WATER VAPOR IN PROTOPLANETARY DISKS

DISS. ETH NO. 21355

WATER VAPOR IN PROTOPLANETARY DISKS A PROBE OF PHYSICS AND CHEMISTRY OF PLANET FORMATION CONDITIONS

A dissertation submitted to

ETH ZURICH for the degree of

Doctor of Sciences

presented by

Andrea Banzatti Dott. Magistrale in Fisica Universitá degli Studi di Milano (Italy) born on Oct. 27th, 1984 citizen of Milan, Italy

accepted on the recommendation of

Prof. Dr. Michael R. Meyer, examiner Dr. Klaus M. Pontoppidan, co-examiner

Zurich, 2013

Cover illustration:

Impression of a water-emitting protoplanetary disk around a young star. (Adapted from an illustration by: NASA/JPL-Caltech, R. Hurt)

Abstract This thesis is devoted to a study of the conditions and evolution of the terrestrial planet formation region in young circumstellar disks (disk radii ≈ 1–10 AU from the central star), by means of spectroscopic observations of molecular gas emission. The main focus of this work is the infrared spectrum of water (H2 O), which provides thousands of emission lines tracing the warm and dense gas inward of the snowline in disks. Aside from that, the analysis includes emission from some organic molecules that trace the carbon chemistry, C2 H2 (acetylene), HCN (hydrogen cyanide), and CO2 (carbon dioxide), as well as emission from OH (hydroxyl) that is connected to the formation and destruction of the water molecule. Particular consideration in this work is given to variable accretion phenomena occurring during the T Tauri phase of young stellar objects, which are used as a tool to better understand the origin and evolution of the molecular gas in inner disks. A pioneering contribution of this work is the study of the change in molecular gas emission observed toward the strongly variable T Tauri star EX Lupi (Chapter 2 in this book). Mid-infrared spectra obtained previous to and during a recent accretion outburst were compared, and found that the gas emission changes remarkably between the two phases. The spectrum in quiescence is composed of the typical molecular lines that dominate mid-infrared spectra of T Tauri systems, showing a dense forest of lines from water, OH, C2 H2 , HCN, and CO2 . These lines are observed in emission and attributed to warm gas layers at a few AU in the disk from the central star. In outburst, water emission increases in strength, new OH lines are detected, and emission from organics disappears. These changes have been interpreted as due to a larger emitting area of the warm gas in outburst (probably linked to a recession of the snowline), when the disk temperature increases after the increase in accretion luminosity from the star. In addition, enhanced UV radiation is found to produce OH from photodissociation of water. The behavior of organics remains unclear from the limited set of data used in this study, but could be due to changes in excitation, chemistry, or both. All molecular emissions are difficult to analyze and interpret, because of strong blending between individual lines (from the

iii

ABSTRACT

same as well as different molecules). This is particularly true for water, which has thousands of lines scattered all over the infrared range. Encountering this complexity motivated a careful analysis and a detailed modeling with particular attention to the applicability of slab models that assume thermal equilibrium for the observed emission. The remarkable findings of this study motivated follow-up investigations in two directions. One direction was taken to better understand the role of accretion variability in shaping the conditions of the molecular gas in inner disks during the T Tauri phase. A new-concept monitoring program was performed observing the T Tauri system DR Tau with two high spectral resolution spectrographs at the ESO Very Large Telescope (Chapter 3 in this book). VISIR (resolving power R = 20000) was used to resolve two individual mid-infrared water lines from the disk, while X-shooter (R = 9000–17000) was used to simultaneously monitor the UV–NIR spectrum (0.3–2.5 µm) and measure changes in accretion luminosity. Three epochs of simultaneous observations were successfully taken, where the accretion luminosity onto the star changed to within a factor of 2 (decreasing in the second epoch, increasing in the third). Water emission from the disk was found to be stable (within 10%) over these three epochs, in contrast with the increase in water emission observed in EX Lupi for a change in accretion luminosity of a factor 40. Comparison of DR Tau and EX Lupi suggests a scenario where variable accretion phenomena have two effects on the inner disk: one is to heat the disk, but considerable variations can be produced only when the accretion keeps higher over long enough timescales for the thermal structure of the disk to change (& weeks, as in EX Lupi). A second effect is to photodissociate water by means of UV radiation, which is the main component of the accretion luminosity spectrum. This latter process is very fast and probably dominates the changes in molecular emission during the short-term accretion variability typically found in the T Tauri phase. A second direction was taken to tackle another fundamental problem: the origin of water vapor and its connection to processes happening inside disks. Competing theories provide two different perspectives, where water is produced by evaporation of icy solids migrating inward of the snowline, or formed in situ via gas-phase reactions. It is important to distinguish which process dominates in order to understand what we learn from observing (or not observing) water vapor in inner disks. One way to do that is by measuring the abundance of water vapor in the inner disk, and compare it to the oxygen abundance available to form water in situ (typically assumed to be solar). In this thesis, for the first time, a systematic rotation diagram analysis has been applied to infrared water emission (Chapter 4 in this book). The analysis established a link between the spread of the rotational scatter and the water abundance in the inner disk, where a large rotational scatter would provide evidence for the migration scenario. A de-blending procedure was developed to extract a large number of emission lines from a dozen of high signal-to-noise Spitzer spectra of protoplanetary disks, so to attempt a comprehensive interpretation of the observed emission. Large rotational scatters are indeed observed in some of these

iv

disks, supporting water vapor enrichment from evaporation of icy migrators, but the intrinsic limitation of the Spitzer data allows us to provide only a tentative confirmation. Measurement of resolved optically thin lines within the forest of optically thick lines is proposed to be a key tool to address the (still open) question, and future higher-resolution observations will provide important answers on the origin of water vapor and its connection to disk evolution and planet formation processes.

v

ABSTRACT

vi

Riassunto Questa tesi è dedicata allo studio delle condizioni e dell’evoluzione della regione di formazione dei pianeti terrestri in giovani dischi circumstellari (a raggi di 1–10 AU nel disco), per mezzo di osservazioni spettroscopiche di emissione gassosa. Questo lavoro si focalizza principalmente sullo spettro infrarosso del vapore acqueo (H2 O), che fornisce migliaia di righe di emissione dal gas caldo e denso presente nelle regioni interne dei dischi. Assieme a questo, l’analisi include lo studio dell’emissione di alcune molecole organiche, C2 H2 (acetilene), HCN (acido cianidrico), e CO2 (anidride carbonica), cos`ı come l’emissione di OH (ossidrile), che è collegato alla formazione e distruzione della molecola d’acqua. Particolare attenzione viene data ai fenomeni di accrescimento non costante durante la fase T Tauri di oggetti stellari giovani, che vengono utilizzati come strumento per comprendere meglio l’origine e l’evoluzione del gas molecolare nei dischi. Un contributo pionieristico di questo lavoro è lo studio della variazione di emissione di gas molecolare osservata verso la stella EX Lupi, una T Tauri fortemente variabile. Spettri nel medio infrarosso ottenuti precedentemente e durante un recente forte aumento d’accrescimento sono stati confrontati, trovando che il gas cambia notevolmente tra le due fasi. L’emissione in quiescenza è composta da righe molecolari che tipicamente dominano lo spettro di sistemi T Tauri nel medio infrarosso, mostrando una fitta foresta di righe d’acqua, OH, C2 H2 , HCN, e CO2 . Queste righe sono osservate in emissione e sono attribuite a strati di gas caldo nel disco, ad alcune AU dalla stella centrale. Durante l’aumento di accrescimento, l’emissione d’acqua diventa pi´ u intensa, nuove righe di OH sono detettate, e l’emissione da sostanze organiche scompare. Queste alterazioni sono interpretate come dovute a una pi` u estesa area di gas caldo, quando il riscaldamento del disco aumenta per l’aumento dell’accrescimento. Inoltre, una maggiore radiazione UV svolge un ruolo chiave nella produzione di OH da fotodissociazione di acqua. Il comportamento delle sostanze organiche è reso poco chiaro dai limiti dei dati utilizzati in questo studio, ma potrebbe essere dovuto a variazioni in eccitazione, chimica, o entrambe.

vii

RIASSUNTO

Tutte le emissioni molecolari sono difficili da analizzare, a causa della forte confusione tra righe individuali. Ciò è particolarmente vero per l’acqua, che ha migliaia di righe sparse su tutto l’infrarosso e per cui modelli semplici in equilibrio termico locale possono solo approssimativamente riprodurre l’emissione (suggerendo effetti non termici nell’eccitazione). I notevoli risultati di questo studio hanno motivato indagini successive in due direzioni. Una direzione è stata presa per comprendere meglio il ruolo della variabilità di accrescimento nel modellare le condizioni del gas molecolare nei dischi durante la fase T Tauri. Un monitoraggio di nuova concezione è stato effettuato osservando il sistema T Tauri DR Tau con due spettrografi ad alta risoluzione spettrale del Very Large Telescope (VLT) dell’ESO. VLT / VISIR (con risoluzione spettrale R = 20000) è stato utilizzato per risolvere le singole righe d’acqua nel medio infrarosso, mentre VLT / X-shooter (R = 9.000–17.000) è stato utilizzato per monitorare contemporaneamente l’UV–NIR (0.3–2.5 µm) e le variazioni in luminosità d’accrescimento. Tre epoche di osservazioni simultanee sono state prese con successo, ma purtroppo l’accrescimento non è cambiato notevolmente (solo di un fattore 2). Una leggera tendenza è stata vista nell’emissione d’acqua, mostrando righe pi` u deboli per una maggiore luminosità di accrescimento, ma il cambiamento è stato troppo basso per essere significativo al di sopra del rumore. Questa ulteriore indagine ha potuto affrontare solo una piccola parte delle questioni sollevate dallo studio di EX Lupi, ma ha permesso di focalizzare meglio una prospettiva interessante. Le analisi di questi due sistemi, combinate, sono coerenti con uno scenario in cui i fenomeni di accrescimento non-stazionario hanno due effetti sul disco interno: uno è quello di riscaldare il disco, ma notevoli variazioni possono essere prodotte solo quando l’accrescimento aumenta su scale temporali abbastanza lunghe perché la struttura termica del disco possa cambiare (& mesi, come in EX Lupi). Un secondo effetto è quello di fotodissociare acqua per mezzo di radiazioni UV, che sono la componente principale della luminosità d’accrescimento. Quest’ultimo processo è molto veloce e probabilmente domina le variazioni di emissione molecolare nelle fasi di rapida variabilità d’accrescimento tipicamente trovate nei sistemi T Tauri. Una ulteriore epoca di osservazioni su DR Tau sarebbe sufficiente per confermare (o smentire) questo scenario, nel momento in cui cambiamenti in accrescimento di almeno un fattore 4 avvenissero. Una seconda direzione è stata presa per affrontare un altro problema fondamentale, l’origine del vapore acqueo nei dischi protoplanetari e la connessione con la loro evoluzione. La molecola d’acqua in stato gassoso viene attribuita a processi fondamentali che avvengono nel disco. Teorie in competizione offrono due prospettive diverse, in cui l’acqua è dovuta a evaporazione di solidi ghiacciati che migrano verso l’interno del disco, o formata in loco tramite reazioni nel gas. Mentre la migrazione e l’evaporazione di solidi ghiacciati farebbero dell’emissione d’acqua un buon tracciante dell’evoluzione del disco e delle condizioni di formazione dei pianeti, la formazione in loco in un disco statico fornirebbe solo informazioni locali sull’atmosfera del disco, che può essere o meno collegata a ciò che accade negli strati pi` u profondi.

viii

` quindi essenziale distinguere quale processo domini per capire cosa impariamo dal E detettare (o non detettare) vapore acqueo nei dischi. Un modo per farlo è quello di misurare l’abbondanza di vapore acqueo nel disco interno, in grado di fornire la prova di arricchimento dal disco esterno prodotto da evaporazione di migratori di ghiaccio. In questa tesi, per la prima volta, un’analisi sistematica tramite diagramma di rotazione è stata applicata all’emissione d’acqua su un campione considerevole di giovani dischi circumstellari. Una procedura è stata sviluppata per estrarre un gran numero di righe di emissione da spettri non risolti, cos`ı da consentire una comprensione pi` u completa dell’emissione osservata. L’analisi ha indicato la possibilità di densità superficiali d’acqua pi´ u grandi di quanto si pensasse, il che potrebbe essere interpretato come una prova dello scenario migratorio. Tuttavia, l’attuale stato di dati e modelli non permette di trarre forti conclusioni, soprattutto per la degenerazione tra abbondanza d’acqua e il rapporto tra gas e polvere assunti nei modelli. Misura di righe otticamente sottili risolte all’interno della foresta di righe otticamente spesse viene proposto essere uno strumento fondamentale per affrontare la questione (ancora aperta), e osservazioni future a pi` u alta risoluzione potrebbero fornire risposte importanti sull’origine del vapore acqueo e sulla sua connessione con l’evoluzione del disco.

ix

RIASSUNTO

x

Contents Abstract

iii

Riassunto

vii

List of Figures

xvi

List of Tables

xvii

Preface

xix

1 Introduction 1.1 A tale of star and planet formation . . . . . . . . . . . . . . 1.1.1 Observed properties of circumstellar disks . . . . . . 1.1.2 Protoplanetary disks and their evolution . . . . . . . 1.1.3 Planets: the remarkable “leftovers” of star formation 1.2 Water in Protoplanetary Disks . . . . . . . . . . . . . . . . . 1.2.1 Basics of infrared spectroscopy of the water molecule 1.2.2 Excitation conditions . . . . . . . . . . . . . . . . . . 1.2.3 Estimation of emitting gas properties . . . . . . . . . 1.3 Outline of this thesis . . . . . . . . . . . . . . . . . . . . . . 2 EX Lupi in outburst: changes in the disk molecular 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 2.2 Observations . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Spitzer IRS archival spectra . . . . . . . . . . 2.2.2 Data reduction . . . . . . . . . . . . . . . . . 2.3 Spectral line analysis . . . . . . . . . . . . . . . . . . 2.3.1 Molecular and atomic emission . . . . . . . . 2.3.2 Estimation of emission line fluxes and errors . 2.4 Constraining gas properties . . . . . . . . . . . . . . 2.4.1 An LTE approach . . . . . . . . . . . . . . . .

