Thesis Is Closely Related to This Question

UvA-DARE (Digital Academic Repository)

The processing and evolution of dust in Herbig Ae/Be systems.

Bouwman, J.

Publication date 2001 Document Version Final published version

Link to publication

Citation for published version (APA): Bouwman, J. (2001). The processing and evolution of dust in Herbig Ae/Be systems.

General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

Download date:10 Oct 2021 Thee processing and evolution of dustt in Herbig Ae/Be systems

Jeroenn Bouwman Thee processing and evolution of dust in Herbig Ae/Be systems s

Thee processing and evolution of dust in Herbig Ae/Be systems s

Dee evolutie van stof in Herbig Ae/Be systemen.

Academischh Proefschrift

terr verkrijging van de graad van doctor aann de Universiteit van Amsterdam, opp gezag van de Rector Magnificus prof. dr J.J.M. Franse, tenn overstaan van een door het college voor promoties ingestelde commissie, inn het openbaar te verdedigen in de Aula der Universiteit

op p

dinsdagg 25 september 2001, te 12:00 uur

door r Jeroenn Bouwman

geborenn te Zaandam PROMOTIF.CO.MMISSIH. .

PROMOTORFSS prof. dr L.B.F.M. Waters prof.. dr J.W. Hovenier CO-PROMOTORR dr A. de Koter OvF.RICHH LFDHX prof. dr E.P.J. van den Heuvel drr A. Natta prof.. dr T. de Jong prof.. dr P.M.A. Sloot prof.. dr A.G.G.M. Tielens prof.. dr C. Waelkens

Sterrenkundigg Instituut "Anton Pannekoek' Faculteitt der Natuurwetenschappen Universiteitt van Amsterdam

ISBNN 90-9015117-6

Cover/Omslag:: Scattering and absorption or light by silicate grains. TSJA, , ditt is r dan.

Contents s

11 From dust to planets 1 1.11 Introduction 1 1.22 Pre-Main-Sequence evolution: T Tauri and Herbig Ac/Bc stars .... 3 1.33 Protoplanetary disks and dust processing 5 1.44 The geometry or the circumstellar disks 8 1.55 Conclusions & outlook 10

22 Dust physics, radiative transfer and spectral analysis 13 2.11 Radiative transfer 13 2.1.11 An optically thin medium 15 2.1.22 Radiative transfer in axisvmmetric geometry 16 2.22 Chemical composition and grain structure of circumstellar dust .... 18 2.2.11 Silicates 18 2.2.22 Other dust species 20 2.2.33 Dust chemistry 21 2.2.44 The grain structure 23 2.33 The optical properties of dust grains 24 2.3.11 The refractive index 24 2.3.22 Basic scattering theory 25 2.3.33 Extrapolation of the refractive index to short and long wavelengths 27 2.44 The spatial distribution of the dust 28 2.55 I he strategy of modeling 29

33 ISO spectroscopy of circumstellar dust in 14 Herbig Ae/Be systems: towards ann understanding of dust processing. 33 3.11 Introduction 34 3.22 Targets and Observations 36 3.2.11 Spectral Hnergy Distributions 38 3.2.22 Properties of the cold grains: mass and grain size 43 3.33 ISO-SWS spectra and their wealth of features 43 3.3.11 I he 2-7 micron wavelength region 45 3.3.22 The PAH bands .../../ 46

i i CONTENTS S

3.3.33 The silicates 47 3.3.44 The 23 micron feature 47 3.3.55 Notes on individual sources 47 3.44 Discussion 49 3.4.11 Geometry of the disc and its effects on the SED 50 3.4.22 Evidence for grain growth 53 3.4.33 Influence of the stellar age on dust properties 53 3.4.44 The amorphous silicate behaviour 54 3.55 Conclusion 55

44 Processing of silicate dust grains in Herbig Ae/Be systems 57 4.11 Introduction 58 4.22 Dust composition and spectral analysis 59 4.2.11 Adopted dust components and grain shapes 59 4.2.22 Spectral analysis 62 4.2.33 Detection of aliphatic hydrocarbons in HD 163296 64 4.33 Results 64 4.3.11 Correlations 68 4.3.22 Correlation to the overall Spectral Energy Distribution .... 73 4.44 Discussion 75 4.4.11 Explanation for the change of the 10 ^m feature 75 4.4.22 The chemical composition of the silicate dust 76 4.4.33 Deviating objects 77 4.55 Conclusions 79

55 The composition of the circumstellar dust around the Herbig Ae stars ABB Aur and HD 163296 81 5.11 Introduction 81 5.1.11 Geometry of the circumstellar dust 82 5.1.22 The onset of near-IR emission 84 5.22 Method and assumptions 85 5.2.11 Model approach and assumptions 85 5.2.22 Adopted chemical composition of the dust 87 5.33 Results 91 5.3.11 The 2—8/jm spectral region 93 5.3.22 The 8-30 /urn region 94 5.3.33 The cold dust component 97 5.3.44 PAHs and crystalline silicates 100 5.44 Discussion 101 5.55 Summary 104

ii i CONTENTS S

66 The formation of crystalline silicate dust: From HD100546 to Hale-Bopp 105 6.11 Introduction 106 6.22 The deviating dust composition and spectral energy distribution of HDD 100546 107 6.33 Modelling 110 6.3.11 Size and shape properties of grains Ill 6.3.22 Chemical composition of grains 112 6.44 Results H2 6.4.11 HD 100546 113 6.4.22 Hale-Bopp 121 6.55 Discussion 124 6.5.11 The mass temperature distribution 124 6.5.22 The disk structure of HD 100546 and the origin of the crystallinee silicates 128 6.66 Summary 131

77 Constraints on HAEBE disk geometry from Spectral Energy Distributions 133 7.11 Introduction 133 7.22 The disk model 134 7.33 Discussion 13o

Nederlandsee samenvatting 141

Bibliographyy 146

iii i CONTENTS S

IV V CHAPTERR 1

Fromm dust to planets

1.11 Introduction

nee of rhe most intriguing questions in astronomy was, and is, how our solar system,, the sun, planets, and objects like comets were formed. The work pre- O sentedd in this thesis is closely related to this question. In the following chapters severall studies of young stellar objects (YSOs) are presented, which are believed to be in thee process of forming a planetary system. Though the process of star formation is far fromm completely understood, the most generally accepted view on how stars form was putt forward by Shu et al. (1987). Star formation can be divided into five distinct phases. Thee first phase is the formation of dense molecular cores inside a giant molecular cloud,, such as for instance the Orion nebula. Typically these cores have masses in the orderr of a few solar masses and sizes of less than a parsec. (see for instance Evans 1999, forr an overview of the core properties). The cores are initially supported against gravity byy magnetic fields and turbulence, but eventually will collapse due to processes such ass ambipolar diffusion (i.e. the drift of the magnetic field relative to the neutral gas), reducingg the magnetic field. This marks the beginning of the second evolutionary phase. Ass the slowly rotating core collapses, a central protostar and surrounding disk will form, stilll deeply embedded within an infalling envelope of gas and dust. In the third phase, as thee central star accretes matter through the disk, a strong stellar wind can develop along thee rotational axis of the system, creating a bipolar outflow. In the fourth phase, this stellarr wind will disperse the surrounding gas and dust terminating the infall of matter, leavingg the protostar and circumstellar disk. Finally, the dust and gas in the circumstellar diskk will be dispersed or incorporated into a planetary system. Observationallyy the last four phases in the formation of a star as described above, can bee identified by the spectral energy distribution (SED) of the YSOs. Lada and Wilking (1984)) suggested a classification based on the properties of the SED into three classes, whichh were shown to correspond to different evolutionary stages of the protostar (Adams ett al. 1987). This classification was expanded with an additional class (Class 0), after the discoveryy of a new group of objects by Andre et al. (1993) representing an earlier evolutionaryy phase. Fig. 1.1 shows a schematic representation of this classification. 1 he panels onn the left show the tvpical SED of objects within a group, the panels on the right show thee corresponding geometry or spatial distribution of material in the system. Class 0 objectss represent second phase of star formation, and are only visible at far infrared (1R) andd millimetre wavelengths. Class I objects represent the third phase in which the star iss still embedded, resulting in deep absorption features. Class II objects represent the

1 1 CHAPTERR 1

\ \

\\t \\t M M CLASSS II >"7 7 %% -8 n n /^ ^ -9 9"" , / / , ;

122 13 14 15 5 Logg v (Hz)

Figuree 1.1: Schematic overview of the various stages in the formation of a star evolving from aa Class 0 to a Class III object (after Nana 1999). The left panels show the typical SED's (solid lines),wheree the dashed lines represent the emission from the central protostar. In the right panels thee system geometry is displayed showing the distribution of circumstellar gas and dust. Idle light andd dark shaded areas depict low and high densities, respectively. The arrows indicate infall or outfloww of material. In Class III objects, cometary sized objects have formed in the flattened, gas depletedd circumstellar disk.

2 2 FROMM DUST TO IT.AN HIS fourthh phase of the formation of a protostar, where the surrounding material has been largelyy dissipated, leaving a protostar with an optically thick circumstellar disk. The last groupp of objects, Class III, are comprised of objects with almost no IR excess. The circumstellarr disk in these systems is being dissipated, leaving a debris disk or possibly a planetaryy system. The systems which are presented in this thesis, a sub-set of the group off Herbig Ae/Be stars, belong to Class II.

1.22 Pre-Main-Sequence evolution: T Tauri and Herbig Ae/Be stars s

Spectroscopicallyy a class of stars with late type absorption spectra was discovered, show- ingg strong emission lines of hydrogen, calcium and iron. These stars showed also vari- ationss in brightness and were often associated with nebulous regions. Called after the brightestt member T Tauri, Walker (1956) suggested that these T Tauri stars are low- masss (VU < IMQ) pre-main-sequence (PMS) stars. Herbig (1960) identified a class of Aee and Be stars with similar properties as the T Tauri stars, now referred to as Herbig Ae/Bee (HAEBE) stars. The criteria used to classify stars as HAEBE are: (1) spectral type AA or B with emission lines, (2) presence of IR excess due to circumstellar dust, and (3) luminosityy class III to V (see also Waters and Waelkens 1998). Herbig (1960) suggested thatt these stars were the high mass counterparts of the T Tauri stars. Many studies have sincee confirmed the young PMS nature of HAEBE stars. Hipparcos parallaxes (e.g. van denn Ancker et al. 1998) allowed accurate positions in the Herzsprung-Russell (HR) di- agramm to be obtained, and from comparison to PMS evolutionary tracks, masses of the orderr of 2 to 10 Mr and stellar ages, calculated from the birthline, of typically 106 to 100 yrs were found. Shownn in Fig. 1.2 is the HR diagram for a sample of HAEBE stars (van den Ancker ett al. 1998). Also plotted in this figure are the evolutionary tracks of PMS stars and two birthliness (Palla and Stahler 1993). Observationally, the birthline marks the positions inn the HR-diagram where the stars first become optically visible during their evolution towardss the zero age main sequence (ZAMS). It marks the phase in the protostellar evolutionn where the accretion of mass on the protostar drops to very low values, and the surroundingg envelope, left over from the star forming process, is largely dissipated. I he exactt position of the birthline is determined by the mass accretion rate before the accretionn stops. Plotted in Fig. 1.2 are the theoretical birthlines for mass accretion rates of 10~4 andd 10~^ MQyr"1 . The phase where the accretion rate has dropped considerably and the masss of the protostar is basically fixed is generally called the pre-main-sequence phase. Thee star is then in hvdrostatic equilibrium and slowly contracts on a Kelvin-Helmholtz timescalee towards the ZAMS, where core hydrogen burning starts. The theoretical curves forr this evolutionary phase are plotted in Fig. 1.2 (solid and dashed lines) for different stellarr masses. The fundamental stellar property that separates T Tauri and HAF.BE stars iss the presence of a large convective layer in T Tauri stars, while HAEBE stars evolve alongg radiative tracks.

3 3 CHAPTERR 1

15.00 ML

9.00 M3 3

oo 3 5.00 M J J c c

44 6 3.6 6

Figuree 1.2: Hertzsprung-Russell diagram of Herbig Ac/Be stars (taken from van den Ancker ett al. 1998). The positions of the HAKBE stars are marked with filled circles. Plotted are the- oreticall evolutionary tracks for pre-main-sequence stars (solid and dashed lines) and the birth liness for 10 and 10~^ MoyL (upper and lower dotted lines, respectively) by Palla and Stahler (1993).. The evolutionary tracks are calculated trom the birthlines. The solid lines have as start- 5 ingg point the birthline for an accretion rate of 10~ M(. yr~' . They are extended (dashed lines) withh tracks starting at the birthline for an accretion rate of 10~4 MQyr . These accretion rates aree rates at which matter was accreted by the protostar before the dissipation of the surrounding envelope,, effectively halting the accretion.

HAEBEE stars form a less homogeneous group than the I Tauri stars. Hillenbrand ett al. (1992) made a similar classification of HAEBE systems based on the properties off their SED as had been suggested for T Tauri stars by Eacla and Wilking (1984). However,, Fuente et al. (1998) pointed out that the correspondance between the SED andd evolutionary PMS phase of stars of spectral type earlier than B5 differs from that off later spectral types. They explained this by a much more efficient removal of the surroundingg envelope material around the early type stars. In this thesis, however, we willl study HAEBE systems with spectral type B9 or later. The PMS evolution of these

4 4 FROMM DUST TO IT.ANITS starss closelv follows that or the lower mass T Tauri systems (apart from the radiative or convectivee envelope distinction). For systems more massive than ~10 MQ no optical PMSS phase can be identified. Such massive stars contract much more rapidly than the intermediatee or lower mass stars, leading to the ignition or hydrogen during the accretion phase.. Massive stars are consequently born on the ZAMS. Ourr research efforts are rocussed on the circumstellar disks surrounding PMS stars. Typically,, these disks have sizes in the range of ~-100 AU and disk masses in the order of :: l(r M0(e.g. Mannings and Sargent 1997, 2000; Natta et al. 2000a). The disk masses comparedd to the stellar mass vary little for stars with spectral type between A0 and M7. Earlierr tvpe stars show smaller values for their relative disk masses. Observations show thatt both gas and solid state components (i.e. dust) are present in these systems. Im- portantt questions we hope to answer are: What is the composition of the disk material? Whichh processes determine this composition and its evolution within the circumstellar disk?? Further we would like to know the spatial distribution of the circumstellar material andd which parameters determine this. In this thesis we use the dust as a diagnostic tool too study the composition and geometry of the disk. Thee best way to study the compositional properties of dust grains is by thermal infraredd spectroscopy. This spectral region contains the most important ro-vibrational transitionss of abundant dust species and is particular)' well suited to obtain detailed informationn about the composition, shape and structure of the dust grains. By observing dustt in various environments, a lot of knowledge about the formation history and processingg of this material can thus be obtained. For this purpose we have obtained spectra off 14 HAEBE stars observed with the Short Wavelength Spectrometer (SWS; de Graauw ett al. 1996) and Long Wavelength Spectrometer (SWS; Clegg et al. 1 996) onboard the In- fraredfrared Space Observatory (ISO; Kessler et al. 1996). We have chosen our sample using the followingg selection criteria: (1) Because we want to study the composition and geometry off circumstellar disks around young stellar objects, the IR emission should be dominated byy the disk, i.e. we focus on stars with low present-day accretion rates that are optically bright:: Class II objects (Fig 1.1). (2) They have to be observable with the SWS and LWS instruments,, which makes the study of T Tauri stars difficult. These latter objects are intrinsicallyy less luminous than HAEBE stars and are often below the detection limit of SWSS and LWS. (3) They have to be isolated, i.e. not inside a star forming region, so that confusionn with other IR sources, like nearby PMS stars or background emission from thee molecular cloud, is minimalised. A detailed description of the sample of HAE.BE systemss studied in this thesis is given in Chapter. 3.

1.33 Protoplanetary disks and dust processing

Thee formation of planets has been a puzzle, occupying astronomers for centuries. This becamee even more true after the first detection of an extrasolar planet around a solar type starr (Mayor and Queloz 1995). The disks around the HAEBE stars are believed to be the sitee of ongoing planet formation (from hereon we will refer to these disks as protoplanetaryy disks). The sub-micron sized dust grains present at the formation time in these disks

} } CHAPTERR 1

250 0

200 0 HDD 100546

>;; 150 Comett f ' ' l l 100 0 V V

. - 50 0 1-1 1 / / Forsterite e 0 0 0 0 20 0 30 0 40 0 Mn mni i

Figuree 1.3: The ISO-SWS spectra of the Herbig Be star HD 100546 and the solar system comett Hale-Bopp (after Malfait et al. 1998b). Also plotted in the figure is the thermal emission spectrumm of the crystalline silicate forsterite at a 160 K. Note the similarity between the spectra of bothh objects, and the emission properties of the crystalline silicate, implying similar compositions withh a high mass fraction of crystalline material.

cann coagulate to form larger objects and eventually planets (e.g. Weidenschilling 1997). 11 he HAEBE stars may be the precursors of Vega-type systems like (3 Pictoris. These are main-sequencee stars with a debris disk. In such a disk, most of the gas has been dissipated andd is no longer dynamically important. The observed dust in these disks is produced by collisionss between planetesimals (e.g. Weissman 1984), not unlike Kuiper belt objects. Resently,, planets have been detected in Vega-type systems (Trilling et al. 2000; Pantin ett al. 2000), directly linking the formation of planets to the disks from which they were born.. So far, no planets could be detected around PMS stars, which is probably due to observationall difficulties. Wee can, however, study the onset of planet formation by observing the dust, the buildingg blocks of planets, in the protoplanetary disks. By deriving the dust composition, andd identifying the processes governing the dust chemistry and growth, valuable insights cann be gained in the planetary formation process. The results of these analyses can be comparedd directly with observations of solar system objects like comets, meteorites and interplanetaryy dust particles (IDPs). A subset of these latter particles, loosely bound dust aggregatess believed to originate from comets, are left over from the early history of the solarr system and provide us with a record of the processes involved in the protoplanetary diskk phase of our own solar system.

6 6 FROMM DUST TO PLANF.TS

Thee most direct link between the HAEBE stars and our own solar system was found byy comparing 1SO-SWS observations of the Herbig B9Ve star HD 100546 and the solar systemm comet Hale-Bopp (Malfait et al. 1998b). As can be seen from Fig. 1.3, the spectra aree remarkably similar, implying identical dust compositions. Many of the features in the spectraa can be identified with thermal emission from forsterite dust grains, a crystalline silicatee also commonly found on earth (see Chapter 6). Att the formation time of the protostar, dust grains are already present in the parental cloudd from which the star forms. It is believed that the bulk of the dust in interstellar spacee is produced in the outflows of stars in the late phase of stellar evolution. Depend- ingg on the elemental C/O ratio, 'oxygen-rich' dust species, such as the silicate whose spectrumm is plotted in Fig. 1.3, or 'carbon-rich' dust species such as carbonaceous grains aree formed. The grains eventually mix with the material already present in the interstellarr medium (ISM). Within the ISM considerable grain processing takes place (see for instancee Tielens 1998; Henning 1999). Crystalline silicates as seen around evolved objectss (e.g. Molster 2000), are not seen in the ISM, indicating that these grains are either destroyedd or converted to amorphous grains in the ISM. The dust present at the early evolutionaryy phases of the protoplanetary disks will consequently consist of amorphous dustt of a mixed oxygen-carbon composition. Detailedd analysis of the ISO spectra of AB Aur, as presented in Chapter 5 show indeedd such a composition. However, as can be seen from Fig 1.3, the presence of crystallinee silicates in the ISO spectra of HAEBE stars show that considerable dust processing iss taking place in protoplanetary disks. Several chapters of this thesis deal with the compositionn of the dust in HAEBE star disks. We have found strong evidence for substantial changeschanges in the composition of the dust in the protoplanetary disks we studied, compared too the dust found in the ISM. The presence of metallic Fe, Si02, and FeO or FeS has beenn demonstrated for the first time. These components have not been found in the ISM,, and we discuss in Chapters 4 to 6 the chemical processing that may lead to the formationn of these materials. Inn Chapter 4 we analyse the silicate composition, both amorphous and crystalline, off our sample. We find that the presence of crystalline silicates is common in HAEBE starss (although not all stars show evidence for its presence) and we find evidence that itss presence correlates with the presence of SiO^. Such a correlation is expected for thermall annealing of amorphous silicates; however, this mechanism is only effective in the innerr regions of the disk where temperatures are high enough, and cannot explain the cooll crystalline silicates in HD 100546. As we will see in Chapter 6, the most likely sce- narioo for the formation of the crystalline silicates in this system is through the collisional destructionn of differentiated protoplanetary objects, i.e. objects in which under the influencee of gravitation the heavier elements like iron have settled towards the core, while thee mantle contains magnesium rich, iron poor minerals as forsterite. The collisions are mostt likely induced by the formation of a "proto-Jupiter", providing indirect evidence forr the formation of a planet in a HAEBE svstem. Thee change in the composition of the circumstellar dust we observe in HAF'BE systems,, from dust with properties similar to the dust present in the ISM, towards highly processedd dust, resembling cometary dust, shows the evolution of the dust from the ISM

7 7 CHAPII IR 1 towardss plancrarv systems. However, at which timescaie the processing occurs is unclear. Xoo correlation with stellar age and dust composition could be round, indicating that thee timescales at which star and circumstellar matter evolve are not strongly linked. 1 his suggestss that the evolution of circumstellar dust is most likely strongly influenced by systemm parameters such as disk mass and size, and not by stellar parameters. Wee should note, however, that the sample of HAT.BT. stars studied in this thesis spans aa relatively small range in stellar parameters. This makes the identification between the evolutionn of the central star and the protoplanetarv disk difficult. Natta et al. (2000a) showw that within a small interval in stellar mass, the derived disk masses can vary within 2 orderss of magnitude. If the disk mass has indeed an influence on the timescale at which thee processing of the circumstellar dust proceeds, large differences in dust processing betweenn HAEBK systems with similar central stars can be expected.

1.44 The geometry of the circumstellar disks

Ass mentioned above, circumstellar disks are expected to form during the formation of protostars.. However, the exact spatial distribution of the circumstellar matter has turned outt to be a problem intensely debated in literature. The problem in part arises because thee SKI) does not uniquely define the geometry of the circumstellar matter, lo analyse thee ISC) spectra we have used detailed dust radiative transfer models, which can handle bothh spherical and 2D {'disk like1) geometries. Combined with a detailed prescription off the dust opacities these models provide a powerful analysis method. An extensive discussionn of the models we have used in our analysis of the ISO spectra is presented in Chapterr 2. The SKI) of a HAHBK system can be reproduced equally well by a spherical distributionn of optically thin circumstellar material as used in Chapters 5 and 6 or, as presentedd in Chapter 7, by a distribution in a disk geometry. The best way to settle the argument,, of course, is bv direct imaging, which has provided evidence for the presence off circumstellar disks. Millimeter wave observations of Herbig Ae stars have convincingly demonstratedd that cold CO gas is in a flattened, rotating geometry, consistent with a Kepleriann disk, while the distribution of cold dust deviates from spherical symmetry (e.g.. Mannings and Sargent 199". 2000). Near-IR interferometric observations are less conclusive,, and are consistent with a more spherical distribution of the hot material near thee star (Millan-Cabet et al. 2001). Recent Hubble Space Telescope NICMOS images showw a flattened distribution of material in some HAKBK stars (Augereau et al. 2001). Thee evidence for disks in the more luminous Herbig Be stars is weaker. Thee disks in the early stages of protostellar evolution, when the accretion rate is high, aree well described by stationary accretion disk models as developed by e.g. I.ynden-Bell andd Pringle (1974) and Shakura and Sunvaev (1973). These accretion disks are mainly heatedd by viscous dissipation. Triedjung (1985), however, noted that in some cases the usee of a stationary accretion disk to explain the observed Shi) required an accretion rate whichh was too high to comply with observations. Triedjung (1985) argued that for lower accretionn rates, not viscous dissipation but the radiation held of the central star was the mainn source of heating. Typically, for HATBT systems, this latter process will dominate

8 8 FROMM DIST TO PI ANK IS

iff the accretion rate drops below ~ 10" MQvr~ , which is the case for the sample of HAEBEE systems studied in this thesis. Thee SEDs of T Tauri stars have been successfully modeled with passive (i.e. no accretion)) circumstellar disks which re-radiate (or reprocess) the UV and optical light of thee protostar into infrared and millimetre radiation (e.g. Kenyon and Hartmann 1987; Chiangg and Goldreich 1997). These models also have been applied to reproduce the SEDss of HAEBE stars (Natta et al. 2001; Chiang et al. 2001). However, as noted by thesee authors, these disk models are able to reproduce the SEDs of HAEBE stars from the mid-IRR to millimetre wavelengths, but fail to reproduce the observed near-lR emission. Ass demonstrated in Chapter 7, the geometry of disks in HAEBE systems deviates from thee standard disk model used to model the SEDs of T Tauri stars, in the sense that the scale-heightt of the inner disk regions is larger than that in T Tau disk models. Dullemond ett al. (2001) have constructed a hydrostatic equilibrium disk model which indeed shows aa large inner disk scale-height, and have shown that this is linked to the development of ann inner hole in the disk structure. This inner region empty of dust has been revealed by recentt near-IR interferometric observations (Millan-Gabet et al. 2001). Apartt from the spatial distribution of the dust in the inner parts of the disk, producingg the near-IR emission, the disk geometry at larger distances of the star is still uncertain.. When the surface of a passive reproducing disk, being isothermal in the directionn perpendicular to the midplane of the disk, can be irradiated directly by the central star,, the disk will flare, i.e. the opening angle of the disk will increase with radial distancee from the star. Such a geometry will produce a large IR excess with a maximum at ~600 jjm. In Chapter 3 we will argue that the SED of a sub-group of HAEBE systems iss well represented by such a disk geometry. However, for a number of sources the SED cann not be reproduced with this disk model. The SEDs of these latter systems can be modeledd assuming a 'flat' disk geometry. What the cause is of the difference in geometry betweenn the systems is still unclear. Bell et al. (1997) showed that shielding the surface of thee disk from stellar illumination can cause a collapse of a flaring disk into a flattened geometry.. Another possibility was suggested by Miyake and Nakagawa (1995). They point outt that the disk structure is determined by the gas pressure, but that the opacities and, consequently,, the thermal emission is dominated by the dust. When dust grains grow byy coagulation, they can de-couple from the gas and can settle towards the midplane by gravitation,, no longer supported by the gas pressure, producing a 'flat disk' SED. Noo obvious correlation between stellar age and disk geometry, i.e. flaring or flat, has beenn found. As noted in the previous section for the dust composition, this could reflect thee narrow range in stellar parameters within our sample of HAEBE stars. I his could, however,, also reflect the possibility that the disk geometry does not evolve monotonie withh time. As discussed in Chapter 4, three of the four systems studied that show a flaring diskk geometry have very processed dust. The fourth system, however, does not show- anyy significant processing compared to the ISM. I he systems with a flat disk geometry showw amounts of processing intermediate between the flared systems. It is tempting too speculate about a correlation between disk geometry and amount of processing. We suggestt that the following scenario might apply: (1) First one has a protoplanetary disk containingg relatively small grains, resembling ISM dust. Coagulation has not proceeded

9 9 CHAFFERR 1 farr and the dust is coupled to the gas. (2) Coagulation proceeds, larger grains de-couple fromm the gas, settle towards the midplane and produce a Hat disk SEI). (3) Protoplanets formm that gravitationallv stir the disk, inducing collisions between large cometarv sized objects,, as a result producing small grains. These small grains are again coupled to the gass producing again an SED consistent with a flaring disk geometry.

1.55 Conclusions & outlook

Wee can conclude that IR spectroscopy has given us valuable new insights in the compositionn and evolution of the circumstellar dust in HAEBE systems. Quantatitive spec- troscopyy in combination with imaging imposes strong constraints on the properties of protoplanetaryy disks and provides us with exciting links to the formation history of our ownn solar system. While the ISC) data, combined with other observations, have allowed uss to make considerable progress, many questions still remain unanswered. Below we brieflyy discuss some future developments, both in theory and observations, that will im- provee our very limited knowledge about the evolution of protoplanetary disks. Theoretically,, a more consistent modeling of the disk structure including gas-grain interactionss and grain growth is needed, as well as a treatment of radiation pressure on thee dust grains. This will provide better estimates of the change of disk structure with timee and it will provide a test whether flat disks are older than flaring disks, or that some otherr parameter (disk mass, radius, ...) determines its structure. I hese calculations need too be coupled to chemical nerwork calculations, taking into account gas-grain chemistry. Also,, a better treatment of dust scattering and thermal infrared emission is needed, which cann cope with grains of arbitrary size and shape. This will improve identification of solid statee species and improve estimates of chemical composition and mass. Observationally,, it is important ft) increase the sample for which high quality spectraa and images are available. Such enlarged samples will allow us to address questions relatedd to the timescale of dissipation of the disk and the conversion of dust from "ISM'1 too "proto-planetary" composition. We can study the effect of the environment on these processes:: it is important to realize that none of the stars studied in this thesis is in a clusterr or massive star forming region! Direct observation of the disk structure at all scaless will put strong constraints on disk models and on the way planet formation may affectt disk structure (e.g. the formation of disk gaps, the effects of planet migration and thee collisional evolution of disks after gas and dust have de-coupled). Fortunately, new observatoriess will become available in the coming years that will provide many of the observationss that are needed to answer these questions. Instruments like the Very Large TelescopeTelescope Interferometer (VIXI) and the Atacama Large Millimetre Array (ALMA) will providee us with observations with which the disk geometries can be determined; the suc- cessorr to ISC), NASAs Space InfraRed Telescope Facility (SIRTF), will greatly enlarge the samplee of stars for which high quality infrared spectra at intermediate spectral resolution willl be available. Thee work presented in this thesis is focused on the dust in protoplanetary disks. However,, it is important to realize that most of the mass in such disks in contained in

10 0 FROMM DUST TO PLANETS thee gas, which will consequently determine the disk dynamics. However, it has proven too be difficult to determine the gas mass and its spatial distribution. With instrumenrs likee Herschel and ALMA we will be able to study the gas with great precision. Inn summary, a bright future lies ahead for this exciting topic!

11 1

CHAPTERR 2

Dustt physics, radiative transfer and spectrall analysis

Inn this chapter, we will discuss details of the physical processes taken into account in ourr models of circumstellar dust environments. We will assume that the gas opacities are muchh smaller than the dust opacities and can be ignored, essentially, the physical pro- cessess tall into three categories. First, the temperatures that dust particles adopt. Second, thee spatial distribution of the dust. Third, the composition and grain structure of the dustt particles. Also,, we will discuss the extent to which observations can constrain models of circumstellarr dust. This is an important point as fitting of SEDs may give ambiguous resultss when trying to determine die properties and spatial distribution of circumstellar solidd state material.

2.11 Radiative transfer

Inn our models we assume that the flux of photons emitted by the central star can be absorbedd and re-emitted or scattered by solid state particles in the ambient medium. We assumee that in the dust medium the energy in the radiation field is conserved, i.e. no otherr sources or sinks of energy are present, such as viscous processes or shocks. The constraintt of radiative equilibrium determines the equilibrium temperature of the dust particles.. We start with introducing some basic equations of radiation transfer, leading too a discussion of the temperature determination of the dust. Inn arbitrary geometry the equation of transfer is given by

11 ?) d /v(r,, t, s) = Tiv(r, t, s) -itvfr, t, sj/v(r, t, s) (2.1) CC <)t OS wheree c denotes the speed of light and Iy(r,t,s) is the specific intensity of radiation att position r, traveling in direction S, with frequency V, at time t. The emission and extinctionn of energy by dust grains is given by the coefficients f|v and }(v, respectively. 11 he extinction coefficient can be split in two contributions, namely

Xv(r,/,s)) - Kv(r,r,s) + ov(r,r,s) (2.2) wheree Kv is the absorption coefficient and Gv is the scattering coefficient. We assume thatt the radiation field is time-independent and the thermal emissivitv and scattering are

13 3 CHAPTERR 2

isotropic.isotropic. Then Eq. (2.1) reduces to

*//v(r,s) ) == -Xv(r)[/v(r,s)-5v(r)] (2.3) ds ds

wheree we have introduced the generic source function

\.(r)) = Xv(r)) (2.4)

== [Kv(r)5v(r) + av(r)/v(r)]/[Kv(r) + av(r)].

Heree By is the Planck function and/v defines the mean intensity of radiation

/v(r)) = ^y"/v(r,sWO (2.5)

whichh is the specific intensity averaged over all solid angles. It is because of the assumptionn of isotropic thermal emission and scattering that the source function can be written inn the form of Eq. (2.4). The second equality in Eq.( 2.4) is only valid when the grain materiall can be represented by a unique grain temperature T. In general, this will not be thee case, as size, shape and composition of the particles may result in different absorption propertiess and consequently, different grain temperatures. In this more general case, the thermall emissivity is given by

Tiv(r)) = XKv./(r)flv(7Kr)) (2.6) wheree i labels the different grain components. Integrationn of Eq. (2.3) over all solid angles and all frequencies, and substitution of thee generic source function yields for each grain component

V-J(T)V-J(T) = 4K[ Kv(r)[ZMr(r))-/v(r)]

%{r)=%{r)=

(4TTT /0 Kv/V dv) equals the total amount of energy that is emitted in the same time interval,, from the same volume element (4TT \^' KyBvd\'). Note that the scattering extinction

14 4 RADIATIVEE TRANSFER MODELING doess not feature in this equation, which is to be expected as scatterings do not affect thee energy contents of a volume element in any way. A global representation of radiative equilibriumm can be derived from the total flux. Assuming e.g. spherical coordinates, such thatt the divergence of the total flux can be written as \/r^}){r'J)/'dr, one finds

Anr~Anr~ J{r) = constant = L (2.9) wheree L is the luminosity of the star.

2.1.11 An optically thin medium Too give an impression of what to expect for the temperatures of dust grains, we derive thee temperature structure in an optically thin medium. In this type of medium the intensityy incident on the grains is unattenuated star light, i.e. extinction processes can be neglected.. Let us assume the radiation incident on the grains is star light with a uniform intensityy /* = By{T*), i.e. represented by a Planck function at a stellar surface temperaturee T*. For the mean intensity at distance r from the star one then finds

Mr)=W{r)BMr)=W{r)Bww{l\){l\) (2.10) wheree W{r) is the geometrical dilution factor

mr)mr) = -i l-\ 1- - >, (2.11.

whichh in this case may be interpreted as being the solid angle subtended by the central star,, as seen from distance r, over the total solid angle. Simple physical reasoning (or 2 substitution)) yields W{Rif)= 1/2, while at distances r > /?*, W(r) = \/A{RJr) . Noww let us assume that the absorption cross section of a specific spherical dust grain iss given by

2 2 Clbs,vv = na Qahs,v = na Q0 I — J , (2.12)

wheree a is the radius of the grain and Qabs,v is the efficiency factor of absorption (see also

Sect.. 2.3.2). The absorption coefficient Kv is recovered by multiplying Eq. (2.12) with thee particle number density. Substitution of the above three formulae in Eq. (2.7) yields

T(r)T(r) = [W(r)]u^'%) 7; '^ ( M - 7„ (2.13) 4// V r

forr the temperature distribution of the grains. In case/? = 1, it follows that 7 (r) °= r~ '. Notee that for this description of Qjbw trie grain temperature is independent of grain size.. This need not to be so in a more general case. CHAPTHRR 2

Inn an optically thin medium, the actual spatial distribution of the dust is not relevant

inn so hir that only A/jUsl(;-)rfV, the total dust mass in the radial distance interval (t; r + dr), needss to be specified. The quantity that determines the dust emission is therefore the dis-

tributionn of mass over temperature, i.e. A/a.USI( 7>/7', and we will use this diagnostics in severall chapters of this thesis. The monochromatic flux at a distance d, can be expressed inn the mass over temperature distribution as

(2.14) ) wheree 7mm and 7m.lx denote the minimum and maximum temperatures that grains can assume,, X\[, the thermal emissivity per unit mass and K', = Kv/p the mass absorption coefficientt where p denotes the density of the medium. In Herbig Ae/Be (HAHBF) stars, thee maximum temperature of grains may be identified with the temperature at which theyy evaporate when coming too close to the central star. If this evaporation temperature iss denoted by /LV.ip, then the radius at which this occurs is given by

/'+-+ + ) ) '' cvap »» R* '' cvap >",v, = R*{\ - -i/: : PP i i !.15) )

AA typical upper limit for the temperature of refractory dust grains is 7'cvip = 1 500 K. For aa HAKBE star with T„ = 9000 K and R* = 2.5/?, this yields r,vap = 227?, or 0.5 AU. Insidee this region no dust can exist. In addition, chemical reactions can set limits on the maximumm or minimum temperature of grains (see Sect. 2.2.3). Ass discussed in Chapter 1, there is growing evidence for circumstellar disks in HAFBFF systems. However, the spectra of the HAFBF systems studied in this thesis show evidencee for a prominent optically thin dust component. Since the emission of such a componentt only depends on A/C|1IU(/'), we can use Fq. (2.14) independent of geometry too fit the spectra. Furthermore, part of the circumstellar dust in HAFBF systems may be distributedd in a more or less spherical shell. For instance a remnant low density halo may bee left from the Class I phase (cf. Fig 1.1). Also, hydrodynamical effects may create an innerr halo or 'puffed up' inner region of which the density mav roughly be represented byy a spherical distribution of material. For the systems discussed in this thesis, spheri- calcal optically thick dust distributions can be ruled out by observations. For instance, the centrall stars in these systems are bright and hardly reddened. For the studies presented inn this thesis a full radiative transfer calculation for a spherical geometry is therefore not required.. Fot that reason, we will not discuss computational methods used to solve the radiativee transfer equation fora spherical geometry.

2.1.22 Radiative transfer in axisymmetric geometry

Ass pointed out previously, one expects the dust in HAFBF systems to be distributed in a disk-likee geometry. If we assume translational symmetry in the ^-direction of a Cartesian

16 6 RADIATIVHH TRANSFER MODELING coordinatee system, transfer equation (2.3) can be written as

«== W(«=.ïv|r)[Wr,8).Wr)], (2.16) asas ox ay wheree s = (sx,sr) and r = (x,y). This equation can be solved by constructing sets or rays, covering;; all directions, along which formal solutions (i.e. with given source function) aree performed that end in all points r; = (.v„y,) defined within the Cartesian system, for whichh the specific intensity is wanted. If the grid points r, are chosen sensibly and if averagingg procedures are employed, one needs only to solve for transfer along portions off these long rays as one 'sweeps' through the grid. The method of solving the integral onlyy along small portions of the ray is generally referred to as Short Characteristics. Inn Chapter 7, we will use a radiative transfer model based on a Short Characteristics methodd defined in spherical coordinates (Dullcmond and Turolla 2000), to study the diskss of HAEBE systems. Whatt to expect for the temperature structure of dust distributed in a disk? To give an impression,, let us assume a disk which is optically thick to the stellar radiation (i.e. at ÜV andd visible wavelengths), and geometrically flat. In this case, unattenuated stellar light strikingg the disk at an angle 0 is fully absorbed by the dust. Because of this angle, the absorbedd energy at r is only sin 0 times the energy absorbed in an optically thin medium. Forr sufficiently large distances from the central star, it holds that sin0 ~ RJr. In a derivationn very similar to the one discussed in section 2.1.1, we find for the temperature distribution n T-M-W^ïï (2.17)

whichh is simply the optically thin result, multiplied by a factor RJr. In case/? = 1, this reducess to 7 (r) <* T~il(\ This implies that in an optically thick and geometrically thin disk,, the temperature shows a faster decline with radial distance than in an optically thinn medium. The flat disk assumption is an oversimplification for the actual disk structuree of HAEBE systems. More complicated disk geometries, based on hydrostatic and radiativee equilibrium, show that the disks are expected to be flared, i.e. that the angle 00 increases with increasing distance from the central star (Chiang & Goldreich 1997; Dullemond,, Dominik & Natta 2001). A simplified treatment of radiative transfer in thesee disks defines two regimes: first, an optically thin surface layer where dust is irradiatedd by direct stellar light; and second, an optically thick interior in which dust only receivess processed photons emitted from the surface layer. Because of the flaring of the disk,, the dust temperature profile in the surface layer is in between that of an optically thinn medium and an optically thick flat disk. Still, the dust in the surface layer is heated too higher temperatures relative to dust in the interior. Inn Chapter 5 we will show that if we assume the dust medium to be optically thin, onee typically finds that two geometrically separated dust regimes are needed to explain thee spectral energy distribution. An inner regime containing relatively hot grains, and a moree distant regime containing relatively cool grains. One cannot exclude on the basis

17 7 CHAPTKRR 2

of-- SED fitting alone that such a bi-modal temperature distribution is the result of a physicall gap in the disk of HAKRE systems due to the clearing out of dust by a proto- planet.. However, in view of the discussion above, it mav be more likelv that it results fromm the presence of a hot surface layer and a cool disk interior occupying essentially the samee radial /one.

2.22 Chemical composition and grain structure of circumstellarr dust

Ass briefly mentioned in Chapter 1, dust grains are formed in the stellar winds of stars in thee late phases of stellar evolution. Depending on the atomic C/O ratio, either oxygen richh dust species like silicates, both crystalline and amorphous, are formed (e.g OH/IR starss Sylvester et al. 1999), or dust species typical for a carbon rich environment, such ass amorphous carbon or graphite, or magnesium sulfides (e.g carbon rich ACB stars Begemannn et al. 1994). The dust species formed in the stellar wind will be injected into thee interstellar medium (ISM), where they will mix with dust grains already present. Withinn the ISM substantial processing takes place (e.g. Henning 1999; 'Helens 1998). 11 his processing leads to the formation of dust grains of a mixed carbon and oxygen chemistry.. Since no crystalline silicates are observed in the ISM, the processing must alsoo lead to the amorphisation of the crystalline silicates initially formed in outflows. It iss this dust composition that is present at the onset of star formation. At later phases, substantiall dust processing takes place in the disks surrounding the protostars. Using the spectrographss onboard the Infrared Space Obsevatory (ISO; Kessler et al. 1996) satellite, enormouss progress has been made in determining both the dust composition as well as thee processes that govern this composition in pre-main-sequence systems. By comparing thee ISO spectra with laboratory spectra of dust material (e.g. see Fig. 1.3), the individuall dust species contributing to the ISO spectra can be identified. Dust species such as silicates,, both crystalline as well as amorphous, metallic iron, iron oxide, carbonaceous grainss and water ice, have been identified in the ISO spectra of Herbig Ae/Be (HAEBE) systemss (e.g. van den Ancker et al. 2000; Bouwman et al. 2000a; Malfait et al. 1999, 1998b).. In the following we will discuss the properties and chemistrv of the species presentt in the protoplanetary disks around HAEBE stars.

