periodes tedetecteren, heb je data nodig die over verschillende jaren is gemeten. Het probleem wordtnog verder bemoeilijkt als de ster ook pulseert, omdat dit de baansignatuur veel moeilijkerdetecteerbaar maakt.


Figure A.3 — Twee sterren die duidelijk afwijken van de sferische vorm. Links: De ’Rode Rechthoek’,een post-AGB dubbelster omgeven door een schijf en met nevel die het sterlicht op zo’n manier re-flecteert dat het een rechthoek lijkt te vormen. Rechts: De ’Vlinder Nevel’, een planetaire nevel meteen bipolaire verdeling van het gas en stof.

De dubbelster bestaat uit een post-AGB ster en een hoofdreekstster die op een afstand vanongeveer 1 Astronomische Eenheid (1 AE is de afstand Aarde-Zon) van mekaar staan. Dezebaan is echter veel te klein om een volgroeide AGB ster bevat te hebben. Blijkbaar heeftde sterke binaire interactie in het systeem ervoor gezorgd dat de evolutie van de AGB steris ingekort. We zien ook dat de banen zeer eccentrisch zijn, wat wil zeggen dat ze eerderellipsvormig zijn dan cirkelvorming. Dit is verbazend omdat theorie aangeeft dat zulk eendubbelsterbaan heel vlug cirkelvormig zou moeten worden door de sterke getijdenwerking.Blijkbaar speelt ook hier de binariteit weer een grote rol.

Deze binaire post-AGB sterren worden gekenmerkt door een aantal observationele kenmerken,die wijzen op de aanwezigheid van een stabiele stofschijf rond de centrale dubbelster. Zo ob-serveren we bijvoorbeeld heet stof en grote stofkorrels rond deze sterren, wat niet kan verk-laard worden door materiaal dat vroeger door de ster zou zijn weggeblazen. Een schijf, diedicht bij de ster zit en dus heet is, lijkt een veel betere verklaring. De dichtheden zijn er ookgroot genoeg om grote stofkorrels te maken, wat niet het geval is in een ijle sterwind. Ook defotosfeer van de ster vertoont eigenschappen die wijzen op de aanwezigheid van een schijf.De fotosfeer wordt gekenmerkt door een depletie (afname) van metalen (Fe,Mg,Y,Zr), terwijlandere elementen (zoals C,N,O,S en Zn) wel in de verwachte abondantie te vinden zijn. Deverklaring hiervoor is dat in de circumstellaire omgeving de metalen in stofkorrels gaan zit-ten, terwijl de andere elementen in de gasfase blijven. Vervolgens treedt er een ontkoppelingop waarbij het gas terugvalt op de ster, terwijl het stof rond de ster blijft zitten.

Hoe de schijf juist werd gevormd is nog onduidelijk, maar er wordt aangenomen dat vormingenkel mogelijk is door sterke interactie in het dubbelstersysteem. Tijdens de sterevolutie, toende primaire ster in de reuzenfase zat, moet er een fase van sterk massaverlies geweest zijn,waarbij het materiaal ingevangen werd, en uiteindelijk in een schijfstructuur terecht kwam.Deze schijf zit rond het volledige dubbelstersysteem, en niet rond een van de sterren. Tussende sterren in is het namelijk te heet om stof te kunnen vormen.

Door specifiek te zoeken naar deze observationele kenmerken werd de groep mogelijke bi-naire post-AGB sterren, omgeven door een schijf, sterk uitgebreid tot een zeventigtal sterren.Er werden zelfs soortgelijke objecten gedetecteerd buiten onze Melkweg, meerbepaald in de

A.2. Schijven rond post-AGB dubbelsterren 163

Figure A.4 — Schets van een schijf met een opgeblazen binnenrand, schaduw en uitwaaierende buiten-rand.

Grote Magellaanse Wolk (Large Magellanic Cloud).

Infrarood spectra van de schijven, die gevoelig zijn voor de stofeigenschappen, laten zien datde schijven rond deze geevolueerde sterren zeer gelijkend zijn op de proto-planetaire schijvenrond jonge sterren en kometen in ons Zonnestelsel (Figuur 1.12). Dit is merkwaardig, daardeze schijven een zeer verschillende vormingsgeschiedenis hebben.

In deze thesis zullen we in detail enkele eigenschappen, zoals de geometrische vorm ensamenstelling, van de schijven rond geevolueerde sterren bestuderen, en vergelijken met watwe zien bij jonge sterren.

A.2.1 Schijfstructuur

De schijven die we in deze thesis bestuderen zijn passieve schijven: schijven die geen sterkeinteractie meer hebben met de centrale ster. De schijf bestaat uit gas en stof, waarbij het stofwordt verwarmd door straling van de ster. Als de dichtheid in de schijf hoog genoeg is, zullenhet gas en het stof dezelfde temperatuur hebben. Als de gastemperatuur nu toeneemt, zal degasdruk verhogen en als gevolg zal de schijf uitzetten. Dit kan dan het effect van gravitatie,dat alle materiaal naar het midden van de schijf wil trekken, tegengaan. Zo blijft de schijf inhydrostatisch evenwicht.

Zulk een stabiele schijf is de ideale omgeving voor stofgroei. Grote stofkorrels zullen on-der invloed van de gravitatie naar het midden van de schijf zakken. Zo krijgt de schijf eenverticale distributie in korrelgroottes en een binnenste van koele grotere korrels. Tussen debinnenstraal van de schijf en de ster zit een groot gat, daar het dicht bij de ster te heet is omstof te vormen. De schijf zal dus pas beginnen daar waar de temperatuur laag genoeg is, zo’n1000 Celsius.

De binnenste rand van de schijf vangt al het sterlicht rechtstreeks op, waardoor dit gebiedsterk zal opwarmen en hierdoor gaan uitzetten. De rand kan zelfs zo sterk uitzetten dat het eenschaduw werpt op de gebieden er net achter, die daardoor zullen afkoelen en inzakken. Verder


weg valt de schijf buiten de schaduw, zodat de buitenste regionen weer kunnen opwarmen enuitwaaieren. Hierdoor krijg je dus een schijf met een zeer specifieke structuur, schetsmatigweergegeven in Figuur A.4. Voor zulke schijven bestaan er theoretische modellen die toelatengeobservationeerde data te modelleren, wat we voor enkele bronnen in dit proefschrift zullendoen.

A.3 Kosmisch stof

Bij het horen van het woord stof denken de meeste mensen meteen aan de vuile grijze laagop de vloer en meubelen, nadat het stofzuigen en afstoffen weer enkele weken vertragingheeft opgelopen. Er wordt echter niet vaak de link gemaakt met de vaak spectaculaire as-tronomische prentjes die de Hubble telescoop ons bezorgt. En toch heeft ook het stof in onseigen huis een astronomische achtergrond. Alle vaste stof hier op Aarde, ook de Aarde zelf,is ooit zijn leven begonnen als klein kosmisch stofkorreltje, ergens in de ruimte. Dus doordit astronomisch stof te bestuderen, kunnen we iets leren over onze eigen geschiedenis en dievan ons Zonnestelsel.