. . . . . . . . .

. . . . . . . . .

emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . .

1 2 3 5 8 10 12 14 15 16

. . . . . . . . .

19 20 22 22 22 25 25 28 32 32 xi

CONTENTS

. . . . . . . . . . . . .

33 34 35 35 45 47 48 50 50 52 52 53 54

. . . . . . . . .

57 58 60 62 65 66 66 67 70 73

4 Searching for signatures of icy migrators in protoplanetary disks 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 An answer from the water abundance . . . . . . . . . . . . . . 4.2 Rotation diagrams of water vapor emission . . . . . . . . . . . . . . . 4.2.1 LTE slab models . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Non-LTE slab models . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 The inner disk water abundance . . . . . . . . . . . . . . . . . 4.3 Application to Spitzer spectra . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Line flux extraction . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Rotation diagrams of Spitzer spectra . . . . . . . . . . . . . . 4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Origin of large water abundances . . . . . . . . . . . . . . . . 4.4.2 Confirmation of large water abundances . . . . . . . . . . . . 4.4.3 Future observations . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75 76 77 81 81 84 84 86 87 91 92 92 93 96 97

5 Conclusions & Outlook

99

2.5

2.6

2.7 2.8

2.4.2 Fitting Spitzer spectra of H2 O emission with LTE models . 2.4.3 Fitting method adopted in this work for H2 O emission . . Results of modelling molecular emission . . . . . . . . . . . . . . . 2.5.1 H2 O rotational lines . . . . . . . . . . . . . . . . . . . . . 2.5.2 OH rotational lines . . . . . . . . . . . . . . . . . . . . . . 2.5.3 C2 H2 , HCN and CO2 rovibrational branches . . . . . . . . 2.5.4 Summary of changes in emission from model results . . . . Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6.1 H2 O and OH emission between quiescence and outburst . . 2.6.2 The lack of C2 H2 , HCN and CO2 in outburst . . . . . . . . 2.6.3 Note on H I and H2 in outburst . . . . . . . . . . . . . . . Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix: a single-slab model of gas in LTE . . . . . . . . . . . .

3 Exploring the link between water and accretion variability 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 VISIR spectra . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 X-shooter spectra . . . . . . . . . . . . . . . . . . . . . 3.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Water line properties . . . . . . . . . . . . . . . . . . . 3.3.2 Accretion luminosity . . . . . . . . . . . . . . . . . . . 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Summary and conclusions . . . . . . . . . . . . . . . . . . . .

CV xii

. . . . . . . . .

. . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . .

103

CONTENTS

Acknowledgements

107

Bibliography

120

xiii

CONTENTS

xiv

List of Figures

1.1 1.2 1.3 1.4 1.5

. . . .

1 4 7 9

1.6 1.7 1.8

Star-forming region in the Carina nebula . . . . . . . . . . . . . . . Spectral energy distribution of disks . . . . . . . . . . . . . . . . . . Accretion history in star formation . . . . . . . . . . . . . . . . . . Mass-radius diagrams for exoplanets . . . . . . . . . . . . . . . . . Sketch of a protoplanetary disk and the regions traced by molecular emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Structure of the water molecule . . . . . . . . . . . . . . . . . . . . Energy-level diagram for the water molecule . . . . . . . . . . . . . LTE spectrum of water vapor emission . . . . . . . . . . . . . . . .

. . . .

10 12 13 15

2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11

Spitzer spectra of EX Lupi in quiescence and outburst . . . . Cross-ladder OH transitions detected in EX Lupi in outburst . Multi-gaussian fit to blended line complexes in EX Lupi . . . . Comparison of LTE line fluxes to data: quiescence . . . . . . . Comparison of LTE line fluxes to data: quiescence . . . . . . . Comparison of LTE line fluxes to data: outburst . . . . . . . . Comparison of LTE line fluxes to data: outburst . . . . . . . . Confidence limits on Tex and Nmol for water and OH emission Rotation diagram of OH lines detected in outburst . . . . . . Quiescent spectrum and molecular LTE models . . . . . . . . Outburst spectrum and molecular LTE models . . . . . . . . .

. . . . . . . . . . .

24 26 30 38 39 40 41 42 46 48 49

3.1 3.2 3.3

Telluric correction of VISIR spectra . . . . . . . . . . . . . . . . . . Three epochs of VISIR data of DR Tau . . . . . . . . . . . . . . . . Relation between airmass and conversion factor Jy/(ADU s−1 ) as derived from photometric standard stars . . . . . . . . . . . . . . . Three epochs of X-shooter spectra of DR Tau . . . . . . . . . . . . Derivation of accretion luminosity from Balmer Jump fitting . . . . Relative epochal variations . . . . . . . . . . . . . . . . . . . . . . .

3.4 3.5 3.6

. . . . . . . . . . .

. . . . . . . . . . .

. . . . . . . . . . .

. 61 . 62 . . . .

64 65 69 71 xv

LIST OF FIGURES

3.7

Comparison of DR Tau and EX Lupi . . . . . . . . . . . . . . . . . . 72

4.1

Degeneracy from slab models fits to water emission in Spitzer spectra of young protoplanetary disks I . . . . . . . . . . . . . . . . . . . . . Degeneracy from slab models fits to water emission in Spitzer spectra of young protoplanetary disks II . . . . . . . . . . . . . . . . . . . . . Rotation diagrams of infrared water emission using slab models in LTE Rotation diagrams of water slab models in LTE, showing line opacity and intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rotation diagrams of RADLite models of water emission from protoplanetary disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FWHM of the instrumental line profile of the Spitzer IRS in highresolution mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Extraction of water line fluxes from Spitzer spectra . . . . . . . . . . Rotation diagrams of water emission observed in water-emitting T Tauri disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model and observed spectra of water vapor in inner regions of protoplanetary disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9

xvi

78 80 82 83 85 88 90 91 95

List of Tables 2.1 2.2 2.3 2.4 2.5

Summary of EX Lupi data from the Spitzer archive (IRS staring mode) OH detections in EX Lupi . . . . . . . . . . . . . . . . . . . . . . . . Flux estimates of emission lines in the Spitzer -IRS spectra of EX Lupi H2 O lines used to constrain the emission observed in EX Lupi . . . . Constraints on H2 O and OH emission from single-slab LTE models .

23 27 31 43 44

3.1 3.2

Three epochs of VISIR observations of DR Tau . . . . . . . . . . . . 60 Properties of water lines observed with VISIR . . . . . . . . . . . . . 67

4.1

Water-emitting disks sample considered in this work . . . . . . . . . . 87

xvii

LIST OF TABLES

xviii

“[THE ARTIST] FEELS THAT NOTHING IS PERFECT UNLESS IT IS PERSONAL.” GILBERT K. CHESTERTON, THE EVERLASTING MAN, 1925

Preface “Far away in some strange constellation in skies infinitely remote, there is a small star, which astronomers may some day discover. At least I could never observe in the faces or demeanor of most astronomers or men of science any evidence that they had discovered it; though as a matter of fact they were walking about on it all the time. It is a star that brings forth out of itself very strange plants and very strange animals; and none stranger than the men of science. That at least is the way in which I should begin a history of the world if I had to follow the scientific custom of beginning with an account of the astronomical universe. I should try to see even this earth from the outside, not by the hackneyed insistence of its relative position to the sun, but by some imaginative effort to conceive its remote position for the dehumanized spectator. Only I do not believe in being dehumanized in order to study humanity. I do not believe in dwelling upon the distances that are supposed to dwarf the world; I think there is even something a trifle vulgar about this idea of trying to rebuke spirit by size. And as the first idea is not feasible, that of making the earth a strange planet so as to make it significant, I will not stoop to the other trick of making it a small planet in order to make it insignificant. I would rather insist that we do not even know that it is a planet at all, in the sense in which we know that it is a place; and a very extraordinary place too. That is the note which I wish to strike from the first, if not in the astronomical, then in some more familiar fashion.” (G. K. Chesterton, The Everlasting Man, 1925) When I read this opening paragraph in a book recently, I immediately decided to use it for my thesis. I hope that its distinguished author will forgive an astronomer for doing that. And I hope that none of my “twenty-five”1 readers will feel offended by that. But I could not find any better words to introduce a perspective that has been developing during my PhD, which is more than a mere matter of numbers 1

This one is dedicated to my mother, and to all who love the masterpieces of our Italian literature.

xix

PREFACE

describing some distant regions in our Galaxy. Apart from the attribution of Earth as a star, which should be forgiven at least in regard for Galileo (who addressed the four Jupiter’s satellites in the same way in his Sidereus Nuncius), I cannot disagree with the statement about the strangeness of the man of science. In honesty, I should admit that I have met many astronomers that confirm that statement. I should also admit that I myself may appear strange, although for different reasons. This author might have been surprised, if I had had the chance to tell him that the reason why I am an astronomer has more to do with Earth (and poetry) than with the ability for scientific abstract reasoning. Every time that I found my thoughts at unhealthy distances from the real world (for instance after hours and hours looking at numbered wiggles on a screen), I had the healthy fortune of looking again at the extraordinary place where we walk about all the time. This has allowed me also to look back at the wiggles on the screen with renewed motivation. I am not born astronomer, I never had toy telescopes nor star charts when I was younger. I was born in an Italian Catholic family where poetry and arts (and Beauty in general) were the daily bread. I am the only one in my family adventuring into science, and I have always kept my artistic heart. The reader is warned that the introductory chapter of this thesis may read sometimes unusually poetic and less quantitative, if compared to other theses in astronomy. My astronomical vocation is very late, and is due to two things above others: big telescopes, and the Moon. When I first looked at them, I was attracted to both. I could sit and watch the Moon for hours, without losing any bit of the initial attraction. And big telescopes reminded me of an intrinsic curiosity and opening to the infinite Heavens that is structurally a part of the human person, something that is particularly strong in me. My approach has always been to stay close to the core of things and keep my hands busy, following the opportunity that is given to men in this world to be able to build good things. In other words, my attitude is observational (experimental) rather than theoretical. If my PhD has not been a time of major scientific discoveries (although for some time I and Michael thought it was), it has been a time of major achievements. Someone could define these achievements as “personal”, which I think are as essential as any technical skill considered typical for a scientist. Actually, to be honest I consider them to be even more essential. Some technical skills can in principle be learned at anytime, while other things are not learnt so easily and not all phases in one’s life favor that. I will try to sketch out what I mean in the following; the reader who is not interested can leave the Preface now and go directly to the other chapters of this thesis. The key point was initially suggested to me during a conversation with a famous astronomer, who made a major discovery that revolutionized astronomy over the last 20 years. During his visit at the ETH in 2011, I asked him to tell us the story of his discovery, which he made as a PhD student (we were at the traditional “beer with the students” after the research seminar, and the audience could not have been more appropriate). “I often wonder what would have happened if instead of me, a PhD student with

xx

nothing to lose, at the telescope there had been a more senior professor. Ask yourselves: how do you react when you find something unexpected? I myself first thought that it was noise, or that something was wrong with my data analysis. Only after many tests came I to believe that it was real. My supervisor said: why not? This is a good supervisor! One who tells you “why not?” and waits and works until the data is saying something undoubtedly. And the interesting thing is that other people had found what we found, before us. But they did not believe what they saw because it was UNEXPECTED! Because theories could not predict what we saw. So don’t worry when you find something unpredictable. Theories are just theories...” These words shepherded my thoughts and my motivation for the rest of my PhD. There is something valuable in the data, something that we can understand. To discover it, we need to put ourselves in the humble listening attitude, regardless of who tell us that this is foolish. It seems obvious to say that, but every serious scientist is aware of the temptation of hiding what is unexpected to “defend” what is already understood (those who are not serious may not even realize to have such a tendency). The key attitude in science is curiosity, open to the unexpected and ready to change one’s position even when previous ideas felt more “secure”. A big question that later arose in me, during the frequent tough periods, was, what sustains such an open approach when the data seem to be dumb, when you feel like walking in the darkness? Hints toward an answer came later when I happened to read the proceedings of an exceptional seminar by a very distinguished French mathematician, Laurent Lafforgue. I will quote here only one point, although I believe that reading the full text of that seminar may be healthy for most scientists today. “All things deserve to be studied with the most careful attention to exactness, with the attention to understand things as they are, to listen to their delicate truth, to be every day ready to revise the interpretation in order to decipher, with a greater loyalty, their speechless language.” In explaining why this religiously respectful approach was developed as a foundation of the medieval University by the Latin Church, he continued saying that the reason is “because all things, created by God, tell us something about their creator, who is infinitely greater than us and does not deceive us.” Once more, I found an indication that being honest with the data and open to the unexpected is worth more than any different approach in science. I heard a similar perspective from another distinguished scientist who I admire, who is professor at the ETH. Answering to a question I posed on what method afforded him the greatest discoveries, he said: “I always tell my students: when something does not make sense focus on it. Nine out of ten times you are wrong, and Nature gives you the opportunity to be corrected. But one out of ten, Nature is bringing you in a new unexplored direction that can lead to a major discovery. When things do not make sense it is Nature that is telling you something.” These three examples are all interesting ways to acknowledge that there is an invisible (and mysterious) dialogue happening between Someone and us, which is mediated by very normal things. All three are a bright examples of being open to xxi

PREFACE

the new, even when it is unexpectedly different from our thoughts. In the end, all real scientific discoveries demonstrate that it is convenient to be honest with the data, and that striving for the truth is better than remaining stuck in personal interpretations. But this is something that any “Sherlock Holmes” knows, and any true researcher as well. From this point of view, science may be better than other human fields, in being dependent on something that we cannot easily make up: the data. I often look at this as a great relief. In this regard, I could even thank the anonymous referee who made me suffer so much to publish my first paper. I was pushed to stick more to the data, to respect what they say, and not to try to make them say what they may not, until it becomes obvious. Extraordinary claims need extraordinary evidences to be proved. But in general any claim needs good evidence, or it is not serious. What can I now say, in conclusion? It might be hard to guess it, but the paper I have enjoyed the most working on during my PhD is the monitoring study of DR Tau (see Chapter 3), which I led from the very beginning to the end. The idea was born in a conversation with Michael, and I planned, requested, obtained, and analyzed the VISIR observations in all details. It is the paper in which I understand best the data, in which I best know on which ground I stand and in which direction I can move. It is where I best learnt how to “think like a photon” (as my supervisor used to say) that comes to the telescope through the atmosphere, is collected on the detector, and is reproduced in the image that the astronomer studies. It is where being respectful towards the data as they are, resisting the temptation of bringing them in other directions, has undoubtedly proven to be the best approach a scientist should take. In short, it is simply an example of a good job. I think that any thoughtful scientist would agree that the value of such a thing exceeds greatly its scientific impact in terms of citations. A.B. Uitikon Waldegg, Switzerland June 24, 2013

xxii

“THAT IS THE NOTE WHICH I WISH TO STRIKE FROM THE FIRST, IF NOT IN THE ASTRONOMICAL, THEN IN SOME MORE FAMILIAR FASHION.” GILBERT K. CHESTERTON, THE EVERLASTING MAN, 1925

1

Introduction

Fig. 1.1 — The Carina nebula, a star-forming region seen toward the constellation of Carina (RA: 10 44 13.45, Dec: -59◦ 380 29.7700 ). The region spans 260 light years across and is located at 7500 light years from Earth. Diffuse interstellar gas and dust are carved by radiation of young and old stars. Young stars are seen in clusters (e.g. to the center-right), and close to them new stars are forming in bright/dark pillars of gas and dust. The image is a composite of three filters (B, V, and R bands) collected with the 1.5-m Danish telescope at the ESO La Silla Observatory. Image credit: ESO/IDA/Danish 1.5 m/R.Gendler, J-E. Ovaldsen, C. Thöne, and C. Feron.