2.2.11 Silicates

11 he most abundant rock forming dust species are silicates, a class of materials also com- monlyy found on Earth (i.e. 90 % of the Earths crust consists of silicates). The fundamen- tall unit on which the structure of all silicates is based consists of four 02~ ions forming aa regular tetrahedron, with at its center one Si'^ ion. The bond between the oxygen and siliconn ions is about 50 % ionic and 50% covalent. So, although the bond arises in part fromm the attraction of oppositely charged ions, it also involves sharing of electrons and inter-penetrationn of the electronic clouds of the ions involved. Each 0:~ ion can poten- tiallyy bond to another silicon ion and enter into another tetrahedral sjroupine In this

18 8 RAHIATIYKK TRANSFER MODELING

Single e tetrahedral l

unitt

Tetrahedral l Single e network k tetrahedral l (SiOo)0 0 ,, 4 chainn ISioO/J

Figuree 2.1: Schematic representation of the lattice structure of silicates (after Klein and Hurlbut 1993).. Shown are the single (SiOzj)4- tetrahedral unit as in olivine (upper left panel), the single chainn structure with two bridging oxygens atoms as in pyroxene (right panel), and the framework inn which all oxygens atoms are shared as in silica (lower left panel). wayy oxygen is used to bridge different tetrahedral groups. The silicates can be divided intoo several families based on their chemical composition and structure. We will discuss thee most important silicates, which are the ones used in the modeling presented in this thesis. .

Olivine e

Iff no oxygen ions are shared, implying that the silicate is made up of independent (SiOzj)4"" groups, the silicate is called a nesosilicate. A schematic representation of this structuree is shown in Fig. 2.1. A member of this class of silicates is olivine. The tetra- hedrall units in olivine are bound to each other only by ionic bonds from interstitial (i.e.. in the spaces between the tetrahedral units) cations (Fe + and Mg"+), forming an orthorhombicc crystal. The chemical composition of olivine is given by Mg2xFe2_2*Si04, wheree x is between 1 and 0. Olivine forms a complete solid solution series from forsterite

(Mg2Si04)) to fayalite (Fe2Si04).

19 9 CHAPTERR 2

Pyroxene e

Iff two of the four oxygen ions in the SiO.( tetrahedron are shared, the silicate is called ann inosilicate. The tetrahedral units in such a silicate form a chain, resulting in a bulkk atomic ratio Si:0 of 1:3. This structure is also plotted in Fig. 2.1. A member of thiss class of silicates is pyroxene. The chemical composition of pyroxene is given bv Mglv^-lv^Od'' where .v is between 1 and 0. Pyroxene forms a solid solution series

fromm enstatite Mg2-Si:06 to ferrosilite FcvSnOf,. The crystal structure of these silicates iss orthorhombic. I he iron and magnesium cations can be substituted bv other ions like Ca-\\ forming a mineral like diopside (CaMgSnOf,), which has a monoclinic lattice structure. .

Silica a

Anotherr important class of silicates, which for instance make up for about 64 % of the Earthss rocky crust, are minerals built about a three dimentional frame work of linked SiO.-ii tetrahedra. These minerals belong to the tectosilicate class in which all the oxygen ionss in each tetrahedron are shared with neighboring tetrahedra. This results in a bulk Si:QQ atomic ratio of 1:2. A schematic representation of this structure can be seen in

Pig.. 2.1. A member of the tectosilicate class is silica, consisting only of SiC)4 tetrahedra, withh the bulk chemical composition given by SiO^. Silica has nine polymorphs among whichh are quartz, tridimite and cristobalite, differing in crystal structure. While of these silicatess quart/, is the most common form found on Earth, the common form found in interplanetaryy dust particles (IDPs) is tridimite (Rietmeijer 1988).

Amorphouss silicate

Thee minerals discussed above were ordered by their lattice structure. However, apart fromm forming crystals, silicates can also be amorphous, i.e. have neither lattice order norr orientational order. Such materials are glasses or smokes. As will be discussed in the followingg chapters in this thesis, the bulk of the silicate dust seen in space is in this form.

Inn an amorphous silicate, such as a glass, the basic unit is still the Si04 tetrahedron, but eachh such tetrahedron has a different number of brids;in" oxygen ions randomly linked too other tetrahedra. The resulting distribution of bridging oxygen ions per tetrahedral unitt is peaked around the average bulk crystalline value (Parnan et al. 1992).

2.2.22 Other dust species

II hough the silicates make up for most of the dust mass in HAEBP! systems, several other dustt species have been identified, and the}- can dominate the ISO spectra within specific- frequencyy intervals.

20 0 RADIATIYHH TRANNFHR MODF.UNC;

Carbonaceouss grains

Ass mentioned previously, the dust seen in HAEBE systems is a mixture of oxygen rich andd carbon rich materials. The ISO spectra show emission bands which can be attributed too polycyclic aromatic hydrocarbons (PAHs; see for instance Chapter 3). These are large moleculess with an aromatic structure transiently heated by the central star. As will be discussedd in Chapter 4, evidence for the presence of aliphatic hydrocarbons can also be found.. These molecules are, contrary to the PAHs, most likely incorporated in carbonaceouss grains. Modeling of these molecules is beyond the scope or this thesis. We will, however,, include carbonaceous grains in the dust mixture used in our models. A form off carbonaceous material is graphite. In this material the carbon atoms are arranged in layerss with a covalent bonding, while the layers themselves are held together by van der Waalss bonds. This type of material is a high temperature condensate. If carbonaceous materiall has been formed within the cold ISM or at lower temperatures in outflows, thee carbonaceous material will be amorphous in nature, i.e. will not have such layered structure. .

Iron n

Anotherr spectroscopicaiiy important dust species is iron or iron containing maiciial. As wee wil! see in following chapters, these dust grains can dominate the near-IR emission seenn in HAEBE systems. Apart from being incorporated in silicates, iron can also be in thee form of metallic iron, iron oxide, or iron sulfide. Both iron oxide and iron sulfide havee been used to identify an emission band around ^23/Jm. In this thesis we will use FeOO to explain this band. However, the possibility that this emission is due to iron sulfide cannott be ruled out.

Waterr ice

Att temperatures below 1 50 K, water will be in the form of ice. The water molecules in thee ice are bonded to each other by hydrogen bonds. In total nine polymorphs of ice aree known. Water ice can be both crystalline as well as amorphous. The ice crystals can bee cubic or hexagonal. Spectroscopicaiiy, however, there is little difference between the differentt crystalline forms.

2.2.33 Dust chemistry

Thee grain species discussed in the previous section do not exist independently but are partt of a complex chemical network, in which dust species are transformed into one another.. The reactions depend on temperature and to a lesser extent on pressure. 'I hough wee do not include a detailed chemical network in our models we do include a temperaturee dependencv of the dust composition. Important processes that are determined by thee dust temperature are crystallisation and grain evaporation by either heating the grain abovee the the thermal stability limit or by chemical sputtering. Crystallisation occurs at

21 1 CHAPTERR 2

sufficientlyy high temperatures, when the atoms in the amorphous material start to re- arrangee themselves into energetically more favorable positions and orientations within thee lattice structure. Silicate dust grains will crystallise at temperatures above 1100 K at timescaless much shorter than any dynamical timescale (e.g. Hallenbeck et al. 2000; Gail 1998).. In our calculations we will therefore adopt this temperature as the maximum temperaturee of the amorphous silicate dust. The evaporation temperature of the crystalline silicatess is higher than the crystallisation temperature, and depends on chemical compositionn and lattice structure. The most stable silicates, the pure magnesium end members off the olivine and pyroxene silicate family, forsterite and enstatite respectively, will evapo- ratee at temperatures of ~ 1400 K. We will adopt this temperature as the maximum grain temperaturee of the silicates. Ass we will see in Chapter 4, when we discuss the crystallisation process of the silicate forsterite,, if the bulk composition of the amorphous silicate differs from the crystalline silicate,, also silica will be formed. Assuming an amorphous silicate with a serpentine compositionn (Mg3Si20) forsterite and silica will be formed by the reaction

2Mg3Si2077 -> 3Mg,Si04 + Si(32. (2.18)

Forsteritee and silica form a meta-stable system. Depending on the amount of reaction surfacee between both minerals, they will form enstatite by the reaction

Mg2Si044 + Si02 -» MgSi03. (2.19)

Olivinee and pyroxene can incorporate iron in their lattice structure. However, at highh temperatures of ~800 K in a reducing environment (i.e. low oxygen pressure), iron willl be removed from the lattice. For olivine dust grains this recation is

2MgFeSi044 + 2H2 -> Mg2Si04 + 2Fe + SiO, + 2H20. (2.20)

Thiss process will form the pure magnesium end member forsterite as well as silica and metallicc iron grains (e.g. AJlen et al. 1 993). The evaporation temperature of metallic iron iss comparable with that of the silicate dust. We have adopted in our models a maximum grainn temperature for the metallic iron grains of 1 500 K. At temperatures below 720 K metallicc iron can react with sulphur containing molecules to form iron sulfide

Fe+H2SS ^ FeS+H2; (2.21) orr oxidize below 400 K to form iron oxide.

Fee + H2O^FeO + H2. (2.22)

Inn principle, carbonaceous grains may be heated to ~1500K before they will ther- mallyy decompose. However, Finocchi et al. (1997) point out that at temperatures above 1000KK the carbonaceous grains will be destroyed by chemical surface reactions. This

22 2 RADIATIVEE TRANSFER MODELING wouldd also imply that amorphous carbon grains will not reach sufficiently high temperaturess to form graphite, but will be destroyed before that can happen. We will therefore adoptt 1000 K as the maximum grain temperature for the carbonaceous grains. Givenn the gas pressure conditions in the circumstellar disks surrounding the HAEBE stars,, water ice will vaporize at temperatures in excess of 1 50 K. The water ice will be crystallinee if formed at temperatures above ~12() K. If formed below this temperature, itt will be amorphous.

2.2.44 The grain structure

Apartt from dust processes discussed above, the dust grains in the protoplanetarv disks are alsoo subject to coagulation and shattering by grain-grain collisions (e.g. Weidenschilling 1997).. Grain growth can proceed either directly from the gas phase or bv aggregation off smaller grains. This latter process will lead to "fluffy" chemical inhomogeneous grains resemblingg IDPs. Also the processes described in the previous section can lead to chemi- callyy inhomogeneous grains. Incomplete crystallisation of an amorphous grain will obvi- ouslyy lead to a grain that is partially crystalline and partially amorphous. Crystallisation reactionss such as Eq. (2.18) will also lead to inhomogeneous grains. Furthermore, water willl most likely freeze out onto the grain surface at temperatures below 150K, leading too a core mantle structure. Itt is, however, debatable whether inhomogeneities in grain structure can be identi- fiedd spectroscopically. If for instance the dust grains in the protoplanetarv disks around HAEBEE stars bear resemblance to cometary grains this could well be the case. Theo- reticall considerations suggest cometary grains to be extremely fluffy, implying that the spectroscopicc properties of the grains will be dominated by the smaller compact units makingg up the porous aggregate (Greenberg and Hage 1990; Brucato et ah 1999). The largerr aggregate would spectroscopically behave as a cloud of smaller compact grains. Also,, if the timescales on which crystallisation proceeds are sufficiently short, any phase inn which the grain is inhomogeneous and unequilibrated might not be observable. Sim- ilarly,, if ice condensation is an efficient process, leading to an ice mantel much larger thann the core, it could effectively shield the core and consequently the particle will behavee spectroscopically as a homogeneous grain. Inn our modeling we will therefore treat the dust grains in a first order approximation ass single, chemically homogeneous and compact. The effect of grain coagulation and/or shatteringg is incorporated in the models by assuming a power-law grain size distribution. Thee number density of grains with (volume equivalent) radii between a and a + da is givenn by

n{a)dan{a)da = A { — I da, (2.23) '' m i n wheree the grain radius a is limited between a minimum and maximum value, am-m and tfmax'tfmax' respectively. The normalisation constant A is determined bv

4 "" -TLii-TLii i p&,n{a)da (2.24) 3 3 Ol.M'ïii R 2

wheree p is the density of the medium due to the dust species and pju the bulk density of tliee dust grains.

2.33 The optical properties of dust grains

II he absorption, scattering and thermal emission properties of a dust grain are determinedd hv its chemical composition, lattice structure, shape and size. Once these strain propertiess are known, it is in principle possible to calculate the interaction of an incom- ingg electromagnetic wave with the grain. Depending on its wavelength, radiation will bee able to interact with the electrons, or it will excite vibrations in the material. F.lec- tronicc transitions are mostly found at ultraviolet and optical wavelengths: conductors andd isolators show different behaviour depending on the availability of free electrons. Vibrationall modes are found in the mid- and far-infrared. 1 his wavelength range is particularlyy rich in strong vibranon.il resonances of abundant molecular bonds in solids, suchh as the Ni-O fundamental vibrational band at about I()jt/m, seen in amorphous silicates.. 1 he lattice structure strongly influences the vibrational resonances that can be dis- tinguishedd obscrwuionallv. lor instance, in crystalline silicates the ordered lattice structuree causes well-defined wavelengths of resonances and thus a multitude of sharp bands, whilee in amorphous silicates manv of these bands merge into two broad resonances at ~~ 1 0 and ~ 1 8 /mi, while other bands disappear all together. Whetherr these resonances actually show tip in the infrared spectrum (either as emissionn or absorption bands), depends on the grain size: if the grain dimensions are large comparedd to the wavelength of the resonance, phonons generated in the solid will be scatteredd inside the particle, resulting in black body emission of the grain. This implies thatt we cannot determine the chemical composition of large (typically > ]()ji/m) grains. Inn our analyses, we have assumed that the grain composition of large grains is similar to thatt found for the small gram population. Inn addition, the wavelength of some resonances in materials such as be() and MgO, aree vcrv sensitive to grain shape. 1 his occurs because of the generation of surface phonon modess whose wavelength depends on grain shape. I he fact that bands shift due to grain shapee effects complicates the identification of these materials in astronomical spectra. II heretore, it is important to characterize the grain population not onlv in terms of compositionn and lattice structure, but also in terms of grain size and shape (see also Sect. 2.2.4 forr the effects of grain aggregates on optical properties).

2.3.11 The refractive index

II he optical properties of a material can be characterized by the complex refractive index

\\\ - nL^n"t: (2.25) wheree ;/ and //" are the real and imaginary part respectively, also referred to as the optical constants,, and A is the wavelength. A plane wave traveling in a direction c through a

24 4 RADIATIVEE TRANSFER MODELING

Figuree 2.2: Schematic representation or scattering of light bv a single particle. mediumm which has a refractive index N, has the form

2Kn"z 2Kn"z Hun'Hun' k EE - Eoexp (2.26) ) X X exp p i(üt i(üt wheree CO is the angular frequency. As one can immediatly see from Eq. (2.26), the imag- inaryy part of the refractive index, n", determines the exponential decay of the wave as it movess through the medium. This directly links n" to the absorption coefficient of the material.. The real part of the refractive index, «', determines the phase velocity v = c/n', wheree c is the speed of light. Thee refractive index can be linked to the microscopic properties of a material by treatingg the ions, electrons and lattice as simple harmonic oscillators (e.g. see Chapter 99 of Bohren and Huffman 1983). From laboratory measurements of the refractive in- dexx for a specific material, the extinction behaviour of dust grains can, in principle, be calculated. .

2.3.22 Basic scattering theory

Inn this section we will discuss some basic concepts of light scattering. For details we refer too van de Hulst (1957) and Bohren and Huffman (1983). Consider a single particle of arbitraryy size and shape, illuminated by a plane harmonic wave traveling in a direction z. AA schematic representation of the situation is given in Fig. 2.2. The incident plane wave cann be written as i(kz-utt) i(kz-utt) E,-== (Ei)lë\\1 + t] ë^!)e (2.27) ) wheree k is the wave number in the medium surrounding the particle, and ê,- and ê^, aree basis vectors parallel and perpendicular to the scattering plane, i.e. the plane containingg the directions of the incident light and the light scattered in a particular direction. Firstt we assume that the particle is spherical, i.e. a homogeneous isotropic sphere. The scatteredd field at a distance r of the particle in a direction making an angle d with the

25 5 CHAPTERR 2 directionn of the incident light can he represented bv

/(kr-Wt) ) E,, = S(i})—-—. (2.28) —ikr —ikr Heree S(i3) is a 2x2 complex matrix, called the amplitude scattering matrix, which dependss on the scattering angle i3, and on the size as well as the refractive index of the sphericall particle. Eqs. (2.27) and (2.28) can be combined to give the relation between thee incident and scattered fields. The result is

EE]I]I\_<^(SM\_<^(SM 0 \(FV\ -ik,- v ° sm) [-lJ> wheree .V,- are the elements of the amplitude scattering matrix. For the forward direction

(dd = 0) wchave51(0)=52(0). Thee extinction cross section of the particle is given by

4TT T Cxtt = 7TM^i(0)} (2.30J andd its scattering cross section by

2 2 Q-aa = ^p{\Sm\ + |52(Ö)| }sinö^. (2.31)

Thee absorption cross section can be calculated from

Q,ss = Cext-Csca. (2.32)

Thee extinction efficiency factor is defined as

0™™ = ^. (2-33)

wheree a is the radius of the spherical particle. The scattering efficiency factor QsCa and absorptionn efficiency factor, Qa[1s, are defined analogously. For scattering by a collection off spherical particles one can add the cross sections of the individual spheres. Thee exact solution of Maxwell's equations for homogeneous isotropic spheres is re- ferredd to as MIE theory after Gustav Mie (1908). Using this theory we can compute thee scattering and absorption properties of spherical particles for all values of refractive index,, size and wavelength needed for the studies described in this thesis. Computationss and experiments to obtain the scattering and absorption properties off non-spherical particles are, in general, much more difficult than for spherical particless (e.g. Mishchenko et al. 2000). A notable exception is, however, Rayleigh scattering, whichh is a good approximation for particles whose size is small compared to the wavelengthh inside and outside the particle. To estimate the effects of non-spherical grain

26 6 RADIATIVHH TRANSI KR MODFI i\c;

shapess on the absorption cross section we will use a statistical approximation in the Rayleighh domain. This is based on the assumption that cross sections of collections of randomlyy oriented irregular particles can be approximately calculated bv averaging over aa uniform distribution of ellipsoidal shapes of a collection of ellipsoids with random orientation.. I his method is usually referred to as continuous distribution of ellipsoids (CL)E).. If the ellipsoids have the same volume and a is the radius of an equal-volume sphere,, we find for the absorption cross section in the Ravleigh domain (see Bohren and Huffmann 1983, Section 12.2)

8JTVV , , Q,ss = ^-/w{a} (2.34)

whei i NN1 1 aa = 2— Lo«/V--2 (2.35) NN::-\-\ * iss the polarizability per unit particle volume, in which I.og/V: denotes the principal value off the natural logarithm of the complex number N1.

2.3.33 Extrapolation of the refractive index to short and long wavelengths

Ideallyy one would like to have the measured values of the refractive index for the entire wavelengthh range, i.e. from UV to far-IR, required for radiative transfer calculations. However,, often the laboratory measurements do not cover this entire wavelength interval.. By using the results from the oscillator model for the refractive index outside of the measuredd wavelength ranges we can make the following extrapolations. At wavelengths shorterr than those at which the electronic transitions occur, i.e. in the far-UV, the refrac- tivee index given by Eq. (2.25) can be extrapolated far from resonances bv using

nn ~ 1 - —L (2.36) 2(0--

ww ^ ~r, (2.37) 22 CO1

withh 0) the angular frequency, 0)p the plasma frequency, and ya damping factor of the electromagneticc wave (see e.g. Bohren and Huffman 1983, Sect. 9.1.2). Both insulators andd metals show the same behaviour at these wavelengths. These extrapolations are reasonablee for wavelengths shorter than ~0.2jtvm and can thus be used to compute the cross sections.. At long wavelengths (sub-millimetre and millimetre) there is a marked differencee in optical properties between grains of different solid-state structure, composition andd shape. In this regime, we extrapolate the cross section (see also Sect. 2.3.2),

1 1 Q-xt/^—— (2.38) ^^ L- - f u wheree w is between 1 and 2 for small grains, depending on lattice structure and grain shape. . CHAl'THRR 2

2.44 The spatial distribution of the dust

Thee exact geometry of the circumstellar dust in HAEBK systems is still a matter or controversy.. The emission features seen in the ISO spectra, however, suggest that a substantiall optically thin medium must be present. As discussed in previous sections, this promptedd us to use an optically thin, spherically symmetric model. The dust grains are distributedd in a single shell or in multiple shells. The density structure within each shell iss given by WW *>(£)". (2.39) wheree r is the radial coordinate measured from the central star and limited between the innerr radius R-in and outer radius Roul of the dust shell, and po the density at the inner radius. . Iff the medium is optically thin, both disk as well as spherical geometries, provided theyy have the same radial density distribution, will produce identical thermal emission spectra.. As discussed in Sect. 2.1.1, a more useful diagnostic than the actual spatial distributionn is therefore the mass-temperature distribution, A/just('/'), which determines the resultingg dust emission. The derived M,\lM(T) is actually a quite robust result that is expectedd to reproduce well the actual mass over temperature distribution in disks that aree optically thick in radial directions through the disk. When seen at an inclination for whichh along the line-of-sight the medium is optically thin, the derived A/a.UM(7T still providess the correct result. In a detailed study of HD104237 (Chapter 7) we indeed foundd that at almost all wavelengths the disk is indeed optically thin for the inclination att which the system is observed. However, even if A/dllst( T) is identical, the actual spatial masss distribution, A/ju,t0"), of a radial optically thick disk can deviate significantly from thee optically thin model. To investigate the effect of a disk like geometry on the model- ingg results compared to spherical models, we will use a simple parameterised disk model. Inn this model the density in polar coordinates is defined as follows:

p(r,e)) = p(r)p(B) (2.40) wheree the functions p(r) and p(9) are defined by

P.„„ = p„(£)"

p(9)) = .v+(l-.v)cos'/(9). (2.42)

Heree 0 is the angle between r and the mid-plane, .v = p7/p« is the ratio of the density at

thee poles, p/1( and the density in the mid-plane, p,, and p„ = p(#m,0) is the mid-plane densityy at the inner radius of the disk. An example of this density structure is plotted in RADIAÏIVF.. TRAXSH-R MOnt-I.INC".

2.55 The strategy of modeling

11 he goal of modeling HAEBE systems is to recover the properties of the circumstellar dust.. The main diagnostics that one has available, at least in principle, are the infrared spectrall energy distribution as well as ground-based near- cv mid-IR and millimetre imaging.. The HAEBE systems studied in this thesis are relatively near-by and have been welll studied from the ground as well as with the ISO satellite. For a sample of fourteen HAEBEE stars we have obtained IR-spectra in the range from 2-200/.im using the Short WavelengthWavelength Spectrometer (SWS; de Graauw et al. 1996) and Long Wavelength Spectrometer (LWS;; Clegg et al. 1996) instruments onboard the ISO satellite. Several attempts have beenn made to image the HAEBE stars in the near-IR. However, these data are of limited usee as the characteristic emitting region of the hot dust responsible for the near-IR flux iss limited to a few AU from the central star. Given a typical distance of 150pc for our sett of HAEBE stars, this implies that the near-IR radiation originates from within the orderr of a few 0.01 arcsec from the central star, which is below the angular resolution limitt of direct imaging of '--0.5 arcsec. At this time, the only way to resolve the inner diskk structure, where the hot dust is located, is using interferometric instruments. How- ever,, near-IR observations of this type only use a very small number of baselines (two) andd provide no phase information, which does not allow to reconstruct a detailed image. Still,, these observations are useful as they provide constraints on the location of the inner edgee of the circumstellar dust disk by using the visibility data (Millan-Gabet et al. 2001). Imagingg at longer wavelengths, i.e. in the far-IR, millimetre and radio regime suffers lesss from the angular resolution problems that exist in the near-IR. First, this is because thee flux at these wavelengths is emitted by cold dust located at relatively large distances fromm the star, i.e. up to order 100 AU. Second, the interferometric instruments operating att these wavelengths have a larger number of independent baselines and provide phase informationn which makes the reconstruction of an image possible. These two properties combinedd allow, at least in principle, to image the spatial distribution of the cold dust. Indeed,, several of the sources studied in this thesis have been resolved at millimeter wavelengthss (e.g. Mannings and Sargent 1997). Thee above discussion shows that at least the modeling of the inner disk relies almost solelysolely on spectral fitting. The great challenge is to constrain as many dust properties as possiblee from this fitting. These include the chemical composition and lattice structure off the grain material, the size and shape of the particles, and their spatial distribution. Becausee of the many ways in which these dust properties can affect the infrared energy distribution,, it is found that the geometry and properties of the material cannot be determinedd uniquely, if only the spectrum is available. Too some extent this problem is independent of the optical thickness of the material.. Here, we will illustrate this ambiguity using predictions based on the assumption off an optically thin medium. Four types of degeneracies mav readily be identified. I he firstfirst three relate to the degeneracy in the mass-temperature distribution (cf. Sect. 2.1.1). Moree specific, for a fixed grain composition, the shape and strength of the SEL) is determinedd by an average mass-temperature profile. I he grain temperature depends on the radiall distance to the star and on grain size, where the smallest grains have the high-

29 9 CHAPTERR 2

1 1 3C C a b . ) ) . )) - I!! 1

.. : : _ _ _ _ 520.. n

\\\\ _ . _ 1 : ~ 1 00/_i] 1! [Ii 1.5 20 0 " " --

10 0 11yy / ^\ v,- -- _J J -- J J " "

100 20 30 10 20 30 AA [„,„]

Figuree 2.3: Model spectra of amorphous olivine (Dorschner et al. 1995). The solid line in bothh panels indicates a model with a power-law grain size distribution with power m = -2.8 (see Eq.. 2.23) and spherical grains with radii between 0.01 and 5.0 jjm. The density distribution is proportionall to r~ and an inner and outer bound of 1 60 and 1000 stellar radii, respectively, has beenn adopted. Panel a: Variations of the grain-size distribution. Panel b: Variation of the inner- andd outer boundary of the dust region.

estt temperatures. This implies that the mass-temperature profile will be determined by averagee grain size and spatial distributions. i)i) Different grain size distributions with the same average temperature will produce identicall spectra if the density (gr cm-'') is appropriately scaled. Fig. 2.3a shows this degeneracy,, using a power-law grain size distribution for the emission of amorphous sphericall olivine grains. The test indicates that in this way minimum and maximum grainn sizes may be varied by about an order of magnitude! ii)ii) One may also play this game using the inner- and outer boundary of the dust region.. Fig. 2.3b shows an example in which the size distribution is kept invariant, but wheree the geometrical extent of the dust region is modified. This shows it is also possible too change the dimension of the dust region by almost an order of magnitude without modifyingg the flux distribution. Hi)Hi) Similar average mass over temperature profiles may also be reached by placing smalll grains further out or large grains closer to the central star. This ambiguity, however, onlyy plays a róle in certain regimes of the grain size. For instance, in the case of metallicc iron (Fig. 2.4a) a change in grain size of over two orders of magnitude is possible

30 0 RADIATIVEE TRANSFER MODELING

. .

\\ [fj.ni]

Figuree 2.4: Model spectra of metallic iron (Henning et al. 1996) (left panel) and amorphous olivinee (right panel). In the panels two models with similar average mass over temperature profiles aree compared with small grains further out and large grains closer to the central star. Indicated in thee figures are the radii of the spherical grains and radial extent of the dust region.

withoutt significantly changing the emergent NIR flux distribution. For a large part this samee result can be obtained for amorphous olivine grains (Fig. 2.4b), although when a certainn maximum grain size is reached (~ 1/jm) the feature amplitude suddenly changes dramaticallyy even after tuning of the extent of the dust region. iv)iv) A further problem arises when no clear spectral signature, such as the 9.7 /vm silicatee feature, is present but only a broad continuum. The possibility of confusion off the relative contribution of the individual dust species then exists. Fig. 2.5 shows thee continuum dust emission from dust species other than silicates for three different models.. The predicted NIR fluxes are almost indistinguishable, although the adopted chemicall compositions vary significantly. Itt may seem that the above test results sketch a somewhat gloomy picture for our chancess of improving our insight in the properties of the ambient medium or HAEBE starss if only spectra are available. Fortunately, the situation is not as bad as it may seem. Somee degeneracies in the modeling may be resolved using physical constraints such ass dust destruction temperatures. For example, the test case shown in Fig. 2.5 in which ann identical near-IR flux is produced by three distinctly different chemical compositions mayy be resolved using 7"evap (see also Sect. 2.2.3). The temperatures required to emit efficientlyy in the 1 to 4 /vm region are of the order of 1 500 K. As argued in Sect. 2.2.3,

31 1 CHAPTERR 2

ii > i i i , \ i \ i 22 4 6 8

Figuree 2.5: Continuum dust emission from dust species contributing to the NIR between 1 andd 9 ;i/m for three different models. The solid line indicates a model consisting of amorphous carbonn (Preibisch et al. 1993), metallic iron and iron oxide, with mass fractions of 0.55, 0.2 and 0.255 respectively. The dashed line indicates a model with only amorphous carbon and metallic ironn with mass fractions of 0.53 and 0.47. The dotted line indicates a model where the entire NIRR emission is due to graphite (Laor et al. 1993). thee only stable dust species at such high temperatures is metallic iron. This leaves only thee second model (dashed line) as a physically realistic solution, where the emission at wavelengthss longer than ~5 jura is dominated by the amorphous carbon grains, with temperaturess slightly less than their destruction temperature of a 1000 K. Forr crystalline materials such as forsterite, having multiple resonance bands in the IRR which are relatively narrow and with well known wavelengths, confusion with other dustt species is almost excluded. This allows for a unique spectroscopic identification. Chemicall processes may also be used to constrain the dust formation history. An examplee is the expected coexistence of forsterite and silica as a consequence of the crystallisationn process at high temperatures (see Eq. 2.1 8). This thermal annealing argument iss used in Chapters 4 and 6 to establish or exclude high temperature crystallisation as the formationn mechanism of forsterite. Inn this thesis we will demonstrate that imposing physical constraints on spectral analysiss results - even in the absence of imaging data - provides a powerful tool to learn moree about the properties and processing of dust around HAEBE stars.

32 2 CHAPTERR 3

ISOO spectroscopy of circumstellar dustt in 14 Herbig Ae/Be systems: towardss an understanding of dust processing. .

G.. Mceus, L.B.F.M. Waters, J. Bouwman, M.E. van den Ancker, C. Waelkens, K.. Malfait

AstronomyAstronomy & Astrophysics 2001, 365, 476-490

ABSTRACT T Wee present Infrared Space Observatory (ISO) spectra of fourteen isolated Her- bigg Ae/Be (HAEBE) stars, to study the characteristics of their circumstellar dust. Thesee spectra show large star-to-star differences, in the emission features of both carbon-richh and oxygen-rich dust grains. The 1R spectra were combined with photometricc data ranging from the UV through the optical into the sub-mm region.. We defined two key groups, based upon the spectral shape of the infrared region.. The derived results can be summarized as follows: (1) the continuum of thee IR to sub-mm region of all stars can be reconstructed by the sum of a power- laww and a cool component, which can be represented by a black body. Possible locationss for these components are an optically thick, geometrically thin disc (power-laww component) and an optically thin flared region (black body); (2) alll stars have a substantial amount of cold dust around them, independent of thee amount of mid-lR excess they show; (3) also the near-IR excess is unrelated too the mid-IR excess, indicating different composition/location of the emitting material;; (4) remarkably, some sources lack the silicate bands; (5) apart from amorphouss silicates, we find evidence for crystalline silicates in several stars, somee of which are new detections; (6) PAH bands are present in at least 50% off our sample, and their appearance is slightly different from PAHs in the ISM; (7)) PAH bands are, with one exception, not present in sources which only show aa power-law continuum in the IR; their presence is unrelated to the presence of thee silicate bands; (8) the dust in HAEBE stars shows strong evidence for coagulation;; this dust processing is unrelated to any of the central star properties

33 3 CHAPTERR 3

(suchh as age, spectral type and activity).

3.11 Introduction

AA circumstellar (CS) disc is expected to be a natural byproduct or the star form ing processs (e.g. Shu et al. 1987). This theoretical expectation has obtained wide support from opticall (e.g. McCaughrean and Ockll 1996), infrared (e.g. Marsh et al. 1995) and millimetree observations of young stars (e.g. Mannings and Sargent 1997). The CS disc is expectedd and observed to gradually disappear, but remnants are still round around severall Main-Sequence (MS) stars, such as Vega (Aumann er al. 1984). F.arlier and recent modellingg of 1 latiri discs have shown that the most successful models are flaring passive discss (Kenyon and Hartmann 1987; Chiang and Coldreich 1997).

Herbigg Ae/Be stars (hereafter HAFBFs), first described as a group by Herbig (1 960), aree believed to be the more massive analogues of T Tauri stars. They are seen as the progenitorss of Vega-type stars (for recent reviews, see Waters and Waelkens 1998; Natta ett al. 2000a). They are characterized by large IR excesses due to thermal re-emission of CSS dust, show emission lines in their spectrum due to CS gas and have masses between 2 andd 8 M. (Herbig 1960). Infrared spectroscopy offers a unique opportunity to scrutinize thee composition and characteristics of their CS dust. Recent ISO (Kessler et al. 1996) studiess have revealed a large variety in the properties of the dust around HAFBF.S, from whichh it became clear that their dust is significantly different from that in the interstellar mediumm (Waelkens et al. 1996; Malfait et al. 1998b; Malfait 1999; van den Ancker 1999). .

Thiss paper is one in a series of papers based upon ISO-SW'S observations of HAFBF stars.. In this study, we compiled a set of data which include, next to the ISC) spectra, also UV,, optical, IR and sub-mm photometry of a large sample of isolated HAF.BH stars. AA similar study was already presented by Sylvester et al. (1996) for a sample of Vega- likee systems. Their ground-based observations in the IR with UKIRT are restricted to 2 ranges:: 7.5-13.5 jum and 1 5.8-23.9 /.im. Some of their sources (HD135344, HD139614, HD142666,, HD144432, HD169142 and 51 Oph) are also part of our HAF.BH sample, andd it is interesting to compare their results with ours. In this paper we give an overview off the IR features in our sample, together with a description of the Spectral F.nergv Distributionss (SFD) and we propose a global model to explain the SHDs. In Sect. 2, we describee our sample stars and their observations. We also present the SHDs (see Fig. 3.1) andd indicate observational trends. ISO-SWS spectra and an inventory of solid state and PAHH bands are shown in Sect. 3, where the individual sources are discussed as well. In Sect.. 4 we propose a global model, and discuss grain processing. Our conclusions are summarizedd in Sect. 5. In a forthcoming paper, detailed radiative transfer models of somee of the sources will be presented (Bouwman et al. in preparation).

}4 }4 ISC)) SPECTROSCOPY OF 14 Hl'RBIG Ah/Bh SPARS

-—-- * oc c—, —\£ £ rj j -v-r r ^r r N N C; ; rf~; rf~; oc c\ C C IT-IT- ^C C xr r ^r r l/^ ^ — —\C C IT IT -sf f fT*, , ~~ ~ ~-~-r-i r-i ri ri — —"-~ r_ "-~l l — — o o V V __ _ r] ] h-- r-i r-i ir r o o r^ ^ r-J J r-J J '"T', , QCC - — — — — — *— — ''—^~ — ^~ ~~ << ^ r] ] rr ^ ^ ^ ^ . . _ __ _ . . „ „ „ „ „ „ „ „ „ „ „ „ „ „ „ „ „ „

^^ "* -r-a "'^ >> > 3~~ -^ ^J J o o ^r r xr r ^J J rrrr'. -3'.-- xT T r r OC C ~-- ^c c—' — ^ ^ryfc 3EE i—' A A . . V V -JT--

j--

r--- o o oc cO O ^3 3 o o — —^ ^ L^ ^ ^; ;vy -- ^: :r~ ) ) rr* rr* —— > -^r r ~_vv I ££ £ UJ J

i i i i rr~-rr~-do d o . . fT^ ^ -- r^-. . sC C IT-IT- i/~** tj " o o o o . . -- O O LT-- ir-- O O \C C \C C U^ ^ A A 06 6 // \ — — ^r r r«-^ ^ 0 0 ,-> > , ,^ r r 11 ,m m o o "sC C 0 0 ;—. .r- J J ,, , S^^ H r r O O o o r r OC C ir-- xr r 1— — O O \jr\ \jr\rs l l rr-. rr-. r-J J <~X| | — —rsl l A A

o o u-- o o 0 0 l/~-- u^ ^ u-- LT-- \r \r ir-- \r \r o o 0 0 o o i/^ ^ ' ' - > "-j j

?? r < < = = :j j ^ ^ > > ^ ^^ ^^ ^^ ^ £> > £ £ x-£f f ££ £ir^ ^ 2OC C2 rf-- jj r , )>% [\ \ c^ ^ r-- ^r r c^-- o o »v v ^ ^ ^\ ^\ — — — — — —< <— —< < < < < << << << << << <

^c ci \ \ OC C «~. x. r r -~. > c ^ ^ vy^ ^ U~ ~ *~N N~ rr-# # T^ ^ \3 3 ri i —'' — 3C ^v v ^r r 1 o o ~ > /—v v c*i i /-»-- ;; *V /"> > > > /"> >/-* /-*/"^ ^ /—^ /-^^ ^/~> >/-> > = = = = < <~ ~I I — —I I X XX X — — — — —^ ^ l/ ^ ^

r r — — _ _ ,_ _ --; ; r~ ~ ~~ ~

35 5 CHAPÏTRR 3

AA \ir I AA l/< 1 Figuree 3.1: ISO spectra of the 14 sample stars, superimposed on their spectral energy distributions.. Crosses: observations; lull line through the optical data: Kurucz model atmosphere; other fullfull line: 1SO-SWS/LWS observations; arrows indicate upper limits. The data are normalized to thee V band.

3.22 Targets and Observations

II he objects we selected are fourteen so-called isolated HAEBE stars. These HAEBE stars aree not located inside a star-forming region, but show all the other characteristics of a HAEBEE star and are presumably the somewhat more evolved members of the HAEBE group.. I hese objects are best suited for our purpose, which is to discuss the evolution of thee CS disc, and offer the additional advantage that the spectra are not strongly affected

36 6 ISOO SPECTROSCOPY OF 14 HERBIG AE/BE STARS

,/vv \ \

x x ^^-rw^-w* *

-- V'"v,

-- . . . .

Figuree 3.2: The ISO-SWS spectra or our programme stars. Iogether with HD100546, we also showw the spectrum of comet Hale-Bopp (Crovisier et al. 1997) for comparison. byy loosely bound remnants of the star formation process, but only show the emission of thee disc. In Table 3.1 we list the programme stars and among others their main parameters:: spectral type, effective temperature and logg. Thee sources have been observed with the ISO Short Wavelength Spectrometer (SWS; dee Graauw et al. 1 996) in mode AOT1. The spectra cover an interval from 2 to 45 /./m. Somee stars have also been observed with ISO-LWS (Clegg et al. 1996), which covers a rangee from 45 to 200 /am. These spectra were discussed by Malfait et al. (1 998b, 1999); Malfaitt (1999). In this study we will concentrate on the SWS data. The spectra were reducedd in a standard way using the ISO-SWS Interactive Analysis (IA) tool containing

y y OlAPTF.RR 3 pipelinee processing steps of OI.P version 8.5, and the ISO Spectral Analysis Package (ISAPP version 1.6a). In Fig. 3.2 the reduced SWS spectra are shown. For some sources (HDll 35344, HI)ISO 193 and 31 Oph), the ratio signal-to-noise is so low at longer wavelengthss that we had to leave out the part longwards of 28 /Jm.

3.2.11 Spectral Energy Distributions

Wee also collected photometric data in the literature (Malfait et al. 1998a and references therein;; sub-mm data from Sylvester et al. 1996, Mannings and Sargent 199"7, 2000; Walkerr and Butner 1993, and Henning et al. 1998), and composed for each star an SP'D,, ranging from the UV until the sub-mm region. In Fig. 3.1, we show for each of the fourteenn sources its SF.D, combined with their respective ISC) spectrum. An appropriate Kuruczz (1993) model atmosphere was fitted through the optical data, representing the photosphericall contribution; it emphasizes the shape and the amount of the excess in thee 1R and sub-mm region. I he shortest wavelength at which an excess is discernible iss listed in Fable 3.1 as A,,nstI, with an uncertaintv of 0.2 /./m. Also shown in Iable 3.1 iss the derived fractional luminosity of the dust, Li^/L*, which is the ratio of the energy radiatedd by the dust to the stellar luminosity. 1 his ratio was calculated as follows: first, we convertedd the data into the F\ versus X scale. Then we integrated the Kurucz model over itss entire wavelength range to calculate L*. To obtain L||,>, we first subtracted the Kurucz modell from the observations, and then integrated this curve longwards of A,onsct. Sylvester ett al. (1996) have already calculated this ratio for six of our sample stars, and their results agreee very well with four of our stars, while they agree less well for HD135344 (0.64) andd HDl 444432 (0.48). The values we obtained for FIR/L* are consistent with a passive reprocessingg disc, except for the star HD142527 (L[R/L* = 1.06). Thee dust continuum behaves very different!}' from source to source, especially in the mid-IRR (15 to 45 /vm). In some stars it is rising, in other stars it is rather flat or even descending.. Also the strength of the dust continuum in these objects is verv diverse: the 122 jt/m excess ranges between 3.5 and 7 magnitudes, the 60 jjm excess ranges between 4.55 and 1 2 magnitudes and the 1.3 mm excess between 1 0 and 13 magnitudes. A second importantt observational fact is the strong variation of the strength of the 10 /vm silicate featuree from one object to another. In some objects (e.g. HD144432) this feature is very strong,, in others (e.g. SI Oph) it is less so, and in several objects (e.g. HDl 69142) it is evenn absent. Notwithstandingg these large differences, the overall structure of the dust discs seems too be similar. It is possible to decompose the spectra into at maximum three components: aa power-law, a black body (BB) and the solid state bands. As a first step, we fitted the IR continuumm of the stars showing a fiat continuum with a power-law. Actually, the deter- minationn of the continuum is non-trivial and should be taken with some caution. After somee experiments, we found that the continuum can be best determined by plotting the spectrumm as log F, versus log X. As an example, we show in Fig. }3, upper panel, how thee continuum determination was done for HDl 501 93. It is surprising to see that for at leastt six sources (HD10423", HD142666, HD144432, HDl 501 93^1)163296 and 511 Oph), the continutim can be fitted verv well with a power-law. We have to remark

38 8 ISOO SPECTROSCOPY OF 14 HERBIG AE/BE STARS

, ,

. .

\ \ : :

< <

\\ \l' I

-- '

AA | Ar-]

Figuree 3.3: The determination of the continuum tor the sources HD150193 (group Ha) and HD1792188 (group la) in log F^ versus log X space. Full line: SWS spectrum, straight dotted line:: power-law continuum fit; curved dotted line: sum of a power-law and a BB (T ~ 190 K) continuum;; asterisks: IRAS colour-corrected fluxes. heree that we did not remove the photospheric component, since it is only a negligible (< 100 %) fraction of the total flux. Only for one source, 51 Oph, the dust component is less dominant,, and the photospheric contribution to the total flux in the IR is much more important.. We therefore did not determine the power-law continuum for this source. Wee then proceeded to fit the sources with a rising mid-IR continuum, assuming wee could apply a similar power-law continuum fit lor these sources. This assumption iss supported by the similarity in the near-IR region for all our sample stars. For these sources,, such as HD179218 (see Fig. 3.3, lower panel), an additional black body' (BB)

Thiss BB component is usually a single-temperature BB, but it also can be a set of typically 2-3 BBs withh different temperatures, to which we refer to as 'the' BB component for convenience, as an analogue to thee power-law continuum.