A.3.1 De detectie van stof

De enige manier om stof in de ruimte te detecteren, is door de interactie van de stofdeelt-jes met licht te bestuderen. Straling die door stof geabsorbeerd wordt, zal de stofkorrelsopwarmen, die dan op hun beurt beginnen te stralen. Deze thermische straling is zichtbaarin het infrarood gebied van het golflengtespectrum, en voor kleine stofdeeltjes zullen ookspecifieke roostervibraties zichtbaar worden als karakteristieke absorptie- en emissiebanden(zie Fig. A.5). Dus, om stofeigenschappen te bestuderen, hebben we instrumenten nodig diegevoelig zijn in het infrarood. De Spitzer-IRS (Fig. 1.15) spectrograaf bleek ideaal om hetstof in de schijven rond onze post-AGB dubbelsterren te bestuderen.

Hoe deze emissie- of absorptiebanden eruit zien hangt af van van de temperatuur, samen-stelling, vorm en grootte van het stofdeeltje. Door de vergelijking te maken tussen syn-thetisch berekende stofspectra en geobserveerde spectra kunnen we zo achterhalen wat deeigenschappen zijn van het stof in onze schijven.

De stofsamenstelling beschrijft uit welke atomen het deeltje is opgebouwd en op welkemanier de atomen met mekaar verbonden zijn. Zo kunnen we bijvoorbeeld een onderscheidmaken tussen kristallijn materiaal, waar de atomen in een perfecte roosterstructuur geordendzitten, en amorf materiaal, waar de atomen zeer ongeordend in hun rooster zitten. Ook dedeeltjesgrootte is belangrijk. Over het algemeen geldt dat: hoe groter het deeltje, hoe min-der uitgesproken de emissiebanden. Als het deeltje een bepaalde grootte heeft bereikt, zal hetzelfs geen banden meer vertonen, maar eerder een vlakke continuumstraling. Voor de beschri-jving van de deeltjesvorm maken we ruwweg een onderscheid tussen sferische (bolvormige)deeltjes en deeltjes met een meer grillige vorm.

Synthetische spectra voor al deze vormen worden berekend door de interactie van licht metde stofdeeltjes te modelleren. Een sferisch deeltje is zeer makkelijk te modelleren, maaris helaas niet zo’n goede weergave van een echt stofdeeltje (Fig. 1.7). Het modelleren van

A.3. Kosmisch stof 165

Figure A.5 — Het verschil tussen de geobserveerde emissielijnen in het infrarood spectrum van kristal-lijne (links) en amorfe silicaten (rechts) (Figuur genomen van Molster 2009, PhD.thesis).

de realistische deeltjesvorm neemt dan weer teveel computertijd in beslag. Daarom werkende meeste modellen tegenwoordig met de statistische benadering, waarbij men probeert eengrillig gevormd deeltje zo goed mogelijk te benaderen met een verzameling van simpeleredeeltjesvormen.

A.3.2 Astronomische stofsoorten

Stof tussen en rond de sterren, interstellair en circumstellair stof, bestaat voornamelijk uitzuurstofrijke en koolstofrijke stofsoorten, zoals bijvoorbeeld silicaten, polycyclische aroma-tische koolwaterstoffen (PAK) en siliciumcarbide (SiC). Typische voorbeelden van deze stof-soorten op Aarde hiervan zijn glas en zandkorrels, wat amorfe silicaten zijn, en roet, eenPAK.

Silicaten, verbindingen met silicium (Si) en zuurstof (O), vinden we zowel terug in amorfeen kristallijne vorm. Ze hebben zeer verschillende emissiespectra en we kunnen ze dus goedvan mekaar onderscheiden in onze waarnemingen (Fig. A.5). Amorfe silicaten vertonen zeerbrede banden, terwijl kristallijn stof zeer nauwe scherpe banden vertoont. Om kristallijn stofte maken is een zeer hoge temperatuur nodig, van wel meer dan 1000 Celsius. Schijven rondsterren zijn de ideale plaats om zulke kristallijne deeltjes te maken, daar ze dicht genoeg bijde ster zitten om voldoende hoge temperaturen te kunnen ontwikkelen.

Silicaten komen veelvuldig voor in het heelal, en zijn onder andere te vinden in kometen, hetinterstellair midden, in schijven rond jongen sterren, in sterwinden van geevolueerde sterrenen in planetaire nevels. De grootste fractie geobserveerde silicaten blijken die te zijn die Si


Figure A.6 — De kristallijne silicaten forsterite (links) en enstatite (rechts), zoals ze gevonden wordenhier op Aarde.

en O binden met magnesium (Mg) en ijzer (Fe), meerbepaald olivijn (Mg2xFe2(1−x)SiO4)en pyroxeen (MgxFe1−xSiO3). Hierbij geeft 0 ≤ x ≤ 1 de magnesiumfractie weer. Vreemdgenoeg worden de kristallijne silicaten enkel geobserveerd in de magnesium-rijke component,namelijk als forsterite (Mg2SiO4) en enstatite (MgSiO3) (Fig. A.6).

PAKs zijn verbindingen van koolstof (c) en waterstof (H), gestructureerd in een benzeen-ring. Zo een ring bestaat uit zes koolstofatomen met waterstofatomen gebonden aan de rand(Fig. 1.10). Deze benzeenringen vormen de bouwstenen van de PAKs, die kunnen bestaanuit verbinden van wel meer dan 20 zulke ringen.

In dit proefschrift zullen we de infrarode spectra van onze sterren gebruiken om de verschil-lende stofsoorten te identificeren en te karakteriseren. Door de geobserveerde emissiebandente modelleren leren we zo iets over de specifieke soorten, deeltjesgrootte en temperaturen vanhet stof in de stabiele schijven rond onze sterren.

A.4 Deze thesis

In dit werk presenteren we een gedetailleerde studie van de samenstelling en geometrie vande circumbinaire structuur rond geevolueerde dubbelsterren, en dit zowel voor galactischeals extragalactische objecten. We gebruiken hoge-resolutie infrarood observaties om de ex-acte stofsamenstelling te bepalen en fotometrische en interferometrische hoge-ruimtelijke-resolutie data om meer te weten te komen over de vorm, grootte, massa, dichtheid en tem-peratuur van de schijf. Door de resultaten voor een groot aantal objecten te vergelijken,kunnen we algemene tendensen en individuele verschillen opsporen. Op deze manier hopenwe eventuele relaties tussen de circumstellaire omgeving en de eigenschappen van de cen-trale ster, de binaire baan en/of evolutionaire fase te detecteren. Hieronder geven we een kortoverzicht van de verschillende hoofdstukken.