1

CHAPTER 1. Introduction

1.1

A tale of star and planet formation

Star formation is a story that begins with a breathtaking beauty, and ends in an unexpected way. In between, much is still a mystery. And as it is for all great mysteries, it is a somewhat veiled story, where the entire truth is not immediately apparent, and investigations require passion, bravery, risk, and sometimes an entire life of dedication. The story that I report here, in the following, is based on a personal attempt to see the big picture from a mosaic of hints inherited from decades of astronomical observations. I will sketch out only some of the main steps that are believed to happen in star formation, while providing my own perspective in highlighting what I find most interesting. With increasing telescopic capabilities, it turned out that the interstellar space is not as black and empty as it looks to human eyes from Earth. Diffuse gas and dust are present everywhere in our Galaxy, with tenuous emission that can be best seen with modern large telescopes. Some of these regions are primarily made of molecular gas (mainly H2 ) and are called molecular clouds. Driven by observational evidence, the currently accepted theory proposes that stars are formed by collapse of “droplets” of material in these regions. Molecular clouds are mainly composed of gas (99%), but dust plays a disproportionately key role. It dominates in opacity at ultraviolet (UV) and optical wavelengths, so to shield the gas from photodissociating radiation and allow for complex molecules to proliferate. It is from the interplay of gas and dust and the surrounding radiation that amazing figures are carved deep in the interstellar space, providing undoubtedly some of the most beautiful images of the local Universe (e.g. Figure 1.1). Cold dense cores (∼0.1 pc in size), which are usually found as fragments along extended or filamentary structures, at some point become gravitationally unstable and rapidly collapse. By angular momentum conservation, any small initial rotation is increased during collapse of gas and dust towards the core center. From the envelope, infalling material therefore ends up in a flattened rotating system, a disk, which in turn accretes material onto a forming star (a protostar ) in the center. Of these two mass accretion processes, envelope collapse ends first (in a few 105 yr) while feeding the disk at larger and larger radii, while disk accretion onto the star lasts for at least 10 times longer (a few Myr, see Figure 1.3). The time when the latter process dominates is usually regarded as the T Tauri phase. This is the phase when the components of the system become visible (i.e. are not hidden by the infalling envelope): a dusty disk and a protostar accreting mass through it. With time, within another factor 10 (∼ 107 yr), disks disappear and the newborn star is ready for its mature phases on the main sequence, burning heavier and heavier elements in its core and eventually releasing them back again into the interstellar space. But this is another story, and we shall not get distracted. As it is right now, at the end of our story, that we unexpectedly notice something that was not previously under focus, when we were eagerly waiting for the star to appear. The circumstellar material has not completely disappeared: some spherical bodies, solid or gaseous, rotate in orbits flattened on a plane that resembles the initial disk. Our quantitative understanding of disk physics would led us believe that the 2

1.1. A tale of star and planet formation

circumstellar matter is quickly accreted onto the star or dispersed by high-energy radiation. Instead, at the end of the star formation process we find planets. The reader could at this point gently remind me that this is not so surprising, and that if that were not true, we would not be here concerned with these matters (nor with any other one). And he/she would demonstrate a remarkable common sense, and be perfectly right. However, it was only very recently that planets were recognized to be the natural outcome of star formation, and theoreticians are still struggling to find a way to make this happen in their models1 . Circumstellar disks are therefore linked not only to the origin of stellar masses, which set most of the stellar evolution, but also to the origin of planets. This is why disks are a major theme in astronomy today. This is also the reason why we tell this story and focus on them in this work: to unveil at least some of what they secretly do within themselves in order to build planets.

1.1.1

Observed properties of circumstellar disks

The first observed property of circumstellar disks was their existence. Emission in excess of stellar photospheres was observed toward young stars at UV and infrared wavelengths (Mendoza V., 1966; Herbig, 1970; Strom, 1972). These excesses were proposed to be the evidence of circumstellar accretion disks by the seminal work of Lynden-Bell & Pringle (1974). In their scenario, the circumstellar material (particularly small dust grains, which dominate the opacity) adsorbs the stellar radiation and re-radiates it at longer wavelengths (from infrared to radio, 1 µm to 1 cm). At the other side of the spectrum, excess UV continuum emission was attributed to hot material accreting onto the stellar surface. Several evidences later confirmed this scenario, showing that the material is accreted nearly at free-fall velocities (producing emission line widths of 200-300 km/s) by means of the stellar magnetosphere. The correlation found between UV and near-infrared excesses supported the idea that indeed there is a strong connection between circumstellar disks and accretion (Hartigan et al., 1990), and that one implies the other. Therefore, disks are studied by looking “blue” (towards shorter wavelengths) and “red” (towards longer wavelengths) from the peak of the stellar photosphere of young stars. Today, disks are observed at many wavelengths, from optical to radio, using a multiplicity of techniques, providing an overall good view of their general properties. Spectroscopy at several wavelengths has provided means to study accretion processes (from line profiles in the UV-optical) and disk gas dynamics. Scattered light in the optical and near-infrared, as well as interferometry in the radio, have provided images of disks (especially bright massive disks) confirming rotation, showing (large) inner holes, and suggesting non axis-symmetric, possibly spiral structures (e.g. Quanz et al., 2013b). The reconstruction of the spectral energy distribution 1

If science were only a matter of common sense, we would miss the most fun of it... but we should nonetheless admit our difficulty in understanding things that Nature apparently makes very easily.

3


Fig. 1.2 — Left: median SED of K7-M2 T Tauri stars in the Taurus-Auriga star formation region (from D’Alessio et al., 2001). The stellar blackbody is shown in dotted-dashed line, while two flared disk models are shown to account for the infrared excess: using small sub-µm-size grains only (dotted line), and including large mm-size grains (solid line). UV excess emission due to accretion is barely visible at wavelengths shortward of ∼ 0.4µm. Right: model SED showing the disk regions typically traced in different parts of the spectrum (from Dullemond et al., 2007).

(SED), from photometry and spectroscopy over a wide range of wavelengths, have allowed global studies of these systems and their different components (star, accretion, disk, see Figure 1.2). SEDs provide evidences for disks being heated mostly by irradiation (at least beyond ≈1 AU) and having flared geometries with multiple vertical layers, where settling of dust towards the disk midplane depletes the upper layers. SEDs trace different disk regions at different wavelengths (Dullemond et al., 2007). The inner disk surface (. 10 AU) dominates the emission at infrared wavelengths, where the small (1 µm) dust grains re-emit the adsorbed stellar radiation. A typical feature that is observed is the 10 µm silicate emission (Cohen & Witteborn, 1985), from warm small grains in the disk surface. At wavelengths longer than ∼ 0.1 mm the outer disk ( 10 AU) and the midplane dominate, where most of the solid mass is though to be. The steep fall-off of SEDs in the Rayleigh-Jeans limit requires an optically thin disk with somewhat larger grains (1 mm) than in the disk surface (D’Alessio et al., 2001), and provides measures of disk masses (Beckwith et al., 1990). Typical masses of T Tauri disks are found to be ∼ 1% of the stellar mass, or ∼ 10−2 M . This is approximately the “minimum mass solar nebula”, the solar-composition mass that is needed to make the solar system planets (Hartmann, 2009). As it is unlikely that all disk mass at this stages goes into planets, this is a signature that disk masses measured from mm data are very likely underestimates of 4


real values or that a significant fraction of the mass is already in large planetesimals that do not contribute to the dust opacity, if T Tauri systems are analogs of the solar nebula. On the other hand, disk masses estimated during earlier phases (still embedded in the envelope) can be 10–20% of the stellar mass (Natta et al., 2000; Eisner, 2012), indicating that T Tauri disks are already one order of magnitude less massive than the initial mass available for star and planet formation. Despite the many ways by which circumstellar disks are observed, detailed studies of their properties have been elusive for long time an many processes thought to happen in disks are still largely unknown. Turbulence, and transport processes in general, are very difficult to constrain, if at all possible (Armitage, 2011). Disk masses and dispersal mechanisms are uncertain, especially because most of the gas mass is in H2 that is very hard to observe. SED fitting to recover disk structures is very degenerate, and combination with other techniques is difficult. Resolved images are likely to provide great improvements in the future, but they are not free from problems. Questions remain on what we learn about the midplane from observing disk surfaces. All this is particularly true in regard of the terrestrial planet formation region, within 10 AU from the central star, which to date remains the most elusive region of circumstellar disks.

1.1.2

Protoplanetary disks and their evolution

There is clear evidence that accretion disks undergo strong evolution, as in a time frame between 1 and 10 Myr most of them disappear (Hernández et al., 2007, and references therein). Planet formation is only one means of disk clearing, together with accretion onto the star and photoevaporation, and yet it is probably the most remarkable. So much that circumstellar accretion disks are today mostly referred to as “protoplanetary”2 to indicate their potential to form planets, even though only in a very few cases we may have direct evidence for that. It is clear that the circumstellar material, after a few Myr, is not simply inherited from the parental molecular cloud and that it is being processed in disks (or “re-processed” if we consider that some processing of dust and gas may already happen in molecular clouds). Evidence is provided by observations of both dust and gas properties. Dust grains change considerably from the interstellar medium, showing crystallinity and larger sizes in disks (Mannings & Emerson, 1994). This requires thermal processing and coagulation in high-density regions, which are not common in molecular clouds. Relative abundances of gas molecules are modified as well (Carr & Najita, 2008), suggesting an active chemistry again favored by the temperatures and densities found in disks. In the end, the change in temperature and density from molecular clouds 2

A search for paper titles using the ADS online services shows that the term “protoplanetary” has overwhelmed in the last decade both “circumstellar” and “accretion”, as referred to disks around stars; the rate is 506:170:88 between 2010 and 2012, while 10 years before (2000-2002) it was 137:147:76, 20 years before (1990-1992) it was 30:46:44, and 30 years before (1980-1982) it was only 5:2:23.

5


(T ∼ 10 K, n ∼ 103 cm−3 ) to disks (T up to a few 1000 K, n up to 1015 cm−3 ) must have strong effects. A fundamental question in planet formation concerns the extent and nature of these effects, particularly in regards of planets that may be formed in the disk. Are disks the places where the fundamental physical and chemical processing of the ingredients building planets takes place? A general consensus on the positive answer is growing, but most of the processes that are thought to occur still need to be understood (and constrained) in a quantitative way. Observations of these processes are hindered by a variety of problems, and the story of planet formation still presents somewhat large gaps. Disks are expected to evolve viscously from the envelope infall phase to disk clearing, transporting angular momentum outward and material inward onto the star. Disk evolution appears to proceed from the inner regions (which are the first to be depleted from material) outward (Meyer, 2009). During this evolution, material from different radial/vertical regions in the disk comes together, providing new conditions for physical and chemical processes. The formation of molecules is favored at higher densities, on grain surfaces or at temperatures that exceed a few 100 K (or even 1000 K, see the case of water in Section 1.2), and protoplanetary disks are therefore potential molecular factories. Moreover, at high enough densities small dust grains can efficiently stick and coagulate, growing in mass and size. As grains grow they feel less support by the gas and settle toward the disk midplane, encountering other grains and growing even further. This is regarded as the first step towards planets in the “core accretion” model (Lissauer, 1993), which finds confirmation in the large (mm-cm) dust grains found in disks (e.g. Banzatti et al., 2011). However, large grains would also be transported by viscous drag in the radial direction (Weidenschilling, 1977), and would be expected to fall onto the star very quickly. One of the main challenges today is to understand how Nature keeps the grains in the disk long enough to grow up to km sizes and build planets. The presence of favorable zones for planetesimal formation, provided by local pressure maxima within the disk, is currently the preferred solution (Johansen et al., 2007; Chiang & Youdin, 2010; Pinilla et al., 2012). In addition, pressure gradients produced by snowlines (temperature and density dependent boundaries separating regions where molecules are in the ice vs gas phase) in the disk could also favor the growth of large bodies (Dodson-Robinson et al., 2009). The understanding of conditions that may allow material to coagulate and grow to planetesimal (km) sizes is slowly growing. Yet, the details are widely unknown, not to talk of the chemistry that participate to the build-up of protoplanets. On top of the general disk evolutionary trend depicted above, the role played by initial conditions and non-steady accretion processes is still not clear, but might be important. For instance, the scatter in disk masses estimated in coeval samples (e.g. in Taurus), as well as in disk dissipation timescales, might be the result of different initial conditions, although there are no proofs of that (Meyer, 2009; Hartmann, 2009). On the other hand, accretion in young systems is non steady and not homogeneous, and may present individual differences. In this regard, interesting