39 9 CHAPTERR 3

% %

i i

Figuree 3.4: ISO spectra of the fourteen sample HAEBE stars, ordered by group. Group la: ABAur,HDD 100546, HD142527, HD1 "9218; group lb: HD1 00453, HD1 35344, HD139614 andd HD169142; and group Ha: HD104237, HD142666, HD144432, HD150193, HD163296 andd 51 Oph.

!() ) ISOO SPECTROSCOPY OF 14 HERBIG AE/BE STARS

Figuree 3.5: The spectrum of HD179218 compared to that of HD100453. The spectrum of HD1792188 can be obtained by adding silicate emission bands to the spectrum of HD100453. onn top of a power-law is needed to fit the continuum. Remarkably, with just these two componentss (a power-law and a BB), the continuum of all sources can be fitted. The slopee of the power-law fits are listed in Table 3.2.1, together with the average temperature off the respective BBs, when needed. In the sub-mm region, we observe a turn-down in thee slope of the continuum. This is because sub-mm wavelengths are longwards of the peakk of the black body for the coolest grains in the disc (thus we observe a Rayleigh-Jeans taill in the SED). Wee have accordingly classified the sample: group I contains sources for which the continuumm can be reconstructed by a power-law and a black body, and group II sources onlyy need a power-law to fit their continuum. The groups can be further subdivided accordingg to the presence or absence of solid state bands. In group Ia/IIa sources solid statee bands are present, while group Ib/IIb sources are without solid state bands. Our samplee does not contain a source which would fit in group lib, i.e. there is no star in ourr sample with a pure power-law continuum which lacks solid state bands. This can be- ann observational selection effect, as sources without a BB continuum are already fainter thann others. We thus have three distinct groups in our sample. In Fig. 3.4 we display the combinedd SWS spectra of the fourteen HAEBE objects arranged in these three groups. Iff we neglect the silicates, then the shape of the IR spectra of group la and lb sources

41 1 CHAPTERR 3

Group p Source e IR R BB B sub-mm m slope e Temp.. (K) slope e Ia a ABB Aur -1.20 0 93 3 -4.28 8 Ia a HD100546 6 -1.17 7 170 0 *-3.08 8 Ia a HD142527 7 -1.15 5 73 3 -3.60 0 Ia a HDD 179218 -1.09 9 195 5 -4.28 8 Ib b HDD 100453 -1.03 3 148 8 -- Ib b HD135344 4 -1.43 3 -- -3.28 8 Ib b HDD 139614 -1.31 1 169 9 -3.20 0 Ib b HDD 169142 -1.25 5 -- -3.24 4 Ha a HDD 104237 -1.10 0 absent t *-2.81 1 Na a HDD 142666 -1.03 3 absent t -2.91 1 Ha a HDD 144432 -1.32 2 absent t -2.65 5 Ha a HDD 150193 -0.95 5 absent t *-2.67 7 Ha a HDD 163296 -1.07 7 absent t -2.94 4 Ha a 511 Oph -- absent t --

TABLEE 3.2: Spectral slope of the power-law continuum fits in the IR region and the slope of the continuumm in the sub-mm region. The slopes were measured in a log \FK versus log A scale. For 511 Oph, we could not determine a power-law continuum in the [R. '1 he temperature of the BBs neededd to fit the remains of the power-law subtracted continuum are listed as well. HD 100453 andd 51 Oph have no sub-mm slope listed because there are no 3 G sub-mm measurements availablee for these stars. For stars indicated with a *\ we had to include an IRAS flux (100/vm flux forr HD100546 and HD104237; 60 ^m flux for HD1 50193) in the determination of the sub- mmm slope, as there is only one sub-mm measurement available for those stars.

iss very alike. They both have a prominent cool dust component, rising in the mid-IR. Whatt remains of the group Ib spectra after continuum subtraction is essentially the same ass for those of group la sources without their silicate features. This is shown in Fig. 3.5, wheree we compare HD17921 8 (group la) with HD100453 (group Ib). Adding the right amountt of amorphous and crystalline silicate components, we can obtain the spectrum off HD 179218 starting from HD 100453. It is striking that both groups seem to have a similarr continuum, yet the silicates behave completely different in both sources.

II o summarize these observational data, it first appears that both the near-IR and sub-mmm excesses are similar for all stars in our sample; the mid-IR flux, on the contrary, showss large source to source differences. Furthermore, the spectra can be decomposed intoo at maximum three parts: a power-law, a BB and the solid state bands.

42 2 ISOO SPECTROSCOPY OF 14 HERBIG AE/BE STARS

3.2.22 Properties of the cold grains: mass and grain size

Thee sub-millimetre flux is substantial in all our programme stars, with the exception per- hapss of 51 Oph. This implies that a large amount of cold, large grains must be present furtherr away from the star. The group-averaged sub-mm excesses are comparable. To comparee the mass of the cold dust berween different groups, we calculated Faoü/miD^, aa normalization of the cold dust mass which assumes that the stars are at the same distance.. These data are also listed in Tabic 3.1. The amount of cold dust does not differ substantiallyy between the three groups. AA first indication of the size of the cold grains can be obtained by inspecting the far- IRR to sub-mm region. All our programme stars show a turnover in their SED at far-IR wavelengths,, indicating that at sub-mm wavelengths we observe the Rayleigh-Jeans (R-J) SEDD of the ensemble of cold grains present in the disc. If these grains are large compared too the wavelength at which they radiate, the spectral slope will be XFA °= X~~ , while for smalll grains this slope will be equal to XFx °c Ar^', where p is defined from Qv °c v^, it iss the slope of the emissivity law of the dust at sub-mm wavelengths. Estimating the size basedd on spectral indices should be done with some caution, since the grain emissivity andd the temperature distribution of the grains in the disc also affect this slope. Spectrall sub-mm slopes have been determined for our sample and are shown in Tablee 3.2.1. Our sample shows a fair similarity in spectral slope, except for the stars ABB Aur and HD179218. For AB Aur, the determination of the slope is more accurate thann for HD 179218, because the last source has only two sub-mm points at a small X- interval.. As was already noted by van den Ancker (1999), AB Aur has a very steep spectral slopee {XF\ °c X~ ; both stars jump out when compared to the average of our sample (occ A~3 2); the slope of AB Aur and HD 179218 is significantly steeper than that of the tail off a black body («= ^3Ü). Bouwman et al. (2000a) show that in AB Aur the 10 jjm silicatee band and the mid-IR continuum are dominated by micron-sized grains, while the millimetree continuum is produced by grains with typical sizes of several 100 microns. Thesee grains are optically thin at sub-millimetre wavelengths, contrary to millimetre- sizedd grains around other HAEBE stars (such as e.g. HD 163296, van den Ancker et al. 2000).. Possibly the grain size distribution in HD 179218 is more similar to that of AB Aur,, with somewhat larger grains (but still in the range of ~ 100 /jm size) producing thee sub-millimetre flux. Sylvester et al. (1996) already noted from their mm/sub-mm photometryy that the dust grains around most of their Vega-like systems are much larger thann those found in the interstellar medium (sub-/vm sized, Mathis et al. 1977).

3.33 ISO-SWS spectra and their wealth of features

Wee describe general trends in the appearance of the solid state bands we have detected in ourr sample of stars. We have carefully inspected the individual spectra in order to verify thee reality of solid state bands. This can be done by analyzing the individual detector scans,, and by inspecting the two independent scan directions at which the data were taken.. We list the solid state bands and their possible identifications in "Iable 3.3. The

43 3 CHAPTERR 3

C-- ^A -Sv _£v 3 —— — — > '-~~ — r-. -t. t-- — ^ C* IJ l-ii J\ JI —— r-j ^v KK vj ^

r, <. .

<.. ' ' <.

<. .

cc >

<. . <-.. ' -<-.

—— C- ,<.. - %<*<. .

r.. ^ <. . <-. .

<-.<-. .

<.. ' ' '<. <.. ' <.

'<.-<.<.. - <.

<.. ' <,

<. .

<.. .

<.<><. .

44 4 ISOO SPECTROSCOPY OF 14 HERBIG AE/BE STARS

-- . .

' '

Figuree 3.6: Continuum subtracted spectra of the 10 pm silicate feature. We also show the spectraa of the M supergiant jj Cep and the galactic center for comparison. The galactic center spectrumm has been converted from absorption to optical depth, which is plotted on the y-axis. Amor- phouss silicate, as seen in the Galactic Center spectrum, peaks at 9.7 pm, while crystalline silicates peakk at longer wavelengths (~ 1 1 /.im). Notice the quasi omnipresence of an 1 1.3 fim band, due too crystalline silicates and/or PAHs in the spectra of the Herbig Ac/Be stars. Only AB Aur has noo crystalline signatures at all. Croup lb sources are not shown due to a lack of band structure in thiss region. featuress for which we have an unsure identification are listed with a question mark. In particular,, the PAH feature at 11.2 /jm can be easily confused with the olivine feature at 11 1.3 jjm. Caution should also be taken around 1 2 f/m, where an SWS band jump occurs. Inn Figs. 3.6 and 3.7, we show the continuum subtracted spectra in the 6 to 14 pm wavelengthh region and the 1 5 to 30 pm wavelength region respectively, highlighting the solid statee bands found in these spectral regions. Below, we briefly discuss the different solid statee components.

3.3.11 The 2-7 micron wavelength region

Fromm Fig. 3.4, one can observe immediately that there is a striking similarity between thee spectra from 2 to 7 pm in all three groups, indicating a similar composition and temperaturee distribution of the material emitting in this range. I his : -IRR excess must be

45 5 CHAPTERR 3

Figuree 3.7: Continuum subtracted spectra of the 1 5 to 30 fim region. We also show the spectra orr the M supergiant j.1 Cep (amorphous silicates) and comet Hale-Bopp (crystalline silicates) for comparison.. From a comparison with ^ Cep, it is clear that most sotirces need an additional band longwardss of the 18 pmsilicate feature to explain their spectra. Group lb sources are not shown duee to a lack of band structure in this region. causedd by small hot grains, most likely metallic Fe (or FeO), or a carbonaceous component,, e.g. graphite or amorphous carbon (van den Ancker 1999; Bouwman et al. 2000a), ass other materials would not survive the high temperatures (<; 1000 K) close to the star.

3.3.22 The PAH bands

PAHH bands are present in group la and lb sources, and only weakly in one source of groupp II (HD142666); they are seen in at least 50% of our total sample. This number mightt still go up as weak features can be lost in the noise. PAHs are strongest in early typee sources, as can be expected (since they are excited by UV radiation). That the PAHs mostt probably belong to the stellar environment can be ascertained from e.g. ISOCAM dataa of HI) 179218, a source with is not extended (Siebenmotgen et al. 2000). Another interestingg observation is their different appearance, at e.g. X ~ 6.2 /ym. PAHs in the ISMM and in HI I regions peak at X ~ 6.20 jjm, while at our stars they rather peak at X ~ 6.255 /Jm. A preliminary result in this context is that non-extended sources tend to have PAHss at 6.25 pm (Van Kerckhoven, private communication). A mote detailed study will

46 6 ISC)) SPECTROSCOPY OF 14 Hl RBIC AH/BH SPARS followw (Van Kerckhoven C, in preparation; Peeters E., in preparation).

3.3.33 The silicates

Inn Hg. 3.6 and 3.7, we show the regions surrounding the strongest amorphous silicate bands.. We also included a spectrum of the M supergiant/./ Cep (Kemper, private communication)) and the galactic center (Kemper et al. , in preparation), as prototypes of amorphouss silicate. I he spectra or group lb sources are nor shown, as they do not show silicatee bands. The 8-12 ftm silicate features are significantly different from the so-called astronomicall silicate', in the sense that they peak at 1 1 fjm rather than at 9.7 /jm. The onlyy exception is AB Aur, a group la source. The peak shift is attributed to a change inn the grain size distribution towards larger (/./m-sized) grains, and/or the presence of crystallinee olivine, causing a peak at 1 1.3 /./m (Bouwman et al. 2001b). Inn some sources such as HI) 100546, there is a large amount of crystalline silicates (Malfaitt et al. 1998b), while in other sources, such as HD163296, it is less so. That both ABB Aur and HDl 00546 are members of group la, shows that the crystallization degree off the silicate material is independent of the shape of the overall spectrum. The inter- pretationn from sub-mm data that AB Aur has the least processed dust (smallest grains at sub-mmm wavelengths) is here further supported by the lack of crystalline features in its spectrum.. We, however, derived a similar slope for the sub-mm region of HD179218 (seee Sect. 2.2), a source wirh a large amount of crystalline silicates. The determination of thiss slope was less accurate, but it is for sure steeper than the other sources. From this we cann infer that coagulation and crystallization processes occur on different time-scales.

3.3.44 The 23 micron feature

Thee f 15-30] fjm spectra (see Fig. 3-7) of group la and I la sources are even more diverse thann the [6-14] /./m spectra. From the spectrum of fj Cep, we can see that the amorphous silicatess alone can not account for the broad bands seen in other sources, e.g. AB Aur and HDD 142666. An additional component, around 22-28 /jm must be present as well. 1 his componentt can be attributed to Fe() (van den Ancker et al. 2000) and/or crystalline silicates.. 5 1 Oph shows two clearly separated bands, supporting our interpretation of the twoo component broad band. The same region for group lb sources can be fitted very welll with black bodies with a temperature ranging between 1 50 and 170 K. Also here we seee no indication for silicate bands in group lb sources.

3.3.55 Notes on individual sources

Groupp la:

ABB Aur: 1 he 10 fjm region of AB Aur shows a broad silicate feature superposed on a risingg conrinuum. 1 he feature peaks at 9.7 jt/m, which is typical for amorphous silicate. Itt is the only source in our sample to show merely pure amorphous silicates. I he 200 fjm region shows a broad band, probably due to a combination of silicates and iron

47 7 CHAPrIRR 3 oxide.. PAH bands are moderately strong, and present at 6.2, 7.~\ 8.6 and 11.2 /./m. 1 his starr has the steepest spectral slope at sub-mm wavelengths. For a thorough analysis of thiss star, we refer to van den Ancker (1999) and Bouwman et al. (2000a). HD100546:: This star is thoroughly analyzed by Malfait er al. (1998b). Amorphous silicatee emission bands are visible at 10 and 18 /Jm. It has the largest amount of crystallinityy in our sample, and shows strong PAH bands at 3.3, 6.2, "7 "7 8.6 and 11.22 jiim. The similarity with the SWS spectrum of comet Hale-Bopp is remarkable (Mairaitt et al. 1998b). 1 he comet Hale-Bopp is an end result or grain processing in our solarr system. Given the large similarities in spectral appearence with HD100546, we cann assume that the same kind or dust processing has taken place in both sources. HD142527:: The silicate feature at 10 fjm is shifted to longer wavelengths, pointing to thee presence or crvstalline silicates. I he continuum rises from longer wavelengths on thann most sources, pointing to colder dtist. Weak PAH bands are present at }.3, 6.2 and 11 1.2 jt/m. We refer to Malfait et al. (1999) for more details. Phis star is special because itt has a large I.IR/I.* ratio, tvpical for stars with an active disc. But these stars also show evidencee for outflow, what is not observed in HI) 142527. Another explanation for the largee IR luminosity is the presence of a more embedded companion, but rhis still needs too be investigated. HD179218:: The overall shape of this spectrum is quite similar to that of HOI00546. Itt shows the 10 /Jm silicate feature, and has a rising continuum. The abundance of crystallinee silicates is less, but still quite high. This source has a higher abundance of crystallinee pyroxenes than crystalline olivines, unlike observed in other sources (Malfait 1999).. Strong PAH bands are present at 3.3, 6.2, 7.7, 8.7 and 11.2 jum.

Groupp lb:

HD100453:: I his star does not show silicate emission bands. It has a rising continutim andd moderately strong PAH bands are present at }.5* 6.2, "77 and 1 1.2 //m. HD135344:: In this source the silicates are absent. The spectrum shows a rising continuum,, but beyond 27 /jm the spectrum is no longer usable dtie to low S/N. Weak PAHH bands are present at 53, "7"7 and 1 1.2 jam. HD139614:: Also here there are no silicate bands. I he spectrum shows a rising featurelesss continuum, PAH bands are absent. HD169142:: Moderately strong PAH bands are present at 3.3, 6.2, "\7, 8.7 and 11 1.2 jum. The spectrum shows a rising spectrum without silicate emission bands.

Groupp Ila:

HD104237:: I he 10 f.tm spectrum shows a verv strong silicate feature, superimposed onn a flat continutim. I he feature peaks longwards of 10 //in. so that crvstalline silicates and/orr larger silicate grains must be present. PAH bands are absent.

48 8 ISC)) SPECTROSCOPY OF 14 HFRBIC AE/BF. STARS

HD142666:: The spectrum shows a flat continuum, upon which the 10 /jm feature is superimposed,, peaking at 10.3 jjm. Very weak PAH bands are present at 3.3, 6.2, 7.7, 8.66 and 11.2 /jm. The 20 jum region shows a broad band, probably consisting of silicate andd FeO. HD144432:: The spectrum of HD144432 is very similar to that of HO 142666, but PAHH bands are absent. Interestingly, these 2 sources are in general quite similar, the mostt important difference being the inclination of their disc (Meeus et al. 1998). The 200 jum region shows a broad band, probably consisting of silicate and FeO. HD150193:: This source has a very strong 10 /jm feature, superimposed upon a flat continuum.. The shape of the spectrum around 20 fjm is very peculiar, and we are not sure iff it is an artefact or real. PAH bands are absent. HD163296:: The 10 /jm silicate feature is superposed on a flat continuum. The 200 jjm region shows a broad band, probably consisting of silicate and FeO. PAH bands mightt be present, but we cannot ascertain this. We refer to van den Ancker et al. (2000) andd Bouwman et al. (2000a) for more details and modelling. 511 Oph: This source is the most extreme, in the sense that its turn-over point towards thee R-J tail already starts at ~ 30 fjm; from which we can conclude is has a smaller amountt of dust. It also has a very rich spectrum, with both gas (COi at 4.2 /urn and

H:00 at 5-6 /jm) and solid state bands. The 10 jum silicate feature is strong and imposed onn a descending continuum. The 20-30 /jm region shows two dearly separated bands, whichh we can attribute to amorphous silicate (18 /jm), and to FeO or crystalline silicate (233 /jm). 'The evolutionary status of 51 Oph is not very clear. Some authors classified thiss object as a Be star (Slettebak 1982), while others consider it as a Herbig Ae/Be star (Malfaitt et al. 1998a) or even a Vega-type star (Sylvester et al. 1996). A more detailed studyy of this object will be presented elsewhere (van den Ancker et al. 2001).

3.44 Discussion

Twoo important results of the comparative study of the SEDs of the programme stars concernn the near-IR and the sub-mm excess: 1. a similar near-IR (1-8 jum) excess is observedd for all stars. Small hot grains close to the star must be replenished because they aree continuously destroyed by the UV radiation of the star. The overall IR spectrum beingg so diverse, but the near-IR so similar indicates that the material close to the star is homogenized;; 2. the sub-mm excess is substantial for all stars in our sample. This implies thatt large grains already formed when the star formation process comes to an end, and thatt large grains remain present around the stars during their evolution towards Vega- typee stars. In a survey of T Tauri Stars (TTS), Beckwith et al. (1990) also found that the discc mass does not decrease with increasing stellar age. Warm grains, however, seem to disappearr on a shorter time-scale. Thee third remarkable observation is the strong variation in strength of the silicate feature:: for some stars it dominates the ISO spectrum, for others it is moderately strong, andd for some stars it is even absent. It is surprising that there is no relation between the silicatee feature at 9.7 ^m and the near-IR excess, although both emission features must be

49 9 CHAPTERR 3 causedd by hor material, presumably located in the same region, and consisting; of grains of aa similar size. We can already exclude inclination angle effects since our sample includes twoo objects, HD142666 and HD144432, with a very different inclination (Meeus et al. 1998)) but with almost identical ISC) spectra, both showing a prominent 10 /.im silicate band.. There are several possible explanations to our observations, and these scenarios1 willl need to be confirmed by more detailed observations and bv careful modelling of the CSS material. In the following subsection we propose a global model.

3.4.11 Geometry of the disc and its effects on the SED

Ass stated before, the main difference between group I and II sources is the amount of mid-IRR excess, which is dominant and rising for group I sources, while moderate and ratherr descending for group II sources. Both groups also differ as far as the total IR lu- 7 minosityy is concerned: L|R/LX is on average 0.32 and 0.1" for group I and II respectively (seee I able 3.1). Croup II sources thus have the smallest emitting surface, and probably thee smallest mass of warm dust. On the other hand, the solid state bands are present withh equal average strength in group la and Ha sources: the solid state bands thus must bee formed in yet another region. Natta et al. (2000b) calculated the silicate 10 jjm featuree intensity with the models of Chiang and Coldreich (1997), and they needed to addd a power-law component to the emission of the disc atmosphere to fit their 10 mi- cronn spectra of 1 Tauri Stars; this finding supports our distinction between the region inn which the solid state bands are formed and the region from which the power-law emissionn originates. 'II he classification of our sample of HAEBE stars into two main groups could be explainedd with the following simple physical picture, which is shown schematically in Hg.. 3.8 (upper panel: a group I source, lower panel: a group II source) and consists of thee following components: I) a (partially) optically thin inner part (~ 10 AU in size); II)) an optically thick, geometrically thin disc which forms in an early stage; and III) a Haringg part, exposed to stellar radiation. Similar geometries were suggested in studies of protostellarr accretion discs by e.g. Bell et al. (1997) and Nelson et aï. (2000). There it is shownn that the protostellar discs can have an inner disc (~ 10 AU in size) with increased scalee height (i.e. "puffed up") which can shield the outer parts of the disc so that no flaringg occurs. One can explain the SEDs of group II sources bv assuming that the inner partt (component I lower panel Fig. 3.8) is partially optically thick, shielding the outer partss of the disc from direct stellar radiation, which prevents the disc from flaring (hence onee does not observe a BB component). SEDs of group I sources then can be explained byy an additional Hared component outwards of an optically thin inner disc which does nott shield. 1 he non-occurence of a group lib source could be a selection effect, since onee would expect such sources to have the lowest flux levels.

II he emission of an optically thick, passive disc varies as 'kh') °c /L ' ' (Hriedjung 1983).. I he average IR slope for the sources in our sample is -1.2, verv close to the valuee for the slope of optically thick discs. There is, however, some dispersion in our slopes,, but nevertheless the assumption of an optically thick disc to explain the power- laww component seems to hold when confronted with theoretical models.

30 0 ISOO SPECTROSCOPY OF 14 HERBIG AE/BE STARS

II I

Figuree 3.8: Schematic presentation of the model. The disc consists out of three parts: I) a (partially)) optically thin inner part (~ 10 AC); 1!) a geometrically thin, optically thick midplane; andd III) a flaring part. The upper panel shows an entirely optically thin inner disc, where the star illuminatess the disc's surface and causes the flaring. The lower panel shows the case where the surfacee of the disc is shielded from direct stellar radiation by an optically thick region, so that it doess not flare (Bell et al. 1997; Nelson et al. 2000).

Inn what follows we discuss the separate components of our model and how they appearr in the SED:

the geometrically thin, optically thick midplane consists of large grains, and is responsiblee for the IR power-law and the sub-mm continuum emission. Here resides thee bulk of the material (Bouwman et al. 2000a). The small hot grains in the inner partt of the disc are responsible for the near-IR excess. The similarity of the spectra in thiss wavelength range suggests a very similar chemical composition and temperature distribution. .

hot grains in a (partially) optically thin 'disc atmosphere' cause the 10 and 188 /./m amorphous silicate emission bands. That the solid state bands are indeed

51 1 CHAPTHRR 3

likelyy not to originate from the flared region around the disc is evidenced by: 1) our groupp I la observations (solid state bands are present, but the BB caused by the flaring iss absent); 2) the fact that silicate grains causing the 10 /vm emission are too large to bee transiently heated, so they have to be located close to the star; and 3) observations off evolved objects by Molster (2000), who notice that disc sources have a high abun- dancee of crvstalline silicates. Where this opticallv thin region is located exactlv is not clear,, but it certainly must be close to the disc and to the star.

the geometrically thick, optical!}' thin, flared dust layer below and above the mid- planee is well mixed with the gas and contains small, warm grains that cause the 100-2000 K black body component. Only a small amount of the total disc mass is lo- catedd in this flared region. In group II sources this region must be verv small or even absent,, since we only see a power-law continuum in the IR. This conclusion is also supportedd bv theoretical models from Kenyon and Hartmann (1 98""), which give an upperr value for EJR/E* for flat discs of 0.25, while flaring discs have larger IR luminosities.. When comparing the values we determined for L|[>/L* (see I able 3.1) with theirr predictions, we see that group II sources indeed have I - [R/I * values arguing for a flatflat disc. The scale-height of the flaring determines the amount of mid-IR excess: the largerr the flaring, the stronger the excess. The temperature of the BB is determined byy the distance of the onset of the flaring to the central star: the further away the flaringflaring starts, the colder the BB. If we e.g. compare HD179218 with HDl4i527, thenn the flaring must start further away for the latter object, as evidenced by its BB temperaturee (see Ttble 3.2.1). That PAHs are only present in group I sources, with thee exception of HD142666 (group Ha, but only very weak PAH features), suggests thatt PAHs are most likely located in the flared region, where they are exposed to thee stellar UV radiation. I hey are small enough to be transient]}' heated, without needingg to be located close to the star.

Otherr geometries interpreting the ShDs of HAEBEs have been proposed, e.g. Miroshnichenkoo ct al. (1999) propose an envelope in addition to a disc to model the dustt emission from HAEBEs. We note, however, that the highly abundant crystalline silicatess in HD 100546 (that have a temperature of 200 K or less', Malfait et al.' 1998b) aree most probably formed in the disc (Molster et al. 1999b), suggesting that the grains att this temperature are not in a loosely bound envelope, but are intimately connected to thee disc. We suggest that the 200 K black body component in this star, and by analogy inn other HAEBE stars in our sample, is associated with the (flared) disc and not with an envelope.. Waelkens et al. (1994) and van den Ancker et al. (1997) interpret the apparent broadd dip around 10 micron in the SEDs of HAEBEs as a physical gap in the radial distributionn of the CS dust (there is simply no dust at a certain distance to the star, correspondingg with the region where the 10 micron flux should come from). To cause a physicall gap, another body surrounding the central star must be present as well. In the lightt of these different possibilities, detailed spatial information is essential to disentangle thee location of the different spectral components.

52 2 ISOO SPF.CTROSCOPY OF 14 HKRBIG AF./BI. MARS

3.4.22 Evidence for grain growth

Evidencee for grain growth in the discs or HAEBEs has been found by several authors. Radiativee transfer modelling by e.g. Bouwman et al. (2000a) shows that the dust grains aroundd the Herbig Ae stars AB Aur and HD163296 are much larger than those of the ISM.. Furtheron, (irady et al. (1996) observe accreting CS gas in HAEBEs and attribute thiss to large infalling objects. A population of large (~ 0.1-1 mm) grains is needed to explainn the observed sub-mm fluxes in otir sample. These observations show that grains growthh indeed takes place around HAEBE stars, and that it is an on-going process which wee can observe indirectly by looking at a large sample of objects. Wee now consider the cause of the difference in SED between group I and II sources. Iff our interpretation of a flared region above and below the disc causing the BB componentt is correct, its absence in group II sources may imply that these small, warm grains havee been removed due to coagulation and/or to radiation pressure exerted on the grains bvv the central star, so that the small warm grains have slowly disappeared and their scale- heightt has diminished accordingly. The absence of PAH bands in group II stars already supportss the assumption that small grains are removed in the extended region. If we ex- pectt grains to grow during the star's evolution towards the MS, then the excess as a whole shouldd decrease. This is consistent with the amount of IR luminosity we derive from the SEDs:: stars from group 11 show a smaller LJR/L* ratio than group I stars, so that group III sources may have the most evolved dust grains. This simple evolutionary assumption iss supported by observations of TTSs by Beckwith et al. (1990), where it is shown that olderr discs tend to be colder and less luminous. The grain-growth assumption to explain thee differences between group I and II also holds in the sub-mm region: from Table 3.2.1 itt is clear that group II sources have a less steep sub-mm slope than group I sources (on averagee -2.8 versus -3.6), which means that the latter have smaller grains radiating in this region. . Too conclude, the SEDs are consistent with a disc model in which the differences betweenn group I and group II can be explained by a different extent of the warm flared laverr above and below the mid-plane. Both the E[R/L* ratio and the sub-mm slope suggestt that group II sources have larger grains than group I sources. These observations are consistentt with an evolution from group I to group II sources. However, a larger sample andd spatial information are needed to prove such an hypothesis.

3.4.33 Influence of the stellar age on dust properties

Thee age of the sample stars was derived by van den van den Ancker et al. (1998) using Hipparcoss parallaxes, and is listed in Table 3.1. Unfortunately, we do not dispose of agess for group lb stars. We are aware of the fact that our sample is biased towards the moree evolved sources, since our sources are isolated. However, we can already reach somee conclusions. There is no clear trend between age of the central star and amount of crystallinee material: the star HD1 00546 (showing the largest amount of crystalline dust) iss probably the most evolved one, while HD179218 (probably the youngest source) also showss a substantial amount of crystalline dust. AB Aur, on the other hand, is also already

53 3 CHAPTERR 3 moree evolved, but shows only evidence for amorphous silicates. The ran ye in aye is very similarr for group la and group I la stars; from a confrontation between both groups we cann conclude that the stellar age has no (or little) influence upon: 1) coagulation, since groupp Ha stars have a less steep sub-mm slope than group la stars; 2) amount of material inn the flared region, as this is much less or even absent in group I la stars; and 3) presence off PAH bands, as they are merely absent in group I la stars. We thus conclude that the timescaless on which star and disc evolve are not strongly coupled for the sample of HAEBEss studied here.

3.4.44 The amorphous silicate behaviour

Itt is surprising that the silicate feature is absent in group lb sources, as the rest of the NEDD argues for similar disc properties as for group la sources. It is not unreasonable too assume that grains grow during the stellar evolution towards the MS. Therefore, the absencee of the silicate feature could be easily explained by the absence of small grains (withh average sizes less than a few/./m). However, the presence of a near-lR excess points too the presence of small, hot grains. Furthermore, almost all of the group lb sources show PAHss in their spectra, which are also caused by very small particles. It thus seems that thee small silicate grains evolve differently than other small particles. The explanation forr group lb sources could be either that the inner part does not exists (e.g. could be geometricallyy thin, optically thick), or that there are no small (^ 50 /7m) silicate grains. Remarkably,, in an atmospheric abundance analysis by the authors and another analysiss by Dunkin et al. (1997), a silicon depletion around 3 out of 4 of the group lb HAEBEE objects was revealed, while sources from group la and Ha were shown to have solarr abundances. The photospheric silicon depletion for group lb stars mav further supportt that around these stars, silicates behave differently. There are three possibilities to removee the small silicate grains selectively: 1) a composition effect; 2) a size-effect: the silicatess are too large to be seen in that wavelength-region; or 3) aggregates with other materials.. In what follows we will discuss these different effects. 1)) A composition effect: We expect the strength of the silicate emission to be related too (among others) the amount of silicates, as the emitting dust around our sources is opticallyy thin. However, it is unrealistic to assume that the sources which do not show silicatee emission have no silicates at all in their disc. This would mean that there were no silicatess in the material from which the group lb stars are formed. "This hypothesis is most unlikely,, since the material from which stars are formed is the ISM, which is relatively uniformm in composition. There is no reason to assume that the silicates were not there in thee beginning around some stars and were there around others. Besides, the presence or absencee of silicate emission does not depend on the initial composition: two very young

HAEBEE stars, I.kH(/ 224 and EkHu 225 are located close to one another, so must be formedd out of the same material. Surprisingly, the first object does not show any silicate band,, while the second object does show silicate absorption (van den Ancker 1999). Thiss observation favours a scenario where the presence or absence of silicate emission is determinedd by other characteristics than only the chemical composition of the CS disc. 2)) A size effect: Since we found evidence for hot grains (from the 2-10/ym excess)

->4 4 ISOO SPECTROSCOPY OF 14 HF.RBIC AF./BH STARS capablee of producing a strong 1 0 jjm feature if they are (partially) composed of silicates, wee must come to the conclusion that the silicate grains around e.g. HD100453 and HI)) 169142 (both group lb objects) do not produce a silicate bump because these grains aree on average larger than the wavelength (i.e. 10 (Am). Large silicate grains result in ann emission resembling a black body without strong spectral signatures reflecting the chemicall composition. Also Hanner et al. (1994) conclude that the absence of small silicatee grains is the cause for the weak silicate emission features in comets Austin and Okazaki-Levy-Rudenko. . 11 he absence ol the 10 (im silicate feature does not necessarilv mean that also the 188 /im silicate feature must be absent: e.g. in NGC 6302 (Molster et al. 2001), the 100 /urn silicate feature is supressed because of the cold temperature of the silicate dust and thee dominant emission of the C-rich dust, but a 18 ^m feature is observed. Therefore, wee searched the group lb sources for silicate features at 18 (Am. But unlike group la sources,, group lb sources do not require solid state bands in addition to the BB+power- laww to fit their continuum. We cannot fully exclude the possibility that small silicates are presentt in the discs of group lb stars, but they must be so by a much smaller amount and/orr colder than in group la and 11a objects. Sylvester et al. (1996), however, claim to detectt silicates (around 18//m) in the spectra of HD 169142 and HD 135344, indicating thatt these stars do have larger silicate particles; from our data, we cannot confirm this observation,, however. Modelling should determine how much silicate material can be presentt in the dust without being revealed in the spectrum. 3)) Another possibility is that, after coagulation, small silicate particles are locked up intoo larger grains, composed of both small silicate grains and some other material (as is seenn in interplanetary dust particles (IDPs)). This would make them invisible if the man- tlee surrounding the silicate is sufficiently thick. If these coagulated grains are transported towardss the star, the silicate material will start to evaporate when close enough, while Fe orr carbonaceous material can survive at higher temperatures. This scenario can account forr both the presence of a near-IR excess and for the absence of hot silicate grains.

3.55 Conclusion

ISO-SWSS spectra have shown that there is a large diversity concerning IR spectral featuress and shapes in Herbig Ae/Be stars. The results from this ISO-sample can be summarizedd as follows:

1.. our sample of 14 HAKBF stars can be classified into two main groups, based upon thee shape of the continuum (flat or rising); a division which is further supported bv thee decomposition of the continuum into a power-law and a BB

2.. the disc geometry is as follows: a geometrically thin disc being responsible for the power-laww continuum, and a flared region of warm dust around the thin disc, causing aa rising continuum in the IR. An optically thin inner disc causes the solid state bands. Hott dust in the inner part of the disc is responsible for the near-IR excess. This disc- geometryy has also been proposed for TTS (Chiang and Goldreich 1997)

")") ) (".HAP!! IR 3

3.. the ncar-IR spectral region is very similar and the sub-mm emission substantial aroundd all sample sources, indicating homogeneity or the hot material and survival off large grains throughout the whole pre-main sequence evolution towards the MS

4.. the mid to far-lR region, on the contrary, is wry diverse, and we attribute this to the amountt of Haring in the disc

3.. group I sources may evolve into group II sources; the latter have evidence for larger grainss and lack the flared region present in group 1 sources

6.. the presence of PAH bands cannot be correlated to any of the stellar parameters, butt thev are only present in stars with a large amount of warm dust (group I). Thev aree most probably located in an extended region around the disc, where thev are irradiatedd by the star. Our PAH bands differ trom PAH bands in the ISM

~.. surprising is the independent behaviour ot the silicate grains: although other small particless are still present, small silicate grains seem to be absent around several stars. II his poses an intriguing problem: what happens to the silicates, what causes them to behavee so differently? It is important to get a definitive answer on the warm silicates, too know to which extent they can be hidden. Only then can the observations and proposedd model converge to a consistent picture. Therefore, the next step in our studyy will be a detailed modelling of some of the sources

Acknowledgements.Acknowledgements. We would like to thank \\. Yanclcnbussche tor assisting with the data reduc- tionn MM\ 11)1.; R. Svlvcster, C. Dominik and A. de Koter for discussions on CS discs; and S. Honvv and (.. \an Kerckhoven tor fruitful discussions about PAHs. CA1 acknowledges financial supportt from the Hemish institute for fostering scientific and technological research in indus- tryy tlWT} under grant IWT/SI3/9S ] (Id-. l.BI'MW acknowledges financial support from NWO pionierr "rant number 616.0~8.333.

So o CHAPTERR 4

Processingg of silicate dust grains in Herbigg Ae/Be systems

AstronomyAstronomy & Astrophysics 2001, in press

J.. Bouwman, G. Meeus, A. de Koter, S. Hony, C. Dominik, L.B.F.M. Waters

ABSTRACT T Wee have analysed the 10 jum spectral region of a sample of Herbig Ae/Be (HAEBE)) stars. The spectra are dominated by a broad emission feature caused byy warm amorphous silicates, and by polycyclic aromatic hydrocarbons. In HI)) 163296 we find aliphatic carbonaceous dust, the first detection of this materiall in a HAEBE star. The silicate band shows a large variation in shape, due to variablee contributions of three components: (i) a broad shoulder at 8.6 /7m; (ii) aa broad maximum at 9.8 /iin; and (iii) a narrow feature with a broad underlying continuumm at 11.3 /Jm. From detailed modeling these features can be identified withh silica (SiOi), sub-micrometer sized amorphous olivine grains and micrometerr sized amorphous olivine grains in combination with forsterite (Mg^SiO.J, respectively.. Typical mass fractions are 5 to 10 per cent of crystalline over amorphouss olivine, and a few per cent of silica compared to the olivines. The detec- tionn of silica in emission implies that this material is heated by thermal contact withh other solids that have a high absorptivity at optical to nearTR wavelengths. Thee observed change in peak position of the silicate band in HAEBE stars from 9.77 /Jm to 11.3 /Jin is dominated by an increase in average grain size, while- changeschanges in composition play only a minor róle. The HAEBE stars, [3 Pic and thee solar system comet Halley form a sequence of increasing crystallinity. We findfind that the abundance of SiÜ2 tends to increase with increasing crystallinity. Thiss is consistent with the compositional changes expected from thermal annealingg of amorphous grains in the inner regions of the disk. We confirm earlier studiess that the timescale for crystallisation of silicates in disks is longer than thatt of coagulation. Our results indicate that the processes that governed grain processingg in the proto-solar nebula, are also at work in HAEBE stars.

57 7 CHAP'i'KRR 4

4.11 Introduction

II his paper is one in a scries in which we study the circumstellar environment around Herbigg Ae/Be (HAEBE.) systems as observed with the Short Wavelength Spectrometer (SWS;; de Graauw et al. 1996) on board of the Infrared Space Observatory (ISC); Kessler ett al. 1996). HAEBE stars represent the final stage of pre-main-sequence (PMS) evolutionn of intermediate-mass stars (~ 2-10 MQ). As a consequence of the star formation processs these stars are typically surrounded by a gas and dust envelope and/or disk. They mayy be the precursors of young main sequence [3-Pictoris and Vega-type stars (see Waters andd Waelkens 1998, for a review). These latter systems are surrounded by circumstellar debriss disks, which perhaps contain planetary bodies (e.g. Aumann et al. 1984; Beust ett al. 1996). This would imply that the environment around HAEBE stars represents ann early phase in the formation of planets. Millimetre interferometry and observations inn CO emission lines show indeed disks around a number of these stars (Mannings and Sargentt 1997, 2000). Furthermore, infalling circumstellar gas observed in a number of HAEBEE systems with similar characteristics as in the (3-Pic system, is consistent with infallingg and evaporating planetesimals (Grady et al. 1997, 1999). ISO spectroscopy of isolatedd HAEBE stars has also strengthened the link between HAEBE stars and planet formation.. Analysis of the ISO-SWS and Eong Wavelength Spectrometer (LWS; Clegg ett al. 1996) spectra of HD 142527 revealed the presence of hydrated silicates (Malfait ett al. 1999). These types of silicates are also found in a major fraction of interplanetary dustt particles (IDPs; Sandford and Walker 1985). The remarkable similarity between the ISOO spectrum of the Herbig Be star HD 100546 and the solar system comet C/1995 Ol (Hale-Bopp;; Crovisier et al. 1997; Malfait et al. 1998b), both showing a high abundance off crystalline silicates and similar dust composition, also strengthens the interpretation off HAEBE stars as sites of planet formation. Heree we want to study the compositional properties of the circumstellar disks of HAEBEE stars. We have therefore selected isolated objects, i.e. not inside a star forming region,, such that confusion with other sources is minimised. Our sample consists of 14 objectss which are observed with ISO-SWS in mode AOTl. The spectra were reduced inn a standard way using the ISO-SWS Interactive Analysis (1A) tool and the ISO Spec- trall Analysis Package (ISAP). For a detailed description of the observations we refer to Meeuss et al. (2001, hereafter Paper I). The ISO spectra of a number of stars in our sample,, HD 100546 (Malfait et al. 1998b), HD 142527 (Malfait et al. 1999), HD 163296 && AB Aur (van den Ancker et al. 2000; Bouwman et al. 2000a, hereafter B2000) have beenn presented and analysed in previous studies. This, however, is the first time that a compositionall analysis of the entire ISO sample of isolated HAEBE stars is made. Using aa simple and identical analysis method for the entire sample, the derived grain compositionn of the individual systems can be compared directly. This enables us to determine a sequencee in the amount of silicate grain processing. Inn Paper I, we presented the infrared spectra of our sample and made a classification intoo two groups depending on the shape of the spectral energy distribution (SEP)). We suggestedd an explanation for the difference between the groups in terms of disk geometry andd grain size. In this paper we focus on the silicate grain processing in these systems

58 8 GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS.

(makingg up the bulk of the dust) and give a quantitative analysis of the observed solid statee emission. We focus our analysis on the 10 jjm spectral region, the main reasons beingg the occurrence of strong silicate resonances and the possibility of making a good estimatee of the emission of other solid state species in this region. The emission in this regionn is dominated by grains with temperatures in the range of ~ 500 - 1000 K. These temperaturess are reached in the inner parts of the protoplanetarv disks. The derived dust compositionn is thus representative for the silicate dust within ~ 10 AU from the central star. . II his paper is organised in the following way: In Sect. 2 we will introduce the main dustt components and the method tised to analyse the ISO spectra. In Sect. 3 modelling resultss are presented and in Sect. 4 we will discuss results and inferences for the evolution off the dust in the circumstellar environment of HAEBF. stars.