In Hoofdstuk 2 voeren we een pilootstudie uit op een van onze Spitzer bronnen, namelijk deRV Tauri ster RU Cen. Door verschillende observationele data en technieken te combinerenbestuderen we in detail de dubbelstereigenschappen van deze ster. Ook vergelijken we dezester met de reeds goed bestudeerde RV Tauri ster AC Her, meerbepaald gaan we na of erverschil is in de stofsamenstelling in de schijven rond beide sterren.

A.4. Deze thesis 167

In Hoofdstuk 3 vergroten we onze studie en bestuderen we de mineralogie en stofverwerkingin de circumbinaire schijven van 21 post-AGB dubbelsterren. Hiervoor gebruiken we hoge-resolutie infrarood spectra, genomen met de 0.85 meter Spitzer Ruimtetelescoop en de 3.6meter ESO telescoop, gelegen in La Silla, Chili. Door de geobserveerde spectra te modellerenmet synthetisch berekende spectra kunnen we zo de verschillende stofsoorten identificeren enhun eigenschappen bepalen, zoals grootte, vorm en temperatuur.

In Hoofdstuk 4 kijken we in wat meer detail naar twee van onze 21 post-AGB dubbelsterren,namelijk EP Lyr en HD 52961. We vonden namelijk in Hoofdstuk 3 dat deze twee sterrenzeer specifieke stof- en gasbanden in hun spectra vertoonden, die we moeilijk konden mod-elleren. Bovendien zenden net deze twee sterren zeer weinig infrarood straling uit, wat kanwijzen op een andere schijfstructuur. In dit hoofdstuk bestuderen we of de unieke gas- enstofeigenschappen van deze twee sterren gelinkt zijn aan hun specifieke schijfeigenschappen.

In Hoofdstuk 5 verruimen we ons onderzoek, en verlaten onze eigen Melkweg. In plaats daar-van bestuderen we de circumstellaire omgeving van 4 extragalactische RV Tauri bronnen inde Grote Magellaanse Wolk. De parameters van de centrale dubbelster en de chemie wordenbepaald aan de hand van hoge-resolutie optische UVES spectra. Dit instrument bevindt zichop een van de 8.2 meter telescopen op de VLT (Very Large Telescope) site in Paranal, Chili.Doordat we de afstand tot de Grote Magellaanse Wolk weten, zo’n 1.6 × 1018 kilometer of165 000 lichtjaar, kunnen we de exacte lichtkracht van de ster bepalen. Dit, gecombineerdmet de chemie en parameters van de ster, bevestigt de post-AGB geevolueerde status van dezesterren. Opnieuw gebruiken we Spitzer infrarood spectra om het stof in de schijven te bestud-eren, en we vergelijken de resultaten bij deze extragalactische bronnen met de bevinden vooronze eigen Melkweg.

In Hoofdstuk 6 bespreken we de algemene conclusies en blijvende vragen van dit thesiswerk.We kijken ook al een beetje in de toekomst door in Hoofdstuk 7 een overzicht te geven vanhet nog komende onderzoek.

Appendix BList of acronyms

AGB Asymptotic Giant BranchAMBER Astronomical Multiple BEam RecombinerAU Astronomical Unitc2d Cores to DisksCDE Continious Distribution of EllipsoidsCE Circumstellar EnvironmentCGRO Compton Gamma Ray ObservatoryCXO Chandra X-ray ObservatoryDHS Distribution of Hollow SpheresE-AGB Early Asymptotic Giant BranchESO European Southern ObservatoryFEPS Formation and Evolution of Planetary SystemsFT Fischer-TropschFWO Fonds voor Wetenschappelijk OnderzoekGRF Gaussian Random FieldsHERMES High Efficiency and Resolution Mercator Echelle SpectrographHST Hubble Space TelescopeIR InfraRedIRAC InfraRed Array CameraIRAS InfraRed Astronomical SatelliteIRS InfraRed SpectrographISM InterStellar MediumISO-SWS Infrared Space Observatory - Short Wavelength SpectrometerISW Interactive Stellar WindJD Julian DateLH Long-High Spitzer-IRS modeLL Long-Low Spitzer-IRS modeLMC Large Magellanic Cloud

170 Chapter B. List of acronyms

MIDI MID-infrared Interferometric instrumentMIPS Multiband Imaging Photometer for SpitzerNASA National Aeronautics and Space AdministrationPACS Photodetector Array Camera and SpectrometerPAH Polycyclic Aromatic HydrocarbonPAK Polycyclische Aromatische KoolwaterstofPDM Phase Dispersion MinimalisationPMS Pre Main SequencePNe Planetary NebulaPPNe Proto Planetary NebulaPSF Point Spread FunctionSAAO South African Astronomical ObservatorySAGE Surveying the Agents of Galaxy’s EvolutionSED Spectral Energy DistributionSH Short-High Spitzer-IRS modeSL Short-Low Spitzer-IRS modeS/N Signal to NoiseSRF Spectral Response FunctionSSC Spitzer Science CenterTIMMI2 Thermal Infrared Multi Mode Instrument 2TP-AGB Thermally Pulsating Asymptotic Giant BranchUV UltraVioletUVES Ultraviolet and Visual Echelle SpectrographVLTI Verly Large Telescope InterferometerYSO Young Stellar Object

Appendix CList of refereed publications

“Spectral analysis of the peculiar post-AGB stars EP Lyr and HD 52961”Gielen, C.; Van Winckel, H.; Matsuura, M.; Min, M.; Deroo, P.; Waters, L.B.F.M.;Dominik, C.Astronomy and Astrophysics, accepted for publication

“SPITZER survey of dust grain processing in stable discs around binary post-AGB stars”Gielen, C.; Van Winckel, H.; Min, M.; Waters, L.B.F.M.; Lloyd Evans, T.Astronomy and Astrophysics, Volume 490, Issue 2, 2008, pp.725-735