6


Fig. 1.3 — Accretion history of a forming star, from the collapse of a dense core in molecular clouds to planet formation and dispersal of the circumstellar accretion disk around the newborn star. From Hartmann (2009). perspectives are provided by the variability of young stellar objects. Young stars have been known for their brightness variability for several decades (Joy, 1945), and part of this variability has been found to be related to changes in accretion rates. Accretion in the disk and onto the star does not decrease with time smoothly, but rather present sudden variations of many orders of magnitude on top of a decreasing trend (see Figure 1.3). Early in the evolution, when the star is still embedded in the infalling envelope, these accretion “FUor outbursts” are thought to be due to imbalances between the material accreted onto the disk from the collapsing core and the material accreted onto the star from the disk, or thermal instabilities (Hartmann, 2009; Hartmann & Kenyon, 1996). These events are still poorly known, and have been observed only in a few objects that have not yet exit from outburst (duration &100 yr). Intermediate accretion phenomena between FUors and the typical T Tauri variability, in terms of strength and duration, are called “EXors” (Herbig, 1989, 2008). They seem to happen in systems that are less embedded (i.e. later in evolution), where the star and the disk are (generally) clearly visible. At least in the prototype of this class of variables, EX Lupi, ongoing processing of dust and ´ gas has been directly observed during a strong outburst in 2008 (Abrah´ am et al., 2009; Banzatti et al., 2012, see Chapter 2). These works pioneered the idea that the accretion history undergone by individual systems can be relevant not only as being 7


part of the overall disk evolution, but also as a phase of specific processing of dust grains and molecular gas in the planet formation region of disks. It is not yet clear how much disks offer observational evidences for processes directly related to planet formation, or rather be a thick blanket veiling everything that happen inside (like cocoons with butterflies). Only very recently observations are providing compelling evidences for protoplanets embedded within disks (Quanz et al., 2013a), starting to fill in the remaining chapter in the story from disks to planets. Less direct evidences of planet formation are being chased looking for discontinuities in molecular abundances that may trace snowline locations and radial transport of icy solids. This is one of the main efforts that have been undertaken in my PhD research (see Chapter 4), building on the heritage of attempts to put our Solar System into context (e.g. Meyer et al., 2007).

1.1.3

Planets: the remarkable “leftovers” of star formation

Although they are not the focus of this PhD dissertation, planets deserve at least a few words at the end of this section. It is remarkable that from the leftover material of the star formation process, complex objects like planets are made. An idea of this complexity could already be established by studying the Earth and the other planets in the Solar System. But today, this complexity is made even greater by the detection of several hundred planets, as well as thousands candidates (Borucki et al., 2010, 2011), around stars other than the Sun (“extrasolar planets”, or “exoplanets”), which show properties very different from the Solar System (e.g. Chiang & Laughlin, 2013). It is still premature to translate the few observed planet properties (masses, radii, orbital radii) into detailed descriptions of planetary surfaces and atmospheres, but their bulk compositions are at least as remarkably diverse as the exoplanet galore (see Figure 1.4). Questions in planet formation are therefore multiplied. Does the diversity of exoplanets relate to different conditions (chemical, physical, and/or evolutionary) found in the parent disk? What are the main processes responsible for setting planetary compositions, and do they happen early or late during the formation of a planet? Do planets take their seeds only locally in the disk, or mixing and large-scale migration dominate in planet formation regions? Will it ever be possible to provide some prediction on the outcome of planet formation (architecture and composition of planets) from the observed properties of circumstellar disks? Are Earth-like planetary surfaces, and conditions for the development of life, common in exoplanets? It goes beyond the scope of this dissertation to address any of these questions, but spectroscopy of the planet formation region of disks promise to be a powerful tool to reach at least some answers in the future.

8


Fig. 1.4 — Top: mass-radius diagram for all exoplanets detected so far (from exoplanet.eu). Bottom: same diagram (in Earth units) for a sample of low-mass planets (mostly from the Kepler mission), showing models of different bulk compositions (from Sohl et al. (2012)).

9


ission

em ganics and or r e t a w Warm

lin ow

sn ter Wa

e

Icy solids migration

Ice evaporation Gas diffusion and condensation on forming planets

Reservoir of gas molecules 1 AU

Formation of terrestrial planets

Reservoir of icy solids 10 AU

100 AU

Fig. 1.5 — Representative sketch of a protoplanetary disk, with some processes thought to be involved in planet formation: icy solids migrate inward from the outer colder disk regions, and evaporate when crossing the snowline; the strong enhancement in gas density inward of the snowline causes diffusion outward, followed by freeze-out on forming rocky planetesimals. Background illustration credit: NASA/JPL-Caltech, R. Hurt

1.2

Water in Protoplanetary Disks

The terrestrial planet formation region of disks (1–10 AU) is to date the least understood and the most difficult to study. Constraining dynamic processes and especially the role and the behavior of the gas in this region is of major importance to understand both disk evolution and planet formation (Armitage, 2011). It is in this context that the warm water emission observed toward disks comes into help. The second part of this introductory chapter is dedicated to set the basis to understand why that is the case, while deeper insights will be provided in the main chapters of this PhD thesis. The water molecule (H2 O) could in principle be the most abundant molecule in the Universe after H2 . With its two hydrogen and one oxygen atoms, it is made of the first and third most abundant elements (the second is Helium). Molecular clouds, as mentioned at the beginning of this chapter, do host a variety of molecules, but water is certainly not the most abundant. This leads us directly to a key problem: the origin of water. Common mechanisms to form molecules are: radiative association, 10

1.2. Water in Protoplanetary Disks

neutral-neutral and ion-neutral reactions, and grain surface chemistry (e.g. Stahler & Palla, 2005). The first two mechanisms require high densities and temperatures that are usually not found in molecular clouds. Specifically, the high energy needed to form water through neutral-neutral reactions exceeds 1000 K 3 . This is likely the reason why water vapor is observed in shocked regions and outflows, but generally it is not found diffusely in molecular clouds (Bergin & van Dishoeck, 2012). The third mechanism is a viable efficient path to molecule formation, but relies on the availability of ions in ionized regions (close to strong X-ray/UV sources, e.g. massive stars). The fourth involves the catalytic action of dust grain surfaces, followed by release in the gas phase (thermal or photo-evaporation). In several regions water is observed in the solid phase with abundances approaching ∼ 10−4 relative to H2 (Gibb et al., 2004; Sonnentrucker et al., 2008). This may be an indication that it is efficiently formed on dust grains, and suggests that water is probably delivered to circumstellar disks primarily as ice (Bergin & van Dishoeck, 2012), at least until temperatures do not exceed 100 K (which is true for most of the initial stages of star formation). Once in disks, mid-infrared observations show that it is abundant in the gas-phase (& 1018 cm−2 Carr & Najita, 2008, 2011; Salyk et al., 2008, 2011). The vapor is hot (300-1000 K) and observed in emission, as coming from hotter disk layers (probably from a hot surface on top of a colder midplane). In addition to the temperature, line profiles (although difficult to interpret so far) suggest an inner disk origin within a few AU (Pontoppidan et al., 2010b). There are two preferred ways to explain water vapor in the inner regions of disks. One proposes that water is evaporated from ice grain mantles when they cross the snowline (Cuzzi & Zahnle, 2004; Ciesla & Cuzzi, 2006), and that water concentration enhancements are expected to happen early in the disk evolution as a consequence of strong waves of migrators. A second interpretation proposes that water is formed in the disk atmosphere via gas-phase neutral-neutral reactions from the ingredients available in situ (oxygen and molecular hydrogen) (e.g. Glassgold et al., 2009). Constraining the relative importance of these two mechanisms is linked to our fundamental understanding of the regions probed by the observed emission. As the solar oxygen abundance is roughly ∼ 10−3 relative to H2 (Asplund et al., 2009), and it is found to be roughly solar in F to K stars (Petigura & Marcy, 2011), one could expect to find in disks water abundances of the same magnitude, if water is formed in a static disk of homogeneous solar material. The water abundance is difficult to estimate from the mostly optically thick emission observed in disks, but could be as high as ∼ 10−2 relative to H2 if the gas to dust ratio is 100 (see Chapter 4), consistent with the migration and evaporation scenario (a maximum enhancement of a factor 10 compared to solar values was predicted in the simulations by Ciesla & Cuzzi, 2006). Whatever its origin be, water is the main gas-phase oxygen carrier (and probably also the main line coolant, see Pontoppidan et al., 2010a) in disk regions where the temperature exceeds 300 K, if carbon is mostly kept into grains. This makes water 3

A common way to express energies is in temperature units E/kB , where kB is the Boltzmann constant.

11


Fig. 1.6 — Structure of the water molecule, with one oxygen and two hydrogen atoms opening an angle of 105◦ from the oxygen. The electric dipole moment µ is on axis B, directed toward the hydrogens. From Stahler & Palla (2005). an exceptionally good tracer of disk structure and evolution in the inner 10 AU (e.g. Zhang et al., 2013), where the snowline is believed to be and where processes relevant for terrestrial planet formation likely happen (see Figure 1.5).

1.2.1

Basics of infrared spectroscopy of the water molecule

In a field where most information is provided by radiation, spectroscopy has proven to be essential to study the properties of astronomical objects. A lot of what is known in star formation is due to studies of atomic and molecular emission. Identification of emission/absorption lines reveal the chemical composition of molecular clouds, dense cores, stars, and disks. Resolved line profiles trace the dynamics of rotating, infalling, or accreting gas (from the width, shape, and Doppler shift of lines), and the presence of winds and outflows. Relative strengths of lines hold information on the excitation conditions (from the relative population of different quantum levels), and can provide estimates of gas temperatures and densities. While cold regions (10 K) are probed at radio wavelengths, where the emission peaks, hotter gas emits strongly at shorter wavelengths due to Wien’s law (λpeak ∝ 1/T ). In particular, concerning disks, high-resolution (R & 1000) infrared spectroscopy opened the way to study the conditions of the gas in the planet formation region, and is therefore the main technique utilized in this work. In the following, I will focus on rotational transitions (and not describe electronic and vibrational) as they dominate the water vapor emission observed toward circumstellar disks. For a simple diatomic molecule (e.g. H2 and CO), the energy of rotation in quantum mechanics is described by the moment of inertia I and the rotational 12


Fig. 1.7 — Energy-level diagram for the water molecule, showing levels with J up to 7 and Eu up to 1400 K. Modified from Stahler & Palla (2005). quantum number J, as ~2 J(J + 1) , (1.1) 2I where rotational levels correspond to positive integers of J, and rotational transitions between levels follow fixed rules for ∆J. The higher the I the closer in energy the levels, and in turn the closer in energy and the easier to populate the transitions. The peculiarity of the water molecule in this regard comes directly from its highly asymmetric structure with two hydrogen atoms opening an angle of 105◦ from the oxygen. This asymmetric top structure has a large permanent electric dipole moment (in the direction of the two hydrogens) and three different moments of inertia along the three axes (see Figure 1.6). The rotational energy of the water molecule has therefore three components like equation 1.1, but they being unequal the solution is a very complicated function of three rotational terms (Polyansky, 1985). For this reason the water spectrum has been for long time a real challenge, from the theoretical point of view to solve and predict the energetics, from the observational point of view to identify and interpret spectra (e.g. Wallace et al., 1995). Rotational levels are described by three quantum numbers (for the total angular momentum and the projection on the axes): J, KA , and KC . The large dipole moment allows for fast dipole transitions (∆J = 0 , ±1) to occur, with large Einstein coefficients for spontaneous emission Aul (up to a few 100 s−1 ), while KA and KC can change by Erot =

13


±1 or ±3. This, in addition to the three unequal moments of inertia, provide a large series of transitions that are easy to populate and very close in energy. Moreover, given the nuclear spin of the two hydrogens, J levels can be ortho (total spin unity, KA + KC = even) or para (total spin zero, KA + KC = odd). This produces two separate branches of levels where transitions in between are forbidden (see Figure 1.7). The consequence of its peculiar energy structure is that, under the necessary conditions (see Section 1.2.2), the typical spectrum of water vapor has a very dense forest of emission/absorption lines where J levels are scattered all over the range, particularly at infrared wavelengths. An example is showed in Figure 1.8.

1.2.2

Excitation conditions

The excitation temperature Tex of a given transition between two levels is defined starting from the Boltzmann relation ∆E gu nu = exp − , (1.2) nl gl kB Tex where n and g are respectively the population and the statistical weight (i.e. the degeneracy) of the upper (u) and lower (l) levels, and ∆E is the energy of the transitions (the energy difference between the levels). At high enough densities the gas is in local thermodynamic equilibrium (LTE), where the level populations are determined by collisions between molecules and Tex is equal to the kinetic temperature of the gas (Tkin ). Under such conditions, all transitions obey to the same relation and Tex is the same for all. This implies that from the relative population of levels (i.e. the relative strength of transitions) it is possible to derive Tex = Tkin . When the gas density is not high enough, the gas is not thermalized (non-LTE) and Tex 6= Tkin . If levels are populated by radiative pumping Tex = Trad , where Trad is a temperature associated to the radiation. In this regime, Tex can differ from transition to transition and the gas excitation conditions are more complicated. What distinguishes these two regimes (collisional/radiative) is the critical density ncrit , which is defined as Aul (1.3) ncrit = Cul where Cul is the collisional rate. The critical density is specific to every transition. If the gas density n is larger than ncrit , the transition is thermalized and Tex = Tkin . In the opposite case, the transition is sub-thermal. Typically, in sub-thermal (non-LTE) conditions transitions are weaker than in the LTE regime (i.e. subpopulated). When ncrit varies largely from transition to transition, as in the case of water, the spectrum can be composed of a mixture of thermal and sub-thermal lines and therefore be difficult to interpret (see Chapter 2). This also implies that, in general, a given transition will be more reflective of physical gas properties where n ∼ ncrit . If n ncrit the transition will be sub- (or non-) populated, if n ncrit the transition will not produce a photon that can escape the gas but rather excite other molecules. 14


Fig. 1.8 — Top: LTE spectrum of water vapor emission, having T = 300 K and N = 1019 cm−2 (details on the model are given in the Appendix of Chapter 2). To illustrate how the transitions are scattered over wide ranges of wavelengths, rotational lines from levels with the same rotational quantum number J (here we show J = 5) are marked with red arrows (cf. Figure 1.7). Bottom: zoomed-in region to illustrate the dense clustering of water lines, as observed with two resolving powers, R = 720 (representative of the Spitzer -IRS, see Chapter 4) and R = 20000 (representative of VLT/VISIR).