4.22 Dust composition and spectral analysis 4.2.11 Adopted dust components and grain shapes

Inn order to arrive at a sensible choice of dust components and grain shapes, we use thee outcome of detailed radiative transfer analysis on two of our programme stars. HI)) 163296 and AB Aur. The stars show a similar dust composition with the near- IRR and midTR dominated by solid state emission from small grains, with sizes between 0.011 and ~ 5//m (B2000). 7 he bulk of the material (~ 70%) consists of amorphous iron magnesiumm silicates (olivine) dominating the 10 and 18 jjm region. The nearTR fluxes aree mainly due to iron or iron-oxides and carbonaceous grains which make up ~ 15% off the total mass. HI) 163296 also shows emission bands due to crystalline magnesium silicatee (forsterite) grains. Similar results are found for HI) 142527 and HD 100546 (Malfaitt et al. 1998b; Mal fait et al. 1999; Bouwman et al. , in preparation) although thee latter system is found to have much higher abundances of crystalline silicates. In thiss study we will focus on the silicate dust component and will subtract the smooth contributionss from other (metallic iron, iron-oxide, carbon) dust species (see Sect. 2.2). Thee residue spectrum of HD 163296, obtained by subtracting the observed ISO- SWSS spectrum from our best radiative transfer model Ht (B2000), revealed an emission componentt at ~ 8.6/vm not included in the dust composition adopted by B2000, which wee tentatively attribute to silica. The presence of silica in the circumstellar dust around HAH.BEE stars is suggested by the presence of large silica grains in interplanetary dust par- tidess (Rietmeijer 1988). Thermal annealing experiments of magnesium silicate smokes showw that during the annealing process silica can form together with forsterite (Rietmei- jerr et al. 1986; Hallenbeck and Nuth 1997; Fabian et al. 2000). Another possibility to formm silica is the reduction of iron-rich silicate grains by H^ (Allen et al. 1993). The dustt components we use in this study are listed in Table 4.1 together with the grain composition,, size, shape distribution and bulk density. Wee investigated the sensitivity of the spectroscopic signature of the grains to changes inn particle shape and size. Shape effects were studied by comparing predictions for sphericall grains (Mie theory), ellipsoidal grains of the same form, and a continuous distribution

59 9 CHAPTERR 4

AA 1/itTl]

Figuree 4.1: The emission features due to crystalline silicates in the spectra of HD 1 501 93 (top), HDD 179218 (middle), and HD 100546 (bottom). The dashed-dotted lines indicate our best fits too the continuum subtracted spectra using a continuous distribution of ellipsoids; dotted lines aree best fit models assuming spheres. The CDE distribution yields by far the best fit result.

60 0 GRAINN PROCESSING IN HI-'.RBK; AH/BI SYSTI-.MS.

TAIUTT 4.1: The «rain species used to Ht the ISO-SW'S spectra. listed are composition, grain size,, assumed shape, and bulk density or the material.

Species s Composition n Size e Shape e [//m] ] [grcm°] ]

Olivine' ' [Mg,Fe]2SiO, , 0.11 & 2.0 Spheres s 3.71 1 -- Forsterite Mg:SiOH H 0.1 1 CDE E 3.33 3 Enstatite1 1 MgSiO, , 0.1 1 CDE E 2.80 0 1 1 Silica SK)2 2 0.1 1 CDE E 2.21 1

Rd'-:(( 1 )I)orscliner et al. (199S); ClServoin and I'iriou (19 3); !3)Jager et al. (1998): (4)Npuzei and Klein- mann (1960)

off ellipsoids (CDE; Boh ren and Huffman 1 983, for a review on these methods). The cal- culationss for ellipsoids were performed in the Rayleigh limit, which yields spectroscopic propertiess independent of grain size. Thee amorphous olivine grains show little dependence on shape, but display a strong dependencee on size. In genetal, for larger grains the contrast of a feature relative to the continuumm becomes less, especially when the grains have radii a > 2^m. By comparing thee optical properties of these particles, assuming they are spheres, for the size range whichh dominates the 10 /jm silicate feature (0.01 to ~ 5 ,um), we find that the emission cann be characterised bv two typical grain sizes: 0.1 ji/m for grains with sizes < 1 /Ltm, and 22 jjm for grains > 1 /./m. We will investigate the contribution of small and large particles onn the emission properties of olivine grains by adopting these two typical sizes. Contraryy to the amorphous olivine, the spectroscopic properties of crystalline silicatess are very sensitive to the adopted grain shape. Before discussing this, we mention thatt also the chemical composition of these grains strongly affects the emission features, specificallyy the Fe over Mg ratio. Detailed analysis of disk sources has shown that the crystallinee silicates of the olivine and pyroxene family are pure or nearly pure magnesium silicatess (e.g. Molster et al. 1999a). Here we will assume that these crystals are the pure magnesiumm silicates forsterite (\1g-,SiO J and/or enstatite (\tgSiO3). 'lo determine the shapee of the silicate crystals we fitted the multiple features of forsterite and enstatite in the 155 to 30 /jm range in the continuum subtracted spectra of HD 150193, HD 179218 andd HD 100546. These stars display the most prominent crystalline bands. Fig. 4.1 showss these spectra. The continuum subtracted spectrum of HD 150193 is dominated bvv emission from SiO^ grains, giving rise to an emission complex around 20 /./m. 1 he spectrumm of HD 100546 is dominated by forsterite emission, while HD 179218 also showss features due to enstatite. I he figure also shows the best fit to the silicate bands forr different shape properties. We obtained emissivities by multiplying the absorption

61 1 CHAI'THRR 4

coefficientss with two black bodies of 350 and 80 K respectively, which gave the best fit too the entire wavelength range. This simplified representation of a temperature distributionn does not significantly affect our conclusions with respect to particle shape. These conclusionss are that spherical grains fail in predicting the correct peak positions as well ass in producing the width of the features (they are too narrow). F.llipsoidal grains on thee other hand produce a fair match to both the location and the width of the features, thee best results being achieved using the CDF distribution. We therefore adopted CDF forr our modelling effort. As we treat the CDF particles in the Ravleigh limit, we can nott study grain size effects. However, in analogy with spherical grains we anticipate a diminishingg contrast of the features when a > l/nm. The calculated optical properties of thee dust species adopted are listed in "Fable 4.1 and are plotted in Fig. 4.2.

4.2.22 Spectral analysis

loo determine the composition of the circumstellar dust in the HAFBF systems, we con- structt model profiles by making linear combinations of the absorption profiles of the adoptedd dust species. The resulting model profile is given bv

iWk.i&)) = $>/K,a) (4.D

wheree K,(A) are the absorption profiles of the dust species listed in Table 4.1 and a,a, the multiplication factors which are listed in 'Fable 4.3. We have several reasons for applyingg this relatively simple modeling approach. It is clearly a completely valid method iff the dust medium is optically thin, as suggested by the fact that all observed solid statee features are in emission. Still relatively modest line-of-sight self shielding effects mayy play a role, which would require full 2-dimensional radiative transfer modeling to properlyy sort out. However, our approach is to first gain a better understanding of the generall qualitative trends in grain composition, shape and processing observed in the ISOO spectra. I hese results may then help in constraining parameter space when making detailedd 2D modeling efforts. We note that preliminary full 2D computations of disk emissionn using identical dust composition as for rhe optically thin model show that the resultss presented in this paper are in qualitative agreement, which adds to the validity of thee present approach (Bouwman et al. 2001a, submitted). Wee focus on the region centered around 10 f.im because one of the main resonances off the amorphous silicates is located at ~ 9.8 jam. Within the wavelength range from 6 too 14 //m a clear distinction between the emission from silicate dust grains and other dust speciess producing a smooth continuum can be made, allowing us to subtract the smooth continuumm contribution (see previous section). This distinction is problematic when usingg the other main resonance of the amorphous silicates near 1 8 ,t/m, where resonances fromm other dust species such as iron-magnesium oxides or iron sulfides are important. Oftenn this resonance is located on the edge of a rising continuum caused by the cold bulk materiall present in these HAFBF systems, making it difficult to determine the shape andd strength of the 18 /.im band. I his makes it impossible to determine the amount of

62 2 GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS.

Figuree 4.2: The emission properties of silicate dust grains. Plotted are the normalised absorptionn coefficients against wavelength. 'I he top panel shows the absorption coefficients of quartz andd amorphous olivine for two grain sizes. The bottom panel shows the crystalline dust species enstatitee and forsterite. I he normalizing constants in cm gr ' for the silica. 0.1 and 2.0 ;Um amorphouss olivine are 3.5 - 10 , 1.7-1 0 and 2.2 I 0 , respectively. For enstatite and forsterite thee normalisation is 7.3 -10 and 7.6 10 , respectively. amorphouss silicate dust. A meaningful fit to these longer wavelengths can only be made byy applying a lull radiative transfer method. It is important to realize that the choice of thee 10 fum region implies that the derived dust compositions are representative only for

63 3 CMAPTHRR 4 thee inner parts or the protoplanetary disks in the HAF.BK systems. Onee or the difficulties in fitting the silicate features is that the main resonance of forsteritee at 1 1.3 /./m (shown in Fig. 4.2) coincides with a strong emission feature usually attributedd to polycyclic aromatic hydrocarbons (PAHs; Allamandola et al. 1989). PAH emissionn is observed in a large fraction of our sample of HAFBF. systems (see Paper I). To determinee the band strength of forsterite we first have to make an estimate of the PAH emissionn at 1 1.3 jum. We have three objects where we know the contribution from PAHs well;; these are shown in Pig, 4.3. Marked in this figure are the PAH bands near 6.2, 7.8, 8.66 and 1 1.3 jum. Both HI) 100453 and HD 169142 do not show any emission due to silicates,, so a direct determination of the PAH band strength is possible. AB Aur does showw silicate emission, but only from amorphous material, which has resonances much broaderr than the PAH bands. From derailed modelling (B2000) we can determine the silicatee contribution, which we subtracted from the 1SO-SWS spectra. Thee position and relative band strengths of the 6.2 and 1 1.3 jt/m bands for all three sourcess are equal within the error margins. We determined an average spectrum of the threee sources and used this as a template to determine the PAH contribution in the other HAEBFF systems. This template spectrum is also shown in Fig. 4.3. I o estimate the PAH contributionn at 11.3 ^m, we scale the PAH template spectrum to the strength of the 6.22 jum feature, which is clearly visible and well isolated, such that contamination effects aree minimised. We use the template spectrum as a separate spectral component in the fitss presented in Sect. 3.

4.2.33 Detection of aliphatic hydrocarbons in HD 163296

Onee system, HD 163296, shows a band at 6.9 fJm due to aliphatic hydrocarbons, which iss the first detection of this species in a HAEBF. system (Hony et al. in preparation). 11 hese types of hydrocarbons are likely to be incorporated into grains unlike rhe PAHs (e.g.. Ciuillois et al. 1996). It is clear that for HD 163296 we cannot use the PAH template,, and we must find other objects with aliphatic bands to serve as a template. The carbon-richh post-asymptotic giant branch star SAO 34504 has one of the most pro- nouncedd 6.9 jt/m bands observed in any evolved carbon-rich star (Fig. 4.3), and we use itss spectrum as a template for HD 163296. Note that the position and shape of the bandss in SAO 34504 are quite different from those of the PAH template. We adopt a temperaturee of 250 K for the aliphatic carbonaceous dust, derived from the 6.2 to 6.9 jjmjjm band strength ratio in HD 163296.

4.33 Results

Fig.. 4.4 shows the continuum subtracted spectra of rhe programme stars. Of the four starss that do not show anv silicate emission bands we onlv plot HD 100453 and HDD 169142, as these are the only ones that do show PAH features. We have also in- cludedd some comparison spectra of other objects showing the 10 /jm band. 1 he idea iss that these reference objects provide limiting cases for the composition of the dust m

64 4 GRAINN PROCESSING IN HFRBIG AE/BI: SVSTF.MS.

rr ir i— yr

-- < vv v w v VV V ^^ ^ Z'-s-- V Vdv" vv v

.—— , _ -^r

vvvv v

-- ^ ur- DC vrf- .. _ c; ^ 3C ,r- : ~~ ~ " ^ ^ t ——; NX, r^"J uh-:~ u^^ ^^^ ? ^ .—:: _: i/~ ^ ~" " VV V ^ ~ V V

xr r ^^ ,—. C" r""-) O xf \t~-

VV ^ V x vr- - x |C " ^ ^ ^ ^ ^ y ^ V 'V V V JJ C .— r--

yy _g ;Ü -a ^

- C" .. ^ "—' ^ X N O O i,^^ «^ .—3 . r—C rTrr\^ II^Cu""\ Ü ^-: :—, — — ^- — v^v v £22 ;i *£ ~ So

—,, — „„u~ ^o „ \r ^S§o^ -- ^ "i & -s V V ^^ w V V V^ 'tt ,? Ï*

rv-,, O jj- ££ »A cS^Ï^Sy-i^vrVVVV << -

££ -j ^ ^ rjj 2: > > > > >> > -> oo **** *> --* ,> j** £ -; oo ^r N u" << < < —

££ x

oc c o o ^ ^^ j j ^t t ~^r r — — ~v~ ~ ^-u~ ~ " " — — ~

zz 1 ^

65 5 CHAPTERR 4

SAO34501 1 15 5 'V V

10; ;

0 0 ' ' 88 ï) 100 11 AA [firn

Figuree 4.3: The infrared emission bands of PAHs and aliphatic hydrocarbons. Plotted are the continuumm subtracted spectra of SAC) .34504, HD100453 and HD 169142 and the model sub- tractedd spectra of AB Aur. Indicated in the figure are the bands at 6.25, 7.8, 8.6 and 1 1.3 fim due too PAHs (dotted lines) and the 6.9 //m band due to aliphatic hydrocarbons (dashed line). The bottomm curve shows the template spectrum used to estimate the relative PAH contributions.

thee HAEBE systems. The silicate profile of the ISM, observed with ISO in the direction off Sgr A (Kemper et al. 2001, in prep.), shows the material in the form in which it is expectedd to be present during the first phases of star and disk formation. The spectrum off the red supergiant/./ Cep is representative for this class of objects and shows dust with

66 6 GRAINN PROCF.SSINV, IN HF.RBK; AH/BK S^STI.MS.

remarkablyy similar properties as the silicates in the ISM. Contrasting these eases where onlyy amorphous silicate dust material is observed, we also show three cases in which the dustt is expected to be highly processed. Thee spectra of the comets Hale-Bopp and 1 P/Halley show the end results of processingg of solar system dust grains before being incorporated into cometary bodies. These typess of grains may be representative for the late stages of dust evolution in the protoplanetaryy disks around HAEBE stars. 1 he spectrum of Halley shown here was observed whenn the comet was at a heliocentric distance of 1.3 AU (Bregman er al. 1987). We show twoo spectra of Hale-Bopp, one observed in October 1996, when the object was a 2.8 AU (Crovisierr et al. 1997), and one observed in April 1 997, when it was at a distance of 0.97 AUU (Hayward et al. 2000). Also representative for the late stages of dust evolution in protoplanetaryy disks may be the spectrum of (3 Pictoris (Pantin et al. 1999), which has a collisionallyy dominated debris disk and may be an end stage of HAEBE evolution. Thee spectra in Fig. 4.4 are ordered (from top to bottom, left to right) according to thee peak-position of the silicate feature. I he feature shifts from 9.8 /vm, the peak position off the ISM silicate band and indicated with the dashed line, towards 1 1.3 /./m. The latter wavelengthh corresponds to one of the strongest resonances of forsterite and a PAH band (seee Fig. 4.2 and Fig. 4.3). Apart from this shift in peak-position the silicate band also broadenss and changes shape from a steep blue wing with a slow red decline towards a profilee which is slowly rising and then drops steeply. To quantify these trends in peak position,, width and skewness we measured the excess flux above continuum at 8.6, 9.8 andd 11.3 /Jm. These positions correspond to the shoulder of the silicate feature at short wavelengths,, the shortest wavelength of the peak position and the longest wavelength of thee peak position, respectively. Inn Fig. 4.5 we show the correlations between the ratios of the excess fluxes at 8.6, 9.8 andd 11.3 /Jm, after removing the PAH contributions. In this and following figures the HAE.BE^^ stars are represented with filled symbols while the reference objects are indicated withh open symbols. The diamond, upside down triangle, star, and circles represent the ISM,, JJ Cep, [3 Pic, and the comets, respectively. Note that the flux ratio F,s0/E11.3 's fairlyy constant for the HAF.BE stars at ~ 0.45, except for three objects. We have marked thee HAEBE systems with a F^.c/Fnj flux ratio of ~ 0.45 with a larger symbol. The exceptionss are HI) 100546 (marked with 2) and the reference object Hale-Bopp (two observationss marked with 1), showing a ratio about twice as small, and HI) 150193 (markedd with 3) which has this ratio a factor of two larger. Leaving out these exceptions, onee finds a reasonably tight correlation between the 8.6 over 9.8 and the 11.3 over 9.88 /am flux ratios. We fitted a relation of the form Y - aX\ resulting in b = 0.93 0.233 and a - 0.43 0.03 (dotted line in top panel). W'c will return to the exceptions inn Sect. 4.4; for the moment we exclude these objects from the correlation analysis we discusss below. Overplottedd in Fig. 4.4 are our best fits, using the analysis method described above. Thee results of our fitting procedure are listed in Table 4.3. The first column lists the spectrall type of the HAEBE stars in our sample, and /.i Cep and P Pic (Malfait et al. 1998a; Dunkinn et al. 1997; Gray and Corbally 1998; Kukarkin et al. 19T1), while columns

67 7 CHAPTERR 4

\\ l/,rr,|

Figuree 4.4: Fit to the 10 /jm silicate feature as observed in our sample of HAEBE stars. Plotted aree the continuum subtracted ISO-SWS spectra and our best fit models. Also plotted for reference aree the silicate bands of the ISM towards the galactic centre, the red supergiant// Cep, P Pictoris, andd the comets (71995 Ol (Hale-Bopp) and 1 P/Hallev. The dashed line indicates the position off the amorphous silicate band as observed for the ISM at 9.8 /jm. Indicated in the panels are thee group to which the individual systems can be classified on grounds of their overall SED (see Sect.. 3.2).

threee to six give the multiplication factors of the absorption coefficients of the included dustt species. The derived mass ratios of the dust components used in our fit are listed in columnn seven to nine. Column seven lists the mass ratio of the 2.0 over 0.1 ^m amorphouss olivine grains, column eight the mass ratio of forsterite over the total silicate mass, andd column nine the mass fraction of silica.

4.3.11 Correlations

Ioo link the derived dust composition with the observed changes in the 10 /./m silicate feature,, we compared the mass ratios as listed in Table 4.3 with the flux ratios used in Fig.. 4.5a. Plotted in Fig, 4.6a is the ratio between mass contained in large amorphous grainss (a = 2 fjm) and small amorphous grains {a = 0.1 jjm). This ratio is indicative for

68 8 GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS.

0.600 A 0.50 0

0.. 10

iff ().:U)

^^ 0.20

0.10 0 0.:S S 2.0 0 FF /F 11 1.:V 9.8

0.600 B 0.50 0

0.. 10

0.30 0

-A A ^^ 0.20

0.10 0 0. . .0 0 11 1.:?

Figuree 4.5: Color-color diagrams. Plotted arc the flux ratios of the excess fluxes at 8.6, 9.8 and 11 1.3 /Jm corrected for the PAH emission. The excess flux is rhe flux minus the assumed continuumm at the specified wavelength. The filled symbols indicate the HAEBE stars, the open symbols thee reference objects. The HAEBE stars belonging to group la are represented by triangles, while thosee belonging to group 11a are indicated with squares. The diamond, upside down triangle, star,, and circles represent the ISM, fj Cep, (3 Pic, and the solar system comets respectively. The dottedd line in the top panel is a least square fit to those HAEBE systems having an 8.6 over 11 1.3 flux ratio of ~0.45, indicated with large svmbols. A tew objects are explicitly marked with aa number from 1 to 5, being Hale Bopp, HD 100546, HD 150193, HI) 179218, and AB Aur, respectively. .

6') ) CHAPTERR 4

M2.(/M0.1 1

rr--

0.10 0 1 1

"."11 \

-- M,,forst ,/' M -, l l

r. r. x x -- 5t t

0.100 0 MM . .,/M ., sio2'' sil M- 5 3 J\# # 1 1

i; ;.3'*9. 8 8

Figuree 4.6: Correlations between 1 1.3 over 9.8 /Jm flux ratio and the derived mass ratios of the dustt components. Flic top panel shows the correlation for the mass ratio between the 2.0 and 0.1 /vmm amorphous olivine grains. The middle panel shows the mass ratio of forsterite over the total silicatee mass, and the bottom panel the SiCT mass ratio. The meaning of the symbols is identical ass in Fig. 4.5. 1 he dotted line indicates a least square fit to the HAEBE systems with F&.(,/¥]].} ~~ 0.45; these are indicated with large symbols.

70 0 GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS.

M M 2.0/ 0.1 1

1.00 0

O.lll l

o.oii A

M,, ,/M .,' ;; 1

A--

T. T.

B B

MM . .,/M io'i77 * sil A T, , V V

0.100 0.50 o.oi) 0.70

F F 8.6/ 9 98 8

Figuree 4.7: The correlation between the 8.6 over 9.8 pm Hux ratio and the mass ratios of the fittedfitted dust components. For an explanation of the symbols see Fig. 4.5.

thee typical grain size. In panel b of the same figure we show the forsterite mass fraction (relativee to the total silicate mass), and in panel c the silica mass fraction. All three quantitiess are plotted as a function of the 1 1.3 over 9.8 jjm flux ratio, which provides aa measure for the peak position of the 10 /Jin silicate feature, which varies between 9.8

"1 1 CHAPTHRR 4 andd 1 1.3 //m as can be seen in Fig. 4.4. the dotted line represents a least square fit to thee objects selected on grounds or their 8.6 over 11.3 f.im flux ratio. In fitting this and alll other correlations presented in this paper, HI) 100546 and HI) 150193 are alwavs excluded.. Also excluded will be objects lor which in the correlation being discussed only upper/lowerr limits are available. Objects not taken into account have been given a small symboll size lor clantv. In figures where no correlation can be found we have not used thiss convention. Forr the HAhBE systems a strong correlation can be seen between the grain size off the amorphous iron magnesium silicates, producing the bulk of the emission, and thee peak position of the silicate feature, hitting the same relation as in big. 4.5a yields bb - 3.96 0.08 a = 4.89 0.05. This strong dependence on grain size of the 1 1.3 over 9.88 jam Mux ratio is a reflection of the emission properties of the amorphous olivine as shownn in big. 4.2. An increase in grain size results m a large increase of the 1 1.3 /./m flux relativee to the 9.8 ;t/m flux. Notice that all solar system objects and (3 Pic fall beneath the derivedd correlation lor the HAEBE, systems. Inn fig. 4.6b we show the relation between the peak position of the silicate band and thee mass fraction of forsterite. Since forsterite has a strong resonance at 1 1.3 /./m, it is expectedd that a high mass fraction of forsterite will also shift the peak to longer wavelength.. We caution that the determination of the forsterite mass fraction suffers from contrastt effects with the amorphous silicate and possible confusion with PAH emission att 1 1.3 ,i/m. Nevertheless, a clear trend can be observed in Fig. 4.6b. However, we do nor findd a significant trend if wc consider only the HAHBE stars, in contrast with the strong correlationn with grain size seen in Fig 4.6a. We conclude that for our sample of HAEBK starss the shift in peak position of the silicate feature is due to a change in grain size, while thee degree of crystallinity plays only a minor róle. Interesting!)',, this situation seems reversed when we consider (3 Pic and the two solar systemm comets in our sample. Fig 4.6a shows that, considering the correlation with grain size,, these objects do not follow the trend set bv the HAEBE stars. On the other hand, Figg 4.6b indicates that the shift in peak position of the silicate band seen in (3 Pic and thee solar system comets correlates well with the degree of crystallinitv. I his suggests that forr these objects the shift in silicate band position is mainly due to a high fraction of forsteritee and is not dominated bv grain size effects. Final!)',, we note that no correlation between SiO^ abundance and peak position is evidentt (Fig. 4.6c). Fig.. 4.^ shows the same mass ratios plotted against the 8.6 over 9.8//m flux ratio. No significantt correlation can be found with the mass ratio of 2.0 over 0.1 /Jtn amorphous olivinee grains or the mass fraction of forsterite. However, a correlation can be observed withh the SiO,] mass ratio. The least square Ht results in b = 1.1 5 0.05 and a - 0.059 0.003. . II he relation between the derived mass fractions is presented in Fig. 4.8. A correlationn between the typical grain size of the amorphous olivine and the other dust componentss is not found as can be seen from Fig. 4.8a and b. Fig. 4.8c, however, does show aa possible correlation between the amount of silica and forsterite. Except lor the objects fll)) ] 00546 and HI) 150193, the HAEBE stars seem to have a relatively larger SiO: GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS. masss fraction with increasing abundances of crystalline magnesium silicates. Again, fittingg a relation of the form Y = aX to the selected objects, indicated in the figure with largee filled symbols, results in b - 0.47 5 and a - 0.082 . This is also consistentt with the upper limits derived for AB Aur (maked with 5), ft Cep and the ISiM, where neitherr crystalline magnesium silicates or the shoulder at 8.6 ^m, indicative for silica, are observed.. We note that comet Hale-Bopp also shows a low abundance of SiC>2 despite itss high fraction of forsterite. In that sense it strongly resembles HD 100546. Though itss FgxJF] i3 flux ratio does not deviate, in the fir we also excluded HD 17921 8 (marked withh 4) due to the detection of enstatite in this system: the derived silica over forsterite masss ratio may not be due to annealing of amorphous iron magnesium silicates, but due too secondary reactions given by Eq. 4.2 leading to the formation of enstatite. We will discusss this point in detail in Sect. 4.3.

4.3.22 Correlation to the overall Spectral Energy Distribution

Inn Paper I we presented a classification of the sample of HAEBE stars discussed here in termss of overall properties of the spectral energy distribution (SED). The SED of so- calledd Group I sources can be represented by the sum of a power law component and aa blackbody component. This blackbody component is needed to account for the large excesss group I sources show at far infrared wavelengths. Group II sources only exhibit the powerr law component. Each of these groups can be further sub-divided into two sub- groupss based on the presence of solid state emission, i.e. essentially silicates. If present, thee group is given the suffix a, if not present a suffix b is added. This classification onn grounds of the overall infrared spectrum is indicated in Fig. 4.4 through 4.8. In thesee figures the filled symbols represent the HAEBE stars. These are further subdivided accordingg to the shape of the SED into triangles (group la) and squares (group Ha). Wee suggest in Paper I that this classification is linked to the spatial distribution off the dust. Both groups feature an optically thick, geometrically thin disk mid-plane responsiblee for the power law component. In group I the disk surface is flared, explaining thee additional far-IR emission. This group also displays PAH emission, which probably originatess in the flaring region that is illuminated by direct stellar UV radiation. It is usefull to investigate the relation between the shape of the SED (reflecting the spatial dustt distribution) and the shape of the 10 ftm feature (reflecting silicate composition). Ass disks evolve, it is likely that both their geometry and dust composition change. Lookingg at Fig. 4.4, group la appears to show the most silicate processing if one excludess AB Aur, which displays the most pristine dust. However, after correction for PAHH emission, clear systematic differences between group la and Ha cannot be found (cf.. Fig. 4.6 through 4.8). This suggests that the amount of processing of the silicate grainss dominating the midTR is not correlated to the overall SED and hence not with diskk geometry if the link suggested in Paper I is correct. Sincee Group lb shows no solid state features, no information on the amount of grain processingg can be derived for this group of stars. A possible explanation for the lack of 100 ftm. band emission could be the grain size. Grains with sizes much larger than the wavelengthh at which they radiate (> 10 ftm) only produce a blackbody continuum. This

73 3 CHAPTERR 4

1.000 0 A A 1,, -

0.100: :

ss 0.010-

0.001 1

J J

== 0.0 100 Ï Ï 0.00100 tr

0.00011 _ 0.01 1 0.100 1.00 10.00 100.00 WM0.1 1

SMI) ) 0.1000: :

0.00 100 E

0.0010; ;

0.0001 1 0.001 1 0.010 0 0.100 0 1.000 0 M,, ,/M ., forstt sil

Figuree 4.8: Correlations between the fitted silicate dust components, a) Correlation between thee mass ratio of the 0.1 and 2.0 jjva sized amorphous olivine grains {mi.o/mo]) and the amount off forsterite. b) Correlation between the amount of silica and m2.0/^0.1 c) Correlation between thee mass ratio of silica and the ratio of the crystalline magnesium silicate forsterite. For an ex- planationn of the symbols see Fig. 4.5. Indicated in Fig. 4.8c with the dotted line is the fitted correlationn of the form Y = aX . The dashed lines represent the measured annealing behaviour off the amorphous magnesium silicates smectite dehydroxylate (SMD; upper curve) and serpentinee dehydroxylate (SD; lower curve). The solid line is the expected annealing behaviour for an initiall mixture of 4% of SMD and 96% of SD.

1A 1A GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS. wouldd implv that group lb has stronglv processed dust compared ro ISM material (based onn this argument we have placed the two members of this group at end or the sequence inn Fig. 4.4). An alternative explanation could be the absence of a sufficiënt amount of opticallyy thin material, resulting in a weak or absent silicate band.

4.44 Discussion

Inn this section we suggest an explanation for the changes in shape of the 10 ^m silicate band,, as seen in our sample of 14 Herbig Ae/Be systems, in terms of a simple physical model.. The change of the profile is dominated by two effects. Hirst, the peak shifts from 9.88 to 1 1.3 /./m; and second, this shift in peak position is accompanied by the appearance off a broad shoulder shortward of 9.8 jum, extending to about eight micron, which we havee characterised using the Hux at 8.6 /nm.

4.4.11 Explanation for the change of the 10 /./m feature Wee have identified two processes that may be responsible for the change in peak position fromm 9.8 /7m, characteristic for ISM material, to 11.3 /Jm, typical for the solar system comett Hale-Bopp:

1.. coagulation of sub-^m sized amorphous olivine grains into micron sized grains; and

2.. annealing of amorphous silicates into crystalline magnesium silicates and silica.

Thee strong correlation between grain size and peak position in HAEBH stars is clearly demonstratedd in Fig. 4.6a, where the Hux ratio ¥ i IJ/FIJ.S is shown against the mass ratio off large over small amorphous grains. Larger particles cause an increase of the 1 1.3 over thee 9.8 /Jin Hux. The flux at 1 1.3 //m also increases when forsterite crystals form from thee amorphous iron magnesium silicate dust (Fig. 4.6b), and seems important for [3 Pic andd the solar system comets, but is not a strong effect in the HAPTiE stars. Compared too HAF.BE stars with a similar Fn^/F^s ratio, the former systems show on average a smallerr grain size with a larger crystalline silicate fraction. Thesee observations are consistent with a picture in which grains in HAFBh stars coagulatee to form larger grains, but do not crystallise on the same timescale. Crystallisation seemss to occur on a longer timescale. 1 his is in agreement with a study by Molster et al. (1999b),, who analysed the degree of crystallisation and coagulation for long-lived disks surroundingg evolved stars. Molster et al. conclude that coagulation precedes crystallisation.. As we will discuss in Sect. 4.2, the relation between forsterite and Si()i (cf. fig. 8c)) suggests thermal annealing to be responsible for the crystallisation of grains in the HAEBFF systems. [3 Pic and Hallcy also obey this relation, implying the forsterite found inn these objects to have a similar origin. Hale-Bopp and, interestingly, HI) 100546 lack SiO;>> emission but do show abundant forsterite indicating a different mechanism may bee at work in these objects. In Sect. 4.3.1 we speculate this mechanism may be related too differentiation of large parent bodies. It is important to stress that the HAF.BE stars.

^ ^ CHAPTFRR 4

PP Pic, and the comers, lic along a sequence of increasing silicate crystallinity (Fig. 4.6b). Thiss suggests that processing or silicates starts in the protoplanetary disk and leads to an increasedd crystallinity. We may conclude that the processes that governed grain processingg in the protosolar nebula are also at work in HAEBE stars.

4.4.22 The chemical composition of the silicate dust

Inn big. 4.8 we showed that the fraction or SiO^ tends to increase with the degree or crystallinity.. Laboratory studies of thermal annealing of amorphous Mg silicate smokes showw an increase in the fraction of forsterite and SiOi as a function of time, temperature, orr both (Rietmeijer et al. 1986; Hallenbeck and Nuth 1997; Fabian et al. 2000). This impliess that the simultaneous occurrence of forsterite and SiOi in our sample is consistentt with an origin due to thermal annealing in the inner part of the proto-planetarv disk.. The lack of correlation between grain size and changes in composition supports our previouss conclusion that grain growth proceeds independently from these compositional changes. . Wee point out that the temperature of the grains probed in the 10 |/m silicate band (typicallyy 500 K) is well below the annealing temperature of amorphous silicates (about 11000 K). Extrapolation of laboratory experiments indicates that the annealing timescale off such relatively low temperature grains would be prohibitively long compared to the agess of these systems (Hallenbeck et al. 2000). This would imply that in situ formation off the forsterite seen in HAF'BE disks is unlikely. Either extrapolation of laboratory resultss is not permitted and the million year timescale allows for annealing well below thee glass temperature, or radial mixing of processed material from the innermost regions iss responsible. Althoughh there is a considerable uncertainty in the relation between silica and forsteritee as seen in Fig. 4.8c, it is in principle possible from the observed trends to sett constraints on the Mg over Si ratio of the amorphous bulk material from which both forsteritee and silica are formed during annealing. Here we discuss the principle; more accuratee data are needed to provide useful constraints. Too constrain the derived correlation between silica and forsterite we make a compar- isonn with the measured annealing behaviour of amorphous magnesium silicates. Plotted inn Fig. 4.8c are the measured annealing curves of the pure magnesium silicates smec- titee dehydroxylate (SMD; Mg^Si^Oji) and serpentine dehydroxylate (SD; Mg3Si20-). Condensationn experiments show that at low temperatures (I ~ 500 K) only these materialss are formed (Rietmeijer et al. 1999). Annealing of SMD produces forsterite and silicaa in a mass ratio of 1.4; for the SD material this ratio is 7.0 (Rietmeijer et al. 2001, in preparation).. As one can see, the curves tightly constrain the found correlation. This con- strainss the Mg/Si ratio between 0.75 (SMD) and 1.5 (SD). The fitted trend implies that withh increasing abundance of forsterite the silica over forsterite mass ratio decreases. This isis inconsistent with the annealing of an amorphous material with a homogeneous chemical composition.composition. Annealing of such a material would give a constant ratio between forsterite andd silica. A good match with the observed behaviour is obtained if we adopt a material initiallyy composed of 4°o SMD and 96% SD by mass (solid line). In plotting this curve

76 6 GRAINN PROCINSINC, IN HI.RBK; AH/BF SYSTEMS. wee assume that first all SMD is converted to horsterite and silica, before SD starts to anneal.. This difference in annealing time scales is suggested by annealing experiments wheree the thermal and mineralogical development of the SMD component is found to bee ahead of the SD material (Rietmeijer et al. 2001, in preparation). In this way one can producee a flatter curve consistent with the fitted correlation. The bulk Mg/Si ratio of suchh a composition would be 1.46, which is larger than the solar average bulk compositionn of CI and CM carbonaceous chondrite of 1.06 and 1.04 respectively (Brownlee 1978).. However, ir is more consistent with the interstellar value adopted by Snow and Wittt (1996) of 1.34. This latter value is based on a compilation of stellar compositionn data of both field and cluster B and disk F and G stars. The data show that the abundancess of young stars deviate considerably from solar. The authors argue that this deviationn could reflect the enhancement of the proto-solar cloud abundances by a nearby supernovaa event (e.g. Cameron and Truran 1977; Olive and Schramm 1982). Inn the discussion above we used the results of annealing experiments of pure magnesiumm silicates. This is motivated by ISO observations which show that the crystalline silicatess contain little or no iron. It is, however, likely that the amorphous material will containn iron. If so, this could have consequences for the interpretation of the Mg/Si ratio derivedd above. An important question is how iron is incorporated into the amorphous silicate.. This is hard to determine observationally due to the amorphous structure of the silicate.. The iron could be in a soiid solution or consist of a mix of pure magnesium andd pure iron silicates. If forsterite and silica are formed through the annealing of Fe- containingg amorphous material, the iron somehow has to be removed from the lattice inn the annealing process, forming metallic iron or iron oxides depending on the oxygen partiall pressure. During this solid state reduction of iron the Mg/Si ratio does not change andd the value derived above would consequently imply a non-solar composition. Condensationn experiments (Rietmeijer et al. 1999) favor the possibility that the amorphouss material consists of a mixture of pure iron and pure magnesium silicates. Iff this is so, the derived Mg/Si ratio could only be relevant to the amorphous magnesium silicates.. Annealing experiments by Hallenbeck et al. (2000) show that the annealing timee scales for iron silicates are considerably longer than for magnesium silicates. It could welll be that while the magnesium silicates anneal and form forsterite and silica, the iron silicatess stay amorphous. Given a system where the amorphous magnesium silicates have aa Mg/Si ratio of 1.46, if the amorphous iron silicates would contain ~ 30 % of the total amountt of Si available in this system, the bulk Mg/Si ratio would be solar. We stress howeverr that the uncertainties on the derived bulk Mg/Si ratio do not allow us to decide whetherr or not it deviates from Solar.

4.4.33 Deviating objects

Fromm our sample of HAHBH stars, three objects deviate in dust composition from the otherr svstems: HD 100546 and HI) 150193, having a lower respectively larger SiO: masss fraction compared with the amount of forsterite in these systems and HD 179218 whichh is the onlv svstem that shows emission from enstatite. In addition, comet Hale- Boppp deviates in a way which strongly resembles HD 100546. We discuss these objects

// / CHAPÏTRR 4

below. .

HDD 100546 and Hale-Bopp

11 he mineralogieal composition of the silicate dust in HDl 005-46 is stronglv deviant fromm most other HAHBK stars: It has very abundant cool crystalline silicates (Malfait ett al. 1998b). Our analysis indicates that the ratio of forsterite over amorphous silicates att temperatures dominating the 10 //m region is lower than that of the cooler mate-

riall dominating at longer wavelengths. We also find a lack of 8.6 /./m SiO: emission. 11 his suggests that thermal annealing mav not have been the origin of the forsterite in HI)) 100546. AA high fraction or forsterite is also apparent in comet Hale-Bopp (see also Crovisier

eii al. 199?). As in HI) 100546, SiOj is under abundant relative to forsterite. It is tempt- ingg to speculate on the origin of this behaviour. One intriguing possibilitv is that (part of)) the grains in HI) 100546 and Hale-Bopp are secondgtmeratimh i.e. originate from largerr parent bodies - in which substantial alteration of the silicates occurred - that were destroyedd through collisions. The recent discovery of gas in the disk of P Pic (Thi et al. 2001)) shows that collisional processes can dominate in disks still containing some gas. Possiblyy second generation dust can also be produced in the disk of HD 100546. The highh fraction of forsterite in comet Hale-Bopp certainly rules out the possibilitv that itt was formed from pristine ISM dust, and hence is strong proof that comets contain materiall processed in the proto-solar nebula. This processing could either be thermal annealingg (Halley) or crystallisation by inclusion into a large parent body (Hale-Bopp). Itt a separate paper we will extensively discuss the dust composition of HD 100546 andd compare it to I lale-Bopp and other solar system comets (Bouwman er al. , in preparation). .

HDD 150193

Itt is not clear why HI) 150193 has such a high abundance of SiO:. We are confident off the identification of the 8.6 and 20 /urn emission with SiO^ given the quality of the spectrall match (I ig. 4.1). A large amount of silica can form from an amorphous silicate byy annealing if the Mg/Si ratio is lower than derived for the other systems. We can onlyy speculate on why this should be the case. We note that HD 150193 is the onlv knownn binary in our sample. A binary companion can limit the disk size and accretion andd can cause the disk to empty out at much shorter time scales (Calvet et al. 2000). Alsoo the settling and growth of grains can be prevented (Sato and Nakagawa 2000). Indeed,, HD 150193 shows the smallest grains producing the silicate feature and has a largee inner hole of about 0.6 AU (Millan-Gibet et al. 2001). This is much larger than observedd for AB Aur and HI) 163296, which have similar stellar ages (van den Ancker ett al. 1998) suggesting that indeed the protoplanetary disk in HD 150193 is influenced byy the companion star.

~8 8 GRAINN PROCESSING IN HERBIG AE/BE SYSTEMS.

HDD 179218

Fromm the ISO spectra, no conclusive evidence for the presence of enstatite in HAEBE systemss can be found except for one system, HD 179218. This is the only system in whichh large quantities of enstatite have formed. Annealing experiments of a magnesium silicatee smoke done by Rietmeijer et al. (1986) show that the initially formed forsterite andd silica react and form enstatite by the following reaction

Mg,Si044 + Si02 -> 2MgSiO, (4.2)

Itt is unclear why this reaction has not taken place in the other systems. The formationn of enstatite depends on the amount of reaction surface between the forsterite and silicaa units. The dust particles must be reasonably compact for this reaction to occur onn time scales comparable with the evolutionary time scales of the HAEBE systems. An aggregatee can be compressed during collisions if the relative velocities are close to the destructionn velocities (Dominik and Tielens 1997). These velocities, however, are only reachedd with low gas densities when the coupling between the gas and dust becomes inefficient.. We note that with L ~ 300 L.: HD 179818 stands out by about an order of magnitudee compared to the other stars in our sample. This large luminosity can influencee the disk structure through radiation pressure forces and may lead to the dispersal of thee gas at much shorter time scales compared to the other HAEBE stars. This clearing outt of material is evidenced by interferometric observations showing an inner hole of ~~ 1 AU (Millan-Gabct et al. 2001), which is larger than for any of the other systems investigatedd here for which such observations are available.

4.55 Conclusions

Thee results of this study can be summarized as follows:

1.. The observed 10 /./m silicate bands in the ISO spectra of the studied HAEBE systems cann be ordered from a profile peaking at 9.8 /jm - resembling the ISM silicate band - towardss a profile similar to cometary spectra, peaking at 11.3 fJm, We find a relation betweenn the 11.3 jJm. flux and the occurrence of a broad shoulder at ~ 8.6 urn.

2.. We can model the different 10 jjm silicate profiles using three components:

silica (SiO^),responsible for the 8-9 ,üm blue shoulder in the silicate band, forsterite, contributing at 1 1.3 jam, and amorphous olivine with two typical grain sizes of 0.1 and 2.0 fjm.

3.. We identify two main causes for the observed shift in peak position of the silicate band: :

a change in average grain size from small (0.1 ^m) to large (2 /Jm). This is the resultt of the depletion of small grains in the inner region of the disk, due to coagulationn or other effects that preferentially remove small grains.

79 9 CHAPTERR 4

a change in composition from amorphous silicate to a mixture of amorphous andd Mg-rich crystalline silicate (forsterite). This may be the result of thermal annealingg in the inner regions of the disk. Laborator)7 experiments indicate that thermall annealing produces both crystalline silicates and SiO^.

4.. The change in shape of the 10 jum silicate band in HAEBE stars is mainly due to an increasee in average grain size of the dust.

5.. The HAEBE stars, [3 Pic and the solar system comet Halley form a sequence of in- creasingg silicate crystallinity. As the degree of crystallinity increases, the abundance off SiOi also tends to increase. This is consistent with the expected changes in compositionn resulting from thermal annealing of amorphous silicates in the inner regions off the proto-planetary disk.

6.. 1 he observed relation between silica (SiC^) and forsterite abundance implies that thee composition of the amorphous silicate from which these materials are formed cann not be chemically homogeneous. A combination of smectite (4 % by mass) and serpentinee (96 %) gave a good fit result.