“Photometric multi-site campaign on massive B stars in the open cluster Persei (NGC 884)”Saesen, S.; Pigulski, A.; Carrier, F.; De Ridder, J.; Aerts, C.; Handler, G.; Narwid, A.;Fu, J.N.; Zhang, C.; Jiang, X.J.; Kopacki, G.; Vanautgaerden, J.; Stlicki, M.; Acke, B.;Poretti, E.; Uytterhoeven, K.; De Meester, W.; Reed, M.D.; Koaczkowski, Z.; Michalska, G.;Schmidt, E.; stensen, R.; Gielen, C.; Yakut, K.; Leitner, A.; Kalomeni, B.; Prins, S.; VanHelshoecht, V.; Zima, W.; Huygen, R.; Vandenbussche, B.; Lenz, P.; Ladjal, D.; Puga An-toln, E.; Verhoelst, T.; Niarchos, P.; Liakos, A.; Lorenz, D.; Dehaes, S.; Reyniers, M.;Davignon, G.; Kim, S.-L.; Kim, D.H.; Lee, Y.-J.; Lee, C.-U.; Kwon, J.-H.; Broeders, E.;Van Winckel, H.; Vanhollebeke, E.; Raskin, G.; Blom, Y.; Eggen, J.R.; Beck, P.; Puschnig, J.;Schmitt, L.; Gelven, G.A.; Steiniger, B.; Drummond, R.Communications in Astroseismology, Vol. 150, p.191

“Dust-grain processing in circumbinary discs around evolved binaries: The RV Tauri spec-tral twins RU Centauri and AC Herculis”Gielen, C.; Van Winckel, H.; Waters, L.B.F.M.; Min, M.; Dominik, C.Astronomy and Astrophysics, Volume 475, Issue 2, November IV 2007, pp.629-637

172 Chapter C. List of refereed publications

“The ongoing 2005 – 2006 campaign on Cephei stars in NGC 6910 and Persei (NGC 884)”Pigulski, A.; Handler, G.; Michalska, G.; Koaczkowski, Z.; Kopacki, G.; Narwid, A.; Van-hollebeke, E.; Stelicki, M.; Lefever, K.; Gazeas, K.; de Meester, W.; Vanautgaerden, J.;Leitner, A.; De Ridder, J.; Van Helshoecht, V.; Gielen, C.; Vandenbussche, B.; Saesen, S.;Reed, M.D.; Eggen, J.R.; Gelven, G.A.; Desmet, M.; Puga Antoln, E.; Aerts, C.; Schmidt, E.;Huygen, R.; Lorenz, D.; Vukovi, M.; Broeders, E.; Bauwens, E.; Verhoelst, T.; Deroo, P.;Lenz, P.; Dehaes, S.; Ladjal, D.; Steininger, B.; Davignon, G.; Beck, P.; Yakut, K.; Drum-mond, R.; Fu, J.-N.; Jiang, X.-J.; Zhang, C.; Provencal, J.; Decin, L.Communications in Astroseismology, Vol. 150, p.191

“The infrared spectra of disks around binary post-AGB stars”Gielen, C.; Van Winckel, H.Baltic Astronomy, Vol. 16, p. 148-150

“Efficient radiative transfer in a circumstellar disk environment”Baes, M.; Vidal, E.; Van Winckel, H.; Deroo, P.; Gielen, C.Baltic Astronomy, Vol. 16, p. 92-94

Bibliography

Abia, C. & Isern, J. 2000, ApJ, 536, 438 91

Acke, B. & van den Ancker, M. E. 2004, A&A, 426, 151 15, 18, 82

Aden, A. L. & Kerker, M. 1951, J. Appl. Phys., 22, 1242 11, 39, 67

Alcock, C., Allsman, R. A., Alves, D. R., et al. 1998, AJ, 115, 1921 45, 111, 113, 116, 135,138, 140

Antoniucci, S., Paresce, F., & Wittkowski, M. 2005, A&A, 429, L1 7, 101, 126

Armus, L., Charmandaris, V., Bernard-Salas, J., et al. 2007, ApJ, 656, 148 15

Artymowicz, P., Clarke, C. J., Lubow, S. H., & Pringle, J. E. 1991, ApJ, 370, L35 6

Baes, M., Vidal, E., Van Winckel, H., Deroo, P., & Gielen, C. 2007a, Baltic Astronomy, 16,92 17

—. 2007b, Baltic Astronomy, 16, 92 154

Balick, B. & Frank, A. 2002, ARA&A, 40, 439 4

Blocker, T. 1995, A&A, 299, 755 3, 32

Bohren, C. F., Huffman, D. R., & Kam, Z. 1983, Nat, 306, 625 11, 40, 67

Bonacic Marinovic, A., Lugaro, M., Reyniers, M., & van Winckel, H. 2007, A&A, 472, L1112

Bonacic Marinovic, A. A., Glebbeek, E., & Pols, O. R. 2008, A&A, 480, 797 153

Boulanger, F., Cox, P., & Jones, A. P. 2000, in Infrared Space Astronomy, Today and Tomor-row, ed. F. Casoli, J. Lequeux, & F. David, 251–+ 9

Bouwman, J., de Koter, A., Dominik, C., & Waters, L. B. F. M. 2003, A&A, 401, 577 38,41, 46, 49, 82

174 BIBLIOGRAPHY

Bouwman, J., Henning, T., Hillenbrand, L. A., et al. 2008, The Astrophysical Journal, 683,479 13

Bouwman, J., Meeus, G., de Koter, A., et al. 2001, A&A, 375, 950 17, 41, 68, 82

Bradley, J., Dai, Z. R., Erni, R., et al. 2005, Science, 307, 244 10

Bradley, J. P., Brownlee, D. E., & Veblen, D. R. 1983, Nat, 301, 473 13, 39, 61

Brandl, B. R., Bernard-Salas, J., Spoon, H. W. W., et al. 2006, ApJ, 653, 1129 15

Bujarrabal, V., Bachiller, R., Alcolea, J., & Martin-Pintado, J. 1988, A&A, 206, L17 33, 52

Bujarrabal, V., Castro-Carrizo, A., Alcolea, J., & Neri, R. 2005, A&A, 441, 1031 5, 52, 53,86

Bujarrabal, V., Castro-Carrizo, A., Alcolea, J., & Sanchez Contreras, C. 2001, A&A, 377,868 4, 8

Bujarrabal, V., Van Winckel, H., Neri, R., et al. 2007, A&A, 468, L45 8, 52

Cami, J. 2002, PhD thesis, University of Amsterdam 81

Cami, J. & Yamamura, I. 2001, A&A, 367, L1 96, 97, 108

Chesneau, O., Lykou, F., Balick, B., et al. 2007, A&A, 473, L29 82

Chiang, E. I. & Goldreich, P. 1997, ApJ, 490, 368 17

Chihara, H., Koike, C., Tsuchiyama, A., Tachibana, S., & Sakamoto, D. 2002, A&A, 391,267 14

Close, L. M., Biller, B., Hoffmann, W. F., et al. 2003, ApJ, 598, L35 33, 45

Cohen, M., Anderson, C. M., Cowley, A., et al. 1975, ApJ, 196, 179 5, 7, 86

Cohen, M., Barlow, M. J., Sylvester, R. J., et al. 1999, ApJ, 513, L135 86

Cohen, M., Van Winckel, H., Bond, H. E., & Gull, T. R. 2004, AJ, 127, 2362 53, 86, 101,107, 126, 151