1.2.3

Estimation of emitting gas properties

Two typical methods to estimate the properties of an emitting gas from observed spectra are the fit to integrated line fluxes and the rotation diagram analysis. The underlying principle is the same for both, and basically consists in measuring the relative strength of transitions from different quantum levels. As explained above, this 15


is useful to estimate of the excitation conditions of the emitting gas, its temperature and density. However, there are differences that make one method preferred over the other under some circumstances. While fitting line fluxes require a reliable and realistic model that reproduces the observations, the rotation diagram analysis is less dependent on our detailed capability to reproduce the observations and provides a “visual” tool to help understand the excitation. In this thesis, both methods have been used as applied to water and OH emission. Chapters 2 and 4 discuss them in detail, with particular attention to their benefits and caveats. The high spectral density of water lines (see Figure 1.8) immediately points at the need of high resolving power to distinguish transitions from each other, the major issue that has plagued estimates of gas properties. So far, available instruments have posed the problematic choice between large spectral coverage but low resolution (e.g. Spitzer, see Chapters 2 and 4), or high spectral resolution but small spectral coverage (e.g. VLT/VISIR, see Chapter 3). Each approach has its advantages and disadvantages: the complex emission from water cannot be constrained well using a few lines, unless the gas is in ideal conditions of thermal equilibrium and low optical depth. The emission from circumstellar disks is therefore better constrained using large samples of lines. On the other hand, the instruments which provided large spectral coverage so far did not have enough resolution to distinguish nearby transitions from each other and suffered from degeneracies. This is one of the main problems tackled in this thesis, and is discussed in the following chapters.

1.3

Outline of this thesis

In this work, I study the conditions and evolution of the planet formation region in circumstellar disks, by means of spectroscopic observations of molecular gas emission. I primarily focus on the infrared spectrum of water vapor, which provides thousands of emission lines tracing the warm and dense gas inward of the snowline in disks. Aside from that, I also analyzed lines from some organic molecules that trace the carbon chemistry in the inner disk, C2 H2 (acetylene), HCN (hydrogen cyanide), and CO2 (carbon dioxide), as well as lines from OH (hydroxyl) that is connected to the formation and destruction of the water molecule. Accretion variability in the T Tauri phase has been considered as a tool to better understand the origin and evolution of the molecular gas in the planet formation region. In Chapter 2 I present a pioneering study of the change in molecular gas emission observed toward the strong variable T Tauri star EX Lupi. I compared blended infrared spectra obtained previous to and during a recent accretion outburst, and found that the gas changes remarkably between the two phases: water emission increases likely due to a recession of the snowline in outburst, OH emission increases due to water photodissociation, and emission from organics disappears. This work motivated a follow-up study using high spectral resolution at the ESO Very Large Telescope (VLT), which is reported in Chapter 3: DR Tau was monitored simultane16

1.3. Outline of this thesis

ously over the range UV to mid-infrared with two high-resolution spectrographs, in order to investigate the effects of accretion variability on water emission in another (less extreme) system in the T Tauri phase. Chapter 4 tackles another fundamental problem, the origin of water vapor in inner disks. Competing theories provide two different perspectives, where in situ gas-phase formation in a static disk bears only local information on disk properties, while migration and evaporation of icy solids would make water emission a good tracer of disk evolution and planet formation conditions. A rotation diagram analysis of a large number of emission lines, de-blended from Spitzer spectra, provides hints of large column densities in support the migration scenario. In the concluding chapter (Chapter 5), I summarize the major results of my research and the issues that remain unsolved, opening new perspectives for future investigations.

17


18

“QUAND

TU VEUX CONSTRUIRE UN BATEAU, NE COMMENCE PAS PAR RASSEMBLER DU BOIS,

COUPER DES PLANCHES ET DISTRIBUER DU TRAVAIL, MAIS REVEILLE AU SEIN DES HOMMES LE DESIR DE LA MER GRANDE ET LARGE.”

´ ANTOINE DE SAINT EXUPERY, CITADELLE, 1948

2

EX Lupi in outburst: changes in the disk molecular emission

Based on a paper published in The Astrophysical Journal, 745, 90 (2012) 4

A. Banzatti1 , M. R. Meyer1 , S. Bruderer1 2 , V. Geers1 , I. Pascucci ´ , F. Lahuis5 , A. Juh´ asz6 7 , T. Henning6 , P. Abrah´ am8

3

1

ETH Z¨ urich, Institut f¨ ur Astronomie, Wolfgang-Pauli-Strasse 27, CH-8093 Z¨ urich, Switzerland Max-Planck-Institut f¨ ur Extraterrestrische Physik, Giessenbachstr. 1, D-85748 Garching bei M¨ unchen, Germany 3 Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA 4 Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA 5 SRON Netherlands Institute for Space Research, P.O. Box 800, NL 9700 AV Groningen, The Netherlands 6 Max-Planck-Institut f¨ ur Astronomie, Königstuhl 17, D-69117 Heidelberg, Germany 7 Leiden Observatory, Leiden University, P.O. Box 9513, NL-2300 RA Leiden, The Netherlands 8 Konkoly Observatory, Konkoly Thege Miklós 15-17, H-1121 Budapest, Hungary 2

19

CHAPTER 2. EX Lupi in outburst: changes in the disk molecular emission

Abstract We present a comparison of archival Spitzer spectra of the strongly variable T Tauri EX Lupi, observed before and during its 2008 outburst. We analyze the mid-infrared emission from gas-phase molecules thought to originate in a circumstellar disk. In quiescence the emission shows a forest of H2 O lines, highly excited OH lines, and the Q branches of the organics C2 H2 , HCN, and CO2 , similar to the emission observed toward several T Tauri systems. The outburst emission shows instead remarkable changes: H2 O and OH line fluxes increase, new OH, H2 , and Hi transitions are detected, and organics are no longer seen. We adopt a simple model of a singletemperature slab of gas in local thermal equilibrium, a common approach for molecular analyses of Spitzer spectra, and derive the excitation temperature, column density, and emitting area of H2 O and OH. We show how model results strongly depend on the selection of emission lines fitted, and that spectrally-resolved observations are essential for a correct interpretation of the molecular emission from disks, particularly in the case of water. Using H2 O lines that can be approximated as thermalized to a single temperature, our results are consistent with a column density decrease in outburst while the emitting area of warm gas increases. A rotation diagram analysis suggests that the OH emission can be explained with two temperature components, which remarkably increase in column density in outburst. The relative change of H2 O and OH emission suggests a key role for UV radiation in the disk surface chemistry.

2.1

Introduction

EX Lupi is a young star+disk system whose photometric variability was discovered in 1944 (McLaughlin, 1946). Since then the source has been monitored and showed repetitive eruptive phenomena related to significant changes in the amount of accreting material onto the star (Lehmann et al., 1995; Herbig, 2007; Juhász et al., 2011). Spectroscopic observations during the quiescent phases between the recurring outbursts suggest similarity with classical T Tauri stars of M0 type (Herbig et al., 2001; Sipos et al., 2009). The discovery of similar variable sources confirmed EX Lupi as the prototype of a class of young strongly active T Tauri stars, called “EXors” (Herbig, 1989, 2008). This class of objects has been proposed as an intermediate stage between FUors and classical T Tauri stars (Herbig, 1977; Teodorani et al., 1999), exhibiting repetitive outbursts that are both weaker and of shorter duration than FUors (Herbig, 2007, 2008). The 2008 outburst of EX Lupi, that we study in this work, is just the most recent of the series of eruptive events recorded during the half-century-long monitoring of the source, but more interestingly it is the most extreme ever. It showed an increase of ∼5 mag in visual light over seven months (January–September 2008), compared to the typical increase of ∼1–2 mag observed in its previous characteristic outbursts (Aspin et al., 2010) as well as in other T Tauri stars (e.g. Herbst et al., 1994). A comparable optical brightening of 20

2.1. Introduction

the source was previously observed only in 1955, but spectroscopic observations were not performed (Herbig, 1977). The 2008 event, instead, was monitored at several wavelengths over its entire duration, providing a deeper insight into the complex behavior of EX Lupi and the still not well understood mechanism producing its ´ strong outbursts (Abrah´ am et al., 2009; Aspin et al., 2010; Grosso et al., 2010; Goto et al., 2011; Juhász et al., 2011; Kóspál et al., 2011). In the last few years the Spitzer Space Telescope (Werner et al., 2004) has enabled the study of the molecular content in circumstellar disks through their mid-infrared (MIR) emission, unveiling a rich chemistry in the innermost regions of T Tauri disks (Salyk et al., 2008, 2011; Carr & Najita, 2008, 2011; Pascucci et al., 2009; Pontoppidan et al., 2010a). So far, such emission has been found to be common in the T Tauri systems observed (Pontoppidan et al., 2010a) and has been studied by means of simple single-slab models that assume the emitting gas to be in local thermal equilibrium (LTE). A certain diversity in the molecular emission observed toward T Tauri systems is apparent (Pascucci et al., 2009; Pontoppidan et al., 2010a; Carr & Najita, 2011), and disks in more evolved systems as well as around Herbig AeBe stars seem to be depleted in warm molecules compared to T Tauri disks (Pontoppidan et al., 2010a; Najita et al., 2010; Fedele et al., 2011). From the theoretical side, several attempts have been made to explain the formation and survival of molecules (especially water) in disks and their abundance during the early phases of planet formation (e.g. Ciesla & Cuzzi, 2006; Glassgold et al., 2009; Bethell & Bergin, 2009; Vasyunin et al., 2011). How does the warm molecular gas composition depend on stellar+disk properties, and how may the interplay and evolution of the different components enable (instead of hinder) the availability of essential molecules for planet formation? The most recent outburst in EX Lupi has provided an extraordinary opportunity to address this compelling questions. While it is difficult to identify one factor that is the main responsible for the diversity in molecular emission of classical T Tauri systems, due to the many properties that vary from system to system, in the case of EX Lupi we have the opportunity of a system where one parameter (the stellar+accretion luminosity) dramatically increases on timescales that can be monitored. Crystal formation in outburst from ´ amorphous dust particles was recently shown by Abrah´ am et al. (2009) comparing Spitzer spectra of EX Lupi, suggesting that episodic increases in disk temperature during outbursts may contribute significantly to the formation of some ingredients that are now present in our Solar System. Using the same spectra, but focusing on the gas, we show here how H2 O (water) emission varies in outburst together with other molecules like OH (hydroxyl), H i and H2 (atomic and molecular hydrogen), and organics like C2 H2 (acetylene), HCN (hydrogen cyanide), and CO2 (carbon dioxide), giving the opportunity to investigate the conditions of the emitting gas and the availability and survival of these important molecules during the unstable early phases in disk evolution.

21


2.2 2.2.1

Observations Spitzer IRS archival spectra

We collected all archival spectra of EX Lupi obtained with the Infrared Spectrograph (IRS) on the Spitzer satellite (Houck et al., 2004). High-resolution spectroscopy with the IRS is available in two different modules with a resolving power R ≈ 600: the short-high (SH) module in the wavelength range 9.9–19.6 µm and the long-high (LH) module between 18.7 and 37 µm. The IRS low-resolution modules instead cover the range 5.2–38 µm with a resolving power R ≈ 60–120. The IRS observations of EX Lupi span the epoch surrounding its latest outburst in 2008. In a quiescent phase before the outburst two different spectra were acquired: in 2004 (PI: N. Evans, hereafter E04) and in 2005 (PI: G. Stringfellow, hereafter S05), with both lowand high-resolution modules. The outburst phase was then observed in April 2008 ´ (PI: P. Abrah´ am, hereafter A08) and May 2008 (PI: J. Carr, hereafter C08) with the high-resolution modules, while post-outburst spectra during a new quiescent ´ phase were taken in October 2008 and April 2009 (PI: P. Abrah´ am, hereafter A09) only with low resolution. In Table 2.1 all the Spitzer IRS observations of EX Lupi are listed with their relevant observational parameters. Since the resolution of postoutburst spectra (A09) is too low to detect the large majority of the gas emission lines studied in this work, we restricted our investigation to the high-resolution spectra taken before and during outburst. More precisely, we decided to use in the present analysis only the spectra closer in time to the onset of the 2008 outburst, to study the changes in emission that can be related to it: in the quiescent phase the S05 spectrum and in outburst the A08 spectrum. The investigation of variations/correlations in emission over a broader range of timescales during a quiescent period (comparing E04 and S05) and short-term changes in outburst (comparing A08 and C08) may be addressed in a follow-up paper.