7.. The mineralogy of HD 100546 and comet Hale-Bopp is remarkable, with very high forsteritee abundance and no evidence for SiOi (c.f. Crovisier et al. 1997; Malfait ett al. 1998b). This composition, which deviates from that of the other objects, suggestss a different origin for the forsterite, possibly from the destruction of highly differentiatedd large parent bodies. Two other objects, HD 1 50193 and HD 179218 showw very high SiÜ2 and enstatite abundances, respectively. The origin of these de- viatingg mineralogies is not well understood.

8.. Crystallisation timescales appear to be longer than coagulation timescales.

9.. No correlation between dust composition and disk geometry can be observed.

Acknowledgements,Acknowledgements, I he authors would like to thank the referee J. Mathis, for helpful comments thatt have improved this paper, and RJ.M Rietmeijer for constructive discussions. 1 he authors wouldd like to acknowledge the financial support from NWO Pionier grant 600-78-333. AdK alsoo gratefully acknowledges support from NWO Spinoza grant 08-0 to E.P.J, van den Heuvel.

80 0 CHAPTERR 5

Thee composition of the circumstellarr dust around the Herbigg Ae stars AB Aur and HDD 163296

AstronomyAstronomy & Astrophisics 2000, 360, 213-226

J.. Bouwman, A. de Koter, M.E. van den Ancker, L.B.F.M. Waters

ABSTRACT T Wee have analysed the 2-200 jjm. ISO spectra of two bright Herbig Ae stars, ABB Aur and HD 163296, in order to derive the composition and mass over temperaturee distribution of the circumstellar dust. We find that the two stars have a similarr composition of the dust, however also with some important differences. Wee address the onset of the strong near-infrared emission between 1—4 /im. We showw that this emission can be explained with thermal emission from metallic ironn or dust species containing iron. The circumstellar dust surrounding both starss shows two distinct regimes in the mass over temperature distributions with temperaturess of ~ 10 and ~ 10" K. Modelling of the ISO spectra further re- veall a significant deviation of the derived grain sizes compared to the interstellar grainn size distribution. A population of large (~ 1 mm) grains is required to explainn the observed (sub-) millimetre fluxes of HD 163296. The presence of crystallinee silicates and the larger grain sizes in HD 163296 compared to AB Aur suggestss that the former star is a more evolved system. Amorphous silicates make outt the bulk of the dust mass in both stars. It is likely that a significant fraction off the dust is carbon-rich. We find strong evidence for a 20—30 fjm band which wee tentatively attribute to FeO.

5.11 Introduction

HAEBEE stars represent the final stage of pre-main-sequence (PMS) evolution ol intermediate-masss stars (~ 2-10 MQ). AS a consequence of the star formation processs these stars are typically surrounded by a gas and dust envelope and/or disk. They

81 1 CHAI'TI-RR 5

mayy be the precursors of young main sequence (3-Picroris and Yega-tvpe stars. These latterr systems are surrounded by circumstellar debris disks, which perhaps contain planetaryy bodies. This would imply that the environment around HAEBE stars represents an earlyy phase in the formation of planets. ISC) spectroscopy of isolated HAEBE stars has strengthenedd this link between H AFBF stars and planet formation. Malfait et al. (1998b) showedd that the ISO spectrum of the B9Y'e star HD 100546 has a high abundance of crystallinee silicates, and that the composition of the dust in this object is remarkably similarr to that seen in the solar system comet Hale-Bopp (Crovisier et al. 1997). More- over,, the Fe star HI) 142527 shows evidence for hydro-silicates, indicative of aquateous alterationn of silicates in its cold circumstellar disk (Malfait et al. 1999), which is expected onlyy to occur on the surfaces of larger bodies. II his paper is the second in a series in which we will study the circumstellar environ- mentt around the Herbig Ae/Be (HAEBE) stars AB Aurigae (AOVe+sh) and HD 163296 (All Ye). We have choosen to investigate these two stars because (i) their basic properties aree well known, (ii) they are isolated and bright, making them ideal candidates to study theirr circumstellar material, (Hi) they are nearby systems that have been spatially resolved inn CO; HD 163296 has also been resolved in the continuum at 1.3 mm, and (iv) on the wholee the two stars are very similar. The last point implies that differences in spectral energyy distribution may be linked to differences in dust composition and morphology. Thiss can provide important insights in the evolution of the circumstellar dust and may yieldd information on characteristic time scales of dust evolution (from comparison with stellarr age) and/or on the importance of properties of the natal molecular cloud, such as initiall cloud size, mass and angular momentum. Inn the first paper (van den Ancker et al. 2000, hereafter Paper I) we presented new infraredd spectra of these two well studied stars obtained with the Short- and Eong Wave- lengthh Spectrometers on board the Infrared Space Observatory (ISO; Kessler et al. 1996). Inn this paper we present quantitative spectroscopic modelling with the aim to constrain dustt properties such as composition, abundance and size- and shape-distribution. In a subsequentt paper we intend to present a detailed multi-dimensional model for the dust distributionn around these HAEBF^ systems. II he geometry of the circumstellar environment around HAEBF. stars remains a per- sistentt problem. I he key issue is that the observational evidence from spectral energy distributionss (SEDs) and spectroscopy do not define the geometry of the circumstellar dustt in a unique way (see Waters and Waelkens 1998, for a review). We will start out byy summarizing this evidence immediately focusing on AB Aur and HD 163296. The secondd important problem addressed in this paper is the nature of the onset of near-]R emission.. I his will be discussed in Sect. 1.2.

5.1.11 Geometry of the circumstellar dust

Inn view of the similarity to their less massive counterparts the 1 lauri stars, HAF.BE stars aree expected to have optically thick disks. Indeed, in the case of AB Aur high-resolution imagingg together with a de-convolution method (Marsh et al. 1995), indicates a disk off size 36 and 72 AU at 1 1.7 and 1T9 /Jin respectively, adopting a distance of 144

82 2 THKK COMPOSITION OF AB AIR AND HD 163296

II ABI.F. 5.1: Astrophysical parameters or the programme stars.

ABB Aur Ref. . HDD 163296 Ref. .

aa (2000) 044 55 45.79 (1) ) 177 56 21.26 (1) ) 88 (2000) +300 33 05.5 (1) ) -211 57 19.5 (D D dd [pc] 144:^2 2 (2) ) 11 ?^ + l (2) ) Sp.. Tvpe A0Ve+sh h (3) ) AlVe e (8) ) V[m] V[m] 7.03-7.09r r (4) ) 6.82-6.89 9 (4) ) AvAv [m] 6 6 (5) ) 0.311 6 (5) ) s 7'fff [K] 9520* "° ° (5) ) cp3Cry,H) ) (5) ) ;'-JU-i(]i) ) log£ £ 4.07 7 (5) ) 4.10 0 (5) ) 7-,, [U.] 2 2 (2) ) 8 8 (2) ) /?.. [R©] 5 5 (5) ) 5 5 (5) ) NLNL [M@] (2) ) 1 1 (2) ) log(Age)) [yr] 2 2 (2) ) 4 4 (2) ) vs'mivs'mi [km s~ ] 5 5 (6) ) 1 1 (9) ) COO disk size [AV2] 4055 x 100- (7) ) 3100 x 160 (7) ) t[°\ t[°\ 76 6 (7) ) 58 8 (7) ) RA.. n +7T_\ +7T_\(7 ) ) ++ 126^ (7) )

Note:: Older photographic measurements (Gaposchkin et al, 1952) show values up to nip,, = 8.4, Correctedd for revised distance (van den Ancker et al. 199""). References:: (1) Ferryman and ESA (1997); (2) van den Ancker et al. (1997); (3) Böhm and Catala (1 993); (4)) van den Ancker et al. (1998); (5) Paper I, but relevant parameters have been scaled such that they are conformm spectral tvpe A1 Ye and not A3Ve; (6 J Bohm and Catala (1995); ( ) Mannings and Sargent (1 997); (8)) Houkand Smith-Moore (1988); (9) Halbedel (1996).

pc.. Direct observational evidence for a disk-like geometry comes from the kinematical propertiess of °CO gas, observed in the / = 1 —» 0 pure rotational transition in AB Aurr and CO gas, observed in the/ = 2 —> 1 pure rotational transition in HD 163296 whichh seem consistent with a rotating Kepierian disk (Mannings and Sargent 1997; Manningss et al. 1997). The dimension of the CO emitting regimes has been estimated too be - 405 x 100 AU, and ~ 310 x 160 AU for AB Aur andVlD 163296 respectively. Thee aspect ratios imply that we see AB Aur almost edge-on at an inclination angle of 76°° and HD 163296 at 58°. For this last star the continuum emission at 1.3 mm has beenn spatially resolved, showing an elongated structure with dimension ~ 1 10 X 95 AU. Near-infraredd interferometric observations of AB Aur (Millan-Cabet et al. 1999) have resolvedd the inner part of the dust distribution, determining the inner edge of the dust too be at 0.3 AU (see also Sect. 1.2). An upper limit on the inclination in this inner regime iss determined to be 45", suggesting that the inner and outer regions of the disk do not

83 3 CHAPTERR 5 havee the same geometry. Thee shape of the IR spectral energy distribution has often been used as a diagnostic forr constraining the geometry of the HAEBE surroundings. In some cases, the SET) mayy provide firm constraints especially when a lack in balance is found between energy absorbedd in the UV and optical and energy re-emitted in the IR (Meeus et al. 1998). Suchh a discrepancy strongly points to a disk-like structure. One should, however, be very carefull with conclusions relating to the spatial distribution of the dust derived from the SEDD only (e.g. Henning et al. 1998; Bouwman et al. 2000b). Evidencee for the presence of a large optically thin medium comes from the strong 9.77 jum silicate emission (Cohen f 980; Sitko 1981; van den Ancker et al. 2000) observed inn both stars. If the silicates are located in a geometrically thin disk, modelling shows one expectss this disk to be optically thick at optical as well as IR wavelengths. However, it is substantiallyy more difficult (if not impossible) to reconcile the presence of this emission adoptingg an optically thick "disk-only"model compared to assuming an optically thin emittingg region. So, if a disk is expected on the basis of imaging, the silicate emission at leastt suggests the presence of an optically thin region of substantial size above the surface off this disk (e.g. Chiang and Goldreich 1997). The forbidden [OI] A. 6300 emission (Corcorann and Ray 1997), which is narrow 0 kms"1), symmetric and unshifted in ABB Aur, could be formed in such an extended surface layer or disk atmosphere, and be broadenedd by Keplerian rotation. Thee above arguments imply that the geometry of the CSM around these stars is likelyy to be complex, i.e. it cannot be explained by a "disk-only" model. Most likely onlyy the inner part of the disk is optically thick and a substantial thin region, such as an extendedd surface layer or (flared) outer disk region - or both - is present as well.

5.1.22 The onset of near-IR emission

Ann important aspect of the geometry of the CSM of HAEBE stars is the innermost regionn where the hottest dust grains are present and which dominate the near-IR SED. Thee key question here is whether one is able to understand the onset of near-IR emission (fromm ~ 2 jum) in terms of a physically realistic scenario. The remarkable uniformity of thee onset of near-IR emission in HAEBE systems (Meeus et al. 2001) suggests a similar geometryy and/or dust grain population at high (~ 1500 K) temperatures at the inner regionss of the CSM. In the following we will discuss several possible explanations for the near-IRR onset. Hillenbrandd et al. (1992) suggested a model in which the observed near-IR Hux is duee to accretion luminosity from a geometrically thin active accretion disk. The emissionn in the NIR is explained in this model with (high temperature) accreting gas in the innermostt part of the disk. Emission at longer wavelengths would be due to dust in an opticallyy thick disk. The accretion rates Hillenbrand et al. (1992) derive are 1.5- 10"(l andd 1.3- 10"f' M.Tyr ' for AB Aur and HI) 163296 respectively. Radio continuum observationss Skinner et al. (1993) point to much smaller accretion rates of 1.1 10 s and << 9.1 1 (r9 Mfv-.yr-1 for AB Aur and HI) 163296 respectively, which are inconsistent

84 4 THHH COMPOSITION OF AB AL'R ANT) HD 163296 withh the high accretion rates needed in the Hillenbrand model. Also the lack or substantiall veiling at optical wavelengths points towards small accretion rates. Böhm and Catalaa (1993) derived an upper limit for the accretion rate of AB Aur, from the veiling off photospheric lines, of < 7.5 1CT8 \40\T~1 , consistent with Skinner et al. (1993). The strongestt objection against the active accretion disk model comes from interferometric observationss of AB Aur (Millan-Gabet et al. 1999) which can not be reproduced using thiss model. AA different explanation for the near-IR emission could be Polvcyclic Aromatic Hy- drocarbonss (PAHs), of which the emission bands are located at 3.4, 6.2, 7.7, 8.6 and 11 1.3 jum. This makes PAHs a candidate for explaining the near-IR Hux. Due to their smalll size, the characteristic temperature of PAH particles can be high as they are no longerr expected to be in thermodynamic equilibrium. Quantum heating effects result inn such high temperatures that their dominant emission is in the near-IR. To account forr all observed near-IR emission however very high abundances are required (Natta ett al. 1993) which seem unlikely. Even though their narrow characteristic features are presentt in AB Aur, it is not expected that PAHs are also responsible for the broad continuumm contribution starting at ~ 2 //m. This conclusion is strengthened by the shape and strengthh of the near-IR emission in HD 163296, which is identical to AB Aur but shows noo PAH bands (van den Ancker et al. 2000). A different explanation seems required. Onee of the main points of this paper is that observed near-IR emission in the pro- grammee stars can be explained simply by dust in thermodynamic equilibrium. The dust speciess that have strong resonances in this spectral region (~ 1—8 pm) are metallic iron, ironn oxide, and carbonaceous dust grains, either graphitic or 'amorphous'. The dust grainss are heated by stellar radiation in the inner parts of the circumstellar disk up to thee dust destruction temperatures of the individual dust species. We will show that in vieww of the high grain temperatures (~ 1500 K) needed to produce the observed flux at near-IRR wavelengths, metallic iron grains are the most likely candidates to explain the onsett of the near-IR emission (see Sect. 3.1). With our model we can also reproduce the interferometricc observations by Millan-Gabet et al. (1999) of AB Aur, showing an inner holee of 0.3 AU. Thiss paper is organized as follows: in Sect. 2 we discuss the chosen approach to model thee ISO spectra and describes the model we used. Sect. 3 gives the results of the model fittingfitting and a discussion over the implications of these results is presented in Sect. 4. We summarizee our results in Sect. 5.

5.22 Method and assumptions

5.2.11 Model approach and assumptions

Inn this studv we have chosen to model the GSM of the programme stars using a sim- plee optically thin dust model. In view of the above discussion it is clear that this is a simplifiedd approach. Still, we have adopted this approach for the following reasons: (i)(i) A main focus of this paper is to identify' and model the solid-state features present inn the rich ISO spectra of AB Aur and HD 163296. For this, we employ a library of

85 5 CHAPTF.RR 5 laboratoryy measurements of the optical constants of approximateK' fifty different grain materialss of interest. From this collection we selected eight species, listed in I able 2, whichh are most likely present in Herbig Ae/Be systems. I hese species are discussed in the nextt subsection. We attempt to constrain composition, abundances, size distributions, shapee properties and mass oyer temperature distribution of the dust grains present. I his goall requires an extended parameter study, which at the present day cannot be done usingg a consistent multi-dimensional dust transfer model. (ii)(ii) I he aboye discussion on the geometry of the CSM around AB Aur and HI)) 163296 shows that part of the emission originates from an extended optically thin medium,, e.g. the broad emission complex from ~- 14 to 28 /.im and the 9.7 jivm silicate emission,, and which suggests that the optically thin model has at least partial validity. Still,, the next step should be to model these two nearby Herbig Ae systems using a 2D-model.. Any constraints on the chemical properties deriyed in this exploratory study willl then contribute significantly to the feasibility of such a complex follow-up investi- gation.. Below, we will discuss model assumptions and adopted optical properties of the dustt constituents. Inn order to model IR spectra we have developed the radiative transfer code MODUST.. This code treats the transfer of radiation through spherical density distributionss of dust grains. The (multiple) dust shells are irradiated by a central source, cither aa black body or a Kurucz model. Any prescription for the dust distribution in the shell mayy be adopted. In the present analysis, we use this model in the optically thin mode soo as to be able to perform an extensive parameter study. In the models used to fit the spectraa of AB Aur and HI) 163296 we assumed a power law shape for the density p(r) ass well as for the grain size distribution n{d), i.e.

p(r)) = po ( Y } and,

n(a)=A\n(a)=A\ — ) (5.2; "r r

Thee shell is confined between an inner and outer edge. p() is the density at the inner edge

R\R\nn of the dust shell. Cirain sizes range herween a minimum /?n-i;- and maximum size. I he constantt A in the equation for the grain size distribution is a normalization constant de- pendingg on the bulk density and size distribution of the grains. A power-law gives a good descriptionn of a grain size distribution. Observations of interstellar extinction show a size distributionn with m = 3.5 (Mathis et al. 1977). Theoretical work (Biermann and Harwit 1980)) predicts a power-law distribution whenever there is shattering and coagulation off grains through grain-grain collisions. Dust condensation models (e.g. Dominik et al. 1989)) also predict a power law distribution to be applicable over a large range of grain sizes.. We use a multi-component mixture of grain species where the grains are homo- geneouss in composition and in bulk density. Table 2 lists the grain species of interest wee used for our modelling (for a discussion see below). We used the optical constants fromm laborator}' measurements to calculate the extinction properties; for a discussion we

86 6 I'HF.. COMPOMIIOX Ol AB Al R AM) HD 163296 referr to the references given in I able 2. At present our model incorporates spherical dust grains,, for which we use Mie calculations, and a continuous distribution of ellipsoidal dustt grains, for which we use CDF calculations to determine the absorption and scat- teringg coefficients (see Bohren and Huffman 1983, for a full review on these methods), lablee 2 also lists the wavelength ranges in which laboratory measurements were available.. Outside these ranges we made the following extrapolations. At short wavelengths, i.e.. in the UY or optical, the dielectric function £ = e' + /£"" can be extrapolated by,

£'.,-11 .5.3) CO" "

e-7^^ (5.4) CO1 1

Withh 0) the frequency, C0p the plasma frequency, and y a damping factor of the electromagneticc wave. Both insulators and metals show the same behaviour at these wavelengths.. These extrapolations are reasonable for wavelengths shorter than ~ ().2fjm. At longg wavelengths (sub-millimetre and millimetre) there is a marked difference in optical propertiess between grains of different solid-state structure, composition and shape. In thiss regime, we therefore extrapolate the extinction coefficient,

Q„„ * jj; (5-5) withh U. equal to 1 for amorphous and CDE dust grains and CK = 2 for spherical metallic andd crystalline grains. Thee temperature distribution of the dust follows from the equation of radiative equilibrium.. The adopted optically thin limit implies that the dust particles absorb the full radiationn from the direction of the stellar disk only. One can easily show that if the absorptionn coefficient of the dust would behave as a power law, Qdbs ^ v^, then the resultingg temperature structure at sufficient distance from the star is also given by a power law, ( +l, withh slope I'{r) °c r~-' /' i Although in realitv Q,,hs contains resonances, the overal run off temperature is usually well represented by a power law. Therefore, in the remainder off this paper, we will characterize the temperature distribution by providing the values att the minimum and maximum radius only.

5.2.22 Adopted chemical composition of the dust

Fromm the SWS and l.WS spectra one can derive information about the mass over temperaturee distribution of the dust, and by looking at spectral features, about the chemical compositionn of the dust. However, if only spectral information is available the geometry andd properties of the circumstellar material can not be uniquely determined (e.g. Bouw- mann et al. 2000b). To determine such properties as density and grain size distribution onee has to constrain the models either bv supplementarv observations, such as imag- ine;,, or bv using theoretical arguments, e.g concerning dust destruction temperatures, crystallisationn time-scales or grain removal by radiation pressure.

87 7 CHAPTKRR 5

II ABl.h 5.2: Sources of the Optical (Constants.

Species s Solid d Wavelength h T,„ni, , Re e State e hum] ] [K] ]

[Mg,Fe]:SiOH H A A 0.2-S00 0 1400 0 (1) ) FeO O C C 0.2-500 0 1000 0 (2) ) C C G G 0.01-1000 0 1000 0 (3) ) c c A A 0.1-800 0 1000 0 (4) ) FTO O AA dee) 0.05-1000 0 150 0 (5) ) MgjSiO.» » c: : 0.04-3.4 4 1400 0 (6) ) 4-250 0 n n Fc c M M 0.1-11 (P 1500 0 (8) ) FeS S C C 0.1-10"' ' 720 0 (8) )

Abbreviationss used to designate the solid state: A = Amorphous; (.. = ('rvsialline; (i = (iraliiic; M = Metallic. References:: (1) Dorschner et al. (1993); (2) Helming et al. (1993): (3) Laor and Drame (1 993); (4) Freibisch ett al. (1993); (3) Warren (1984); (6) Scott and Dulev (1996); C7) Servoin and Piriou (19""\3); (8) Henning andd Stognienko (1996).

Onee of the main difficulties with the determination of the chemical composition off the circumstellar dust by comparison with laboratory measurements, is that solid- statee resonance bands are rather broad. With different grain species contributing at the samee wavelengths, this will result in a more or less continuous spectra without a clear spectrall signature that can be uniquely attributed to a single grain species. Exceptions too this are for instance the crystalline silicates such as forsterite (MgjSiO.J and enstatite (\lgSiO3)) which have strong and narrow resonances in the IR, and the vibrational modes inn disordered (amorphous) silicate grains around 1 Ofjm. To determine which dust species contributee at parrs of the spectra where no clear spectral feature is present, additional constraintss need to be imposed (e.g. abundance constraints, condensation temperature off the individual dust species), hiking the results of dust nucleation models, one can estimatee which dust species are present. Wee assume that all the Mg, He, and Si is incorporated into dust grains. FMdence forr this comes from abundance studies of the ISM, which show a correlation between heavyy element depletion and density in the ISM (O'Donnell and Mathis 1997). This pointss to a very efficient mechanism for accretion of these species in dust grains. Studies off these elements in the gas phase also show abundances less than solar. This could bee a result of either very hard}' dust grains or could indicate that the elements in the ISMM have abundances less than solar. The observed near solar gas-phase abundance of SS points to a solar composition of the ISM (Fit/patrick and Spit/er 199"7), suggesting Mg,, Fe and Si are in grains which are very difficult to destroy. If so, the ratio of iron pluss magnesium to silicon atoms in the dust (3.4:1) is greater than the maximum ratio

88 8 THEE COMPOSITION OF AB AUR AND HD 163296

15 5

-10 0

15 5

Figuree 5.1: Model fit to the spectral energy distribution (SED) of AB Aur (top) and HD 163296 (bottom).. The solid dots indicate ground-based and IRAS photometry corrected for extinction. Alsoo plotted are the ISO-SWS spectra, and for AB Aur the ISO-LWS spectrum. The dashed lines representt Kurucz models for the stellar photospheres. The dotted lines show the contribution fromm the hot dust component and the dot-dashed lines the contribution from the cold dust component.. Indicated in the figure are the effective temperature of the adopted Kurucz model andd the line-of-sitrht visual extinction at maximum brightness.

89 9 CHAPTERR 5

HDD 163296 ABB Aur Component: : Hot t Cold d Hot t Cold d

Radius s 1-1 1 00 AU 30-55 1 1 AL' 1- 111 1 AL' 28-1755 AL p(r) ) «« ,-i «,--' ' == ' oc,-"1 1 pnn fgr cm"1] 1 15-1 00 ls 2.4-- urUl 1. 3-10-. |s s 4.00 10"1" l 1 Ntju^/M-: : i i .9-- ur ' 2.0-- H)- 2 266 1 ir'' 3.3-- }(r^ TT range 15 00-205 00 K 300-200 K 11 500-200 K 400-500 K 1A lA n(a) ) ««

Dustt species: M.V„-- aa ijt/m] M|,,kk a [jt/m] Mfb, , aa [//m] MtrjL.. a [/im] Olivine e 0.73 3 0.01-6.0 0 0.766 0.1-1260 0.71 1 0.01-5.0 0 O.744 0.01-126 Carbon n 0.13 3 0.01-2.0 0 0.1T** 0.1-200 0.16 6 0.01-2.0 0 0.111 0.01-32.0 Waterr Ice -- -- 0.077 0.1-25.0 -- -- 0.155 0.1-40.0 Iron n 0.03 3 0.01-0.08 8 -- 0.06 6 0.01-0.1 1 -- Ironn Oxide 0.04 4 -- 33 10-"1 0.07 7 -- ' ' F'orsterire e 0.07 7 0.01-2.0 0 22 - 10-3 0.1-5.0 ------

TARLHH 5.3: Model fit parameters of AB Aur and HD 163296. Listed are the parameters defining thee density and grain size distribution, the chemical composition and the mass fraction Mfrat of thee individual dust species in both the hot as well as the cold dust component. Note that the given valuee for p0 = p(R\n), i.e. the density at the inner radius of the shell, assumes that all dust species existt at Rm irrespective of the condensation temperature. In reality, we only have a dust species presentt if its temperature is below Tconj (see 'Fable 2). This implies that for the hot component, thee density at the inner radius is slightly modified, i.e. it is somewhat lower.

thatt can be accounted for by silicate grains alone (2:1). This result seems to imply that a substantiall fraction of the iron or magnesium is in a grain population other than silicates (Fitzparrickk 1997). Theoretical work on dust condensation (Gail and Sedlmayr 1999) showss that metallic iron could be one of the primary condensates in outflows around M-typee giants. Most of the Mg will be incorporated into silicate. Iron can enter the silicatee at a lower temperature forming iron rich amorphous silicates (Gail 1998). In the ISM,, metallic iron can be oxidized to FeO at temperatures below 400 K on a time scales A)xx > U/10nm)108/«nyr (Jones 1990). 'Faking a typical density of n\\ ~ 100 cm-3 in diffusee clouds, small iron grains (0.01-0.1/jm) will be fully oxidized in 106—1 0 year. In vieww of typical lifetimes of diffuse clouds (10 -108 yr), one may expect that all small Fee grains are oxidized. This also seems to be confirmed by mass spectrometer data on Halleyy and chemical analysis of interplanetary dust particles (IDPs; Bradley et al. 1992; Schulzee et al. 1997). These studies show that most Mg is confined to silicates but that Fee is also incorporated in different materials such as metals, oxides and sulphides. Thee nature of the carbonaceous dust is not clear. The only direct observational evidencee for the presence of carbon around the Herbig Ae stars comes from the observed PAHH features around AB Aur and from the presence of carbon ions in their winds and in

90 0 ["HF.. COMPOSITION OI-' AB Al'R AND HD 163296 in-fallingg gas. T he carbonaceous dust in the ISM is modelled with a variety of materials (lorr a overview see Henning 1996). Most models include graphite and a form or disorderedd carbon. Since one is looking at (reprocessed) interstellar dust one can expect these materialss also to be present around young stellar objects (YSOs).

5.33 Results

Inn this section we present the results for the composition and abundance study of the grainn population in AH Aur and HD 163296. We have split this presentation in a discussionn of the near IR properties on the one hand, and of the far IR properties on the otherr hand. However, we start out with a discussion of those properties that we found to bee similar for both stars. Inn big. 5.1 we present the SWS & LWS spectra together with ground-based and IRASS photometry (see Paper I). Overplotted in these figures are the current best model fits.fits. I he parameters of these model fits are listed in I able 5.3. The most striking feature inn these spectra appear to be the presence of two distinct dust components, a relatively hott dust component with a mass averaged temperature of ~1000 K and a relatively cold dustt component with a mass averaged temperature of ^100 K. This bi-modal temperaturee structure in the circumstellar dust surrounding both Herbig Ae stars is best seen in thee spectrum of HD 163296, where for wavelengths shorter than 40 f.im the hot dust componentt dominates the spectrum, while for longer wavelengths the Wien tail of the coldd dust component starts to dominate the SED. In a similar way the cold dust in ABB Aur starts to dominate the spectrum longward of 30 jam. The total dust mass derived fromm the model fits in the cold dust component is ~ 3 10-^ and ~ 2- 10~4 Mg for ABB Aur and HD 163296 respectively. The mass ratio of the hot over the cold dust componentt is ~ 10 '-10 \ so by far most of the mass is contained in the cold component. Inn view of the uncertainties in the modelling discussed in the next sections the derived massess are uncertain by a factor of two to three. A blow-up covering the ISC) SWS and LWSS wavelength region for both stars is shown in Fig. 5.2. Again we include the best modell fits but now we have indicated the relative contribution of all individual dust species. . II he features in the observed spectra of AB Aur and HD 163296 have been described inn detail in Paper I. 1 herefore we only briefly discuss the most prominent ones relevant forr the model fitting. Both stars show a rather flat continuum up to ~ 8 /./m, similar too the observed near-IR fluxes of other HAEBE stars. A prominent broad amorphous silicatee band is present around ~ 10 /Jm due to the Si-O stretching mode. HD 163296 clearlyy shows the presence of a broad emission complex between ~ 20-35 jjm in which olivinee bending modes have an important contribution. In AB Aur this complex does nott appear as pronounced, because the cool component already starts to dominate the spectrumm longwards of ~ 30 /Jm. The main difference in spectral signattire between thee stars is that AB Aur shows emission bands at 6,2, 7.7, 8.6 and 11.2 /jm, usually ascribedd to polycyclic aromatic hydrocarbons (PAH) while HD 163296 does not show thesee bands. Conversely, the spectrum of HD 1 63296 reveals crystalline silicates at 1 1.3,

91 1 CHAPTERR 5

-- ABB Aur 100 0

50 0 -- jS* jS* V V -e ' ^ -- \\

%% —.. », V

.. H20 11 ' IVO O -- Fe e , , "-50 0

// ^\ IIÜÜ 163296 1 1 * * 20 0 1 1 1 // \ \ \ /P ^ ^ // ' w 11 .J*——feii aa v-V 1 1 01 1 +^*~ +^*~ -- - - ' c '

---._ H20

-20 0 FeO O

- - 1 1 10 0 20 0 50 0 100 0 200 0

AA [//in]

Figuree 5.2: Model tit to the ISO spectra of AB Aur and HD 163296. The solid dots indicate groundd based and IRAS photometry-. The dotted lines represent the contributions to the continuumm Hux of the individual dust components. The curves are offset lor clarity. The squares indicatee the zero flux level for each component.

92 2 THEE COMPOSITION OF AB AUR AND HD 163296

ii i i i i i i i i i 22 4 6 8 AA | / n n

Figuree 5.3: Continuum dust emission from dust species contributing to the NIR between 2 andd 8 /./m for three different models. The solid line indicates a model consisting of amorphous carbon,, metallic iron and iron oxide, with mass fractions of 0.55, 0.2 and 0.25 respectively. The dashedd line indicates a model with only amorphous carbon and metallic iron with mass fractions off 0.53 and 0.47 respectively. The dotted line indicates a model where the entire NIR emission iss due to graphite.

16.3,, 17.8, 23.5, 31.3 and 33.5/Jm, while AB Aur does not appear to have any crystalline material. .

5.3.11 The 2-8 um spectral region

Wee first concentrate on modelling the flat near-IR continuum from 2-8 /vm. The presentedd results are essentially equally valid for both AB Aur and HD 163296. We iden- tifiedd several dust species that can reproduce the observed continuum flux. These are metallicc iron, iron oxide and a carbonaceous dust component, either graphite or a more "amorphous"" form of carbon dust. Due to the lack of clear spectral signatures in this wavelengthh region, the possibility of confusion of the relative contribution of individual dustt species exists. Ass an example Fig. 5.3 shows three different models for the near-IR flux of AB Aur which aree spectroscopically indistinguishable from one another. I he chemical composition of thee three solutions is given in the figure caption. It is clear, one has to introduce sec- ondaryy arguments to distinguish between these models. The highest grain temperature

93 3 CHAPTERR 5

requiredd to reproduce the observed Mux at the shortest wavelengths is ~~1500 K, which exceedss the grain evaporation temperature or iron oxide, but is at the condensation temperaturee of metallic iron and carbonaceous grains (e.g. Cail 1998). However, at about —— 1000 K, carbon grains become subject to chemical sputtering (oxidation) and will be destroyedd before reaching the condensation temperature (Duschl et al. 1996; 1 inocchi ett al. 199T I his also applies to amorphous carbon grains (AMC), which cannot re- producee the observed Mux between 2 and 4 //m even when heated up to 1500 K. The onlyy dtist species that can reproduce the observed Mux at these short wavelengths with an acquiredd grain temperature tip to 1500 K is metallic iron. A satisfactory Ht to this part off the spectrum could only be achieved when the temperature distribution of the iron grainss ranged between ~ 700-1500 K. When more grains at lower temperature where included,, we were nor able to reproduce the Hat near-IR spectrum. The continutim longwardss of 4 /7m can be reproduced bv thermal emission of iron oxide and carbonaceous grainss heated to their grain destruction temperatures. Iron oxide however does not easilv reachh temperatures close to its condensation temperature as above — 400 K it is slowlv convertedd into solid iron (H.-P. Cail priv. comm.). In Fig. 5.2 we have indicated the contributionn of each individual dust species to the total Hux. This clearlv shows that metallic ironn is the chemical species responsible for the 2-4 /./in Huxes. For the dust at 7' < 1000 KK we have used amorphous carbon in our model fits, however graphite cannot be ruled outt on spectroscopic grounds. Forr the model ht to AB Aur, additional constraints on the dust properties near the innerr boundary can be obtained using recent interferometric observations (Millan-Cabet ett al. 1999). Fig. 5.4 shows the visibility curves of three models to the NIR spectrum of thee star. I hese visibilities essentially probe the innermost region of the disk, where the emissionn is dominated by the metallic iron grains. "The solid line gives the best Ht model, inn which Fe grains in sizes from 0.01 to 0.1 //m are tised. Note that this model does notnot reproduce the observed visibilities in the K- and Fl-band. Also shown in the figure- aree the resulting visibility curves using models containing onlv single sized Fe grains off 0.1 and 0.32 //m, respectively. Note that both these models vield identical Hts to thee observed spectrum as does the best fit model. The 0.32 //m grain model fits the visibilityy best. The associated inner radius of the iron dust at 0.35 AU is consistent with thatt found by Millan-Cabet et al. (1999). This result is therefore suggestive that grain evaporationn plays an important role in the inner part of the proto-planetarv disk, where onlyy the largest grains, having the lowest temperature, survive to a distance of ~ 0.3 AU off the central star. However, the latter model would lead to a discontinuitv in the spatial distributionn with respect to the other dust species. 1 his is because our current models do nott consistently incorporate the destruction of dust grains.

5.3.22 The 8-30 jjm region.

Substantiall contribution to the Hux around 10 //m in both I lerhig Ae stars comes from disorderedd (amorphous) silicates (see Fig. 5.2). The width of the silicate feature is such thatt it cannot be fitted with a single-size dust distribution, which would vield an emissionn feature which is too narrow. A size distribution is required in which the smallest

94 4 THEE COMPOSITION OF AB AUR AND HD 163296

LOO O K K

RR I .6AU, a O.Olum 1111 ill nun ' __ _ R 0 6AI a 0 1//1H nunn ' :: 35A1 a 0 32um mmm '

* *

: :

100 0

i() i() id d to o 50 0

BB | MA

Figuree 5.4: Visibility curves of three model Hts of the metallic iron component to the NIR spectrumm of AB Aur as a function of baseline length. The black squares indicate the observational dataa with the error bars indicating the rms error. The top panel shows the visibility in the K—band, andd the bottom panel the H-band. The solid line indicates the current best ft, while the dashed andd dot-dashed line indicate models with a single grain size of 0.1 and 0.32 ftm respectively.

95 5 CHAPTFRR 5 grainss contribute most to the short wavelength side of the 10 /am silicate feature, while thee largest grains contribute to the long wavelength rail. Though a clear spectral sig- naturee for the silicates exists and confusion with emission from other species is not a problem,, a different type of degeneracy exists in the modelling. For a fixed grain compositionn the shape and strength of the SEDs is determined by the mass over temperature distributionn and the emissivity of the dust. Fig. 5.5a shows three model fits for the silicate emissionn around AB Aur for different grain size distributions all having the same mass overr temperature distribution. The value given in 'Fable 3 (and represented bv the solid linee in Fig. 5.5a) for the power of the grain size distribution is the mean value between thee limiting cases (dashed and dashed-dotted lines) for which a model fit could be made. II hough an uncertainty in the exact grain size distribution remains, one can however con- cludee that the required distributions deviate substantially from the ISM size distribution

(m=3.5,, /7min — 0.025 and amM 2r 0.25^m). Substantial grain growth/coagulation must havee taken place to produce the required grain size distributions. The maximum grain sizee required to fit the 10 fjm silicate feature (5 /Jin in our best fit model) depends, to a lesserr extent, also on the strength of the underlying continuum at the long wavelength taill of this feature. Assuming the carbonaceous grains to be graphitic results in a slightly higherr continuum contribution at these wavelengths, resulting in a smaller maximum sizee for the silicate dust by a facror of two, compared to the assumption of an amorphous carbonaceouss dust component. Betweenn 14 and 20 jum the spectra are characterized by a strong rise in flux. This rise cann be attributed to O-Si-O bending modes in the amorphous silicates around 18 ^m. However,, between 20 ji/m and the wavelengths where the cold dust starts to dominate the spectra,, an additional solid-state emission component, apart from the silicate, is required too explain the observed fluxes. This broad emission complex is most clearly seen in HD 163296.. Apart from the broad resonances in the near-IR, which may contribute to the observedd flux in this region, iron oxide has a strong spectral feature between 21 and 25 /jm,, depending on grain shape (Begemann et al. 1995). Only by including non-spherical FeOO grains a satisfactory model fit could be obtained. Because we calculated the optical propertiess of FeO entirely in the Rayleigh limit, no information about grain sizes could bee obtained for this species (see also Sect. 2). Howw well constrained are the results listed in Table 5.3 for rhe hot dust component?? Apart from degeneracies in the model fits due to a lack of clear spectral signatures andd the uncertainties in the grain size distribution, both discussed above, another de- generacyy may be identified (Bouwman et al. 2000b). As an example for this degeneracy, Fig.. 5.5b shows two model fits to the silicate component in AB Aur, both having the samee mass averaged temperature, but with different density distributions, resulting in identicall spectra. This example is equally valid for the other dust species around AB Aur ass well as around HO 163296. As discussed in the previous subsection, to explain the observedd NIR fluxes the dust grains have to be heated to their destruction temperatures, whichh determines the inner radius of the dust distribution of the individual dust species. Withh the assumption that all dust species are well mixed and that, therefore, the only spatiall separation between species is caused by differences in grain destruction temperatures,, resulting in different inner radii, the density distributions listed in Ttble 3 are the

96 6 THEE COMPOSITION OF AB AIR AND HI) 163296

JU U

a) ) 00 01 ... b)) ___ 160 [000.. p- 1

_ _ .. m 21 __ p=l ' ' -- A A 5 5 (\ \ 20 0

10 0 . . \ \ -- '' \

W/ W/ NN \ : : \ \ \: :

.. jj 0 0

1 1 3C C

Figuree 5.5: Degeneracies in the model fits. Panel a) shows three model fits to the 10 fim silicate featuree of AB Aur for different grain size distributions. Indicated in the figure are the minimum andd maximum grain size and power m (see Eq. 2). Panel b) shows the degeneracies in the dust densityy distribution. The density at the inner edge is 7.15 10~'1 (solid line) and 2.06- 10" gr CITT33 (dashed line) respectively for the two models. Indicated in the figure are the inner and outer radiii (in stellar radii) between which the dust shells are confined and the power of the density distribution/;; (see Eq. 1). The solid lines in both panels represent the best model fit to the silicate componentt in AB Aur (see also Table 3).

tostt likely ones.

5.3.33 The cold dust component

Fig.. 5.1 and 5.2 also show the ISO-LWS spectrum of AB Aur, the one star for which wee also have LWS observations. For wavelengths exceeding the LWS wavelength range forr AB Aur and for wavelengths exceeding the ISO-SWS wavelengths for HD 163296, wee only have photometric points. This means that far less information about the chemicall composition can be derived at these wavelengths than from the ISO-SWS spectra, alsoo because the LWS spectrum does not show any discrete spectral signatures. To get aroundd this problem we have taken the chemical composition of the hot dust component,, derived from the SWS spectra, and used the same material to model the cold dust component.. The main difference is the inclusion of water ice as component of the cold dust.. We further assume that all iron is locked up in amorphous silicate grains and, as

97 7 CHAl'THRR 5

aa minor constituent, in iron oxide. This is suggested bv the temperature distribution of thee metallic iron in the hot dust component which, having a minimum temperature of ~-- ^00 K, suggests a presence of metallic iron only at the inner edge of the hot dust component.. Also the sharp rise in Hux levels between 40 and 60 /urn points to the absence orr metallic iron in the cold dust, as including this dust species would tend to flatten out thee Shi) in this region, in contradiction with the observations. Evidence for crystalline hh()) ice was round in the Herbig star HD 14252^ (Malfait et al. 1999), and is expected too be present in the outer parts of proto-planetary disks. The inclusion of HiO ice sig- nificantlyy improves the quality of the fit in the 40-80 /.tm wavelength range. Only by addingg hhO ice as a dust constituent the correct slope could be reproduced as a result of thee strong 60 jam feature of water ice. label 3 lists the model parameters of the best fir too the cold dust component. The listed relative abundances are, due to confusion of the relativee contribution of the individual dust species, uncertain bv a factor of two. Inn the current model a gap of ~ 20 AL' is required between the hot and the cold dustt component in order to create the required bi-model temperature structure. We will comee back to this point in the next section. II he spectral energy distribution at sub-mm to radio wavelengths is determined bv thee temperature and density distribution of the cold grains, as well as bv their (average) grainn size. (Sub)-mm and radio imaging constrain the location of these cold grains to distancess less than several 100 AC. Using these constraints, a population of large (~ 1 andd ~ 0.1 mm for HD 163296 and AB Aur respectively) grains is required to reproduce thee observed shape of the SED from sub-mm to radio wavelengths. large grains (> 1 mm)) radiate as black bodies at mm wavelengths, and the Rayleigh-Jeans part of the SED off HD 163296 can naturaly be explained by such grains emitting at temperatures > 20 K.. 1 he spectral shape of AB Aur at these long wavelengths is significantly steeper than thatt of a Rayleigh-Jeans tail (of a black body), hence smaller grains dominate the SED. Ourr model fit gives a maximum grain size of ~ 0.1 mm, a factor often smaller comparedd to HD 163296. As stated above, these results are derived by constraining the dustt distribution with the available image data. However, relating the predicted mass overr temperature distribution to the exact spatial distribution imposed by the imaging dependss on the correctness of the model assumptions, i.e. on whether rhe assumption thatt the medium is optically thin is correct. Shielding of direct stellar radiation bv an opticallyy thick dust distribution will modify the temperature-distance relation of the grains,, and consequently will change the spatial distribution of the dust considerably comparedd to an optically thin mode! with the same mass over temperature distribution (seee Sect. 6 for further discussion). The question therefore is: can one relate the current modell directly to the image data? Adopting the same grain sizes as derived for the hot dustt component, i.e. assuming a maximum grain size of a tew micron, the cold dust of ABB Aur can only be fitted with our optically thin model if the grains are about a factor off 10 further out than imposed by the image data. If considerable optical depth effects aree present in rhe dust around AB Aur the dust could be moved inwards, having the samee temperatures as in rhe optically thin model, thus complying with the observations withoutt the need for large grains. However, in the case of HI) 163296 we want to point outt that it is not possible to Ht the observed slope of the SED at mm wavelengths usin»

98 8 THH I- COMPOSITION OF AB AUR AND HI) 163296

\BB Aur

10 0

PAHH bands -10 0

HH 1 1 1 1 1 1 1 | 1 1 1 1 1 1- II I I I

( 10 0 II ID l(>:i:i )(>

i i

forsteritee bands -10 0

10 0 20 0 30 0 4C C

AA [//m |

Figuree 5.6: Residual spectra of AB Aur (top) and HD 163296 ISO SWS spectra subtracted by thee current best model fit. Indicated in top panel with the thick vertical lines are the positions of thee main features attributed to polycyclic aromatic hydrocarbons (PAH) bands at 3.3, 6.2, 7.7 8.66 and 1 1.3 fim for AB Aur; and in the lower panel are indicated the main features of forsterite att 1 1.3, 16.3, 18.7, 23.4, 27.5 and 33.6 pm for HD 163296. The dotted line represent the best modell fit for forsterite. ourr optically thin model and an identical size distribution as derived for the hot dust component.. As optical depth effects primarily affect the location of the dust and not thee slope of the far-IR spectrum, which in both an optically thin and thick medium is

99 9 CHAPTERR 5

Figuree 5.7: Cumulative dust mass over temperature distribution. The left panels show the distributionn for AB Aur, and the right panels that for HD 163296. The top panels show the hot dustt component and the bottom panels the cold dust distribution. Indicated in the figures are thee relative contributions of the individual dust species. The inset figures in the bottom panels showw the warmest dust or the cold dust component, being dominated by the iron oxide grains. Notee that the scale of the y-axis differs for the bottom panels, which is in units of 10~4M.T- for bothh panels. essentiallyy transparent for locally emitted radiation, this result will also be valid for an opticallyy thick model. Therefore, even if optical depth effects play a role in HD 163296, thee grains needed to explain the sub-mm to radio wavelength part of the SED need to bee of the same size (~ 1 mm) as predicted in our optically thin model.