Crovisier, J., Leech, K., Bockelee-Morvan, D., et al. 1997, Science, 275, 1904 13, 38, 41

D’Alessio, P., Calvet, N., & Hartmann, L. 2001, ApJ, 553, 321 46

D’Alessio, P., Canto, J., Calvet, N., & Lizano, S. 1998, ApJ, 500, 411 17

De Marco, O., Hillwig, T. C., & Smith, A. J. 2008, AJ, 136, 323 4

De Ruyter, S. 2006, PhD thesis, KULeuven 7, 15

De Ruyter, S., Van Winckel, H., Dominik, C., Waters, L. B. F. M., & Dejonghe, H. 2005,A&A, 435, 161 3, 8, 32, 35, 87

BIBLIOGRAPHY 175

De Ruyter, S., Van Winckel, H., Maas, T., et al. 2006, A&A, 448, 641 5, 8, 21, 24, 32, 35,44, 45, 49, 52, 54, 55, 56, 86, 87, 88, 108, 149

Decin, L., Morris, P. W., Appleton, P. N., et al. 2004, ApJS, 154, 408 26

Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B., & Kotzlowski, H. 2000, in Society ofPhoto-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 4008, Society ofPhoto-Optical Instrumentation Engineers (SPIE) Conference Series, ed. M. Iye & A. F.Moorwood, 534–545 113

Deroo, P. 2007, PhD thesis, KULeuven 84, 149

Deroo, P., Acke, B., Verhoelst, T., et al. 2007a, A&A, 474, L45 8, 48, 52, 86, 87, 108, 155

Deroo, P., Van Winckel, H., Min, M., et al. 2006, A&A, 450, 181 84, 86, 89, 102, 104, 108,149

Deroo, P., Van Winckel, H., Verhoelst, T., et al. 2007b, A&A, 467, 1093 8, 48, 52, 92

Dominik, C., Dullemond, C. P., Cami, J., & van Winckel, H. 2003, A&A, 397, 595 5, 7, 101,126

Dorschner, J., Begemann, B., Henning, T., Jaeger, C., & Mutschke, H. 1995, A&A, 300, 50368

Draine, B. T. 1988, ApJ, 333, 848 68

Dullemond, C. P. 2002, A&A, 395, 853 19

Dullemond, C. P., Brauer, F., Henning, T., & Natta, A. 2008, Physica Scripta Volume T, 130,014015 10

Dullemond, C. P. & Dominik, C. 2004, A&A, 417, 159 17, 19, 44, 52, 53, 70, 83, 101, 132,154

Dullemond, C. P., Dominik, C., & Natta, A. 2001, ApJ, 560, 957 17, 19, 44, 101, 132

Dullemond, C. P., Hollenbach, D., Kamp, I., & D’Alessio, P. 2007, in Protostars and PlanetsV, ed. B. Reipurth, D. Jewitt, & K. Keil, 555–572 15, 154

Dullemond, C. P., Natta, A., & Testi, L. 2006, ApJ, 645, L69 154

Fabian, D., Jager, C., Henning, T., Dorschner, J., & Mutschke, H. 2000, A&A, 364, 282 13,61

Fazio, G. G., Hora, J. L., Allen, L. E., et al. 2004, ApJS, 154, 10 112

Feast, M. 1999, PASP, 111, 775 111

Furlan, E., Hartmann, L., Calvet, N., et al. 2006, ApJS, 165, 568 99

Garcıa-Hernandez, D. A., Abia, C., Manchado, A., & Garcıa-Lario, P. 2006a, A&A, 452,1049 91, 92

176 BIBLIOGRAPHY

Garcıa-Hernandez, D. A., Manchado, A., Garcıa-Lario, P., et al. 2006b, ApJ, 640, 829 126

Garcia-Lario, P., Manchado, A., Pych, W., & Pottasch, S. R. 1997, A&AS, 126, 479 3

Garcıa-Segura, G., Langer, N., Rozyczka, M., & Franco, J. 1999, ApJ, 517, 767 4

Garcıa-Segura, G., Lopez, J. A., & Franco, J. 2005, ApJ, 618, 919 4

Geers, V. C., van Dishoeck, E. F., Visser, R., et al. 2007, A&A, 476, 279 15, 18

Gillet, D., Burki, G., & Duquennoy, A. 1990, A&A, 237, 159 33, 35

Gillet, D., Duquennoy, A., Bouchet, P., & Gouiffes, C. 1989, A&A, 215, 316 33

Giridhar, S. & Ferro, A. A. 1989, Journal of Astrophysics and Astronomy, 10, 47 118

Giridhar, S., Lambert, D. L., & Gonzalez, G. 1998, ApJ, 509, 366 32, 54

—. 2000, ApJ, 531, 521 32, 33, 54, 87, 155

Giridhar, S., Lambert, D. L., Reddy, B. E., Gonzalez, G., & Yong, D. 2005, ApJ, 627, 432112, 119

Giridhar, S., Rao, N. K., & Lambert, D. L. 1994, ApJ, 437, 476 6, 32, 54

Gledhill, T. M., Chrysostomou, A., Hough, J. H., & Yates, J. A. 2001, MNRAS, 322, 321 57

Gonzalez, G., Lambert, D. L., & Giridhar, S. 1997a, ApJ, 481, 452+ 32, 54, 155

—. 1997b, ApJ, 479, 427+ 6, 32, 54, 81, 89, 155

Gordon, K. D., Clayton, G. C., Misselt, K. A., Landolt, A. U., & Wolff, M. J. 2003, ApJ, 594,279 132

Grynko, Y. & Shkuratov, Y. 2003, Journal of Quantitative Spectroscopy and Radiative Trans-fer, 78, 319 11, 40, 67

Gutenkunst, S., Bernard-Salas, J., Pottasch, S. R., Sloan, G. C., & Houck, J. R. 2008, ApJ,680, 1206 86

Hadrava, P. 2004, in Astronomical Society of the Pacific Conference Series, Vol. 318, Spec-troscopically and Spatially Resolving the Components of the Close Binary Stars, ed. R. W.Hilditch, H. Hensberge, & K. Pavlovski, 86–94 37

Henning, T. & Stognienko, R. 1996, A&A, 311, 291 68

Herczeg, G. J., Linsky, J. L., Valenti, J. A., Johns-Krull, C. M., & Wood, B. E. 2002, ApJ,572, 310 17

Herwig, F. 2005, ARA&A, 43, 435 2, 3

Higdon, S. J. U., Devost, D., Higdon, J. L., et al. 2004, PASP, 116, 975 26, 117

Hildebrand, R. H. 1983, QJRAS, 24, 267 47, 105

BIBLIOGRAPHY 177

Hillenbrand, L. A., Strom, S. E., Vrba, F. J., & Keene, J. 1992, ApJ, 397, 613 16