2.2.2

Data reduction

We re-reduced the IRS spectra starting from the Spitzer Science Center S18.7.0 products and applying the latest version of the pipeline provided by the “Cores to Disks” Spitzer legacy program, described in Lahuis et al. (2006). This pipeline was developed to reduce IRS pointed observations and reach high sensitivity particularly when observing faint objects. The extraction of one-dimensional spectra from the images is done using an optimal extraction, where the IRS point spread function, defined using sky-corrected high signal-to-noise (S/N) calibrators, is fitted on good pixels only (i.e. excluding known bad/hot pixels). Dither positions are first combined and then the spectrum is extracted, which gives the best resultant S/N. Moreover, this method provides an estimate of the local sky contribution, which was separately observed with Spitzer only in the case of the A08 spectrum. The EX Lupi spectra, like all IRS spectra of bright point sources, are affected 22

2.2. Observations

Tab. 2.1 — Summary of EX Lupi data from the Spitzer archive (IRS staring mode) Obs. ID

Date

Phase

Investigator (id)

172

2004-08-31

quiescence

Evans (E04)

3716

2005-03-18

quiescence

Stringfellow (S05)

477

2008-04-21

outburst

Abraham (A08)

50641

2008-05-02

outburst

Carr (C08)

524

2008-10-10 2009-04-07

quiescence

Abraham (A09)

Module SH LH SL2, SL1, LL1 SH LH SL2, SL1 SH, LH, SL2, SL1 SH, LH (backgr.) SH LH SH (backgr.) LH (backgr.) SL2, SL1, LL2, LL1 SL2, SL1, LL2, LL1

tint (s) 60 120 24 480 240 112 48 48 240 480 120 240 48 48

by systematics that reduce the maximum S/N achievable with long exposures (e.g. uncertainties in pointing correction, flux-dependent calibration, etc.). These uncertainties are difficult to account for and are not included in the errors on individual pixels propagated through the reduction pipeline. In our analysis we chose to derive our own estimate of the noise as follows. We measure in each spectrum the dispersion of pixels from a baseline (first-order polynomial) fitted to the continuum. Unfortunately, given the spectral resolution of Spitzer, a clean continuum is difficult to find and often we can constrain only a pseudo-continuum given by weak emission lines. We use the HITRAN database to avoid ranges where the emission from obvious molecules is likely to be strong (see Section 2.3.1). The pixel dispersion with reference to the baseline is then checked for gaussianity with the Shapiro-Wilk test (Shapiro & Wilk, 1965). We consider four spectral ranges in SH (11.95–12.2, 15.44–15.6, 17.45–17.7 and 19.05–19.2 µm) and four in LH (24.15–24.55, 25.45–25.85, 30–30.2 and 31.3–31.65 µm) as an attempt to account for the variations of noise with wavelength. We discard the rms estimated in a given range if the probability from the Shapiro-Wilk test that the data are consistent with having been drawn from a Gaussian is less than 10%. As a result, in quiescence (S05) we find an rms of ∼3 and ∼9 mJy in SH and LH respectively, while in outburst (A08) we find ∼8 and ∼20 mJy respectively. We assume these estimates as the flux uncertainty on individual pixels in the two IRS modules separately, and we use them in the derivation of line flux errors as described in Section 2.3.2. On the one hand these should be regarded as conservative estimates as the molecular emission in the chosen ranges is expected to be small, but weak features (as well as chemical species that are not considered) 23


may still contribute to the pixel dispersion and thus artificially increase the noise. On the other hand, if high systematics strongly affects Spitzer spectra our estimates might still be underestimating the true uncertainty (cf. Salyk et al., 2011; Carr & Najita, 2011). However, we have some indications that our estimates of the noise are conservative rather than optimistic (see Sections 2.3.2 and 2.5.1).

Fig. 2.1 — Comparison between the quiescent (S05) and the outburst (A08) phases in the Spitzer spectra of EX Lupi: the SH module is shown in the top and the LH module in the bottom. The quiescent spectrum is shifted for comparison in both plots. OH transitions are marked with dashed lines, while other species (C2 H2 , HCN, CO2 , H i, H2 , [Ne ii]) are marked with dotted lines. All unmarked lines are identified as water rotational transitions.

24

2.3. Spectral line analysis

2.3 2.3.1

Spectral line analysis Molecular and atomic emission

The identification of molecular emission features in the Spitzer observations of EX Lupi is based on the comparison between the observed spectra and synthetic models generated using the HITRAN 2008 database (Rothman et al., 2009). The MIR spectra of EX Lupi are dominated by a forest of water emission lines9 , placing this source within the sample of “wet” T Tauri disks recently revealed by Spitzer (Carr & Najita, 2008, 2011; Salyk et al., 2008; Pontoppidan et al., 2010a). H2 O lines are detected in both the quiescent and the outburst phases of EX Lupi and their measured flux with respect to the continuum increases in outburst over the entire spectral range covered by Spitzer, on top of the continuum level that is ∼4 times higher than in quiescence (see Figure 2.1). The water vapor emission is composed of unresolved pure rotational transitions in the ground vibrational state with upper level energies (Eu ) of a few 1000 K. The complex excitation structure of water makes its observational study a real challenge, especially at low spectral resolution. Even small spectral ranges in the infrared can be finely populated by rotational transitions from different Eu , which are resolved only with a spectral resolution much higher than that provided by Spitzer (see e.g. Figure 5 in Pontoppidan et al., 2010a). The criticality of such a limitation on the derivation of the emitting gas properties will be addressed in Sections 2.4 and 2.5. In addition to water, the MIR spectra of EX Lupi reveal a variety of other molecules and atoms with a noticeable difference between the two activity phases of the star. In quiescence the rovibrational Q-branches of C2 H2 at ∼13.7 µm, HCN at ∼14 µm, and CO2 at ∼15 µm are clearly detected, while a weak contribution from OH is identified. In outburst organic molecules are no longer seen, while strong H i 7–6, H i 14–9, H2 S(2), and several new OH lines appear and increase the confusion with water emission, especially in the SH module. The OH emission consists of rotational transitions in the ground vibrational state with Eu from ∼600 up to ∼20,000 K and is detected in outburst throughout the entire spectrum. These transitions are mainly intra-ladder transitions belonging either to the 2 Π3/2 or the 2 Π1/2 pure rotational ladders, but several cross-ladders transitions are also identified, indicating a remarkable population of intermediate and low Eu (see Figure 2.2 and Table 2.2). To our knowledge, this is the first time that cross-ladder transitions are identified in disks, especially those with Eu ∼ 1000-2000 K. Tappe et al. (2008), in the HH211 outflow, were the first to ever report the detection of OH transitions from levels as high as ∼28,200 K, and they also identified two lower-Eu cross-ladder transitions (one of which, at 28.94 µm, is also detected by us in the EX Lupi outburst spectrum). In EX Lupi, additional OH lines from Eu as high as those found by Tappe et 9

In the text we distinguish “lines” from “transitions”. The former refers to the unresolved emission features observed in the spectra, while the latter to the individual molecular transitions that produce the unresolved observed features.

25


Fig. 2.2 — Cross-ladder OH transitions detected in EX Lupi in outburst. Each plot shows, from top to bottom, the outburst A08 spectrum, the quiescent S05 spectrum (shifted in flux for comparison), and an LTE model for OH with Tex ∼ 600 K and Nmol ∼ 3 ×1017 cm−2 . In the OH model, which includes both intra-ladder and crossladder transitions, the position of cross-ladder transitions is indicated with vertical dashed lines. In each plot, observed emission lines that do not match the OH model shown in the bottom are attributed to H2 O, apart from the 14.97 µm feature in quiescence due to CO2 .

al. (2008) might be present in the outburst spectrum, but their detection is hindered by the strong silicate feature at ∼10 µm. We will come back to the analysis of OH emission, and the importance of the simultaneous detection of low- and high-Eu lines in outburst, in Section 2.5.2. 26


Tab. 2.2 — OH detections in EX Lupi λ (µm) 11.06 12.65 13.07 13.55 14.07 14.65 15.00 16.66 18.73 18.85 23.10 23.22 27.42 27.67 28.94

Transitions 1/2 1/2 1/2 1/2 1/2 1/2

(57/2 → 55/2), 3/2 (59/2 → (47/2 → 45/2), 3/2 (49/2 → (45/2 → 43/2), 3/2 (47/2 → (43/2 → 41/2), 3/2 (45/2 → (41/2 → 39/2), 3/2 (43/2 → (39/2 → 37/2), 3/2 (41/2 → 1/2 (17/2) → 3/2 (15/2) 1/2 (15/2) → 3/2 (13/2) 1/2 (13/2) → 3/2 (11/2) 1/2 (29/2 → 27/2) 3/2 (25/2 → 23/2) 1/2 (23/2 → 21/2) 3/2 (21/2 → 19/2) 1/2 (19/2 → 17/2) 1/2 (7/2) → 3/2 (5/2)

57/2) 47/2) 45/2) 43/2) 41/2) 39/2)

Eu (K)

Aul (s−1 )

S05

21,300 15,100 13,900 12,800 11,800 10,800 2400 2000 1550 6250 4100 4150 2900 2960 620

1300 850 780 690 620 540 0.2 0.2 0.2 250 130 130 80 80 0.1

√ √ ? ? ?

A08 √ √ √ √ √ √ √ √ √ √ √ √ √ √

Note. — At the spectral resolution of the Spitzer IRS, all observed OH lines are blends of doublets of intra-ladder or cross-ladder transitions belonging to the two pure rotational ladders 2 Π3/2 and 2 Π1/2 in the ground vibrational state. In this table we provide in parentheses the upper and lower J -level of each doublet. Cross-ladder doublets are reported as a single feature even when two lines can be distinguished (see Figure 2.2), and in the rotation diagram analysis we use the unified flux (Section 2.5.2). The wavelengths we report are approximated at the center of each feature. Molecular data are taken from the HITRAN database. The Aul reported here are the sums over the individual transitions blended in each observed line. The last two columns on the right indicate the detection in the quiescent (S05) and outburst (A08) spectra (S/N > 2σ). A question mark shows lines that are not unambiguously identified as OH.

In Spitzer spectra all the chemical species mentioned above are blended with H2 O lines. This, in addition to the above mentioned blending of transitions from different energy levels, hinders not only the molecular line identification but also the derivation of gas properties. Excluding the organic molecules, whose Q-branches are broader than the gas-phase water lines, it can be misleading to identify molecules from only 1-2 emission lines observed where their emission is expected to be, unless such lines prove to be clearly separated by other species. The case of OH is particularly good as an example. In outburst we can identify all the expected lines from the two rotational ladders in the ground level, thus the molecule is certainly detected. 27


In quiescence, instead, only a few lines are unambiguously identified, at 12.65 and 14.65 µm, which are the most spectrally separated by H2 O lines. All other tentative OH detections are confused with other molecules. The case of OH lines at longer wavelengths (the LH module) is particularly difficult to judge, especially because of the unconstrained H2 O emission (see Section 2.5.1). However, consideration of reasonable models for the H2 O emission suggests that the data in LH in the S05 spectrum can be explained as H2 O emission alone. Therefore the (partial) detection of OH in quiescence is based on two lines in the SH module, while in the LH module it is highly dubious. Other chemical species that have been identified in the MIR in other T Tauri stars may additionally contribute to the emission observed in EX Lupi (see e.g. Najita et al., 2010). However, in the present paper we focus our analysis on H2 O and OH, which largely dominate the MIR emission that we observe in EX Lupi. We assume that the MIR emission from these two molecules originates in the innermost disk atmosphere (∼1–10 AU from the central star), as proposed for other systems in similar recent studies (Carr & Najita, 2008, 2011; Pontoppidan et al., 2010a; Salyk et al., 2011) and confirmed by higher-resolution observations for a few objects (Pontoppidan et al., 2010b; Fedele et al., 2011).

2.3.2

Estimation of emission line fluxes and errors

Given the need for reliable tracers to probe the emission variability, we chose to use only the most trustable emission lines and to measure their flux and error using a robust method based on our rms estimate. We restrict the analysis to strong and/or isolated emission lines which we believe are the least affected by artifacts. For instance, the lines in spectral ranges where different spectral orders overlap, which are affected by drop of signal, and those where the continuum or pseudo-continuum is more uncertain are excluded from the analysis. In Table 2.3 we list the emission lines considered in this paper, indicating the spectral range of each observed feature (including the nearby continuum), the chemical species identified as contributing to the emission, and the flux measured in quiescence and outburst. Our method partly resembles techniques previously utilized for the measure of line fluxes in Spitzer spectra (e.g. Pontoppidan et al., 2010a; Najita et al., 2010), and for the estimate of flux uncertainties (Pascucci et al., 2008). We developed our method in IDL building on publicly available routines as follows. For each observed emission line we select a spectral range including at least three pixels on each side for the continuum (typically a few tenths of a micron). Where a continuum cannot be found in between two or more lines because of their vicinity, we fit them together. We fit the data using a least-squares Levenberg-Marquardt algorithm (Markwardt, 2009) with a baseline plus one Gaussian function for each line considered, with width and area as free parameters. Since we fit the continuum locally, we always assume a first-order polynomial for the baseline even though a higher order could in some cases slightly improve the fit. Each line flux is computed as the area of the 28


locally continuum-subtracted best-fit Gaussian. The observed emission line(s) must be well represented by one (or more) Gaussian functions, and we set an acceptance threshold to the goodness of the fit to χ2red . 1.5. In most cases the χ2red is much lower than that limit, supporting that our noise estimate is indeed conservative (see e.g. Figure 2.3). This method provides the possibility to measure the individual line fluxes even in the case of complexes composed of several blended lines from the same or different chemical species (an example is shown in Figure 2.3), unless the different lines are overlapping too much to be distinguished, as it is often the case in the LH module. This approach is not used for the broad emission features from C2 H2 , HCN, and CO2 , because the shape of the Q-branch, which is at least partially resolved by the IRS, is not Gaussian. We therefore fit only a baseline continuum to nearby regions as described above and then integrate the flux below the continuum-subtracted data. The 1-σ error on each line flux is then estimated using the following Monte-Carlo based approach. In each of the spectral regions considered above, we add normally distributed noise to the observed flux of individual pixels, that we take as the mean. As the noise to be added we use the rms calculated from the supposed line-free regions (Section 2.2.2). The procedure is repeated 1000 times and we measure the line fluxes from each realization exactly as described above. The error on the line flux is then set to a robust estimate of the standard deviation of the distribution of its 1000 realizations (see Figure 2.3). With this method we take into account the noise on single pixels (given by our estimate of the rms) as well as the uncertainty on the local fit to the continuum.

29


Fig. 2.3 — Measure of line flux and error for emission lines in the S05 spectrum. In the first plot (top, left) the best-fit curve (first-order polynomial + three Gaussians) is overplotted to the original data, where the error on individual pixels is set to the measured rms dispersion (see Section 2.2.2). The χ2red of the best fit is 0.2. Such a low value suggests that our rms estimate is conservative. The error on each line flux is set to a robust standard deviation of the distribution of 1000 repetitions of the flux measure (see Section 2.3.2). The first two lines are H2 O lines, while the third is OH.