5.3.44 PAHs and crystalline silicates

Fig.. 5.6 shows the residue of the ISO-SWS spectra subtracted by the best fit dust continuumm for AB Aur and HD 163296. Clearly visible are the PAH bands in AB Aur and thee forsterite bands in HD 163296, where we have over-plotted the model fit to this component.. A forsterite component has to be present in both the hot and in the cold dustt component to reproduce all observed features. The presence of this cold crystalline materiall poses some interesting problems. While in the hot dust component the temper-

100 0 THHH COMPOSITION OF AB AIR AND HI) 163296 aturess are sufficient to have thermal annealing at sufficiently short time scales compared too the dynamic time scales of these systems to produce crystalline silicates (e.g. Gail 1998),, this is not the case for the cold dust component. This would suggest that either extensivee mixing between the cold and the hot regime has taken place, or that crystal- lizationn at low temperatures can take place by means of a non-thermal annealing process off the amorphous silicates (Molster et al. 1999b). Another interesting problem is the differencee in grain size between the forsterite and the amorphous olivine (see Tabic 3) forr our best model fit. The requirement that the forsterite in the cold dust component hass to reproduce the emission feature at 33.6 /jm limits the maximum grain size to ~ 5 jt/m.. This is much smaller than the average size of the amorphous grains (~ 1 mm) in the coldd component of HD 163296. Adopting similar large grain sizes for the crystalline silicatess in the cold component would not produce any observational emission band at 33.6 /vinn due to the large grain size compared to the wavelength. This does not imply, however,, that large crystalline grains can be excluded. Large (> 100 jjm) crystalline grains effectivelyy have the same spectroscopic behaviour as amorphous grains and are thus spectroscopicallyy indistinguishable from amorphous grains. The above does however imply thatt there is no continuous grain size distribution from small (~ 0.1 pm ) to large (~ 11 mm) grains for the torsterite. A bi-modei grain size distribution remains a possibility, wheree the smaller grains (say up to 5 /Jm ) could reproduce the flux between 40 and 100 /jm/jm and a population of large grains would be responsible for the (sub)millimetre fluxes. Thee modelling of the PAH bands observed in AB Aur is beyond the scope of this paper.. The 6.2 /jm band caused by the C-C strech of the bonds in a benzene ring (Schutte ett al. 1993) is at 6.25 ^m in AB Aur, while in most other objects it is close to 6.22 /jm. Furthermore,, the absence of the 3.3 /Jm PAH feature suggests processing of the PAH moleculess (Peeters et al. in prep).

5.44 Discussion

Ann important result of the presented study is that the circumstellar material around both ABB Aur and HD 163296 is characterized by a bi-modal mass over temperature distribution.. "Phis dichotomy is presented in Fig. 5.7. The left panels show the distribution for ABB Aur; the right panels that for HD 163296. Upper panels show the hot component; lowerr panels the cool component. The choice of the vertical axis unit ensures that the integrall of each shaded region (representing different species) reflects the total mass per component.. Note the large difference in scaling of the vertical axis for the hot, respec- tivelyy cold dust. The inset panel shows the warmest dust in the cold component in the samee scaling as the hot component. 11 he presented mass over temperature distribution reflects the dust properties responsiblee for the observed SED and holds irrespective of the assumed model geometry and/or associatedd optical depth. In our optically thin model, the dichotomy in the temperature distributionn can only be explained if we assume a physical gap of ~ 20 AU between the hott and cold component. These two separate dust shells are really necessary, as even a bi-modall grain size distribution, consisting of small hot grains and larger cooler grains

101 1 CHAPIT.RR 5 cannott solve this problem. Ass discussed in Sect. 1. imaging shows a disk like structure around AB Aur and HI)) 163296. 1 he bi-modal temperature distribution mav therefore perhaps be naturally explainedd with the cold dust distributed in an optically thick disk and the hot dust in ann optically thin surface laver or extended atmosphere on top of this disk. In this model, aa discontinuity in the spatial dust distribution mav not be needed. big. 5.8 shows a schematicc outline of such a disk geometry. I he dense, optically thick inner regions of thee disk shield material located somewhat farther out from direct illumination by stellar radiation.. In a dust distribution which is optically thick both to locally and non-locallv emittedd radiation, the temperature as a function of radial distance will drop as 1 °c R 1,"t. Inn an optically thin medium the temperature is expected to drop as I °= A* "~ "^ (assuming thee dust opacity is proportional to A. ). I his means that in the optically thick case, a givenn temperature is reached closer to the star. lig. 5.8 also shows the schematic run of temperaturee for two radial paths through the proro-planetarv disk: path A represents a beamm for which the optical depth I is less than unity, therefore / is relatively high; while pathh B represents a beam for which T > 1, therefore the temperature is relatively low. II he difference in optical depth between the hot and cold component seems consistent withh the difference in mass in both these components, i.e. the hot dust contains of the orderr of ~- 10"1 —lO"1 less mass than does the cold dtist. 11 he above model could in principle reproduce the mass over temperature distributionn of the cold dust closer to the star compared to our optically thin model and thus couldd bridge the gap between the hot and cold component, resulting in a more con- tinuouss dust distribution. Indeed, more detailed modeling of proto-planctarv disks (e.g. Chiangg and Coldreich 199"7; Menshchikov and Henning 199"7) shows that the required bi-modall temperature distribution can occur in an optically thick disk. Still,, the possibility that a hot and cold dust component are phvsicallv separated can nott be excluded. I he presence of a large mass, such as a proto-planet, clearing a part off the disk could be an explanation for such a distribution. I his is also suggested by aa previous srudv of the shape of the energy distribution of a sample of HhABT. stars (Maifaitt et al. 1998a). Iwidence for larger bodies around Herbig Ac stars is suggested by observationss of infalling circumstellar material. I he observed velocities of the infalling gass seem consistent with infalling and evaporating larger bodies (dradv et al. 1999). Thee chemical composition of the dust closest to the central star is determined by thee dust destruction temperatures of the individual dust species. As can be seen from big.. 5.7, the mass at the highest temperatures is dominated by metallic iron. At lower temperaturess the iron is not in the form of metallic iron but is incorporated in iron oxide andd silicates. I his is suggestive of chemical processing of the dust. I he dust, being slowly accreted,, is heated up thereby transforming iron oxide to metallic iron and releasing the ironn hv solid state reduction from the olivine. Several models for the chemical evolution off proio-planetarv disks predict the presence of trolite (bcS) (e.g. ( iail 1 998; Pollack et al. 1994).. I his component is also found in meteorites and interplanetary dust particles (II)Ps).. However, using the optical constants measured hv (Henning and Stognienko 1996),, which show a clear spectral signatures between M) and . no evidence could

102 2 THEE COMPOSITION OF AB AIR AND HI) 163296

logR R

Figuree 5.8: Schematic representation of a proto-planetary disk. The dark area represent the part off the disk where optical depths in excess of one are reached. Area I is optically thin for all paths emanatingg from the central star and can be directly illuminated. Area II is shielded from direct illuminationn by the inner narr of the disk. The inset figure gives a schematic temperature profile forr two beams one through the disk along path B and missing the disk along path A.

bee found for the presence of this species. II he requirement of having dust grains at temperatures up to their destruction or condensationn temperatures poses an interesting problem lor the silicates. Silicate grains withh a temperature in excess ol ~ 800 K, (much lower than the condensation temperature),, can crystallize at time scales shorter than the age of the Herbig Ae stars (Gail 1998).. The relatively small fraction of forsterite compared to that of amorphous olivine inn HI) 163296, and the complete absence of crystalline silicates in AB Aur seems in contradictionn with these crystallization time scales. AA solution to this apparent problem is that we are not looking at a static mass distribution,, but have a situation where the hot dust component is removed (by infall) andd replenished with unprocessed material. The difference in spectral signature between ABB Aur, where no crystalline silicates are detected and HD 163296 where a mass fraction off ~ 0.1 in the hot dust component is in crystalline form, could be explained in terms off different rates of removal of the high temperature silicate grains. Thee amount of forsterite that could be added to the hot dust component of AB Aur withoutt causing a spectral signature is ~ 10~" MQ. TO comply with observations, thiss amount is ~ 10 iMQfor HD 163296. The total mass of silicate grains with aa temperature above the crystallisation temperature is — 10~10 Mo for both AB Aur andd HD 163296. From a comparison of these masses with observed accretion rates off 1.1 x 10~10 and < 9.1 x 10"'1 MQyr1 (assuming a dust to gass ratio of 0.01) for ABB Aur and HD 163296 respectively (Skinner et al. 1993), one can conclude that it is

103 3 CHAPTERR 5 possiblee to accrete the hot amorphous silicate grains at a short enough time scale before rhevv give a significant spectral signature. Thee presence or crystalline silicates in the cold dust around HI) 163296 could be explainedd by vertical mixing in a disk between the hot, optically thin surface laver and thee optically thick inner part, though non-thermal annealing of amorphous silicates is alsoo a possibility (Molster et al. 1999b).

5.55 Summary

1.. We have modelled the energy distribution of AB Aur and HI) 163296 using an opticallyy thin dust model, and have derived the chemical composition and size distributionn of the dust components contributing to the spectrum.

2.. The SKD of the Herbig Ae stars AB Aur and HI) 163296 is characterized by a bi-modall mass over temperature distribution. A low mass (~ 10" Mc ) dtist componentt with temperatures of ~ 1000 K and a cold component (~ 100 K) which containss most of the mass (~ 10 ^ to 10 ^ Mc) is required to fit the ShD.

3.. The onset of the near-IR emission at 2 f/m can be explained with emission from metallicc iron grains.

4.. In both stars substantial grain growth, compared to interstellar dust, has taken place, resultingg in grain sizes up to ~ 0.1 mm around AB Aur and up to ~ 1 mm around HI)) 163296.

5.. The grain growth around HD 163296 has been more substantial than around ABB Aur. Together with the presence of crystalline silicates in the former star, this suggestss that HD 163296 is a more evolved system than AB Aur.

Acknowledgements.Acknowledgements. I he authors would like to acknowledge the financial support from NVC'O PionierPionier grant 6()0-~"8-333 and from \'\VO/\TRA grant \s 1 -"6-01 S. AdK also gratefully ac- knowledgess support from N\X'() Spinoza grant 08-0 to \\\\\. van den Heuvel.

104 4 CHAPTERR 6

Thee formation of crystalline silicate dust:: From HD100546 to Hale-Bopp p toto be submitted

J.. Bouwman, A. de Koter, C. Dominik, L.B.F.M. Waters

ABSTRACT T Wee have investigated the spatial distribution and the properties and chemical compositionn of the dust orbiting the system HD 100546. The infrared spectrumm of this star is very similar to that of C/1995 Ol Hale-Bopp (Malfait et al. 1998b).. Using an identical methodology, we have therefore also studied this solarr system comet. In both systems amorphous silicates and carbonaceous grains makee up the bulk of the dust material. Compared to other HAEBE systems aa large fraction of the crystalline silicate forsterite is present. Also present are waterr ice, metallic iron and iron oxide. The main difference in grain propertiess between the two systems is that in Hale-Bopp the chemical components of thee dust are in thermal contact with each other, while this is not the case in HDD 100546. Wee find that the mass fraction of crystalline silicates in HD 100546 increases withh decreasing temperature, i.e. with larger radial distances from the central star.. This distribution of crystalline dust is inconsistent with radial mixing modelss where the crystalline silicates are formed by thermal annealing above the glass temperaturee in the very inner parts of the disk, and are subsequently transported outwardss and mixed with amorphous material. Wee propose that the crystalline dust is produced by collisional destruction of differentiated,, at least comet sized objects and that these collisions are probably inducedd by gravitational interaction with a proto-Jupiter. Two additional argu- mentss may support this scenario. First, in HD 100546 a component of small (<< 10 /Jm) grains radiating at ~ 200 K is present that is not seen in other well investigatedd HAEBE systems. This grain component may be indentified with thee dust expected to be produced in a collision cascade of large objects. Second,

105 5 CHAPTERR 6

thee fraction of intercepted stellar light that is absorbed and re-emitted in the midTRR is so large (~ "()%) that it requires the disk to be more 'puffed up' at aboutt 10 AL\ where the grains have / ~ 200 K. 1 his mav occur if a proto- Jupiterr clears out a gap at this distance allowing direct stellar light to produce an extendedd rim at the far side of the gap, consistent with our proposed scenario.

6.11 Introduction

Herbigg Ac/Be stars (hereafter referred to as HAEBE stars) were first described as a group bvv Herbig (I960), in a studv which was aimed at finding intermediate mass voung stars.. Many studies have since confirmed the young pre-main-sequencc (PMS) nature- off HAEBE stars. Hipparcos parallaxes (e.g. van den Ancker et al. 1998) allowed accurate positionss in the HR diagram to be obtained, and from comparison to PMS evolutionary trackss masses of the order of 2 to 8 M . and stellar ages of typically 1 0h to 10 yrs were found.. Direct imaging at millimetre wavelengths of several Herbig Ae stars revealed the presencee of rotating flattened structures (Mannings and Sargent 1997, 2000), believed too be the remnant of the accretion disk and the site of on-going planet formation. It is nott clear whether disks are also common around more massive Herbig Be stars. Att infra-red (IR) wavelengths, the Infrared Space Observatory (ISC) Kessler et al. 1996)) has obtained full 2-200 /jm spectra of the brightest HAEBE stars. These spectraa show a wealth of detail, both concerning the gas-phase molecular component, as welll as the thermal emission from the dust in the circumstellar environment. In a se- riess of papers, we have studied the thermal dust emission from HAEBE stars (Waelkens ett al. 1996; Malfait et al. 1998b, 1999; van den Ancker et al. 2000, 2001; Bouwman ett al. 2000a, 2001b; Meeus et al. 2001), focusing on the mineralogical composition of thee dust. These studies indicate that a substantial modification of the dtist composition fromm that in the interstellar medium (ISM) occurs on time scales that are still poorly constrained,, but are generally less than 10 years. We mention grain growth, the formationn of crystalline silicates and of crystalline HjO ice, and of hydro-silicates. I hese modificationss are important clues to the processes that eventually lead to planet formation,, and which can be compared to the records of planet formation as found in solar svstemm objects (meteorites, comets, interplanetary dust particles). Inn this studv we re-examine the ISO spectrum of HI) 100546, first presented by Waelkenss et al. (1996) and Malfait et al. (1998b). This object has an exceptional!}' high fractionn of crystalline silicates, and its ISO spectrum shows a remarkable resemblance to thatt of the solar system comet Hale-Bopp. Malfait et al. (1998b) already suggested the excitingg possibility that we are witnessing the birth of an Oort cloud with a multitude off cometarv bodies being scattered bv a (hypothesised) giant planet. Bouwman et al. (2001b)) analysed the lO^mi silicate band of 14 HAHBH stars including HI) 100546, andd found that the mineralogical composition of HI) 1 00546 deviates substantially from thatt of other HAEBE stars: in addition to a large fraction of forsterite. the average grain sizee of the particles causing the 10 jLim emission is larger than that of other stars. Also, aa lack of silica compared to forsterite was found. This difference in composition can

106 6 HDD 100546 & HAI H-BOPP bee interpreted as a difference in the crystallisation process leading to the formation of forsterite.. We decided that a closer look at the distribution of the different mineralogical componentss in terms of mass and temperature is needed to better understand the nature off HD 100546. We also analyse the ISO spectrum of Hale-Bopp, using the same analysis methodd as for HD 100546. We have already carried out a similar detailed analysis for ABB Aurigae and HD 163296 (Bouwman et al. 2000a), and we will compare the results off that study with those obtained in the present paper for HD 100546. I he observations off HD 100546 used in this paper are taken from Malfait et al. (1998b) and Meeus et al. (2001).. The observations of Hale-Bopp, when the comet had a heliocentric distance of 2.88 AL', where taken from Crovisier et al. (1997). Thiss paper is organised as follows: in Sect. 6.2 we discuss the difference in dust compositionn and spectral energy distribution of HD 1 00546 compared to other HAEBF, systems.. In Sect. 6.3 we discuss the method to fit the spectrum. Sect. 6.4 describes our resultss for HD 100546 and for Hale-Bopp. In Sect. 6.5 we compare our results to those obtainedd for AB Aur and HD 163296, and discuss the implications for the formation processs of the crystalline silicates in HD 100546.Sect. 6.6 summarises the results of our study. .

6.22 The deviating dust composition and spectral energy distributionn of HD 100546

Thee spectral appearance and dust properties of HD 100546 differs from that of other HAEBE,, stars. We will summarize these differences in the following. First,, the ISO-spectra of HD 100546 show strong and pronounced emission, iden- tifiedd with the crystalline silicate forsterite, which in abundance must be far in excess of thatt seen in other HAEBE systems (Malfait ct al. 1998b; Meeus et al. 2001). Second,, Bouwman et al. (2001b) identified an emission component at 8.6^/m in the spectraa of a large sample of HAEBE stars which they attribute to silica (SiO:), and which seemss to be correlated with forsterite. This correlations is interpreted as evidence that thee crystalline silicates are formed by thermal annealing. However, the dust properties inn HD 100546, and in this respect also that of comet Hale-Bopp. does not comply with thiss correlation between silica and forsterite. This suggests a different formation process too be responsible for the large mass fraction of forsterite seen in both objects. Third,, from a detailed analysis of the 1 0/Jm spectral region (Bouwman et al. 2001 b), itt was shown that the typical grain size of the silicate particles emitting at these wavelengthss is larger compared to the sizes found in other HAEBE systems. 1 his suggests the diskk in HD 100546 is more evolved. Fourth,, the spectral energy distribution of HD 100546 shows remarkable differences whenn compared to other HAF.BF. stars. To illustrate the difference, we have plotted in Fig.. 6.1 the SEDs of AB Aurigae (light line) and HD 100546 (dark line) normalized to thee stellar luminosity. Both stars intercept about 50% of the stellar light (see 7able6.2). Itt is easily seen that ABAur re-emits a larger fraction of this intercepted light in the near-IR,, while HD 100546 shows a larger emission at mid-IR wavelengths. I he far-IR

107 7 CHAPTERR 6

0.010 0

CM M Q Q

,< ,<

0.001 1

10 0 100 0 A,, [jam]

Figuree 6.1: Comparison herween rhe energy distriburion of AB Aur (lighter specrra) and HDD 100546. Plorred are rhe combined ISO-SWS and LWS specrra. Nore rhe large luminosityy differences in rhe near and mid-lR herween borh systems.

andd sub-millimetre luminosities are comparable. The difference in the way radiation is redistributedd by the dust essentially implies that the stellar flux is absorbed at different locationss in the circumstellar disk. Dust grains dominating the near-IR emission have temperaturess of at least ~ 1000 K. For a grain to reach such a temperature it has to be withinn ~ 1 AU of the central star. In the case of AB Aur it is within this region that about haltt of the intercepted light is reprocessed (see Table 6.2). In the case of HD 100546 a relativelyy modest amount of radiation is emitted in the near-IR, however, 70% of the totall dust luminosity comes out in the mid-IR. Grains emitting in the mid-IR have typicall temperatures of ~ 200 K. This corresponds to a distance of ~ 10 AU to the

108 8 HD1005466 & HAIJ.-BOPP

l.jllst/L* * I-NIR/kkivt t lAllR/I-dust t Lii [R/l-duM ABB Aur 0.46 6 0.51 1 0.34 4 0.15 5 Hl)) 100546 0.51 1 0.17 7 0.70 0 0.13 3

TABU;; 6.1: Luminosity output of the circumstcllar dust in AB Aur and Hl) 100546. The Hrstt column lists the total dust luminosity as a fraction of the stellar luminosity. The following columnss list the fractional dust luminosities in the near-lR (A < 8/jm), the mid-lR (8 < A < 50jum)) and far-1R (A > 50/jm), respectively. centrall star, which therefore is the regime from which this 70 % of re-emitted light must originate. . Inn estimating the above emitting regions, we have assumed that the dust medium iss optically thin. Let us, for the moment, hold on to this assumption. For T

109 9 Cl]] API i'R 6

thosee observed in other HAKBH systems currently dominate the HI) IOOS46 system. In thiss paper we will try to trace these processes.

6.33 Modelling

II he diagnostic method used is identical lor the Herbig Be star HI) 10()Vi6 and comet Hale-Bopp.. We use the radiative transfer program MODI'S I' to model the circumsrellar dustt using the mode in which the material is assumed ro be optically thin. This is correct forr comet Hale-Bopp, but lor the circumstellar disk of HI) 100546 this mav not be valid.. However, in discussing dust properties we will focus on the mass over temperature distributionn of the material responsible for the infrared emission. This diagnostic does nott depend on the optical depth properties of the medium and therefore the simple approximationn used here is still meaningful, hor both objects the material is distributed inn a spherical shell, which for Hale-Bopp is positioned at the distance from the sun correspondingg ro that of the location of the coma and tail at the time of observation. For recentt applications and descriptions of techniques used in MODI'S I, see e.g. Bouwman ett al. (2000a); Kemper et al. (2001). II he grains are irradiated by the central star, for which we use Kuruc/ (1993) en- erg)-- distributions. The Hipparcos distance to the B9\ ne star HI) 100546 is 103T-7 pc, yieldingg a luminosity /. = 32 Lg (van den Ancker et al. 1998). Comparison with evolutionaryy tracks, using 'I[.u^ 10500 K, places the star on the main sequence having an agee of > 10 Myr and mass of 2.4 MQ. In modelling Hale-Bopp, we use Kuruc/s solar model,, with '/^y- 5 77_7 K. 11 he extinction of energy at mostly ultraviolet and optical wavelengths and re- emissionn in the infrared is consistently taken into account assuming the particles are in radiativee equilibrium, yielding the spectral energy distribution. A kev difference of this methodd compared to other often employed approaches in modelling circumstellar and cometaryy spectra is that both the spectral characteristics responsible for the distinct featuress as well as those responsible for the featureless continuum are modelled simultane- ously.. Other methods introduce an artificial separation between continuum and features, fittingg the spectrum combining empirical temperatures with laboratory extinction effi- cienciess (e.g. Wooden et al. 1999: Mason et al. 2001 ) or with template extinctions based onn measurements of interplanetary dust panicles (II)P's) (e.g. Wooden et al. 2000). Inn the near- and mid-IR. typically carbon and large (> 10//m) amorphous silicate grainss are responsible for the overall continuum, while water ice is a dominant contrib- utorr in de far-IR and sub-millimetre range. Ar 10 //m the molecular vibrational modes off Si-O bonds in small amorphous silicate grains (< 10//m) produce a distinct feature, ass do silicate crystals in the mid-IR part of the spectrum. These latter particles, however, alsoo produce significant flux outside of their resonances, which one can not a prion'dis- tinguishh from continuum only contributors. Continuum subtraction methods therefore aree susceptible to (systematic) errors. This mav significantly affect temperatures, abundances,, and sizes derived for especially the crystalline material. Our method does not sufferr from this problem.

11 10 Hl)ll 00546 & HM t-BoiM'

6.3.11 Size and shape properties of grains

Composition,, size and shape properties of grains in proto-planetarv disks mav provide importantt constraints on the formation history ot circumstellar dust. Wee first focus on shape properties. Yerv little is known about the structure of grains inn circumstellar disks. Particles mav be compact or "Huffs'" and mav be chemicalh' ho- mogeneouss or consist of a mix of different materials. Information mav be obtained from measurementss of grains sublimating from solar svstem comets nearing perihelion. From aa comparison of measurements of the angle dependence ot the scattering albedo of severall bright comers with theoretical predictions, Gehrz and Nev (1992) found the coma grainss to be consistent with "Huffy aggregates of smaller compact particles. Properties off the aggregate, such as temperature and spectroscopic signature, will be affected bv the actuall degree ot fiulfiness of the grain, often expressed in terms ot the porositv factor off the particle (I.isse et al. 1998; Harker 1999). Large porous grains mav heat to much higherr temperatures than compact grains ot the same mass and mav show spectroscopic characteristicss similar to the smaller compact particles constituting the fluffy aggregate. II heoretical considerations suggest cometary grams to be extremely Huffy with porosity factorss 0.93 ^ P ^ 0.975 (Greenberg and Hage 1990), effectively implving spectro- scopicallvv deduced properties relate to the smaller compact units making up the porotts aggregatee (e.g. Brucaio et al. 1999, and references there in). In our modelling we will thereforee concentrate on (small) compact particles, leaving open the possibility that they mavv be part of larger fluffv aggregates. Regardingg the chemical homogeneity of the small particles, Li and Greenberg (1998) pointedd out that in situ mass spectra of cornet P/Halley 1986 111 dust showed that carbonaceouss and silicate materials were mixed on Hue scales, suggesting these two species aree not phvsicallv separated. If this is the case, the components are likely in thermal contact.. I he question of thermal contact is especially relevant with respect to the magnesiumm over iron content of the silicate material, a major constituent of circumstellar dust,, as the absorption properties of olivine (Mgjvbe^ .^SiO,, with .v between 0 and 1) j andd pyroxene (MgA lc ]_vSiO J sensitive!}' depend on this ratio. I he pure magnesium silicatess forsterite (MgvSiO,) and enstatite (MgNiO \) are optically much more transparent thann iron rich silicates, implying that if He-rich and He-poor material coexists as sepa- ratee particles — with comparable shape and size - the He-rich dust will reach significantly higherr temperatures. Wee have opted to treat the different chemical species as physically separated. In the casee of Hale-Bopp this allows to investigate whether the forsterite crystals, of which the spectroscopicc signatures are prominently visible in the mid-IR spectrum, are in thermal contactt with the bulk amorphous silicate material, which is likely to contain a significantt fraction of iron, bs' determining the mass-averaged temperatures of both species. Itt significantly different, one mav assume the components to be phvsicallv separated; it similarr temperatures are found, it is likelv thev are in thermal contact. Horr each particle, we calculate the extinction properties from optical constants determinedd in laborators experiments, as listed in Table 6.3.2. We assume spherical grains, forr which we use Mie calculations, or a continuous distribution of ellipsoidal grains, for

11 11 CHAPTERR 6 whichh wc use CDE calculations, to determine the absorption and scattering coefficients (seee Bohren and Huffman f983, for a full review on these methods). Ass we tread the CDF particles in the Ravleigh limit, we can not studv grain size effects.. For spherical particles this assumption is not required. I he spherical particles of ourr multi-component mixture of grains range between minimum size amm and maximumm size am;lx and are distributed following a power-law, i.e.

«(*)«(—)) • (6.1) V'mrn/ /

11 heoretical calculations predict this type of size distribution whenever there is shattering andd coagulation of grains through grain-grain collisions (Biermann & Hartwit 1980). Too get some feeling for the value of the power-law index, extinction observations imply aa size distribution with ;;/ = 3.5 for interstellar grains (Mathis et al. f 977).

6.3.22 Chemical composition of grains

Thee grain species used to model the spectra of both objects are listed in Table 6.3.2. Thee dust composition is very much similar to that used by Bouwman et al. (2000a) too model the isolated Herbig Ae stars AB Aurigae and HD 163296, though different sourcess for the optical constants of crystalline silicates and water ice are used. Of the twoo isolated Herbig stars mentioned onlv in HD 163296 a small amount of forsterite couldd be identified, modelling of which did not sensitively depend on the adopted op• ticall constants. However, for the two objects investigated here - which show prominent featuress of olivine — differences between the laboratory measurements are important. To modell the crystalline silicate features we tried three sets of measurements: Servoin and Piriouu (1973) measured the optical properties of forsterite (Mg^SiO.^ FolOO). Steyer (1974)) used a natural olivine sample, which was estimated to contain a small amount off iron, i.e. x = 0.91 (Fo91). The sample used by Mukai and Koike (1990) contains ann almost equal amount of iron (.v = 0.90; Fo90). Fhese measurements onlv cover the 1RR wavelengths. To estimate the optical properties of the grains at visual and UV wave• lengths,, where most of the stellar light is absorbed, we used measurements of comparable materialss from several other sources, referenced in Fable 6.3.2. Neitherr in AB Aur nor in HD 163296 the 44 jum feature characteristic for crys• tallinee HiO ice was found, prompting the use of amorphous water ice in modeling their circumstellarr environment. In both HD 100546 and Hale Bopp, however, the 44 fjm featuree appears to be visible, making it more appropriate to assume the water ice in these sourcess to be crystalline.

6.44 Results

Inn this section we present the results of our analysis of HD 100546 and of comet Hale- Bopp. .

112 2 HD1005466 & HAI.K-BOPP

TABLF.. 6.2: Overview or dust species used, their solid state, wavelength interval over which opticall constants are measured, and destruction temperature.

Speciess Solid Wavelength 1 jcsrr Ref. State e 1pm] 1pm] IK] ] [Mg,Fe]SiO., , A A 0.2-500 0 1100 0 (1) ) FeO O C C 0.2-500 0 1000 0 (2) ) C C A A 0.1-800 0 1000 0 (3) ) FTO O C(ice) ) 0.05-103 3 150 0 (4) )

Forsterire e c: : 0.04-3 3 1400 0 (5) ) 3-250 0 (6) )

Olivinee (Fo91) c c 0.04-3 3 1400 0 (5) ) 3-250 0 (7) )

Olivinee (Fo90) c c 0.01-0.3 3 1400 0 (8) ) 0.3-2 2 (9,10) ) 7-200 0 (11) )

Fe e M M 0.1-105 5 1500 0 (12) )

Abbreviationss used to designate the solid state: A = Amorphous; C = Crystalline; M = Metallic. References:References: (1) Dorschncr et al. (1995); (2) Henning et al. {1995); (3) Prcibisch et al. (1993); (4) Bertie et al. (1969);; (5) Scott and Dulev (1996); (6) Servoin and Piriou (1973); (7) Stcver (1974); (8) Huffman and Stappp (19-3); (9) Jones and Merrill (19?6); (10) Rogers et al. (1983) (11) Mukai and Koike (1990); (12) Henningg and Srognienko (1996).

6.4.11 HD 100546

Plottedd in Fig.6,2 is our best model fit to the SFTJ of HD 100546. Fhe resulting model parameterss are listed in Table 6.3. Fhe top panel shows the entire SED. Indicated in the figurefigure is the Kurucz model for the central star (dashed line). The lower panel shows the ISO-SWSS and LWS wavelength region. Indicated are the contributions to the spectrum off the individual dust components as listed in Fable 6.3. To fit the SED a bi-modal grain sizee distribution is required, similar as to that found in previous analysis of HAEBE starss (Bouwman et al. 2000a, 2001a). 'Fhe small (< 10 jjm) grains dominate the SED shortwardd of ~ 40jum, while the large (up to 200 /jm) grains dominate at the longest wavelengths. .

113 3 C'HAP'11 l-.R 6

!^ ^

'S'S \r — • ~- ~

> >

t-- — ^

t^^ — o oc

GCC bc

-fcxx ~ 4^^ J>.

II "

II !

ss z

]] 14 HD1005466 & HALE-BOPP

10 0

'0 0

EE IQ"'

^^ IQ"10

10 0

10 01 2 2

13 3 10 0 1000.0 0

30C C

200 0

11 00

-100 0

Figuree 6.2: The top panel shows our best model Ht to the spectral energy distribution of HD1005466 (solid line). Indicated with the dashed line is the spectrum or the central star. Trian- gless indicate ground-based and IRAS photometry. The thin solid line is the ISO-SWS and LWS spectrum.. The bottom panel shows our fit to the ISO spectra in detail. Also plotted in this figure aree the relative contributions of the individual dust components. Indicated with a solid line are thee contributions of amorphous olivine and forsterite marked in the figure with OI and Fo, respectively.. The contribution of water ice, carbonaceous grains and metallic iron are represented byy the dashed, dotted and dashed-dotted lines, respectively. I he curves of the carbonaceous, forsteritee and metallic iron grams are offset by -40, -75 and -90 Jv, respectively. 1 he inset shows thee 2 to 8 pm region.

I'M M CHAPTHRR 6

Thee small grain component

Thoughh the small grain component contains only a minor fraction of the total dust mass (~~ 1%) it dominates the SF.D at midTR wavelengths. To reproduce the observed fluxes inn this wavelength range, a bi-model density structure is required. As can be seen from Tablee 6.3, the small grains are distributed between 0.3-9.8 AU, and between 9.8-43 AU, fromm here on referred to as zone 1 and zone 2, respectively. I he inner boundary of zone 11 at 0.3 AU is determined by the grain destruction temperatures of the individual dust speciess as listed in Table 6.3.2. A similar dust component as that in zone 1 is also present inn AB Aur, HI) 163296 and HD 104247 (Bouwman et al. 2000a, 2001a). The sudden increasee in density at 9.8 AU, marking the onset of zone 2, is found only in HD 100546. Itt is this additional component that produces the much larger mid-IR luminosity comparedd to the other HAFBF. systems as discussed in Sect. 6.2. The resulting total mass of 6.8-- 10" MQ in small grains is two to three orders of magnitude larger than found in thee studies mentioned.

AA major fraction of the dust emission seen in HAKBH systems is due to amorphous silicates.. 1 hough it is difficult to determine the exact nature of the amorphous material, ann excellent fit can be made if we use the optical properties of a silicate glass with an olivinee stoichiomerry (see also Fable 6.3.2). The grain size of the amorphous silicate is welll constrained by the shape and strength of the 1 0 /Jm silicate feature, and the flux ra- tioo between the 10 and 18 fjm bands. Apart from the amorphous silicates an additional sourcee of continuum emission is required. We added this in the form of carbonaceous grains.. Since this material has no distinct spectral features, the grain size and consequently,, mass fraction, are less well established and are uncertain within a factor of two. Inn our best fit model we assumed an equal grain size distribution as for the amorphous silicate.. 1 he band at 43.3 /Jm in the spectrum of HD 100546 indicates the presence of aa population of small ice grains. Given the ice sublimation/evaporation temperature of 11 50 K, water ice can only be present in zone 2, just outside the "snow limit\ For the smalll ice grains we also assumed the same grain size distribution as for the amorphous material.. Interesting to note is that Smith er al. (1994) showed that the ice band can onlyy shift to the observed wavelength of 43.3 jum if amorphous ice grains where heated too the crystallisation temperature and then cooled to ~50 K. One could interpret this as aa local heating event or local radial mixing.

Ass one can see from the inset in Fig. 6.2, the near-IR fluxes shortward of 4 jjm are dominatedd by the emission from metallic iron grains. These grains are the only refractory dustt species which are stable at the required high temperatures (~ 1500 K), and that havee sufficient opacity at near-IR wavelengths to emit in this spectral region. We had too constrain the spatial distribution of the metallic iron grains to within 4.9 AU. The grainn temperature of the iron grains is between 1500 and 750 K, This suggests that thatt metallic iron is formed at high temperatures, most likely by solid state reduction of ironn bearing silicates heated above the glass temperature at the verv inner parts of the circumstellarr disk.

116 6 HI)) 100546 & HALK-BOIM'

Thee crystalline silicate component

Inn the sample of isolated HAEBE stars presented in Meeus et al. (2001), HI) 100546 showss the most prominent features of crystalline silicates. This enables one, at least in principle,, to determine the exact chemical composition and crystalline structure of the silicates.. We used three sets or laboratory data listed in Table 6.3.2, to determine the bestt fit to the crystalline silicate bands. Plotted in Fig. 6.3 are the 24 and 34 jjm silicate bandss of HI) 100546, compared with our best fit models lor the three data sets. As onee can clearly see, the model based on the measurements ol Servoin and Piriou (1973) providess the best match with the observations. For this model and the model using the Steyerr (1974) data set, the assumption ol a CDF shape distribution (see Sect. 6.3) gives aa much better fit than the assumption ol spherical grains. In the model using the Mukai andd Koike (1990) data, MIE calculations produce the best match. We conclude that thee crystalline silicate leatures seen in the ISO spectra ol HI) 100546 are due to non- sphericall forsterite grains. In our best model fit presented in Fig. 6.2 and Table 6.3 we usedd this latter model for the crystalline silicate. Fhee lorsterite grains also require a bi-modal density structure (zone 1 and zone 2). Ass can be seen horn Fable 6.3 the crystalline mass fraction is much higher in zone 2 than inn zone 1, indicating that the abundance ol crystalline material is increasing outwards. Inn our best model fit the fouteiite grains are not co-spatial with the other dust species. Fhee spatial distribution of lorsterite grains has to extent out to 600 AU, well beyond thee observed disk limit of 380 AU (Augereau et al. 2001). This, however, could be an artilactt of the grain model we used for calculating the optical properties of the forsterite grains.. The CDE approximation assumes the grains to be in the Rayleigh limit, and consequentlyy the optical properties ol the grains do not depend on grain size. However, outsidee the Rayleigh limit, grain size has a strong effect on grain temperature. Farger grainss have to be placed much closer to the centra! star than smaller grains to reach the samee equilibrium temperatures. Fhe grains can however not be much larger than 1 jjm too be able to reproduce the observed spectral features. This is smaller than the typical grainn size ol the amorphous material. Our best estimate fot the effect ol grain size on the spatiall distribution, using the measurements of iMukai and Koike (1990) indicates that thee crystalline grains could be moved inwards to within the observed disk limits, but not larr enough to be co-spatial with the amorphous silicate. Thee grain size distribution also influences the mass fraction of the dust species. The modell assuming a maximum grain size of 2 /./m lor the forsterite grains would increase thee mass in crystalline dust by a factor of five. This would imply that as much as 50 % off the total mass of the small grain component could be crystalline silicates.

Thee large grain component

4oo reproduce the observed slope and fluxes at mm wavelengths a dust component with aa large grain size up to 200 ^m is required. Due to this large grain size, one can not determinee its exact composition from ISO spectroscopy, since such large grains will radiate ass black bodies at these wavelengths. We used the dust species found spectroscopically

117 7 CHAPTERR 6

Figuree 6.3: The 24 and 34 fjm forsterite bands of HI) 100546. Overplotted are the best fit silicatee bands using the data of Servoin and Piriou (1 973) (solid line), Steyer (1 974) (dotted line) andd Mukai and Koike (1990) (dashed line). In this latter model spherical grains where assumed \\\ ith a size of 0.01pm. 1 he other models have a CDE .shape distribution.

too model the large grain component. The mass fractions of the amorphous silicate and carbonaceouss grains listed in Fable 6.3, ate thetefore uncertain. We simply adopted the relativee mass ratios derived for the small grain component to model the large grains. Similar,, it is not possible to determine if these grains contain crystalline material, as this willl also result in black body emission. We therefore have not used crystalline silicates in thee large grain component. However, due to the presence of the emission band at 60 /Jtn attributedd to crystalline water ice gtains, the mass fraction of this dust species can be detetmined.. The maximum gtain size the ice grains can have and still produce a 60 /./m band,, is 25 jjm. This is much smaller than the maximum grain required to fit the slope

118 8 HOII 00546 & H,\l i -Boi'i' off the SKI) at mm wavelengths. 1 here are however some problems with the 60 /vm band ass we will discuss in Sect. 3.1.4. Wee have constrained the spatial extent or the large grains using the disk size observed byy Augereau et al. (2001), with the HST/NICMOS2 instrument, 'lb fit the observations, aa gap of 28 AU is required between the central star and the dust shell containing the large grains.. As noted in previous studies (Bouwman et al. 2000a), this is most likelv an artifact off the assumption made in this study that the medium is optically thin in the radial direction.. As for HI) 100546 we see the disk fairly face on (51° Augereau et al. 2001). 11 his implies that in the line-ot-sight the medium may well be optically thin. Therefore, thee derived mass over temperature distribution T(w) remains meaningful (albeit not the Y(r)Y(r) structure). As suggested by Bouwman et al. (2000a) the large grains have most likely settledd to the disk mid-plane, extending all the vvav to the disk inner ed»e. Recent full 21)) radiative transfer calculations (Bouwman et al. 2001a) show that this could indeed bee the case.