Hines et al. 2005, “FEPS Data Explanatory Supplement”, Version 3.0, Pasadena SSC 26, 117

Hinkle, K. H., Brittain, S. D., & Lambert, D. L. 2007, ApJ, 664, 501 5, 7, 96, 101, 108, 126

Honda, M., Kataza, H., Okamoto, Y. K., et al. 2004, ApJ, 610, L49 68

Hony, S., Tielens, A. G. G. M., Waters, L. B. F. M., & de Koter, A. 2003, A&A, 402, 211 86

Hony, S., Van Kerckhoven, C., Peeters, E., et al. 2001, A&A, 370, 1030 15, 60, 86

Hony, S., Waters, L. B. F. M., & Tielens, A. G. G. M. 2002, A&A, 390, 533 86

Houck, J. R., Roellig, T. L., van Cleve, J., et al. 2004, ApJS, 154, 18 56, 89, 117

Hrivnak, B. J., Kwok, S., & Volk, K. M. 1989, ApJ, 346, 265 3, 52

Izumiura, H., Noguchi, K., Aoki, W., et al. 2008, ApJ, 682, 499 92

Jaeger, C., Molster, F. J., Dorschner, J., et al. 1998, A&A, 339, 904 14, 68

Johnson, J. J., Anderson, C. M., Bjorkman, K. S., et al. 1999, MNRAS, 306, 531 5, 7, 101,126

Jones, A. P., Tielens, A. G. G. M., Hollenbach, D. J., & McKee, C. F. 1994, ApJ, 433, 797 9

Jorissen, A., Frankowski, A., Famaey, B., & Van Eck, S. 2009, ArXiv e-prints 153

Jura, M. 1986, ApJ, 309, 732 3, 32, 87

—. 2003, ApJ, 582, 1032 100

Jura, M., Balm, S. P., & Kahane, C. 1995, ApJ, 453, 721 33

Jura, M., Bohac, C. J., Sargent, B., et al. 2006, ApJ, 637, L45 86, 99, 100, 101, 126

Jura, M., Chen, C., & Werner, M. W. 2000, ApJ, 541, 264 33

Jura, M. & Kahane, C. 1999, ApJ, 521, 302 3, 33, 87

Justtanont, K., de Jong, T., Helmich, F. P., et al. 1996, A&A, 315, L217 13

Justtanont, K., Feuchtgruber, H., de Jong, T., et al. 1998, A&A, 330, L17 96

Kaufl, H.-U., Sterzik, M. F., Siebenmorgen, R., et al. 2003, in Presented at the Societyof Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol. 4841, InstrumentDesign and Performance for Optical/Infrared Ground-based Telescopes. Edited by Iye,Masanori; Moorwood, Alan F. M. Proceedings of the SPIE, Volume 4841, pp. 117-128(2003)., ed. M. Iye & A. F. M. Moorwood, 117–128 22, 27, 56

Keller, L. P., Hony, S., Bradley, J. P., et al. 2002, Nat, 417, 148 18

Kemper, F., Molster, F. J., Jager, C., & Waters, L. B. F. M. 2002, A&A, 394, 679 86

178 BIBLIOGRAPHY

Kemper, F., Vriend, W. J., & Tielens, A. G. G. M. 2004, ApJ, 609, 826 61

Kessler-Silacci, J., Augereau, J.-C., Dullemond, C. P., et al. 2006, ApJ, 639, 275 17, 34, 56,90

Kholopov, P. N., Samus, N. N., Frolov, M. S., et al. 1999, VizieR Online Data Catalog, 2214,0 89

Klahr, H. & Kley, W. 2006, A&A, 445, 747 154

Koike, C., Chihara, H., Tsuchiyama, A., et al. 2003, A&A, 399, 1101 14

Kraemer, K. E., Sloan, G. C., Bernard-Salas, J., et al. 2006, ApJ, 652, L25 99

Kurucz, R. L. 1979, ApJS, 40, 1 56

Kwok, S., Boreiko, R. T., & Hrivnak, B. J. 1987, ApJ, 321, 975 3

Kwok, S., Purton, C. R., & Fitzgerald, P. M. 1978, ApJ, 219, L125 4

Kwok, S., Volk, K. M., & Hrivnak, B. J. 1989, ApJ, 345, L51 53, 86

Lahuis, F. & Boogert, A. 2003, in SFChem 2002: Chemistry as a Diagnostic of Star Forma-tion, ed. C. L. Curry & M. Fich, 335–+ 26

Lahuis et al. 2006, c2d Spectroscopy Explanatory Supplement, Cores to Disks, SpitzerLegacy Team, (Pasadena: Spitzer Science Center) 25, 34, 56, 90

Lisse, C. M., Kraemer, K. E., Nuth, J. A., Li, A., & Joswiak, D. 2007, Icarus, 187, 69 49, 53,82

Little-Marenin, I. R. 1986, ApJ, 307, L15 86

Lloyd Evans, T. 1985, MNRAS, 217, 493 33, 52

—. 1990, MNRAS, 243, 336 86, 91

—. 1991, MNRAS, 249, 409 91

Lloyd Evans, T. 1999, in IAU Symposium, Vol. 191, Asymptotic Giant Branch Stars, ed.T. Le Bertre, A. Lebre, & C. Waelkens, 453–+ 52

Lodders, K. 2003, ApJ, 591, 1220 7, 125

Lugaro, M., Pols, O., Karakas, A. I., & Tout, C. A. 2005, Nuclear Physics A, 758, 725 108

Lyra, W., Johansen, A., Klahr, H., & Piskunov, N. 2008, A&A, 479, 883 154

Maas, T., Van Winckel, H., & Lloyd Evans, T. 2003, in Astronomical Society of the PacificConference Series, Vol. 303, Astronomical Society of the Pacific Conference Series, ed.R. L. M. Corradi, J. Mikolajewska, & T. J. Mahoney, 143–+ 52

Maas, T., Van Winckel, H., & Lloyd Evans, T. 2005, A&A, 429, 297 6, 32, 54, 112, 119,149, 155

BIBLIOGRAPHY 179

Maas, T., Van Winckel, H., & Waelkens, C. 2002, A&A, 386, 504 33, 34, 35, 39, 52, 81

Malfait, K., Waelkens, C., Waters, L. B. F. M., et al. 1998, A&A, 332, L25 18, 49

Mannings, V. & Sargent, A. I. 1997, ApJ, 490, 792 16

Mastrodemos, N. & Morris, M. 1999, ApJ, 523, 357 8, 54, 92, 153

Matsuura, M., Yamamura, I., Cami, J., Onaka, T., & Murakami, H. 2002, A&A, 383, 972 97