30


Tab. 2.3 — Flux estimates of emission lines in the Spitzer -IRS spectra of EX Lupi Spectral Range (µm)

Line (µm)

11.00–11.10 12.22–12.50 12.22–12.50 12.46–12.72 12.46–12.72 12.46–12.72 12.67–12.90 12.95–13.15 13.40–13.60 13.40–13.60 13.40–13.60 13.50–13.80 13.75–14.12 13.95–14.30 13.95–14.30 14.43–14.61 14.55–14.85 14.80–15.05 14.93–15.10 15.67–15.87 15.85–16.10 16.05–16.18 16.53–16.80 16.73–16.99 17.03–17.19 17.14–17.31 17.28–17.46 18.65–18.80 18.73–19.11 18.73–19.11 20.68–20.90 21.70–22.04 22.82–23.37 22.82–23.37 27.20–27.88 27.20–27.88 28.81–29.07 29.72–30.10 30.40–31.05 30.40–31.05 32.78–33.40

11.06 12.28 12.37 12.51 12.58 12.64 12.81 13.06 13.43 13.49 13.55 13.70 14.00 14.06 14.20 14.51 14.65 14.95 15.00 15.74 16.00 16.11 16.66 16.89 17.11 17.22 17.36 18.73 18.85 19.00 20.79 21.84 23.10 23.22 27.42 27.67 28.94 29.84 30.49 30.88 33.00

Quiescence (S05) ... OH+H2 O H2 O H2 O H2 O OH [Ne II]+OH H2 O+OH H2 O H2 O OH C2 H2 +H2 O HCN+H2 O OH H2 O H2 O OH CO2 +H2 O ... H2 O H2 O H2 O H2 O H2 O H2 O H2 O H2 O ... ... H2 O H2 O H2 O ... H2 O ... H2 O H2 O H2 O H2 O H2 O H2 O

1.22 1.51 1.23 0.62 0.82 0.66 0.73 0.90 0.39 2.76 7.48 1.38 1.87 1.10 3.20 1.48 1.43 0.85 1.47 1.34 1.01 1.26 2.14

1.24 2.85 5.51 0.50 1.21 1.23 1.66 1.05 3.12 5.03

... ± ± ± ± ± ± ± ± ± ... ± ± ... ± ± ± ± ... ± ± ± ± ± ± ± ± ... ... ± ± ± ... ± ... ± ± ± ± ± ±

Outburst (A08) 0.55 0.46 0.25 0.23 0.26 0.28 0.38 0.14 0.16 0.29 0.63 0.30 0.20 0.39 0.29 0.17 0.22 0.19 0.18 0.22 0.14 0.15 0.28

0.11 0.28 0.44 0.19 0.35 0.22 0.27 0.26 0.25 0.36

OH H2 +OH+H2 O H i+H2 O H2 O H i+H2 O OH [Ne II]+OH OH H2 O H2 O OH ... ... OH H2 O H2 O OH H2 O OH H2 O H2 O+OH H2 O H2 O+OH H2 O+OH H2 O H2 O H2 O OH OH H2 O H2 O H2 O OH OH OH OH OH H2 O H2 O H2 O H2 O

2.63 ± 0.51 2.87 ± 0.43 6.78 ± 0.59 ... 3.71 ± 0.61 ... 2.59 ± 1.13 1.33 ± 0.51 3.27 ± 0.53 0.86 ± 0.31 1.51 ± 0.44 ... ... 1.22 ± 0.37 1.71 ± 0.56 2.08 ± 0.46 3.51 ± 1.19 3.36 ± 0.77 1.35 ± 0.66 2.04 ± 0.45 4.34 ± 0.95 ... 2.87 ± 0.76 5.76 ± 0.84 1.61 ± 0.66 3.36 ± 0.60 2.70 ± 0.39 1.45 ± 0.55 2.17 ± 0.61 3.37 ± 0.97 2.78 ± 0.74 6.52 ± 0.57 7.26 ± 2.87 5.64 ± 1.64 6.43 ± 1.43 4.58 ± 0.82 1.45 ± 0.56 9.39 ± 1.19 6.88 ± 0.92 4.61 ± 0.51 8.62 ± 0.82

Note. — We give the spectral range of the observed emission features where we apply our method to derive line fluxes and errors (see Section 2.3.2). For each range we indicate the identified species most contributing to the measured flux. In the cases where it was possible to distinguish different lines within a feature (e.g. see Figure 2.3), we report the components separately. The line flux is given in units of 10−14 erg s−1 cm−2 . Where the flux is omitted, emission lines are not clearly identified and/or the measured flux is consistent with zero.

31


2.4 2.4.1

Constraining gas properties An LTE approach

A simple approximation utilized in observational astrochemistry is to assume the emitting gas to be in local thermal equilibrium. In LTE the population of the excited levels simply follows the Boltzmann distribution, i.e. all levels are thermalized to the kinetic temperature (Tk ) of the gas. A transition between two levels can be described as to be in LTE when collisions dominate the excitation/de-excitation, i.e. when the density of the ambient gas is higher than the critical density of the transition (ncrit = Aul /Cul , where Aul is the Einstein-A coefficient and Cul the collision rate). If all transitions are in LTE, the kinetic temperature and the column density of the gas can be derived from the observed flux of spectrally-resolved transitions. The techniques that are most frequently used to derive gas properties are based on the rotation diagram analysis (e.g. Goldsmith & Langer, 1999, hereafter GL99) or the direct fit of transition intensities (e.g. Meijerink et al., 2009, hereafter ME09). In the extremely diverse conditions found in astrophysical environments, however, it is often the case that the ambient density is insufficient to thermalize some (or all) levels and the LTE description falls short of deriving the real gas properties. ME09 showed how the LTE approximation is inappropriate for circumstellar disks, as the wide range of thermodynamic conditions found in the inner ∼ 10 AU of disks (gas temperatures ranging from ∼ 100 to a few 1000 K, and densities from ∼ 103 to 1016 cm−3 ) results in emission produced by a mixture of thermalized and non-thermalized levels. GL99 showed how the rotation diagram analysis can help in distinguishing the effects of optical depth and non-thermal excitation on the typical straight-line behavior of optically thin lines in LTE. However, for this type of analysis spectrally resolved observations are required, to separate different energy levels. With the spectral resolution provided by Spitzer this is possible in the case of OH, but not water. The implementation of such a technique in the case of OH emission will be presented in Section 2.5.2. Now we focus on the method used to derive constraints on the excitation properties of the unresolved and blended water emission. Non-LTE models that perform line radiative transfer and account for the geometrical and physical structure of disks have been explored to some extent (e.g. Pavlyuchenkov et al., 2007), but they are not yet viable alternatives for constraining water MIR line emission because of the incompleteness and uncertainty of collision data (see van der Tak, 2011). This is why the simple LTE approximation is still the first approach utilized when modeling the molecular emission observed in Spitzer spectra (Carr & Najita, 2008, 2011; Salyk et al., 2011). We therefore developed our own simple single-slab LTE model to constrain the molecular emission in EX Lupi. The model is explained in detail in Appendix 2.8, while here we provide only a short description. The free parameters of the model are the excitation temperature Tex , the column density Nmol , and the emitting area A. In the fit results we will report the value of the radius, rp (projected), of the circular area that is equivalent to 32

2.4. Constraining gas properties

A. The excitation and relative strength of different transitions are sensitive to Tex and Nmol , while A only has the effect of a scaling factor common to all line fluxes. Observed and model spectra are compared using the individual fluxes of blended emission lines (given in Table 2.3), and a χ2 test is performed minimizing over the difference between observed and model line fluxes as an ensemble. Despite the fact that the model parameters are sensitive to different properties and that we account for line optical depth effects, the fit results are degenerate in Tex , Nmol , and A. In general, for a given emitting area the fit is degenerate with higher Tex for lower Nmol (and vice versa, see e.g. results in Section 2.5.1). In the extreme optically thick regime (τ 1) the line flux depends on the product A × Tex , thus only those can be derived while the model is insensitive to changes of Nmol . In the extreme optically thin case (τ 1), instead, the line flux is proportional to A × Nmol × Tex , and we can reduce again to two parameters: the temperature and the number of emitting molecules = A × Nmol , yet still suffering of the degeneracy between Tex and Nmol . Ideally, three emission lines that are differently sensitive to variations in Tex and Nmol should be enough to constrain all model parameters together. Water MIR transitions do cover a wide range of opacities and therefore an appropriate selection of them could serve the scope. However, in practice their different opacity cannot be retrieved unambiguously from the blended lines they produce in the Spitzer spectra and the degeneracy between model parameters is somewhat unavoidable.

2.4.2

Fitting Spitzer spectra of H2 O emission with LTE models

Fitting the MIR unresolved water emission in EX Lupi using LTE single-slab models is challenging. One must keep in mind that such simplified models do not account for any disk geometry, structure, nor temperature/column-density profiles, which are complex (see e.g. Bergin et al., 2007) and in the case of EX Lupi might also change from quiescence to outburst. In addition to that, flux contamination from different molecules can also affect the fits, unless different contributions can be individually constrained and removed (which usually require many “clean” lines of the same species and a reliable modeling). In fact, we found that in EX Lupi it is impossible to get a good fit over the wide spectral range covered by the Spitzer IRS with a single set of Tex and Nmol . Yet it is possible to obtain a reasonable fit for sub-sets of lines, that likely share similar excitation conditions. Even in the literature there is still no common agreement on the most appropriate way to fit single-slab LTE models to the water MIR emission from disks observed with Spitzer. Salyk et al. (2011) try to derive average gas properties for 48 T Tauri systems fitting peak-tocontinuum values of many emission lines over the entire 10–35 µm range, but they also note that it is impossible to reproduce entirely the observations using a singleslab LTE model and that the outcome varies if different spectral ranges and/or fitting methods are used. Carr & Najita (2008, 2011) instead focus on fitting lines observed in a small spectral range limited between 12 and 16 µm. Applying this 33


method to a sample of 11 T Tauri systems Carr & Najita (2011) report that while they can reach a very good match over this range, the same model overpredicts lines at longer wavelengths (16–19 µm). Different methods provide different results, so what is the most appropriate way to proceed and how shall we interpret different gas properties obtained fitting different spectral ranges? A viable solution comes from theoretical works. ME09 showed that LTE models generally overpredict the line flux if H2 O is not in LTE, as emission lines excited in non-LTE regimes are sub-thermally populated (see also Pavlyuchenkov et al., 2007, who explored the case of other molecules). In their Figure 8, ME09 show that the difference in line flux of a water transition in the ground vibrational level excited in LTE and in non-LTE conditions is minimal for small values of Aul . If the excitation state of the emission is unknown, a sufficient spectral resolution may allow a selection of lines to fit with LTE models based on the value of Aul . However, they also show that such transitions are widely distributed in wavelength (see Figure 7 in ME09). Therefore, when the resolution is not enough to separate different transitions (as in the case of the Spitzer IRS) it is hard to say in advance how well a blended line can be described assuming LTE excitation. If the emission is originated in a mixture of LTE and non-LTE conditions, which is a reasonable assumption for T Tauri disks (ME09), it is therefore not surprising that LTE models may not be able to reproduce the relative strength of the observed lines. Moreover, as ME09 remark, even finding a good fit (χ2red ≈ 1) to a handful of blended water lines does not necessarily indicate that the LTE model provides a good estimate of the real conditions of the emitting gas. It is clear that some criteria for the fitting method are needed in order to use LTE models and give reliable results that do not misinterpret the data. A sensible choice is to use LTE models to fit only lines at the shortest wavelengths, where the overprediction of the line flux assuming LTE is minimal (see Figure 7 in ME09). This is consistent with the approach utilized by Carr & Najita (2008, 2011). In this approach one should check the value of Eu and Aul of the transitions blended in the observed lines, because transitions with high Eu and Aul can be relevant even at short wavelengths (ME09). Another possibility is to consider the critical density and select lines based on the lowest values of ncrit , which are the first to be thermalized if the ambient gas density is high enough. However, collision rates of water transitions are accurate only within a factor of 3–10 depending on Eu (Faure & Josselin, 2008), which in turn affects ncrit and its use as a selective tool.

2.4.3

Fitting method adopted in this work for H2 O emission

In this work we chose to fit those H2 O lines that are the least likely to be affected by non-LTE effects. Based on the arguments presented in Section 2.4.2, we start fitting over the short wavelengths (12–15 µm) and then we explore changes in the outcomes adding/removing lines in the range 12–19 µm. Each time a best-fit model is found for a sub-set of lines, we check the Eu , Aul , and ncrit of the transitions that contribute for at least 10% to the line fluxes. For ncrit , we take the collision 34

2.5. Results of modelling molecular emission

rates from Faure & Josselin (2008) and we consider collisions with H2 only. We try to avoid lines that include transitions with Eu & 4300 K or Aul & 1 s−1 , as in such cases non-LTE effects are more likely to be strongly relevant (ME09). To check our best-fit models we consider two factors: the value of the χ2red (that we require to be less than 1.5) and the plot of the ratios of observed and model line fluxes against the wavelength. More precisely, we expect the latter to resemble the behavior reported in Figure 7 in ME09 if the fitted lines are in fact the closest to a thermal excitation10 . In other words, an LTE model should be able to reproduce the flux of lines given by transitions with the lowest values of Eu , Aul , and ncrit , while in the case of transitions with higher values of Eu , Aul , and ncrit the line flux should be strictly overpredicted. It is clear that our approach should be considered exploratory. Moreover, the fact that a single-slab LTE model fails to reproduce the observations may imply that the observed emission is indeed excited in non-LTE conditions and/or that it may be produced in several disk regions with different Tex , Nmol , and emitting areas rather than in a single homogeneous layer. Given the limitations of using a single-temperature LTE model, we are able to constrain only the emission that originates in a portion of the disk atmosphere where the local gas density is high enough for thermalization, and which is narrow enough to be approximated with one temperature. However, thanks to the test over the line ratios as explained in the previous paragraph, we are able to obtain indications on whether the emission is mainly to be attributed to non-LTE excitation rather than to different temperatures from different regions in the disk (see Section 2.5.1).