Thee residual spectrum

Plottedd in Fig. 6.4 is the residual spectrum of HI) 100546, after subtraction of the best fitfit model to the ISO-SWS and l.WS spectra. Shown in the top panel is the 2 to 15 ji/m regionn in which the emission bands from polycyclic aromatic hydrocarbons (PAHs; markedd with the vertical lines) are clearly visible. For comparison we also plotted the PAHH spectrum of the [WC] star BD+30 3639, which may serve as a typical example off a generic PAH spectrum. Apart from the bands at 6.25, 7.9 and 11.3 /./m, also PAH featuress at longer wavelengths can be observed (see also Hony et al. 2001). Thee lower panel shows the 35 to 150 jjm region. Marked with shaded areas are residuall emission bands. The band at —100 /vm has been interpreted as evidence for thee presence of hydro-silicates (Malfait et al. 1998b). If correct, this can be seen as evidencee for the presence of larger bodies on which the hvdro-silicates are believed to have formedd by aqueous alteration of liquid water. For comparison we plotted the band of HL)) 142527 which is identified with hydro silicates (Malfait et al. 1999). As one can see thee width and position are similar for both bands. However, a secure identification of this ratherr broad feature strongly depends on the exact run of the underlying continuum. A smalll change in the location of this continuum may change both the strength and width off the feature and may even cause it to disappear. Thee residual at ~55 ^/m could indicate the presence of an emission band of a species nott taken into account in our models. It is however, more likelv that this residtial originatess from the poor fits of the ice bands at 43 and 60 jiim. Though the positions of the observedd ice features coincide well with our ice model, the width and relative strength off the 43 and 60 /7rn bands do not. I his could reflect the limitations of otir simple ice model,, e.g. the assumptions of homogeneous spherical grains. Also the temperature of thee material at which the optical properties where measured (100 K in our models) will influencee the strength and width of the features. The observed position of the ice band att 43.3 jjm indicates that the ice is —50 K (Smith et al. 1994). An additional problem iss that the ice features appear in two separate instruments, LWS and SWS. Indicated

119 9 CHAPTERR 6

20 0

10 0 HDD 100546 LL L ^-Sy^—i'—'" " v"V V 0 0 '•'^^-Ivviw^*y^^ft^ w ^*y^^ft^ Vvww »>• BD+300 3639

-10 0 "•• • .•''

88 10 122 14 >LL [u_m]

hydroo silicate HD142527

AA M 400 50 60 70 80 90100 150 0 XX [|il7l]

Figuree 6.4: Residual spectra of HD 100546 1SO-SWS and LWS spectra subtracted by the best modell fit. 1 he top panel shows the 2 to 15 //m region clearly showing the PAH emission bands (indicatedd with vertical lines). For comparison we also plotted the PAH spectrum of BD+30 3639 (offsett by -12 Jy). I he bottom panel shows the 35 to 150 /Jm region. Indicated with an arrow- iss the connection between the ISO-SWS and LWS spectra. Also plotted in the figure are the absorptionn spectra of several silicates ot interest multiplied bv a 70 K black body, and the emission bandd in the spectrum of HD 14252- identified with hydro silicates (Malfait et al. 1999). The shadedd areas indicate features in the residual spectrum, discussed in Sect. 6.4.1.

120 0 HD1005466 & HAI F-Boi'ï' inn Fig. 6.4 with an arrow is rhc connection between both instruments. I he error in thee absolute flux calibration between both instruments is 10-15 %. Lowering the LWS spectrumm with this percentage would enable us to fit the ice bands without the residue att ~55 /Jin. Interesting to note is that the LWS spectrum would then also agree better withh the IRAS photometry which is now below the LWS spectrum as can be seen from Pig.. 6.2. However, we stress that from the ISO observations there is no strong evidence thatt there is a jump between the ISO-SWS and LWS spectra at ~-45 jum. AA third feature in our residual spectrum appears at 40.3 /./m. This band could be duee to crystalline dust species other than forsterite. Plotted in Fig. 6.4 are the spectra of diopsidee and ortho- and clino-enstatite (Koike et al. 2000). All these species show bands inn the 40 fjm region. The 40 /jm band of diopside coincides nicely with the residual spectrumm but would also produce a band at 65 jam at the location of the water ice band, makingg the problem at 55 /./m even more severe. If one of the suggested materials is responsiblee for the 40 jam band, it has to be cold, i.e. <7() K, as at shorter wavelengths noo evidence for crystalline silicates other forsterite can be found. Finally we mention that thee bands at 40.3 and 43.3 /Jin, coinciding with the ice band, may be due to forsterite. Thiss suggestion is based on mass absorption spectra taken by Jager et al. (1998), which showw evidence for bands at these wavelengths. Though these features are weak, given the overalll strength of the forsterite bands this could be a possibility.

6.4.22 Hale-Bopp

Plottedd in Fig. 6.5 is the ISO-SWS spectrum of comet Hale-Bopp, together with our bestt fit model. Also shown are the contributions to the spectrum of the individual dtist species.. The parameters of the model fit are listed in Table 6.4.2. As has been discussed byy previous authors (e.g. Ma I fait et al. 1998b), the ISO spectrum of Hale-Bopp bears aa striking resemblance to that of I IF) 100546. Our modelling efforts of the spectrum off Hale-Bopp will be focused on the question wether both objects have similar dust compositionss and properties and, if not, what are the differences.

Grainn composition and properties

Inn general we find that the dust composition is similar to HI) 100546 with a few exceptions.. We find no evidence in the spectrum of Hale-Bopp for emission from metallic ironn grains as in the case of HAFBF. systems. 1 his could be a spurious result as in the coldd (160 K) dust of Hale-Bopp iron grains have no distinct spectroscopic feature and theirr emission could blend with that of other smooth continuum sources, for instance fromm carbonaceous grains. However, it could also implv that metallic iron grains are reallyy absent, given the suggested high temperature formation mechanism of this grain speciess (see Sect. 6.4.1). At lower temperatures metallic iron can also react to form iron sulfidess or oxides (e.g (jail 1998). Indeed, we find evidence for an iron oxide component, similarr to that found in AB Aur and HI) 163296 (Bouwman et al. 2000a), though less abundantt (in terms of mass fraction). The possibility exist that the emission component heree attributed to iron oxide is due to iron sulfide which has similar emission properties

121 1 CHAPTFRR 6

200 0

-50 0

•100 0 0 0 10 0 200 30 40 0 50 0 XX \\m)\

Figuree 6.5: Best model fit to the ISO-SWS spectrum of Hale-Bopp. Also plotted in the figure aree the contributions or the individual dust species to the spectrum. I he solid lines show the con• tributionn of the amorphous silicate with an olivine stoichiometry (marked in the figure with Ol) andd forsterite (Fo). The dotted, dashed and dot-dashed lines show the contributions of carbona• ceouss material, ice and iron oxide grains, respectively. The curves of forsterite, carbonaceous, ice andd iron oxide are offset by -10,-50,-70 and -90 Jy, respectively.

inn this wavelength region (Keller et al. 2000). \ \ componentt o,fr f amorphou)h( ( s silicateilicatesss anandd carbonaceous grains similar as in HDD 1 00546, constituting a major fraction of the dust mass, is seen in Hale-Bopp. Inter• estingg to notice is the typical grain size of the cometary dust, which is with a power-law slopee of m = 2.8 much smaller than in HD 100546 (m = 2). Note that for the forsterite andd iron oxide grains we used a CDE shape distribution which does not allow for the de• terminationn of grain size. Spectroscopicallv, the best fit to the forsterite bands is, like for

122 2 HDD 100546 & HAII -Boi'i

l Md™ ™ 4.2-- 10 ' kg

1 Dustt species R[AU] ] Poo Igrr cm - ] M»v.u u Amorph.. silicace 2.8-3.0 0 3.4-- 100 2l' 0.64 4 Carbon n 2.8-3.0 0 1.3-- ïo-' ' 0.23 3 Waterr Ice 0.6-0.8 8 4.0-- ioo •-•' 0.03 3 Ironn oxide 12.6-12.8 8 1.11 •1 00 M 0.03 3 Forsterite e 0.9-1.1 1 3.0-- ioo -l) 0.07 7

TABU-- 6.4: Model Ht parameters of Hale-Bopp. Listed are the parameters defining the density andd grain si/e distribution, the chemical composition and the mass fraction Mt[,K ol the individ• uall dust species, for all dust components we assumed a power law dcnsitv distribution p(r) oc >• andd size distribution n{a) « a~~ 's. The radial extend and grain size range are given in the table.

HOO 100546, achieved by using the measurements of Servoin and Piriou (1973). Similar uncertaintiess apply for the derived forsterite mass fraction as for HI) 100546, discussed inn Sect. 6.4. J. Our best estimates for the effects of a grain size distribution on the de• rivedd mass indicates that the maximum mass fraction of forsterite could be about six timess higher, resulting in a mass fraction of 40 %. A grain size distribution would also resultt in a higher continuum contribution of the forsterite grains, which eliminates the requirementt for an additional component in the form of iron oxide, from the ISO-SWS spectrumm it is difficult to determine the characteristics of the water ice component. We thereforee used the results of I.ellouch et al. (1998), who found from an analysis of the ISO-II WS spectrum that the average water ice grain is 15 /urn. I he mass fraction of the icee is similar to what Lellouch et al. (1998) find, however, they infer a total dust mass off 1.1 • 1011 kg, almost two orders of magnitude larger than we derive. They however, assumedd a grain size of 100/jm in deriving the total dust mass. Lowering this to the sizes wee find would scale their total dust mass to a similar value as found here.

Thermall contact of grain species

Inn order to model the spectrum of comet Hale-Bopp, we found, as expected, that the bulkk of the dust material (i.e. the amorphous silicates and the carbon, together account• ingg for 87 percent of the dust mass) has to be placed at the proper heliocentric distance at thee time of observation (2.8 AU and distributed over 0.2 AU), giving it a mass averaged temperaturee of ~16() K. To fit the near-IR wavelength range, positioning the iron oxide att the correct distance resulted in a temperature for this species too high to properly fit thee spectrum. We determined the best fir temperature for this material by positioning it att a larger distance. This yields I ~- 160 K. 1 he same approach was followed to deter• minee the temperature of the water ice and the crvstalline silicate. Here it was found that duee to their relative!)' poor light absorption properties both materials had to be placed

123 3 CHAPTERR 6

closerr to the star, bur again at such a distance that their hulk mass averaged temperature iss 160 K. 1 his result implies that all dust particles are in thermal contact with one anotherr and that their temperature is essentially determined bv the materials dominating thee thermal energy content of the dust. The latter are the amorphous silicates and the carbonaceouss material.

6.55 Discussion

6.5-11 The mass temperature distribution

11 he Shi) as presented in Fig. 6.2 is determined bv the mass-temperature distribution of thee circumstellar dust. Irrespective of the assumed model geometry, this distribution has too be reproduced for the visible dtist producing the observed emission feattires. Plotted inn big. 6.6 is the derived mass over temperature distribution of the best fit model. The upperr two panels show the mass-temperature distribution of the small grain component, thee lower panel shows the same for the large grains. Indicated in the figure are the individuall contributions of each species. The vertical axis unit is chosen in such a way thatt the integral of the cumulative mass-temperature distribution equals the total dust mass.. Plotted in Pig 6.7 for comparison is the mass-temperature distribution of AB Aur ass derived bv Bouwman et al. (2000a). The first panel shows the mass-temperature distributionn of the small grains, the second panel that of the large grain component. Ass can be seen from comparing Fig. 6.6 and 6.7 the dust in HD 100546 is, like in thee AB Aur system, characterised by a bi-modal mass-temperature distribution. The inset inn the top panel of Fig. 6.6 shows the mass distribution at temperatures above 500 K, producingg the near-IR emission. T his component, dominated by hot iron or iron oxide grains,, is very similar for both objects. Meeus et al. (2001) have shown that the near- IRR spectrum of HAhBh systems is generic, suggesting a similar geometry and/or dust compositionn at the very inner parts of the circumstellar dust region. However, the total masss contained in the small grains in HI) 1 00546 is almost 3 orders of magnitude larger. 11 his excess mass is located in /one 2 (see Table 6.3) of the small grain component, which iss not present in AB Aur and is the cause for the much larger mid-IR excess as observed inn HI) 100546. Focussingg on at the mass-temperature distribution of the small grains, one notices thee difference in the distribution between the amorphous dust and the forsterite grains. Ioo fit the SF.D, the crystalline silicate had to have a much lower temperature. 1 his result holdss irrespective of the adopted laboratory measurements of the crystalline silicate. This mass-temperaturee distribution imposes the forsterire grains to have a larger radial extent comparedd to the amorphous dust. II he large grain component of HD 100546 has a distribution similar to that of ABB Atir. 1 he total mass contained in this component is the same within a factor of two. II he small difference in dust mass between both systems is due to a grain size difference, wheree in AB Aur the maximum grain size is smaller bv a factor of two. 11 he lower panel of Fig. 6."7 shows the mass-temperature distribution of Hale-Bopp. Contraryy to the Pi I) 100546 system, the dust species in Hale-Bopp all have the same

124 4 HD1005466 & HALE-BOPP

111)100.")) K; SMALLL GRAINS ZONKK I

— 0.6—0 0 s s 0.50 0 0.. 10 0.30 0 -- 0.20 0 £ £ii .. I" -- ii.nu u ;nn a.eo 2.90 3.oo 3.10 3.20

SMALLL CHAINS ZONKK •>

LARGEE GRAINS

oli\\ nu' :carbon n HII met allic iron f'orsll fiil c

ironn oxide

Figuree 6.6: Cumulative dust mass over temperature distribution of HD 100546. Indicated in thee figures are the relative contributions of the individual dust species. '1 he first two panels show thee contributions of the small grain component with sizes < fO/Jm located in /.one 1 and zone 2, resp ectivelvp .. The lower panel shows the distribution of the Iars;e grains (< 10 /./m). The inset thee top panel shows the high temperature dust component of HI) 100546.

125 5 CHAPTERR 6

AHH Aur SMALLL (,KAÏNS

o o ii i LARGEE (.RAINS

-- i.(]

££ 0.8

\\ 0.6

i).. I

0.2 2

?:S S Halee —Bopp ]] ,o

lopp T

Figuree 6.7: Cumulative dust mass over temperature distribution of AB Aur and comet Hale- Bopp.. I he two top panels show the distribution for AB Aur for the small and large grains, respectively.. I he bottom panel shows the distribution for the solar system comet Hale-Bopp. Indicatedd in the figures, in an identical manner as in Fig. 6.6, are the relative contributions of the individuall dust species.

126 6 HI)) 100546 & HAII-BOPP masss averaged temperature of approximately 160 K. The difference in the width of the distributionn of the individual dust species results from the assumption of a grain size distributionn for the amorphous dust, giving a broader distribution than for the other dust speciess (forsterite, iron oxide) where a CT)h shape distribution is assumed. Though the dustt species are the same as in HI) 100546, Fig. 6.7 shows the fundamental difference betweenn the dtist Hale-Bopp and that in the Herbig Be system: The grains in Hale-Bopp aree in thermal contact, those in HI) 100546 are not. As discussed in the previous section, givenn the optical properties of the dust species, the mass-temperature distribution derived forr Hale-Bopp could only be achieved if the grain species where placed at different he- liocentricc distances, inconsistent with observations. A co-spatial distribution of all grain species,, complying with the derived mass-temperature distribution can only be obtained iff thermal contact between the grains is imposed. However, forcing the grain species in HDD 100546 to be in thermal contact, the observed difference in the mass-temperature distributionn between the amorphous and crystalline dust could not be reproduced. This leadss to the conclusion that the grains in HI) 100546 are homogeneous, separate entities.

big.. 6.8 shows the relative forsterite mass fraction compared to the total mass in silicatess as a function of temperature for the best fit model. Clearly visible is the rise in thee crystalline mass traction at lower temperatures. A crucial question is which mecha- nismm is responsible for this outward increase of the crystalline fraction. This question is closelyy related to the formation processes of the crystalline silicate. Forsterite will form byy annealing amorphous silicate at temperatures above the glass temperature. At temperaturess above 1 100 K the annealing time-scales for the amorphous silicate become much smallerr than any dynamical time-scale (e.g. Hallenbeck et al. 2000), and will transform fastt into crystalline silicates. This implies that the dust in the very inner parts of the circumstellarr disk, with temperatures above 1100 K, will be entirely crystalline up to the evaporationn temperature of silicate grains at about 1400 K. Several radial mixing models havee been proposed to transport material outwards to the regime where crystalline silicatess are observed well below the glass temperature. Stevenson (1990) predicted that the masss fraction of the mixed material would decline outwards following a power law. Prinn (1990)) argued that the neglect of nonlinear momentum terms in the Stevenson model couldd underestimate the radial mixing. He predicts that the mass fraction of the mixed materiall could almost be constant as a function of radius. This same result is found in recentt models by Bockelée-Morvan et al. (2000). Plotted in Fig. 6.8 (dotted line) is a modell with a constant relative mass fraction of forsterite over amorphous silicate as a functionn of radial distance. Note that we have converted the spatial mass distribution off this model to a mass-temperature distribution as this has proven to be the more rele- vantt diagnostic for the analysis presented in this paper. Between these distributions is an opacityy effect. I he torsterite grains, having lower opacities in the UY and optical comparedd to the amorphous silicate, have a different temperature distribution. I his shift in thee temperature distribution between the amorphous and crystalline dust causes the rise- seenn around ~300 K. As can be seen from this figure, outward radial mixing of material fromm the inner parts of the proto-planetarv disk is not the explanation for the observed distributionn of crystalline silicates.

12"7 7 CHAPTERR 6

1.00 0

C C O O

'•*—> '•*—> o o 03 3 i_ _ w— — c/) ) in in 03 3 0.10 0 CD D

CD D •*—> •*—> en en O O

0.01 1 1.55 2.0 2.5 3.0 3.5 logT T

Figuree 6.8: Relative forsterite mass fraction as a function or temperature. I he solid line indicates ourr best model fit. The dotted line represents a model where the crystalline and amorphous silicatess are co-spatial with a constant relative mass fraction as a function of radial distance.

6.5.22 The disk structure of HD 100546 and the origin of the crystalline silicates s

Ass argued in the previous section the mass-temperature distribution of the forsterite grainss does not comply with the distributions predicted by radial mixing models. Also, Bouwmann et al. (2001b) show that the dust composition producing the 10/,/m silicate feature,, probing grains at the inner parts of the proto-planetary disk, is inconsistent with thermall annealing. Clearly a different mechanism for the formation of the observed crvs- tallinee dust fraction and distribution must be formulated. Also if we compate the energy distributionn of HD 100546 with that of AB Aur, we have to conclude that the protoplanetaryy disk in the former system is not simply a hydrostatic radiative equilibrium disk

128 8 HD1005466 & HAI F.-BOPP suchh as proposed by Chiang and Goldrcich (1997). Inn this section we will speculate that a massive planet has formed in the disk at a distancee of about 10AU from the star, and that this planet is responsible for the striking differencee both in the mass-temperature-distribution and in the chemical composition andd crvstallographic structure of the dust around HD 100546. A distance or 10AU seems reasonablee to assume, in particular since the temperatures in the disk around an A star willl be similar at 10 AU to the conditions at 5 AU in the early solar system. It has been suggestedd that the snow-line, i.e. the line where water ice can condense and remain stable inn the disk, has enhanced the densities in the solar disk locally to speed up the accretion off Jupiter.(Stevenson and Limine 1988). A similar process might have been operating in HD100546. .

Diskk structure

AA giant planet will start to form a gap in the disk (Lubow et al. 1999). II the local gass density starts to decrease, the disk height will not decrease as predicted by passive diskk models (e.g. Chiang and Goldreich 1997). The disk on the back side of the gap willl be in full sight of the stellar light, and the disk height will increase by typically a factorr of two over the height of a flaring disk at the same position (Dullemond 2000; Dullemondd et al. 2001). This surface will absorb more stellar radiation and produce a significantt additional component in the spectrum. We believe that this is the source of thee amorphous grain emission in zone 2 of the small grain component in the disk model.

Debriss dust

Iff a massive planet forms inside a disk, it will start to disturb smaller bodies in its vicin- ityy due to its strong gravitational attraction. These smaller bodies will therefore start movingg on orbits with higher eccentricities and inclinations. Some of the bodies will eventuallyy collide with the massive planet and merge into it. However, the changed distributionn of orbital elements will also start frequent collisions between smaller bodies. Sincee the gravitational potential of these bodies is much smaller than of the big planet, suchh collisions can become destructive. Such collisions can be a source also of small dust grains.. Depending on the size of the colliding parent bodies, the dust produced in such collisionss may be chemically processed. If the parent body was large enough to fully meltt during its accretional phase, it will be differentiated. The crust will be deprived of siderophilee elements like Fe/Co/Ni. The composition of the crust will therefore largely bee a Magnesium-rich silicate, very similar to what the observations indicate. AA massive planet will be able to strongly influence bodies within its Hill sphere which iss given by (Hill 1878) '•(£)"" "

wheree A is the semi-major axis of the planets orbit, M?\ is the mass of the planet and Ah iss the mass of the star. For a Jupiter-mass planet orbiting a solar-mass star, hi A = 0.07.

129 9 CHAI'ITRR 6

rhi'' mass available in this part of the disk must at least be the mass of the giant planett formed there, but there are many reasons to believe that the mass initiallv must havee been a hietor or several higher. During the cleanup phase of a planetarv svstem, aa large fraction of the available material is actually thrown out of the svstem and the assumptionn of a minimal solar nebula does not include this material. Since the core of Jupiterr iu our solar system is assumed to be between 15 and 30 Earth masses, we can assumee that at least 100 Earth masses of solid material is available in the region around aa giant planer during the planetary accretion phase. An important question is now, how muchh reprocessed dust can be expected from a collisional cascade of such bodies. Dohnanyii (1969) has shown that the collisional cascade leads to a size distribution fdi)) x /r-"0 or f I IN) °= ;;/'" (' in which the mass is dominated bv the largest particles. Iff the observations measure all particles up to a certain size limit

130 0 HI)) 100546 & HALE-BOPP

3800 AU

Figuree 6.9: Schematic representation of the HD 100546 system (not on scale). A giant planet hass formed a gap in rhe flaring disk at 10 AU, causing a increase of the disk scale height. An collisionall cascade induced by the giant planet is producing dust and throwing it up out of thee disk mid-plane. The small crystalline grains (gray area) above the disk are blown away by radiationn pressure and rain down further out at a few hundred AU on the surface of the disk. fragmentationn of larger bodies, and are scattered out of the disk. Small grains for which radiationradiation pressure is dynamically important will be driven away from the star and by a combinationn of gravity and disk geometry land further out on the disk surface. This is thee location where we see these grains today. A schematic representation of the above sketchedd disk structure and processes is plotted in Fig. 6.9.

6.66 Summary

Wee can summarise the results of our analysis of the ISO-spectra of HD 100546 and the comett Hale-Bopp as follows:

1.. The ISO/SWS spectra of HD 100546 and comet Hale-Bopp can be fitted with very similarr dust compositions and properties. In both cases, the bulk of the material is dominatedd by amorphous silicates with carbonaceous grains as the second most importantt constituent. In addition, significant amounts of forsterite, crystalline water ice,, metallic iron and iron oxide are present.

2.. There is one major difference in the dust properties of the two investigated objects. Thee chemical constituents of the dust in Hale-Bopp are found to be in thermal contactt while this is not the case for HD 100546.

131 1 CHAPTERR 6

3.. I he fraction of crystal line silicates, i.e. forsterite and/or very iron poor olivines, in HDD 1 00546 increases with decreasing temperature, i.e. with increasing radial distancee from the central star. Such a distribution is not in agreement with predictions byy radial mixing models. In such models the crvstals are formed bv thermal annealing inn the verv inner parts of the disk and arc subsequently transported out and mixed withh amorphous material.

4.. HD 100546 shows a component of small grains (< 10 //m) with temperatures of aboutt 200 K that emits in the mid-IR and that is not present in the spectra of other Herbigg Ae/Be stars.

5.. At about the distance where the bulk dust temperature is ~ 200 K, i.e. at ~ 10 AU, thee vertical extend of the disk must be in excess of that expected in a standard flaring diskk model. This is required to explain the large fraction of stellar light intercepted andd re-emitted by grains emitting in the mid-IR.

6.. On the basis of the above three results, we propose the following: the crystalline dust iss produced by collisional destruction of differentiated objects. These collisions arc- probablyy induced by gravitational interaction with a proto-Jupiter. The grain componentt emitting at ~ 200 K may be identified by the dust expected to be produced inn the collisional cascade that is thought to occur when large bodies collide. This pre- dominantlyy crystalline material may be driven out by radiation pressure and "land" furtherr out on the (flaring) disk surface. The gap cleared out by the proto-Jupiter al- lowss direct stellar light to produce an extended rim or wall at the far side of the gap. Itt is thought to be this rim that intercepts the radiation responsible for the strong mid-IRR component in the spectrum of HD 100546.

7.. I he similarity in spectral appearance between HD 100546 and Hale-Bopp implies thatt similar processes in the earlv historv of our solar svstem mav have been relevant forr the formation of Hale-Bopp. I he comet mav have assembled from debris dust producedd through collisions of differentiated objects.

132 2 CHAPTERR 7

Constraintss on HAEBE disk geometryy from Spectral Energy Distributions s

toto be submitted

J.. Bouwman, C.P. Dullemond, A. de Koter, C. Dominik, A. Nana, L.B.F.M.. Waters

ABSTRACT T Wee present self-consistent 2D radiative transfer calculations of disks surroundingg Herbig Ae/Be stars. We show by example of the star HD 104237 that the entiree spectral energy distribution of Herbig Ae stars can be fitted with a simple diskk model, with no additional components required. Pitting the energetically importantt near-IR bump between 1 and 8 fjm requires a scale height of the innerr disk in excess of predictions by current flaring disk models.

7.11 Introduction

Herbigg Ae/Be stars (HAEBE stars) are intermediate mass pre-main-sequence stars that aree still surrounded by substantial amounts of circumstellar matter, left over from the starr formation process. It is believed that HAEBE stars are excellent candidates for protoplanetaryy systems. A question which has triggered an intense debate in the literature iss the geometry of the gas and dust which surrounds HAEBE stars. Obviously, the best wayy to answer this question is by direct imaging. Millimeter wave observations of Herbig Aee stars have convincingly demonstrated that the cold CO gas is in a rotating geometry, whilee the distribution of cold dust deviates from spherical symmetry (e.g. Mannings and Sargentt 1997, 2000). Near-IR interferometric observations are less conclusive, and arc- consistentt with a more spherical distribution of the hot material near the star (Millan- Ciabett et al. 2001). Recent Hubble Space 'telescope NI CMOS images show a flattened distributionn of material in some HAEBE stars (Augereau et al. 2001). The evidence for diskss in the more luminous Herbig Be stars is weaker. Mostt studies however use the spectral energy distribution of HAEBE stars to de- rivee constraints on the geometry of their circumstellar environment, with contradictory

133 3 CHAPTHRR " resultss however. Several authors have shown that the SEDs or HAEBE stars can be tit- tedd well using spherical envelopes (e.g. Miroshnichenko et al. 199~; Malhiit et al. 1999; Bouwmann et al. 2000a) or in combination with a geometrically thin disk (Vinkovic et al. 2000).. In an important paper Chiang and C.oldreich (199") (hereafter re f e red to as CG) usee a flaring two-layer passive disk model to reproduce the SEDs of T Tauri stars. While thee CG model was not developed lor HAEBE stars, Xarta et al. (2001) and Chiang et al. (2001)) show that it is remarkably successkil in reproducing the observed mid-IR. to mil- limeterr wavelength spectra ot a sample of isolated Herbig Ae stars, whose spectra are not affectedd bv emission from the loose surroundings. Unfortunately,, the model systematically tails to reproduce the observed fluxes in the 11 to 8 ]Lim region bv a factor of 5 to 10. 1 his is a serious breakdown since this spectral regionn represents a substantial fraction of the luminosity of the disk. Clearly not enough stellarr Hux is being absorbed in the inner parts of "standard disk-only models. Other modelss have introduced spherical haloes to match the near-lR (c.f. Vinkovic et al. 2000). Sincee the 1-8 /./m spectra of the isolated HAhBh stars are all very similar, irrespective of thee shape at longer wavelengths (Meeus et al. 2001), this is a generic spectral property whichh much be addressed by models. II his letter presents the first disk model that can succesfully reproduce the entire SKI),, including the near-IR. I he models presented here are based on a self-consistent 21)) radiative transfer code (l)ullemond and Iurolla 2000) which, combined with a detailedd description of dust extinction and emission (Bouwman et al. 2000), allows one to calculatee the full SED of axi-symmetric disks. We will show that a disk is indeed capable off reproducing the full SEI) of the isolated Herbig Ae star HI) 104237, if the vertical heightt of the inner part of the disk is considerably higher than predicted bv current disk models. .

7.22 The disk model

Wee construct a parametrized axi-svmmctric density distribution of dust, given bv:

p(r,6)) = p(1(r/r(y[.Y+(l-.v)sin'"e] (7.1) wheree p() is the density at the inner edge of the disk in the midplane and .v the density contrastt between the equator and the pole. While not hydrostaticallv self-consistent, this approachh has the advantage that it allows LIS to study the effect of the scale-height of the diskk on the resulting Sid). It will allow us to set constraints on the vertical extent of the disk.. A study which investigates physics behind an increased scale height of the inner disk iss in preparation (I)ullemond et al. 2001). In the given geometry we solve the full 21) radiativee transfer since the 2-laver approach introduced bv CG will onlv work if the scale- heightt is continuously increasing with distance from the star. 1 he temperatures of the dustt grains of the various species and sizes are sell-consistently computed bv requiring radiativee equilibrium with the 2-1) radiation field. loo account for the large sub-millimeter flux observed m 1 1AEBE stars we include a significantt component of lar^e grains. 1 hese cold, massive grains are expected to have

134 4 HERBIGG AH/BK DISK GEOMETRY.

300 E

aa [AU]

Figuree 7.1: The disk model is composed of a midplane laver of large grains, dominating the totall dust density far from the star, and a more extended layer of smaller grains, significantly contributingg to the dust density in the inner parts of the disk. The large grains are cold and producee the far-lR and millimeter flux, while the small warmer grams produce the near- to mid- IRR continuum and are responsible for the 10 and 20 jjm silicate emission. The actual geometry shownn is that for our best fit model, discussed in Sect. 7.3. The shading gives the total dust densityy relative to the midplane value at the inner boundary (cf. Table 1).

decoupledd from the gas and to have settled in the midplane. The size of the cold grains is sett by limits on the spatial extent of the millimeter emission in HAEBE stars, and by the slopee of the far-IR and sub-millimeter SED (see Bouwman et al. 2000a, for a discussion). II he density distribution of the large grain component for all models discussed in this letterr is defined by m = 300 and x = 1.0- 10 . The near-IR is dominated by emission fromm small grains, which we distribute in a disk of equal size compared to that of the largee grains, but which has a larger scale-height. The geometry of this setup is shown inn Figure 7.1. We consistent!}' take into account the destruction of grains above the maximumm temperature at which they can exist. For both components, we use a similar dustt composition as derived by Bouwman et al. (2000a) from a fit to the Infrared Space Observatoryy (ISO) spectrum of AB Aur and HI) 163296 (see Table 1). Thee grain size distributions and mass of both the small and the large grains have been tunedd to fit the SED of the A4 IVe+sh star HD 1 04237 - as an example. The observations

135 5 CHAPTKRR ""

II ABLE' "". 1: Model parameters of our best Ht to the SED of HL) 10423". Listed are the densitv structuree (see Fq. 1) and the dust composition or the disks midplane and ot the extended I aver.

Fxtendedd laver Disk midplane Radius: : 0.3-80AU U 0.3-80AU U P P -1.85 5 -1.0 0 poo [grctrT1] 1.00 \0~u •^.766 uru m m 10 0 300 0 6 .V V 1.00 10"' i.oo nr lS ( Mdllsr/iM0 0 1.255 10 1.22 \ir ' n{a) ) -22 -\ occ ,-/"-"*

Species s Mfrac c aa hum] Mr™ ™aa hum] (olivine e 0.72 2 0.01-6 6 0.74 4 0.1-80 0 Carbon n 0.10 0 0.01-2 2 o.n n 0.1-80 0 Iron n 0.18 8 0.01-0.08 8 -- — — H2() ) -- -- 0.15 5 0.1-80 0

aree raken from Meeus er ah (2001), and references therein. I he basic parameters of this starr are adopted from van den Ancker et ah (1998): 7;.(y = 8 500 K; I = 35 U-; M = 2.3 MM . , and d = 116 pc. hor the interstellar extinction we assume Ay = 0.1 2, and a standard extinctionn curve (Savage and iMathis 1979). Usingg these stellar properties, we calculated a number of disk models in which we variedd the scale-height of the small grain component, controlled by the parameter w, whilee leaving the midplane grain component and the grain composition constant.

7.33 Discussion

Figuree 7.2 shows that changing the scale-height of the small grain component has a strongg effect on the shape of the IR energy distribution. For large scale-height, i.e. small ;;;,, the dust intercepts a fairly large fraction of the stellar luminosity, and re-radiates it att mainly near- and mid-1 R (2-20 /jm wavelengths). Reducing the scale-height of the diskk bv increasing ;•;; decreases the emerging IR flux, or, equivalently, the covering frac• tion,, defined as the ratio of the total !R luminosity over the stellar luminosity. This occurss simplv because the extended layer absorbs less stellar light, while the amount off light absorbed in the midplane laver, which is opticallv thick in radial directions at alll frequencies, remains essential])' unchanged. For m = 300 the near-IR flux is at the

136 6 HERBIGG AE/BE DISK GEOMETRY.

Il l -- 0.1 1 1.00 10.0 100.0 0 Logg ,\ [/'m i

Figuree 7.2: The effect of varying the scale-height of the warm, small grain component is shown. Ass the scale-height decreases the IR flux excess reduces considerably as well as the strength of the 11 0 and 20 jjm silicate bands. The far-lR and millimeter continuum are less effected. The models aree labeled with m (see Eq. 1). All spectra are calculated for an inclination of 40°.

levell of the CG model, however the mid-IR flux is far below the CG result since the parametrizedd disk does not flare in the outer parts. In this context, we note that large differencess are observed in the 20-60 ;/m part of the SEDs of HAEBE stars, with other- wisee similar SEDs. Meeus et al. (2001) interpreted the presence or lack of a strong 20-60 jjmjjm bump in the SED as evidence for a strong or weak flaring of the disk, respectively. Indeed,, HI) 104237, which has no strong 20-60 /jm bump, can be fitted with our sim- plee constant scale-height over distance model, i.e. without strong flaring. We would like too stress, however, that irrespective of the disk structure at larger radii (i.e. flaring or not flaring)flaring) an identical density structure at the inner parts of the disk is required.

137 7 CHAPTERR 7

:: too.c [urn] [urn]

Figuree 7.3: Best model fit to the observed SED of the HAEBE star HD 104237. Plotted in thee figure is the ISO SWS spectrum (Meeus et al. 2001), together with visual and near 1R photometricc data (Hu et al. 1989), IRAS photometric data and 1.3 mm continuum measurements (Henningg et al. 1993). The photometric points are indicated with square, we have assumed a ten percentt error on the photometry. The fit to the 1.3 mm point is shown in the inset. The ISO spectrumm is shifted by a factor of 1.4 to match the photometry. The disk is seen at an inclination oll 60°. Parameters of our best model fit are listed in Table 1.

Inn Figure 7.3 we compare the observed SED of HD 104237 with our models. We findfind an excellent fit using the parameters given in Table 1. As the scale-height decreases forr increasing m, one effect is that the slope of the near-IR continuum becomes shallower, reflectingg the fact that optical depth effects first start to play a role at the shortest IR wavelengths.. Although the best fit model has m = 10, modest deviations from this value alsoo yield a satisfactory fit. However, m = 25 clearly results in a near-IR slope which iss too shallow. I he absolute level of near-IR flux sensitively depends on the amount off material in the inner part of the disk that can receive (almost) unattenuated stellar

138 8 HERBIGG AF./BE DISK GEOMETRY.

100.0 0

Figuree 7.4: Extent of the surface layer for our best fit model (lightly shaded area) and a CG modell (dark shaded area), as a function of the midplane distance a. The areas are limited by the stellarr flux-weighted radial I = 1 surface (solid line) and the I = 0.3 surface (dashed line). Note thatt the disk is optically thin from the inner edge at 0.3 AU up to ~ 1.5 AU. The CG model iss scaled to have the same covering fraction at the outer edge of the disk, vielding comparable far-lRR and sub-millimeter fluxes.

light.. Figure 7.4 shows the stellar flux-weighted radial X = 1 and X = 0.3 surfaces for both thee best fit model and a comparable CG model, clearly illustrating that our model is 'puffedd up' near the inner edge of the disk. The more extended surface layer implies that forr more beams of radially streaming photons, measured as a fraction of the total solid angle,, optical depth unit}' is reached close to the star. Therefore, a larger fraction of the incidentt stellar luminosity is absorbed close to the star. While the fit may not be unique, itt does demonstrate that a disk is capable of fitting the entire SFD of a HAEBE star if thee inner disk region is puffed up compared to comparable hydrostatic surface-irradiated disks. . Itt is important to stress that the near-IR shape of the SFD of'isolated' HAEBF. stars iss remarkably similar (Meeus et al. 2001). The onset of the IR excess in these stars occurss in a very narrow wavelength range centered around 1-2 ^um. This suggests a similar innerr disk structure, with a disk inner radius probably dictated by the sublimation temperaturee of the most stable refractory materials, such as carbon-rich grains and metallic

139 9 CHAPTKRR 7 ironn (Bouwman et al. 2000a). This is consistent with the detection of an inner hole in thee dust distribution tor a number of HAEBE systems in our sample which have been observedd with a near-IR interferometer (Millan-Gabet et al. 2001). Therefore the con- clusionn that the scale-height of the inner part of the disk must be increased compared to thatt of the CG model is likely applicable to all HAKBE star disks.

Achiowledgemtmii.Achiowledgemtmii. The authors would like to acknowledge the financial support from WX'O PionierPionier grant 600-78-333. CPD acknowledges support from the European Commission under T.MRR grant ERBFMRX-CT98-0 1 95 (Accretion onto black holes, compact objects and protostars').. AdK also gratefully acknowledges support from NWO Spinoza grant 08-0 to E.P.J, van denn Heuvel.