Matsuura, M., Zijlstra, A. A., Molster, F. J., et al. 2004, ApJ, 604, 791 126

Meeus, G., Waters, L. B. F. M., Bouwman, J., et al. 2001, A&A, 365, 476 15, 17

Meixner, M., Gordon, K. D., Indebetouw, R., et al. 2006, AJ, 132, 2268 7, 54, 112, 113, 156

Men’shchikov, A. B., Schertl, D., Tuthill, P. G., Weigelt, G., & Yungelson, L. R. 2002, A&A,393, 867 52, 86, 101, 107, 126

Menzies, J. W. & Whitelock, P. A. 1988, MNRAS, 233, 697 57

Min, M., Dominik, C., & Waters, L. B. F. M. 2004, A&A, 413, L35 11

Min, M., Dullemond, C. P., Dominik, C., de Koter, A., & Hovenier, J. W. 2009, ArXiv e-prints 17, 154

Min, M., Hovenier, J. W., & de Koter, A. 2003, A&A, 404, 35 11, 14, 39, 40, 67

—. 2005a, A&A, 432, 909 11, 14, 39, 40, 67

Min, M., Hovenier, J. W., de Koter, A., Waters, L. B. F. M., & Dominik, C. 2005b, Icarus,179, 158 38, 49, 82

Min, M., Hovenier, J. W., Waters, L. B. F. M., & de Koter, A. 2008, A&A, 489, 135 68

Min, M., Waters, L. B. F. M., de Koter, A., et al. 2007, A&A, 462, 667 38, 61, 68

Miyake, K. & Nakagawa, Y. 1995, ApJ, 441, 361 45, 101

Molster, F. J., Waters, L. B. F. M., & Tielens, A. G. G. M. 2002a, A&A, 382, 222 7, 15, 38,53, 61, 69, 86, 87

Molster, F. J., Waters, L. B. F. M., Tielens, A. G. G. M., & Barlow, M. J. 2002b, A&A, 382,184 38, 43, 53, 61

Molster, F. J., Waters, L. B. F. M., Tielens, A. G. G. M., Koike, C., & Chihara, H. 2002c,A&A, 382, 241 38, 53, 61

Molster, F. J., Yamamura, I., Waters, L. B. F., et al. 2001, A&A, 366, 923 91, 92

Molster, F. J., Yamamura, I., Waters, L. B. F. M., et al. 1999, Nat, 401, 563 33, 38, 53, 126

Natta, A., Prusti, T., Neri, R., et al. 2001, A&A, 371, 186 17

Nordhaus, J., Blackman, E. G., & Frank, A. 2007, MNRAS, 376, 599 4

180 BIBLIOGRAPHY

Ohnaka, K., Izumiura, H., Leinert, C., et al. 2008, A&A, 490, 173 92

Ohnaka, K. & Tsuji, T. 1999, A&A, 345, 233 91

Oudmaijer, R. D., van der Veen, W. E. C. J., Waters, L. B. F. M., et al. 1992, A&AS, 96, 6253, 52

Peeters, E., Hony, S., Van Kerckhoven, C., et al. 2002, A&A, 390, 1089 15, 60, 81, 86, 98,126, 127

Peeters, E., Mattioda, A. L., Hudgins, D. M., & Allamandola, L. J. 2004, ApJ, 617, L65 96

Pollard, K. R., Cottrell, P. L., Kilmartin, P. M., & Gilmore, A. C. 1996, MNRAS, 279, 94933

Pollard, K. R., Cottrell, P. L., Lawson, W. A., Albrow, M. D., & Tobin, W. 1997, MNRAS,286, 1 34, 35

Pols, O. R., Karakas, A. I., Lattanzio, J. C., & Tout, C. A. 2003, in Astronomical Society ofthe Pacific Conference Series, Vol. 303, Astronomical Society of the Pacific ConferenceSeries, ed. R. L. M. Corradi, J. Mikolajewska, & T. J. Mahoney, 290–+ 6

Preston, G. W., Krzeminski, W., Smak, J., & Williams, J. A. 1963, ApJ, 137, 401 32, 89

Raveendran, A. V. 1989, MNRAS, 238, 945 52

Reimann, H.-G., Linz, H., Wagner, R., et al. 2000, in Presented at the Society of Photo-Optical Instrumentation Engineers (SPIE) Conference, Vol. 4008, Proc. SPIE Vol. 4008, p.1132-1143, Optical and IR Telescope Instrumentation and Detectors, Masanori Iye; AlanF. Moorwood; Eds., ed. M. Iye & A. F. Moorwood, 1132–1143 22, 27, 56

Reyniers, M., Abia, C., van Winckel, H., et al. 2007, A&A, 461, 641 7, 86, 111, 118, 125,155

Reyniers, M. & Van Winckel, H. 2007, A&A, 463, L1 7, 49, 54, 111, 112, 124, 125, 155

Reyniers, M., Van Winckel, H., Gallino, R., & Straniero, O. 2004, A&A, 417, 269 86

Rieke, G. H., Young, E. T., Engelbracht, C. W., et al. 2004, ApJS, 154, 25 112

Roddier, F., Roddier, C., Graves, J. E., & Northcott, M. J. 1995, ApJ, 443, 249 52

Rothman, L. S., Jacquemart, D., Barbe, A., et al. 2005, Journal of Quantitative Spectroscopyand Radiative Transfer, 96, 139 97

Ryde, N., Eriksson, K., & Gustafsson, B. 1999, A&A, 341, 579 96

Sahai, R., Balick, B., Blackman, E., et al. 2009, ArXiv e-prints 4

Sahai, R., Morris, M., Sanchez Contreras, C., & Claussen, M. 2007, AJ, 134, 2200 4

Sahai, R. & Trauger, J. T. 1998, AJ, 116, 1357 4

Sargent, B., Forrest, W. J., D’Alessio, P., et al. 2006, ApJ, 645, 395 17

BIBLIOGRAPHY 181

Savage, B. D. & Mathis, J. S. 1979, ARA&A, 17, 73 35, 56, 57

Schneller, H. 1931, Astronomische Nachrichten, 243, 99 89

Schutz, O., Meeus, G., & Sterzik, M. F. 2005, A&A, 431, 175 82

Servoin, J. L. & Pirou, B. 1973, Phys.Stat.Sol.B, 55, 677 68

Shenton, M., Evans, A., & Williams, P. M. 1995, MNRAS, 273, 906 33

Shkuratov, Y. G. & Grynko, Y. S. 2005, Icarus, 173, 16 11, 40, 67

Sloan, G. C., Jura, M., Duley, W. W., et al. 2007, ApJ, 664, 1144 99, 100

Sloan, G. C., Keller, L. D., Forrest, W. J., et al. 2005, ApJ, 632, 956 18, 99

Sloan, G. C., Kraemer, K. E., Goebel, J. H., & Price, S. D. 2003, ApJ, 594, 483 96