140 0 Nederlandsee samenvatting

Eénn van de meest intrigerende vraagstukken in de sterrenkunde, of in de wetenschap in hett algemeen, is hoe ons zonnestelsel onstaan is. Het onderzoek dat in dit proefschrift gepresenteerdd is, staat in nauw verband met dit vraagstuk. Sterren, zoals de Zon, ont- staann door het onder zijn eigen zwaartekracht ineenstorten van een grote wolk bestaande uitt gas en stofdeeltjes. Een dergelijke wolk , ook wel proto-stellaire wolk genoemd, heeft meestall een massa ter grootte van enige malen de massa van de Zon en heeft een door- snedee van enkele lichtjaren. De proto-stellaire wolk bestaat voor 99 procent uit de ele- mentenn waterstof en helium. Deze zijn gevormd tijdens het ontstaan van het heelal. Slechtss één procent van de massa van de wolk bestaat uit andere elementen, zoals ijzer, silicium,, magnesium en koolstof. Deze elementen, welke gevormd worden in sterren, wordenn in de laatste levensfasen van sterren de ruimte ingeblazen, waarna ze, na onge- veerr een miljard jaar, in een proto-stellaire wolk terecht komen waaruit weer een nieuwe sterr gevormd zal worden. Deze elementen vormen de bouwstenen voor stofdeeltjes ter groottee van ongeveer een micrometer, bestaande uit materialen zoals ijzer, grafiet en silicaten.. Van de stofcomponenten is silicaat het meest voorkomende materiaal, zowel in dee ruimte als op aarde (90 procent van de aardmantel bestaat uit silicaten). In de kern vann de ineenstortende wolk, waar de druk en temperatuur het hoogst zijn, zal zich een dichtee gasbol vormen. In ongeveer één miljoen jaar zal de druk en temperatuur door het verderr ineenstorten van deze gasbol zo hoog zijn opgelopen dat in de kern waterstof tot heliumm kan fuseren. De hiermee opgewekte energie stopt verdere ineenstorting en een sterr is geboren. Rondd de jonge ster zal zich nog een deel van de oorspronkelijke gas- en stofwolk bevindenn waaruit de ster gevormd is. Dit gas en stof kan een schijf vormen rond de pass gevormde ster. Uit een dergelijke schijf kan zich vervolgens, door het aaneenplakken vann stofdeeltjes, een planetenstelsel ontwikkelen. De schijven rond pas gevormde sterren wordenn daarom ook wel proto-planetaire schijven genoemd. Hoee de vorming van planeten precies in zijn werk gaat is één van de grote vraagstuk- kenn in de sterrenkunde. Dit werd bijzonder actueel na de ontdekking van een planeet buitenn ons zonnestelsel in 1995 door Mayor en Queloz. Om inzicht te verkrijgen in de beginfasee van planeetvorming is in dit proefschrift de samenstelling van de stofkorreljes,, de bouwstenen van planeten, in proto-planetaire schijven, en de processen die deze samenstellingg bepalen bestudeerd. Ook de vorm van de schijven is onderzocht in dit proefschrift.. Het gas, dat 99 procent van de totale massa voor zijn rekening neemt, en

141 1 Ni.ni-.Rii AND.SF SAMI.WAITINC; daardoorr zeker dvnarmsch van belang is, is echter niet bestudeerd in dit proefschrift. Dee stofdeeltjes produceren zelf geen energie, maar verstooien en absorberen het licht vann de ster. De stofkorreltjes, verwarmd door het sterlicht, stralen de geabsorbeerde ener- giee weer uit in de vorm van warmtestraling, dus in het infrarood. Het beste spectraal- gebiedd om stofschijven te bestuderen is dus het infrarood. Aangezien de aardatmosfeer slechtss in bepaalde en relatief kleine golflengte-intervallen transparant is voor infrarood straling,, is het beter gebruik te maken van satellieten, omdat de/e geen last hebben van dee atmosfeer. De waarnemingen aan de proto-planetaire schijven welke gepresenteerd enn geanalyseerd zijn in hoofdstuk drie tor en met zeven van dit proefschrift, zijn gedaan mett de Infrared Space Observatory (ISO) satelliet, welke gelanceerd werd op l7 november 19966 en gefunctioneerd heeft tot 8 april 1998. Omm deze waarnemingen te kunnen analyseren en interpreteren hebben wij geavan- ceerdee modellen ontwikkeld. In hoofdstuk 2 van dit proefschrift zijn deze modellen beschreven.. Deze zogenaamde stralingstransportmodellen beschrijven de verstrooiing en absorptiee van licht door de stofschijven en stellen ons in staat, gecombineerd met een gedetailleerdee beschrijving van de stofeigenschappen en dichtheidsverdeling, de waarnemingenn te simuleren. Hierdoor konden de samenstelling en ruimtelijke verdeling van hett stof in de door ons bestudeerde objecten bepaald worden. Inn dit proefschrift bestudeerden wij een specifieke groep van jonge sterren met een proto-planetairee schijf, namelijk zogenaamde Herbig Ac/Be (HAKBE) sterren. Deze sterren,, voor het eerst bestudeerd door George Herbig in de jaren 60 van de vorige eeuw, zijnn jonge zware sterren met een massa van twee tot tien maal die van de Zon. In hoofdstukk 3 staat een overzicht van de ISO-waarnemingen van 14 HAhBK systemen. Door vergelijkingg van deze waarnemingen konden enige belangrijke kwalitatieve conclusies ge- trokkenn worden met betrekking tot zowel de geometrie van de schijf als de samenstelling vann het stof. Het bleek dat HAKBh systemen in twee groepen verdeeld kunnen worden, watt geïnterpreteerd kan worden als berustend op een verschil in schijfgeometrie. De eer- stee groep heeft een schijf waarvan de dikte naar buiten toe toeneemt, oftewel tiitwaaiert, terwijll de tweede groep een geometrisch vlakke stofschijf heeft. De reden van dit ver- schill in schijfgeometrie is nog niet geheel duidelijk, hen schijf zal uitwaaieren als zijn oppervlakk beschenen wordt door de ster. Als echter, door bijvoorbeeld een verdikking aann de binnnenkant, het sterlicht geblokkeerd wordt en het oppervlak in de schaduw vann de/e verdikking ligt. zal de schijf niet uitwaaieren maar geometrisch vlak zijn. Ook dee grootte van de stofkorreltjes kan van invloed zijn. Als stofkorreltjes klein zijn, (b.v. een micrometerr of zelfs kleiner) worden hun bewegingen bepaald door het gas. Kchter, als de stofkorreltjess aan elkaar plakken en zo grotere korrels vormen, kunnen deze ontkoppeld rakenn van het gas. De grotere stofkorrels, niet langer ondersteund door het gas, zullen doorr de zwaartekracht naar het middenvlak van de schijf zakken en een geometrisch vlakkee schijf vormen. Litt de studie blijkt ook dat, ongeacht of de schijven mi uitwaaieren of plat zijn, dee binnenkanten van deze stofschijven een identieke geometrie en/of stofsamenstelling hebben.. Uit de waarnemingen blijkt ook dat het stof in de proto-planetaire schijven aan- zienlijkk afwijkt \.\n stof zoals dat aanwezig was in de proto-stellaire wolk, zowel qua samenstellingg als qua grootte van de stofkorreltjes. De maximale stofgrootte in de HAhBh

U2 2 NlDKRII ANDM- SA.\1L\\'ATT[\(,

systemenn is enige millimeters , terwijl dat in een proto-stellaire wolk slechts enige tien- denn van een micrometer bedraagt. De silicaten die in proto-stellaire wolken voorkomen zijnn amorf-, dat wil zeggen hebben een zeer onregelmatige structuur, zoals bijvoorbeeld glas.. In de proto-planetaire schijven daarentegen, komen ook kristallijne silicaten, dus mett een zeer geordende structuur, voor. Omm beter inzicht te krijgen in de processen in de stofscbijven die leiden tot kris- tallisatiee van de amorfe silicaten hebben wij een gedetailleerde studie gemaakt van de silicaten.. Deze studie is gepresenteerd in hoofdstuk 4. Amorfe silicate!: kunnen kristalliserenn als ze verhit worden rot voldoend hoge temperaturen, in de orde van 1000 keivin. Dergelijkee temperaturen worden bereikt aan de binnenrand van de schijf, dicht bij de ster.. hchter, de temperatuur van de kristallijne silicaten, zoals deze worden waargenomen,, is slechts in de orde van een paar honderd keivin. Deze temperatuur is veel te laagg voor thermische kristallisatie. Dit betekent dat, of een ander soort kristallisatiepro- cess een rol speelt, óf dat de silicaten, verhit tot hoge temperaturen, aan de binnenrand kristalliserenn en vervolgens door dynamische processen naar de buitenkant van de schijf getransporteerdd worden, waar, verder verwijderd van de ster, de stoftemperatuur veel lagerr is. Door nu een gedetailleerde studie te maken van welke kristallijne silicaten ge- vormdd worden en in welke mate, en dit te vergelijken met laboratoriumexperimenten vann het kristallisatieproces, kon het vormingsproces van de kristallijne silicaten bepaald worden.. Het bleek dat in de meeste HAERE systemen de kristallijne silicaten gevormd wordenn door thermische verhitting aan de binnenrand. Echter, één ster, HD 100546, waarvann een gedetailleerde studie beschreven is in hoofdstuk 6, heeft kristallijne silicaten diee op een andere wijze gevormd zijn. Het materiaal in deze ster lijkt sterk op dat van dee komeet Hale-Bopp. Dit toont aan dat meerdere processen een rol spelen in protoplanetairee schijven en dat deze ook een rol hebben gespeeld in de vroege geschiedenis vann ons zonnestelsel. Inn hoofdstuk 5 is een kwantitatieve analyse beschreven van het stof in twee HAEBE systemen.. De stofkorreltjes in deze systemen bestaan uit amorf silicaat, ijzer, ijzeroxide off ijzersulfide, ijs en een amorfe koolstofrijke component. Eén van de svstemen, HDD 163296, heeft ook kristallijn silicaat als subcomponent. Het bleek dat het stof in dezee systemen zowel een bimodale grootteverdeling als een temperatuursverdeling heeft. Omm de ESO-waarnemingen te kunnen reproduceren is zowel een subcomponent met kleinee stofdeeltjes, ter grootte van ongeveer een micrometer, met een hoge temperatuur, alss een koude stofcomponent met een maximale korrelgrootte van ongeveer een milli- meterr noodzakelijk. De totale stofmassa rond deze sterren is ongeveer tien tot honderd maall zo groot als de massa van de aarde. Verreweg het grootste deel van de massa bevindt zichh in de koude, grote stofkorrels. De massa van de kleine, hete stofkorreljes is slechts eenn kleine fractie van de totale stofmassa. Deze hete stofcomponent kan geïnterpreteerd wordenn als de oppervlaktelaag van een proto-planetaire schijf, welke beschenen wordt doorr de centrale ster. De koude stofcomponent, bestaande uit grote stofkorrels, kan geïnterpreteerdd worden als het middenvlak van de stofschijf. De grote korrels, niet langer ondersteundd door het gas en daardoor naar het middenvlak gezakt, zijn afgeschermd van hett sterlicht en daardoor veel kouder dan het stof aan het oppervlak. Eenn vergelijkbare studie als in hoofdstuk 5 is gepresenteerd in hoofdstuk 6, maar

143 3 NHDF.RI.ANDM-- SAMI-WATTING

dann voor de Herbig Be srer HD 100546 en de komeet Hale-Bopp. Hen here en koude subcomponent,, respectievelijk her schijfoppervlak en de middenlaag, is ook aanwezig inn her snor rond HD 100546. Echter, een extra component van kleine, warme stof- deeltjess is nodig om de waarnemingen te kunnen reproduceren. De/e extra component kann niet geïnterpreteerd worden als een schijfoppervlak of -middenlaag. Van alle door onss bestudeerde systemen heeft HD 100546 de grootste massafractie kristallijne silicaten.. In hoofdstuk 4 bleek al dat de samenstelling van de silicaten inconsistent was met thermischee kristallisatie bij hoge temperatuur, aan de de binnenrand van de schijf. In hoofdstukk 6 is aangetoond dat de massa-temperatuurverdeling van de kristallijne silicaten,, in samenhang met die van het amorfe materiaal, ook inconsistent is met model- lenn die de aanwezigheid en ruimtelijke verdeling van kristallijne silicaten verklaren door middell van thermische kristallisatie aan de binnenrand. De resultaten van onze analyse kunnenn echter wel verklaard worden door aan re nemen dat zich een planeet gevormd heeftt in de schijf rond HD 100546. Een dergelijke planeet zal de schijf verstoren en doorr gravitationele interactie een cascade van botsingen van planetoïden induceren, wat eenn grote hoeveelheid kleine, kristallijne stofdeeltjes produceert. Dit kan zowel de extra componentt van warme, kleine stofdeeljes als de verdeling en samenstelling van de kris- tallijnee silicaten verklaren. De opmerkelijke overeenkomst in srofsamenstelling tussen de komeett Hale-Bopp en HD 100546 suggereert dat processen zoals nu waargenomen in HDD 100546 ook hebben plaatsgevonden tijdens het onstaan van ons zonnenstelsel, toen Hale-Boppp gevormd werd. Zoalss al eerder genoemd in deze samenvatting, bleek dat de binnenkanten van de protoplanetairee schijven rond de HAEBE sterren een identieke geometrie en/of samenstellingg hebben. Echter, de standaardmodellen van dergelijke schijven bleken niet in staat dee schijfstructuur te voorspellen welke nodig is om de waarnemingen te kunnen verklaren.. Hoofdstuk 7 bevat echter een schijfmodel met een dichtheidsstructuur die de waarnemingenn wel kan reproduceren. Samenvattend,, de resultaten gepresenteerd in dit proefschrift suggeren het volgende evolutionairee scenario voor de proto-planetaire schijven m HAEBE systemen. De vroeg- stee fase in de evolutie van een proto-planetaire schijf in een HAEBE systeem wordt gekenmerktt door een uitwaaierende schijfgeometne en een srofsamenstelling welke sterk overeenkomtt met die van de proto-stellaire wolk. In de volgende fase zal de gemiddelde stofgroottee toenemen door het aaneenplakken van stofdeeltjes. Deze grotere stofdeeltjes zullenn naar het middenvlak van de schijf zakken wat leidt tot een vlakke schijfgeometrie. Amorfee silicaatstofdeeltjes, verwarmd tot hoge temperaturen aan de binnenrand van de schijf,, kristalliseren en worden door dynamische processen naar buiten getransporteerd. Ditt leidt tot een geleidelijke toename van de massafractie van de kristallijne silicaten. Dee volgende stap in de evolutie van de proto-planetaire schijf is de vorming van proto- planeten,, welke de schijf door gravitationele interactie kunnen verstoren, wat tot bot- singenn van planetoïden kan leiden. Hierdoor zal een grote hoeveelheid kleine deeltjes geproduceerdd worden met een grote kristallijne massafractie. De kleine deeltjes, welke gekoppeldd zijn aan het gas, zullen weer leiden tot een uitwaaierende schijfstruktuur. Uit vergelijkingg van de resultaten van de studie van de HAEBE systemen met bijvoorbeeld waarnemingenn van kometen, die nog sporen bevatten van de vroege geschiedenis van ons

144 4 NF.DERI.ANDSI-- SAMENVATTING zonnestelsel,, kan de conclusie worden getrokken dat de processen die nu waargenomen wordenn in de schijven van de HAF.BE systemen, ook een rol hebben gespeeld tijdens het onstaann van ons zonnestelsel.

145 5

Bibliography y

Adamss F.C, Lada C.|., Shu F.H., 1987. Spectral evolution of young stellar objects. Ap|312C88. . Allamandolaa L,)., Lielens CG.M., Barker ).R., 1989. Interstellar polycyclic aromatic hydrocarbonss - The infrared emission bands, the excitation/emission mechanism, and thee astrophysical implications. ApJS71, 733. Allenn (I.C., Morris R.V., Lauer HA'., McKay D.S., 1993. Microscopic iron metal on glasss and minerals - A tool for studying regolith maturity. Icarus 104, 291. Andree P., Ward-Thompson D., Barsonv M., 1993. Submillimeter continuum observationss of Rho Ophiuchi A - The candidate prorosrar VLA 1623 and prestellar clumps. ApJ406,, 122. Artvmowiczz P., 1988. Radiation pressure forces on particles in the Beta Pictoris system. ApJ355,, 179. Augereauu J.C, Lagrange A.M., Mouillet D., Ménard F, 2001. HST/NICMOS2 coro- nagraphicc observations of the circumstellar environment of three old PMS stars: HD 100546,, SAO 206462 and MWC 480. A&A365, 78. Aurnannn H.H., Beichman C.A., Giliett F.C, et al., 1984. Discovery of a shell around Alphaa Lyrae. ApJ278, L23. Beckwithh S.V.W., Sargent A.I., Chini R.S., Guesten R., 1990. A survey for circumstellar diskss around young stellar objects. AJ99, 924. Begemannn B., Dorschner )., Henning I., Mutschke FL, Thamm F., 1994. A laboratory approachh to the interstellar sulfide dust problem. ApJ423, 171. Begemannn B., Henning T, Mutschke H., Dorschner J., 1995. Magnesium-iron oxides- astrophysicall origin and optical constants. Planet. Space Sci.43, 1257 Belll K.R.,'Cassen PM., Klahr H.H., Henning T, 1997. The Structure and Appearance off Protostellar Accretion Disks: Limits on Disk Flaring. ApJ486, 372. Bertiee J.E., Fabbé H.J., Whalley F„ 1969. Absorptivity of Ice I in the range 4000-30 cmm 7. JChPh 50, 4501. Beustt H., Lagrange A.M., Plazy F, Mouillet D., 1996. The [3 Pictoris circumstellar disk. XXII.. Investigating the model of multiple cometarv infalls. A&A310, 181. Biermannn P., Harvvit M., 1980. On the origin of the grain-size spectrum of interstellar dust.. Ap]24l, F105. Bockelée-Morvann D., (iautier I)., Hersant F, Huré |.M., Robert F, 2000. 'Furbulent radiall mixing in the solar nebula as the source of crystalline silicates in comets. In

14^ ^ BIBLIOGRAPHY Y

AAS/Dii'isiofiAAS/Dii'isiofi of Planetary Sciences Meeting, volume 32, p. 4407. Bohmm T., Catala C, 1 993. A spectral atlas of the Herbig Ae-star AB-Aurigae - the visible domainn from 391 -\M to 874-NM. A&AS 101, 629. —,, 1995. Rotation, winds and active phenomena in Herbig Ae/Be stars. A&A301, 155. Bohrenn C.E, Huffman D.R., 1983. Absorption and scattering of light by small particles. Neww York: Wiley, 1983. Bouwmann J., de Koter A., van den Ancker M., Waters L., 2000a. The composition of thee circumstellar dust around the Herbig Ae stars AB Aur and HI) 163296. A&A360, 213. . —,, 2000b. The Spatial Distribution of Dust around the Herbig Ae Stars AB Aur and HDD 163296. In ASP Conf. Ser. 196: Thermal Emission Spectroscopy and Analysis of Dust,Dust, Disks, and Regoliths, pp. 63-70. Bouwmann J., Dullemond C, de Koter A., Dominik C, Natta A., Waters 1.., 2001a. Constraintss on HAEBE disk geometry from Spectral Energy Distributions. ApJSub- mitted. . Bouwmann J., Meeus G., de Koter A., Hony S., Dominik C, Waters L., 2001b. Process- ingg of silicate dust grains in Herbig Ae/Be systems. A&AIn press. Bradleyy J.P., Humecki H.J., Germani M.S., 1992. Combined infrared and analytical electronn microscope studies of interplanetary dust particles. ApJ394, 643. Bregmann J.D., Witteborn KG., Ailamandola L.J., et ah, 1987. Airborne and ground- basedd spectrophotometry of comet P/Halley from 5-13 micrometers. A&A187, 616. Brownleee D.E., 1978. Interplanetary Dust: Possible Implications for Cornets and Pre- Solarr Interstellar Grains. In IAU C.olloq. 52: Protostars and Planets, p. 1.34. Brucatoo j.R., Golangeli E., Mennella V., Palumbo P., Bussoletti E., 1999. Silicates in Hale-Bopp:: hints from laborator)' studies. Planet. Space Sci.47, 773. Galvett N., Harrmann L., Strom S.E., 2000. Evolution of Disk Accretion. In Protostars andand Planets IV, p. 377. Cameronn A.G.W., Truran J.W., 1977. The supernova trigger for formation of the solar system.. Icarus 30, 447. Chiangg E., Goldreich P., 1997. Spectral Energy Distributions of I Tauri Stars with Passivee Circumstellar Disks. ApJ490, 368. Chiangg E.I., Joung M.K., Creech-Eakman M.J., et al., 2001. Spectral Energy Distri- butionss of Passive T Tauri and Herbig Ae Disks: Grain Mineralogy, Parameter De- pendences,, and Comparison with Infrared Space Observatory EWS Observations. ApJJ 547, 1077. Clcggg P., Ade P., Armand C, Baluteau J.P., Barlow M., et al., 1996. The ISO Long- Wavelengthh Spectrometer. A&A315, E38. Cohenn M., 1980. Infrared observations of young stars. VIII - Spectra in the ten-micron region.. MNRAS1 91, 499. (Corcorann \L, Ray P.P., 1997. Forbidden emission lines in Herbig Ae/Be stars. A&A321, 189. . Crovisierr [., Leech K., Bockelee-Morvan I)., et al., 1997. I he spectrum of Comet Hale- Boppp ((71995 01) observed with the Infrared Space Observatory at 2.9 AU from the Sun.. Science 275, 1904.

148 8 BIBLIOGRAPHY Y dee Graauw I., Haser L., Beintema D., Roelfsema P., van Agthoven H., et al., 1996. Observingg with the ISO Short-Wavelength Spectrometer. A&A31 5, L49. Dohnanyii J.W., 1969. Collisional model of asteroids and their debris. J. Geophys. Res. pp.. 2531-2554. Dominikk G., Sedlmayr L., Gail LLP., 1989. The size distribution of dust particles in a dust-drivenn wind. A&A223, 227. Dominikk G., Helens A.G.G.M., 1997. The Physics of Dust (Coagulation and the Struc- turee of Dust Aggregates in Space. ApJ480, 647. Dorschnerr J., Begemann B., Henning T., Jager (7, Mutschke H., 1995. Steps toward interstellarr silicate mineralogy. II. Study of Mg-Fe-silicate glasses of variable composition.. A&A300, 503. Dullemondd G.P., 2000. Are passive protostellar disks stable to self-shadowing? A&A361, L17. . Dullemondd C.P., Dominik G., Natta A., 2001. Passive irradiated circumstellar disks withh an inner hole. A&AIn press. Dullemondd G.P., Turolla R., 2000. An efficient algorithm for two-dimensional radiative- transferr in axisymmetric circumstellar envelopes and disks. A&A360, 1 187. Dunkinn S.K., Barlow JVI.J., Ryan S.G., 1997. High-resolution spectroscopy of Vega-like starss -1. Effective temperatures, gravities and photospheric abundances. MNRAS286, 604. . Duschll W.J., Gail H.., Tscharnurer W.M., 1996. Destruction processes for dust in protoplanetaryy accretion disks. A&A312, 624. Evanss N.J., 1999. Physical Gonditions in Regions of Star Formation. ARA&A37, 311. Fabiann D., Jager G., Henning T., Dorschner J., JVlutschke H., 2000. Steps toward interstellarr silicate mineralogy. V. Thermal Evolution of Amorphous Magnesium Silicates andd Silica. A&A364, 282. Farnann I., Grandinetti P., Baltisberger J.H., er al., 1992. Quantification of the disorder inn network-modified silicate glasses. Nature358, 31. Finocchii E, Gail H.P., Duschl W.J., 1997. (Chemical reactions in protoplanetary accretionn disks. II. Garbon dust oxidation. A&A325, 1264. Fitzpatrickk E.L., 1997. Gas- and dust-phase abundances of Si, Fe, and MG in the interstellarr medium. In IAU Symp. 187: Cosmic Chemical Evolution, volume 187, p. E7. Fitzpatrickk E.L., Spitzer L.J., 1997. (Composition of interstellar clouds in the disk and halo.. IV. HD 215733. ApJ475, 623. Friedjungg \L, 1985. Accretion disks heated by luminous central stars. A&A146, 366. Fuentee A., Martin-Pintado J., Bachiller R., Neri R., Palla E, 1998. Progressive dispersal off the dense gas in the environment of early-type and late-type Herbig Ae-Be stars. A&A334,, 253. Gaill H.., Sedlmayr PC, 1999. Mineral formation in stellar winds. I. (Condensation sequencee of silicate and iron grains in stationary oxygen rich outflows. A&A347, 594. Gaill H.P., 1998. (Chemical reactions in protoplanetary accretion disks. IV. Multicompo- nentt dust mixture. A&A332, 1099. Gaposchklnn G.H.P., Gaposchkin S., Menzel D.H., 1952. Variable stars in Milton fields 11 to 2T. Annals of Harvard Gollege Observatory 118.

149 9 BIBII IOCRAPHY

(iehrzz R.D., Ncy P.P., 1992. ()."- to 23-micron photometric observations of P/Hallev 29866 III and six recent brighr comets. Icarus 100. 162. CiradvC.A... Perez M.R., Bjorkman K.S., Massa I).. 1999. Transient Infal! Events in the Diskk of AB Aurigae: The beta Picroris Phenomenon at 2-1 Megavears. ApIS1 1, 925. dradyy (..A., Perez M.R., 'lalavcra A., et ah. 1996. The [3 Pictoris phenomenon among Herbigg Ae/Be stars. I V and optical high dispersion specrra. A&AS120, 15^. dradyy (..A., Sitko M.F., Bjorkman K.S., et aL 199"". The star-grazing extrasolar comets inn the HI) 100546 system. ApJ483, 449. Cirayy R.O., Corbally Of, 1998. 1'he incidence of lambda Boons stars via an extension off the MK spectral classification system to very young A-type stars. AJ1 16, 2530. Oreenbergg J.M., Hage J.I., 1990. From interstellar dust to comets - A unification of observationall constraints. ApJ361, 260. (iLiilloiss ()., Nenner I., Papoular R., Revnaud ('., 1996, doal models for the infrared emissionn specrra of proto-planerary nebulae. Ap)464, 810. Halbedell H.M., 1996. Rotational veloeitv determinations for 164 Be and B stars. PASP108,, 833. Hallenheckk S., Nuth j., 1997. Infrared observations of the transition from chaotic to crystallinee silicates via thermal annealing in the laboratory ApcxSS255, A2T. Hallenheckk S.E., Nuth J.A., Nelson R.N., 2000. Evolving optical properties of annealing silicatee grains: From amorphous condensate to crvstalline mineral. Apf535, 24~. Hannerr M.S., Lynch D.K., Russell R.W., 1994. The 8-13 micron spectra of comets and thee composition of silicate grains. ApJ425, 2~74. Harkerr D.E., 1999. Ph.D. thesis, . U. Wyoming, (1999). Harvardd I'.I.., Hanner M.S., Sekanina Z., 2000. Thermal infrared imaging and spec- troscopyy of comet Hale-Bopp (C/l 995 ()]). ApJ538, 428. Henningg T, 1996. Interstellar dust grains - an overview. In IAU Synip, 1 "8: Molecules inin Astrophysics: Probes e?' Processes, volume 1^8, p. 343. —,, 1999. drain formation and Evolution in the Interstellar Medium. In Solid Inter- stellarstellar Matter: The ISO Revolution, les llouches Workshop, Pehruary 2-6, 1998. Edited byby h. d'Heudecourt, (i. jobliu, and A. Jones. TDP Sciences and Springer A erlag, 7 999., p.p. 2^", p. 24", Henningg T., Begemann B., Mutschke H., Dorschner [., 1995. Optical properties of oxidee dust grains, AcVASll2, 143. Henningg T, Burkert A., Faunhardt R., EeinertC., Stecklum B., 1998. Infrared imaging andd millimetre continuum mapping of Herbig Ae/Be and PL' Orionis stars. AcxA336. 565. . Henning'P... Pfau \X', Zinnecker H., Prusti P., 1993. A 1.3-milhmetcr stirvcv of circumstellarr dust around young ("hamaeleon objects. A&A276, 1 29. Henningg I., Stognienko R., 1996. Dusr opacities for protoplanctarv accretion disks: influencee of dust aggregates. A&A31 I, 291 . Herbigg (i.H., I960. I he spectra of Be- and Ae-tvpe stars associated with nebulositv. ApJS4,3.r. . Hilll (i., 18"7S. Researches in the lunar historv. .American |. Math. 1. 5. Hillenbrandd I..A., Strom S.F., Wba P.J., Keciic J., 1992. Plerbig Ae/Bc stars -

150 0 BIBLIOGRAPHY Y

Intermediate-masss stars surrounded bv massive circumstellar accretion disks. Ap|39 ', 613. . Honyy S., van Kerckhoven C Peeters F., 'I'ielens A.G.G.M., Hudgins I).XL, Allaman- dolaa L.J., 2001. I he CH out-ol-plane bending modes of PAH molecules in asrro- phvsicall environments. A&A3~0, 1030. Houkk N\, Smith-Moore \L, 1988. Michigan catalogue ol two-dimensional spectral typess lor the HI) stars,vol. 4 : -26 to -12 degrees. In Michigan Spectral Survey, Attn Arbor,Arbor, Dept. of Astronomy, Univ. Michigan (Vol. 4) (19HHK p. 0. Huu J.Y., The PS., de Winter D., 1980. Photometric and spectroscopic study ol three candidatee Herbig Ac/Be stars - HD 3741 1, HI) 100546, and HD 104237. A&A208, 213. . Huffmann D.R.. Stapp J.L., LG3. Optical Measurements on Solids of Possible Interstellar Importancee (presented by I). R. Huffman). In I All Symp. 52: Interstellar dust and relatedrelated topics, volume 52, p. 297. Jagerr G., Molster F.J.. Dorschner J., Henning 4'., Mutschke H., Waters L.B.F.M., 1998. Stepss toward interstellar silicate mineralogy. IV. The crystalline revolution. A&A339, 904. . [oness A.P., 1990. Iron or iron oxide grains in the interstellar medium? MNRAS245, 331. . Joness T.W., Merriii K.M., 1976. Model dust envelopes around late-type stars. ApJ209, 509. . Kellerr L., Bradley ]., Bouwman J., et ah, 2000. Sulfides in Interplanetary Dust Parti- cles:: A Possible Match to the 25 µm Feature Detected by the Infrared Space Observator}'.. In Lunar and Planetary Science Conference, volume 31, pp. 1860+. Kemperr F, Waters L.B.F.M., de Koter A., Tielens A.G.G.M., 2001. Grystallinity versus mass-losss rate in asymptotic giant branch stars. A&A369, 132. Kenn yon S.J., Hartmann L,, 198"7. Spectra! energy distributions of 1' Tauri stars - Disk Haringg and limits on accretion. ApJ323, 714. Kesslerr M.F, Stein/ J.A., Anderegg M.F., et aL 1996. I he Infrared Space Observatory (ISO)) mission. A&A315, L27.^ Kleinn G., Hurlbut G.S., 1993. Manual of Alineralogy. John Wilev & Sons, New York. Koikee G., Isuchiyama A., Shibai H., et ah, 2000. Absorption spectra of Mg-rich Mg-Fe andd Ca pyroxenes in the mid- and far-infrared regions. A&A363. 1115. Kukarkinn B.V., Kholopov P.N., Pskovskv Y.P., et ah, 1971. I he third edition containing informationn on 2043" variable stars discovered and designated till 1968. In (leneral CatalogueCatalogue of Variable Stars, 3rd ed. (19^1), p. 0. Kuruczz R., 1993. Model atmospheres. GD-ROM Nos. 1-18, Cambridge, Mass.: Smith- soniann Astrophvsical Observatory . Ladaa G.J., Wilking B.A., 1984. I he nature of the embedded population in the Rho Ophiuchii dark cloud - Mid-infrared observations. ApJ28G 610. l.aorr A., Draine B. L, 1993. Spectroscopic constraints on the properties of dust in active galacticc nuclei. Ap|4()2, 441. Lellotichh L., Crovisier J., Lim L, et ah, 1998. I.vidence for water ice and estimate of dustt production rate in comet Hale-Bopp at 2.9 AL from the Sun. A6xA339, L9.

151 1 BIBLIOGRAPHY Y

Lii A., Greenberg J.M., 1998. From Interstellar Dust to Comets: Infrared F.mission from Comett Hale-Bopp (C/1995 Ol). ApJ498, L83. Lissee CM., A'Hearn M.F., Hauser M.G., et al., 1998. Infrared Observations of Comets byy CO BE. ApJ496, 971. Luboww S.H., Seibert M., Artymowicz P., 1999. Disk accretion onto high-mass planets. ApJ526,, 1001. Lynden-Belll D., Pringle J.E., 19"74. The evolution of viscous discs and the origin of the nebularr variables. MNRAS168, 603. Malfaitt K., 1 999. The properties and evolution of the dusty disks surrounding Herbig Ae/Be stars.stars. Ph.D. thesis, AA (Astronomical Institute, KULeuven, Celestijnenlaan 200B, B-30011 Heverlee, Belgium). Malfaitt K., Bogaert E,, Waelkens C, 1998a. An ultraviolet, optical and infrared study off Herbig Ae/Be stars. A&A331,211. Malfaitt K., Waelkens C, Bouwman J., de Koter A., Waters L, 1 999. The ISO spectrum off the young star HD 142527. A&A345, 181. Malfaitt K., Waelkens C, Waters L., Vandenbussche B., Huvgen E., de Graauw M., 1998b.. The spectrum of the young star HD 100546 observed with the Infrared Spacee Observatory. A&A332, L25. Manningss V., Koerner D.W., Sargent A.I., 1997. A rotating disk of gas and dust around aa young counterpart to beta Pictoris. Nature388, 555. Manningss V., Sargent A.I., 1997. A High-Resolution Study of Gas and Dust around Youngg Intermediate-Mass Stars: Evidence for Circumstellar Disks in Herbig AF^ Sys- tems.. ApJ490, 792. —,, 2000. High-Resolution Studies of Gas and Dust around Young Intermediate-Mass Stars.. II. Observations of an Additional Sample of Herbig AE Systems. ApJ529, 391. Marshh K.A., Van Cleve J.E., Mahoney M.J., Hayward TL., Houck J.R., 1995. Spatially resolvedd mid-infrared observations of circumstellar dust around AB Aurigae. ApJ451, 777. . Masonn C.G., Gehrz R.D., Jones 'F.J., Woodward C.E., Hanner M.S., Williams D.M., 2001.. Observations of Unusually Small Dust Grains in the Coma of Comet Hale- Boppp C/1995 Ol. ApJ549, 635.' Mathiss J.S., Rumpl W., Nordsieck K.H., 1977. The size distribution of interstellar grains.. ApJ217, 425. Mayorr M., Queloz D., 1995. A Jupiter-mass companion to a solar-type star. Nature378, 355. . McCaughreann M.J., Ociell C.R., 1996. Direct Imaging of Circumstellar Disks in the Orionn Nebula. AJlll, 1977. MeeusC,, Waelkens C, Malfait K., 1998. HD 139614, HD 142666 and HD 144432: evidencee for circumstellar disks. A&A329, 131. Meeuss C, Waters L., Bouwman J., van den Ancker M., Waelkens C, Malfait K., 2001. ISOO spectroscopy of circumstellar dust in 14 Herbig Ae/Be systems: Iowards an un- derstandingg of dust processing. A&A365, 476. MeiVshchikovv A.B., Henning T., 1997. Radiation transfer in circumstellar disks. A&A318,, 879.

152 2 BIBLIOGRAPHY Y

Millan-Gabett R., Schlocrb P.P., Traub W.A., 2001. Spatially Resolved Circumstellar Structuree of Herbig AE/BE Stars in the Near-Inlrared. ApJ546, 358. Millan-Gabett R., Schlocrb HP, Traub W.A., Malbet F., Berger J P., Bregman J.D., 1999. Sub-Astronomicall Unit Structure of the Near-Inlrared Emission from AB Aurigae. ApJ513,, L131. Miroshnichenkoo A., Ivezic Z., Vinkovic D., Elitzur M., 1999. Dust Emission from Herbigg AE/BE Stars: Evidence for Disks and Envelopes. ApJ520, El 15. Miroshnichenkoo A., Ivezic Z., Elitzur M., 1997. On Protostellar Disks in Herbig Ae/Be Stars.. ApJ475, E41. Mishchenkoo M.I., Hovenier J. W., Travis L.D., eds., 2000. Light scattering by nonspherkal particles:particles: theory, measurements, and applications. Miyakee K., Nakagawa Y., 1995. Dust particle settling in passive disks around T Tauri stars:: Models and IRAS observations. Apj44l, 361. Molsterr E, Waters E., Trams N., et ah, 1999a. The composition and nature of the dust shelll surrounding the binary AFGE 4106. A&A350, 163. Molsterr F.J., 2000. Crystalline silicates in circumstellar dust shells. Ph.D. thesis, AA (As- tronomicall Institute "Anton Pannekoek", University of Amsterdam, Kruislaan 403, 10988 SJ Amsterdam, The Netherlands). Molsterr F.J., Lim T.L., Sylvester R.J., et al., 2001. The complete ISO spectrum of NGC 6302.. A&A372, 165. Molsterr F.J., Yamamura E, Waters E.B.F.M., et al., 1999b. Low-temperature crystalliza- tionn of silicate dust in circumstellar disks. Nature401, 563. Mukaii T, Koike C., 1990. Optical constants of olivine particles between wavelengths of 77 and 200 microns. Icarus 87, 180. Nattaa A., 1999. Star Formation. In Infrared space astronomy, to-day and to-morrow, Eds. F.Casoli,F.Casoli, F.David and J. Lequeux, EDP-Sciences/ Springer- Verlag. Nattaa A., Grinin V., Mannings V., 2000a. Properties and Evolution of Disks around Pre-Main-Sequencee Stars of Intermediate Mass. In Protostars and Planets IV, p. 559. Nattaa A., Meyer M.R., Beckwith S.V.W., 2000b. Silicate Emission in T Tauri Stars: Evidencee for Disk Atmospheres? ApJ534, 838. Nattaa A., Prusti T, Krugel E., 1993. Very small dust grains in the circumstellar environ- mentt of Herbig Be-stars. A&A275, 527+. Nattaa A., Prusti T, Neri R., Wooden D., Grinin V.P., 2001. A reconsideration of disk propertiess in Herbig Ae stars. A&A371, 186. Nelsonn A.F., Benz W, Ruzmaikina TV., 2000. Dynamics of Gircumstellar Disks. IE Heatingg and Gooling. ApJ529, 357. O'Donnelll J.E., Mathis J.S., 1997. Dust grain size distributions and the abundance of refractoryy elements in the diffuse interstellar medium. ApJ479, 806. Olivee K.A., Schramm D.N., 1982. OB associations and the nonuniversality of the cosmicc abundances - Implications for cosmic rays and meteorites. ApJ257, 276. Pall la E, Stahler S.W., 1993. The Pre-Main-Sequence Evolution of Intermediate-Mass Stars.. ApJ4l8, 414. Pantinn E., FT S., Marchis E, Endl M., Kürster M., Sterzik M., 2000. First Detection of

153 3 BIBII lOURAPHY

aa Dust Disk around Iota Horologii, a Southern Star O rh it ted bv an Lxtrasolar Chant Planet.. In AAS/Di vision of Planetary Sciences Meeting volume 32, p. 6533. Pantinn P.. W'aelkens ('.., Malfait K., 1999. SW'S observations of the [3 Pictoris dust disk. P.SAA SP-427: The Universe as Seen bv ISO 42^, 385. Perrymann M.A.C., ESA, 199'. The HIPPARCOS and TYCHO catalogues. Astrometric andand photometric star catalogues derived frinn the PSA HIPPARCOS Space Astronietry Mission.Mission. I he Hipparcos and Iveho catalogues. Astrometric and photometric star cataloguess derived from the PSA Hipparcos Space Astrometrv Mission, Publisher: Noordwijj k, Netherlands; ESA Publications Division, 1997, Series: ESA SP Series vol no:: 1200, ISBN: 929092399^ (set). Pollackk J.B., Hollenbach D., Beckwith S.. Simonelli D.P., Roush 'P., Pong \V., 1994. Compositionn and radiative properties of grains in molecular clouds and accretion disks.. Ap[421,615. Prcibischh 'P., Osscnkopf V., Yorke H.\X'., HenningT. 1993. The influence of ice-coated grainss on prorostellar spectra. A&A2~9, 57_. Prinnn R.G., 1990. On neglect of nonlinear momentum terms in solar nebula accretion diskk models. Ap)348, 725. Rietmeijerr P.J.M., 1988. L'ltrarine-grained mineralog\' and matrix chemistrv of olivine- richh chondritic interplanetary dust particles. In Lunar and Planetary Science Confer- ence,ence, 19th, Houston, 'IX, Mar. 14-18, 1988, Proceedings (A89-36486 15-91). Cam- bridge/Houston,bridge/Houston, IX Cambridge University Press/Lunar and Planetary Institute, vol- umee 19, pp. 513-521. Rietmeijerr P.J.M., Hallenbeck S.P., Nuth J.A., Karner J.M., 2001. Amorphous mag- nesiosilicatee smokes annealed in vacuum: The evolution of magnesium silicates in eircumstellarr and cometary dust. Icarus Submitted. Rietmeijerr P.J.M., Nuth J. A., Karner J.M., 1999. Metastable eutectic condensation in a Mg-Pe-SiO-PT-O^^ vapor: Analogs to eircumstellar dust. Ap|527, 395. Rietmeijerr P.J.M., Nuth J.A., MacKinnon I.D.R., 1986. Analvtical electron microscope off Mg-SK) smokes - A comparison with infrared and XRD studies. Icarus 66, 21 1. Rogerss C, Martin P.O., Urabtree D.R., 1983. 'I he eircumstellar dust of ML' Cephei. Ap)2~2,, r5. Sandfordd S.A., Walker R.M., 1985. Laboratory infrared transmission spectra of individuall interplanetary dust particles from 2.5 to 25 microns. Ap|29L 838. Satoo Y., Nakagawa Y., 2000. Dust particle settling in proroplanetary disks around young starss in binary systems. In I All Symposium, volume 202, p. L83. Savagee B.D., Mathis J.S., 19~9. Observed properties of interstellar dust. ARA&Al",

Schul/ee H., Kissel J.,Jessberger P^.K., 1997. Chemistrv and mincralogvof comet I Ialle\ s dtist.. In ASP Con f. Ser. 122: Prom Stardust to Planetesnnals, p. 39~\ Schuttee W.A., Pielens A.O.O.M., Allamandola L.f, 1993. Theoretical modeling of thee infrared fluorescence from interstellar polwvclic aromatic bvdrocarbons. ApJ4 1 5, 39". . Scottt A.. Dtiley WAX'., 1996, Ultraviolet and Infrared Refractive Indices of Amorphous Silicates.. Ap|SK)5. 401.

154 4 BIBLIOGRAPHY Y

Servoinn J.P., Piriou B., 1973. Infrared reHectivitv and Raman scattering of Mg^SiO., singlee crystal, phys. stat. sol. 55, 6™. Shaktiraa N.I., Sunvaev R.A., 1973. Black holes in binary svstcms. Observational appearance.. A&A24, 33'. Sbuu PH., Adams P.O., Lizano S., 1987. Star formation in molecular clouds - (observa- tionn and theory. ARA&A25, 23. Shuu KH., Shang H., lee F, 1996. Toward an astrophvsical theory of chondrites. Science 271,, 1545. Siebenmorgenn R., Prusti 'P., Natta A., Muller 'P.O., 2000. Mid infrared emission of nearbyy Herbig Ae/Be stars. A&A361, 258. Sirkoo M.L., 1981. Spectral energy distributions of hot stars with circumstellar dust. ApJ247,, 1024. Skinnerr S.I,., Brown A., Stewart R.T., 1993. A high-sensitivity survev of radio continuumm emission from Herbig Ae/Be stars. ApJS87, 21"7. Slettebakk A., 1982. Spectral types and rotational velocities of the brighter Be stars and A-FF type shell stars. ApJS50, 55. Smithh R.G., Robinson G., Hyland A.R., Carpenter O.P., 1994. Molecular ices as temperaturee indicators for interstellar dust: the 44- and 62-mu.m lattice features of H2C) ice.. MNRAS271,481. Snoww P.P., Witt A.N., 1996. Interstellar depletions updated: Where all the atoms went.

APJ468,, L65. Spitzerr W.O., Kleinman D.A., I960. Infrared lattice bands of quartz. Physical Review 121,1324. . Stevensonn D.J., 1990. Chemical heterogeneity and imperfect mixing in the solar nebula. ApJ348,, 730. Stevensonn D.J., Limine J.I., 1988. Rapid formation of Jupiter by diffuse redistribution off water vapor in the solar nebula. Icarus 75, 146. Steyerr PR., 1974. Infrared Optical Properties of Some Solids of Possible Interest in Astron- omyomy and Atmospheric Physics. Ph.D. thesis. The university of Arizona. Sylvesterr R.J., Kemper F., Barlow M.J., et ak, 1999. 2.4-197 mu m spectroscopy of OH/IRR stars: the IR characteristics of circumstellar dust in O-rich environments. A&A352,, 587. Sylvesterr R.J., Skinner C.J., Barlow M.J., Mannings V., 1996. Optical, infrared and millimetre-wavee properties of Vega-1 ike systems. MNRAS279, 91 5. Phii W.F., Blake G.A., van Dishoeck P.P., et ak, 2001. Substantial reservoirs of molecular hydrogenn in the debris disks around young stars. Nature4()9, 60. Helenss A.O.G.M., 1998. Interstellar depletions and the life evele of interstellar dust. ApJ499,, 267. 'Prillingg D.P., Brown R.H., Rivkin A.S., 2000. Circumstellar Dust Disks around Stars withh Known Planetary Companions. ApJ529, 499. vann de Hulst H.C., 1957. Light Scattering by Small Particles. Light Scattering by Small Particles,, New York: John Wiley & Sons, "l 95^. vann den Ancker M., Bouwman J., Wesselius P., Waters P., Dougherty S., van Dishoeck E.P,, 2000. ISO spectroscopy of circumstellar dust in the Herbig Ac svstems AB Aur

155 5 BIBLIOGRAPHY Y

andd HD 1632%. A&A35", 325. vann den Ancker M.E., 1999. Circumstellar Materialin Young Stellar Objects. Ph.D. thesis, AAA (Astronomical Institute 'Anton Pannekoek", University of Amsterdam, Kruislaan 403,, 1098 SI Amsterdam, The Netherlands), vann den Ancker M.E., de Winter D., Tjin A Djie H.R.E., 1998. HIPPARCOS photometryy of Herbig Ae/Be stars. A&A330, 145. vann den Ancker\\l.E., Meeus G., Garni J., Waters L.B.F.M., Waelkens G., 2001. The compositionn of circumstellar gas and dust in 51 Oph. A&A369, 1-17. vann den Ancker M.E., The PS., Tjin A Djie H.R.K., et al., 1997. HIPPARCOS data on Herbigg Ae/Be stars: an evolutionary scenario. A&A324, L33. VinkovicVinkovic D., Miroshnichenko A., Ivezic Z., Elitzur M., 2000. Evidence for Disks and Envelopess Around Herbig Ae/Be Stars. In ASP Conf. Ser. 196: Thermal Emission SpectroscopySpectroscopy and Analysis of Dust, Disks, and Regoliths, pp. 71-75. Waelkenss G., Bogaerr E., Waters LB.KM., 1994. The spectral evolution of Herbig Ae/Bee stars. In ASP Conf. Ser. 62: The Nature and Evolutionary Status of Herbig Ae/Be Stars,Stars, p. 405. Waelkenss C Waters L.B.EM., de Graauw M.S., et al., 1996. SWS observations of youngg main-sequence stars with dusty circumstellar disks. A&A315, 1.245. Walkerr H.J., Burner H.M., 1995. Follow-Up Observations of (3-Pic-like Stars. Ap&SS224,, 389. Walkerr M.K, 1956. Studies of Extremely Young clusters in NGC 2264. ApJS2, 365 + . Warrenn S.G., 1984. Optical constants of ice from the ultraviolet to the microwave. Appl.Opt.23,, 1206. Waterss LB.EM., Waelkens C, 1998. Herbig Ae/Be stars. ARA&A36, 233. Weidenschillingg S.J., 1997. Planetesimals from Stardust. In ASP Conf. Ser. 122: Erom StardustStardust to Planetesimals, p. 281. Weissmann P.R., 1984. The VEGA particulate shell - Comets or asteroids? Science 224, 987. . Woodenn D.H., Burner H.M., Harker D.E., Woodward C.E., 2000. Mg-Rich Silicate Crystalss in Comet Hale-Bopp: ISM Relics or Solar Nebula Condensates? Icarus 143, 126. . Woodenn D.H., Harker D.E., Woodward C.E., et al., 1999. Silicate Mineralogy of the Dustt in the Inner Coma of Comet (71995 01 (Hale-Bopp) Pre- and Postperihelion. Ap]517,, 1034.

156 6 NietNiet het einde, maarmaar een begin

IRSTSJÉI I w w é, é,

Éi^HiïltiÉ--