Sloan, G. C., Kraemer, K. E., Wood, P. R., et al. 2008, ApJ, 686, 1056 86

Sofia, U. J., Gordon, K. D., Clayton, G. C., et al. 2006, ApJ, 636, 753 39, 81

Soker, N. 2000, A&A, 357, 557 6

Soker, N. & Livio, M. 1989, ApJ, 339, 268 4, 153

Space Telescope Science Institute, . & Osservatorio Astronomico di Torino. 2007, VizieROnline Data Catalog, 1305, 0 156

Steenman, H. & The, P. S. 1989, Ap&SS, 159, 189 57

—. 1991, Ap&SS, 184, 9 57

Stellingwerf, R. F. 1978, ApJ, 224, 953 36

Suarez, O., Garcıa-Lario, P., Manchado, A., et al. 2006, A&A, 458, 173 3

Sylvester, R. J., Skinner, C. J., & Barlow, M. J. 1998, MNRAS, 301, 1083 86

Szczerba, R., Siodmiak, N., Stasinska, G., & Borkowski, J. 2007, A&A, 469, 799 3, 52

Thevenin, F. 1989, A&AS, 77, 137 118

—. 1990, A&AS, 82, 179 118

Thi, W. F., van Dishoeck, E. F., Blake, G. A., et al. 2001, ApJ, 561, 1074 17

Thi, W.-F., van Zadelhoff, G.-J., & van Dishoeck, E. F. 2004, A&A, 425, 955 17

Tielens, A. G. G. M. 2008, ARA&A, 46, 289 15, 98, 126

Tielens, A. G. G. M., McKee, C. F., Seab, C. G., & Hollenbach, D. J. 1994, ApJ, 431, 321 9

Tielens, A. G. G. M., Waters, L. B. F. M., Molster, F. J., & Justtanont, K. 1998a, Ap&SS,255, 415 12, 13, 81

182 BIBLIOGRAPHY

—. 1998b, Ap&SS, 255, 415 39, 61

Toon, O. B. & Ackerman, T. P. 1981, Appl. Opt., 20, 3657 11, 39, 67

Trams, N. R., Waters, L. B. F. M., Lamers, H. J. G. L. M., et al. 1991, A&AS, 87, 361 5

van Boekel, R., Min, M., Waters, L. B. F. M., et al. 2005, A&A, 437, 189 13, 18, 22, 27, 34,53, 56, 63, 82, 83

van Boekel, R., Waters, L. B. F. M., Dominik, C., et al. 2003, A&A, 400, L21 18, 63

van de Steene, G. C., Ueta, T., van Hoof, P. A. M., Reyniers, M., & Ginsburg, A. G. 2008,A&A, 480, 775 126

Van Kerckhoven, C., Tielens, A. G. G. M., & Waelkens, C. 2002, A&A, 384, 568 18

Van Kerckhoven, C., Hony, S., Peeters, E., et al. 2000, A&A, 357, 1013 60

Van Winckel, H. 2003, ARA&A, 41, 391 3, 4, 6, 53, 108

—. 2007, Baltic Astronomy, 16, 112 4, 6, 52, 54, 86, 112

Van Winckel, H., Mathis, J. S., & Waelkens, C. 1992, Nat, 356, 500 89

Van Winckel, H. & Reyniers, M. 2000, A&A, 354, 135 86, 118

Van Winckel, H., Waelkens, C., Fernie, J. D., & Waters, L. B. F. M. 1999, A&A, 343, 20232, 52, 87, 89

Van Winckel, H., Waelkens, C., & Waters, L. B. F. M. 1995, A&A, 293, L25 5, 32, 52, 86,89, 155

Van Winckel, H., Waelkens, C., Waters, L. B. F. M., et al. 1998, A&A, 336, L17 6, 7, 32, 33,35, 39, 48, 54, 81, 87, 118

Verhoelst, T., van Aarle, E., & Acke, B. 2007, A&A, 470, L21 54, 82, 153

Volk, K., Kwok, S., & Hrivnak, B. J. 1999, ApJ, 516, L99 86

Waelkens, C., Lamers, H. J. G. L. M., Waters, L. B. F. M., et al. 1991a, A&A, 242, 433 5, 89

Waelkens, C., Van Winckel, H., Bogaert, E., & Trams, N. R. 1991b, A&A, 251, 495 89

Waelkens, C., Van Winckel, H., Waters, L. B. F. M., & Bakker, E. J. 1996a, A&A, 314, L176

Waelkens, C., Waters, L. B. F. M., de Graauw, M. S., et al. 1996b, A&A, 315, L245 13, 53

Wallerstein, G. 2002, PASP, 114, 689 3

Waters, L. B. F. M., Cami, J., de Jong, T., et al. 1998, Nat, 391, 868 5, 52, 86, 96

Waters, L. B. F. M., Molster, F. J., de Jong, T., et al. 1996, A&A, 315, L361 13, 53, 86

BIBLIOGRAPHY 183

Waters, L. B. F. M., Trams, N. R., & Waelkens, C. 1992, A&A, 262, L37 7, 32, 39, 54, 81,87, 136

Waters, L. B. F. M. & Waelkens, C. 1998, ARA&A, 36, 233 16

Waters, L. B. F. M., Waelkens, C., Mayor, M., & Trams, N. R. 1993, A&A, 269, 242 5

Watson, D. M., Leisenring, J. M., Furlan, E., et al. 2007, ArXiv e-prints, 704 82

Werner, M. W., Roellig, T. L., Low, F. J., et al. 2004, ApJS, 154, 1 56, 89, 117

Westerlund, B. E. 1997, Cambridge Astrophysics Series, 29 111

Willacy, K. 2004, ApJ, 600, L87 101

Witt, A. N., Vijh, U. P., Hobbs, L. M., et al. 2008, ArXiv e-prints 5, 86, 151

Yamamura, I., Dominik, C., de Jong, T., Waters, L. B. F. M., & Molster, F. J. 2000, A&A,363, 629 91, 92

Zaritsky, D., Harris, J., Thompson, I. B., & Grebel, E. K. 2004, AJ, 128, 1606 156

Zsoldos, E. 1993, A&A, 268, 149 33

—. 1995, A&A, 296, 122 89

Zsom, A. & Dullemond, C. P. 2008, A&A, 489, 931 10

Katholieke Universiteit Leuven · werden gebracht, al het over-en-weer gesleur met meubelen, elke...

Documents

Transcript of Katholieke Universiteit Leuven · werden gebracht, al het over-en-weer gesleur met meubelen, elke...