2.5 2.5.1

Results of modelling molecular emission H2 O rotational lines

For the sake of comparison with the results obtained following our “LTE-fitting” method explained above in Section 2.4.3, we first report the results obtained fitting lines regardless of the values of Eu , Aul , and ncrit . In quiescence, the results obtained from fitting water lines over the SH module first, and then over the entire SH+LH coverage, are shown in Figure 2.4. The temperature varies from ∼ 520 to ∼ 330 K in the two cases respectively, while the column density stays close to ∼ 5 × 1018 cm−2 . The probability that the data are well described by the models is lower than 0.001% in both cases, i.e. the LTE model is strongly inappropriate as the scatter in Figure 2.4 suggests. We therefore proceeded, as described above, with fitting sub-sets of lines at short wavelengths, and we found a best fit over the range 12–17 µm with Tex = 580 K and Nmol = 3.2 × 1018 cm−2 (see Figure 2.5). The χ2red is much lower than 1, suggesting that our error-bars may overestimate the real errors by a factor of ∼ 5 or more. The lines used in the best fit are predominantly produced 10

Note that we plot the ratio of observed/LTE fluxes while ME09 plot non-LTE/LTE fluxes.

35


by transitions with Eu . 4300 K and Aul . 1 s−1 , and their critical densities range between 108 and a few 109 cm−3 (see Table 2.4). Therefore the chosen lines are the first to thermalize, compared to other lines in the SH module that instead include transitions with Eu up to 7000 K, Aul up to 102 s−1 , and ncrit up to 1011 cm−3 . The behavior of the line flux ratios plotted in Figure 2.5 supports that we are indeed probing the most thermalized lines: while the LTE model can reproduce very well the fitted lines, other lines are instead strictly overpredicted. The emission is close enough to an optically thick regime (τ ∼ 1–10) that the emitting area is well constrained and the best fit is found for rp = 0.58 AU. The confidence limits derived for the best-fit parameters are shown in Figure 2.8 and listed in Table 2.5. The same procedure is applied to the outburst spectrum, and the results are shown in Figures 2.6 and 2.7. Here we have an additional limitation given by the increased confusion of water lines with other molecules/atoms, mainly OH and H i at short wavelengths (see Figure 2.1). Therefore, from the sample of lines that were used in quiescence we exclude those which are the most contaminated by other species (see Table 2.3). Fitting over the SH module gives a temperature of ∼ 370 K and a column density of ∼ 1020 cm−2 , while over the entire spectrum we find a temperature of ∼ 300 K and a column density of ∼ 5 × 1018 cm−2 . Again, the model is not appropriate to reproduce the observations (probability lower than 0.001%), as the scatter in Figure 2.6 shows. The best-fit model using instead only lines closer to thermalization gives a good fit with Tex ≈ 600 K and Nmol ≈ 3 × 1017 cm−2 , but the emission is more optically thin than in outburst (τ ∼ 0.1–1) and it is sensitive to changes in the emitting area only for low values of rp (i.e. high values of Nmol , when the emission gets optically thick). The emission is found consistent with rp & 0.6 AU (68% confidence), but for rp > 2 AU it is so optically thin (τ < 0.1) that the result is completely insensitive to a change in emitting area. Assuming that OH and H2 O emission probes the same disk radii (as recently shown from very high-resolution near-infrared lines by Doppmann et al., 2011), we choose to fix the emitting area for H2 O in outburst to rp = 1.3 AU (the best-fit value of low-Eu OH in outburst, see Section 2.5.2). Then we derive confidence limits on Tex and Nmol for this fixed radius, which we take as our reference case (see Table 2.5 and Figure 2.8). The opportunity of having the same emitting area for OH and H2 O is that we can easily compare their relative change in terms of Tex and Nmol only. In the case that the assumption was found to be wrong and H2 O had instead the same radius found in quiescence (0.58 AU), the model gives Tex ≈ 700 K and Nmol ≈ 1 × 1018 cm−2 to explain the increase in line fluxes. In our reference case, fixing rp to 1.3 AU, the lines considered in the best fit have mostly transitions with ranges of Eu , Aul , and ncrit consistent with the quiescent case, even though the best-fit model shows a relevant contribution from transitions having higher values of Aul , and ncrit (see Table 2.4). The behavior of line ratios discussed above is once again confirmed11 (Figure 2.7). However, it is clear that the sub-set of 11

A noticeable exception is the line at ∼ 13.43 µm, which, despite being one with the lowest Eu and Aul , is underpredicted by the best-fit model. Contamination by another species is a viable

36


lines we end up fitting in outburst is more affected by non-LTE excitation, and that the Tex derived is therefore probably only a lower limit to the real gas temperature of the observed emission. In fact, Tex always lowers when we include in the fits lines that may be more affected by non-LTE according to the criteria explained in Sections 2.4.2 and 2.4.3. It is very interesting that a similar effect is seen in comparing two recent studies. Despite the similarity in the samples of T Tauri stars considered in the two papers, the average gas temperature reported in Salyk et al. (2011), who include in their fits lines at wavelengths up to 35 µm, is ≈ 25% lower than what Carr & Najita (2011) find, who instead limit their fits to the 12–16 µm range. A viable explanation is that the different results are due to a different contribution from non-LTE excitation given by the different choice of the spectral ranges to fit. This is supported by the fact that, using the outburst spectrum, our result obtained from fitting lines over the entire spectrum is consistent with what Salyk et al. (2011) found for EX Lupi in outburst, while when we restrict the fits to lines closer to LTE the results are consistent with the average found by Carr & Najita (2011). As already said, if transitions at long wavelengths suffer more from non-LTE effects (i.e. are sub-thermally populated), then the Tex found using an LTE model fitted to such transitions will be lower than if they were properly modeled in non-LTE. The fact that a lower temperature is always found when lines at long wavelengths are included strongly suggests that the MIR emission we observe is indeed from non-LTE excitation, and that only some lines can be approximated as being in LTE. However, non-LTE excitation is not alone in affecting MIR molecular emission. In fact, the best fits we obtain from fitting lines closer to LTE are clearly not appropriate for the emission lines detected in LH: line fluxes are strongly underpredicted, contrarily to what is found in the SH module (see Figures 2.5 and 2.7). Lines at wavelengths longer than 20 µm are predominantly produced by lower Eu than at shorter wavelengths, typically Eu ∼ 900-1800 K. Therefore an LTE solution with Tex ∼ 200-300 K could in principle account for such low levels. However, the strict overprediction of line fluxes that we find using lines in the SH module cannot be found using lines in LH even if we assume a low temperature. A reason for that can be seen in that the Aul are instead generally higher (Aul & 10 s−1 ) than at short wavelengths, and higher is in turn the effect of non-LTE excitation. This is again consistent with Figure 7 in ME09: at wavelengths longer than 20 µm the ratios of non-LTE and LTE fluxes span the whole range from 1 to 0, such that no line observed with the Spitzer resolution can be approximated as in LTE. Nonetheless, non-LTE excitation alone is not able to account for the observed fluxes at long wavelengths when we assume the same temperature found at the short wavelengths. Given the lower Eu , what we observe in the LH module is likely produced by water vapor emitted from a colder region in the disk, probably at larger radii.

explanation.

37


Fig. 2.4 — Quiescence: ratios of the observed and LTE model water line fluxes for the model fits to lines in SH (left) and over the entire IRS spectrum (right). The model parameters are reported in the box. 38


Fig. 2.5 — Same as Figure 2.4, but showing the best-fit found using only the lines indicated by filled circles, which are the least likely to be affected by non-LTE effects. All other lines, shown as empty circles, are consistent with being closer to non-LTE conditions and are overpredicted by the LTE model, while in LH the best-fit found in SH is clearly not appropriate. 39


Fig. 2.6 — Outburst: ratios of the observed and LTE model water line fluxes for the model fits to lines in SH (left) and over the entire IRS spectrum (right). The model parameters are reported in the box. Water lines that are heavily contaminated by other species are not considered here. 40


Fig. 2.7 — Same as Figure 2.6, but showing the best fit found using only the lines indicated by filled circles, which are the least likely to be affected by non-LTE effects. All other lines, shown as empty circles, are consistent with being closer to non-LTE conditions and are overpredicted by the LTE model, while in LH the best-fit found in SH is clearly not appropriate. The line at ∼ 13.43 µm is strongly underpredicted by this model and is excluded from the fit. 41


Fig. 2.8 — Confidence limits on Tex and Nmol estimated for H2 O in quiescence (S05, left) and in outburst (A08, middle), and for the low-Eu OH component in outburst (right). In the case of water in outburst we show the contours for a fixed emitting area (with rp =1.3 AU, see Section 2.5.1). Contours show the 68%, 95% and 99% two-dimensional confidence regions, while the best-fit locus is marked with a star. The degeneracy between the two model parameters is apparent, giving higher/lower temperatures for lower/higher densities. 42


Tab. 2.4 — H2 O lines used to constrain the emission observed in EX Lupi Line (µm)

12.51 14.51

15.74 16.89

13.49 14.20 14.51

15.74

Transitions (µm)

12.51 12.52 14.49 14.51 14.54 15.74 16.89 16.90 13.48 13.50 14.18 14.21 14.49 14.51 14.54 15.74

Eu (K)

Aul (s−1 )

Quiescence (S05) 4133 1.50 4133 1.50 3234 1.22 3233 1.21 2605 0.15 2824 1.09 2698 0.70 2125 0.26 Outburst (A08) 5499 8.61 3341 0.49 2876 0.30 3951 3.35 3234 1.22 3233 1.21 2605 0.15 2824 1.09

ncrit (cm−3 ) 4 3 3 2 3 2 2 5

τ

(9) (9) (9) (9) (8) (9) (9) (8)

0.6 1.7 2.1 6.2 1.6 11.4 8.2 7.5

2 (10) 1 (9) 6 (8) 8 (9) 3 (9) 2 (9) 3 (8) 2 (9)

0.1 0.2 0.2 0.5 0.2 0.6 0.2 1.1

Note. — For each observed line, we report the transitions that, according to the best-fit model, contribute for more than 10% to the observed flux. The value of the opacity of each transition is dependent on the best-fit model (in outburst we report here the model with rp = 1.3 AU). All other parameters are instead intrinsic of each single transition. Molecular data are taken from the HITRAN database, apart from the collision rates which are taken from Faure & Josselin (2008). Critical densities must be multiplied by 10a , where the exponent a is shown in parentheses.

43


Tab. 2.5 — Constraints on H2 O and OH emission from single-slab LTE models Line Sample

H2 O

low-Eu OH

Parameter

Quiescence (S05)

Outburst (A08)

Tex (K)

580+40 −40

595+130 −140

Nmol (1017 cm−2 )

32.0+26.5 −13.7

2.8+9.8 −1.7

rp (AU)

0.58+0.07 −0.08

(1.30)

Tex (K)

(590)

590+30 −20

Nmol (1017 cm−2 )

. 0.7

2.8+0.6 −0.9

rp (AU)

(0.58)

1.30+0.08 −0.08

Note. — The confidence limits on the parameters are derived using constant χ2 boundaries. In particular we report the ∆χ2 = 1.0 boundaries, which in one-dimensional planes give the 68.3% confidence on each parameter estimate. Values in parentheses are assumed from water and OH emission that are well characterized (see Section 2.5). For OH we consider here only lines with Eu . 6000 K (see Table 2.2).

44


2.5.2

OH rotational lines

OH rotational transitions from different levels are not as intermingled in wavelength as for the H2 O molecule, and the individual lines observed with Spitzer are composed of doublets from a very narrow range of Eu and Aul (see Table 2.2). This enabled us to perform a rotation diagram analysis, using the description in Larsson et al. (2002). In Figure 2.5.2 we show the rotation diagram based on the OH lines detected in outburst. We assume that their measured fluxes are entirely given by OH emission, and we exclude the ∼16.66 µm doublet which is strongly blended with water (see Figure 2.2). It is apparent from the rotation diagram that the detected lines are not well approximated with a single straight line, that would be expected if the emission was optically thin and in LTE. GL99 showed how deviations can be due to different factors: LTE/non-LTE excitation, optically thin/thick emission, the complexity of the molecular structure, and a multi-temperature emission. The interpretation of the emission we observe in EX Lupi is not straightforward because we may have a combination of effects. First of all, the OH molecule does not belong to the simple linear molecules studied in GL99, given its ground-state double-ladder excitation structure. Second, we very likely observe emission that can only partially be approximated to LTE, as in the case of H2 O. Third, the opacity of emission lines is not known a priori and depends on Tex and Nmol . Unambiguously distinguishing between these different effects is challenging. We propose here a viable interpretation. Given the much smaller transition probabilities of the cross-ladder compared to the intra-ladder transitions (see Table 2.2), the former are the most likely to be optically thin. A linear fit to the cross-ladder fluxes alone finds Trot ≈ 950 ± 460 K, but we note that a reasonable fit could also include lines up to Eu ∼ 6000 K. If we extend the linear fit to include such lines we find Trot ≈ 620 ± 50 K (see Figure 2.5.2). It is instead apparent from the diagram that such a low rotational temperature is certainly not adequate for higher-Eu lines. When fitting separately over the high-Eu range (Eu > 10, 000 K) we find Trot ≈ 9500 ± 5200 K (see Figure 2.5.2). Two components can explain the observed line fluxes even when we account for the optical depth using our LTE model presented in Section 2.4.1. A model fit to lines with Eu . 6000 K finds Tex = 590 K and Nmol = 2.8 × 1017 cm−2 over an emitting area with rp = 1.30 AU. The emission is optically thick (although the crossladder transitions are still optically thin) and well characterized (see Figure 2.8 and Table 2.5), so that we take its area to be the same for water in outburst as assumed in Section 2.5.1. On the other hand, also the high-Eu lines (Eu > 10, 000 K) can be reproduced reasonably well by a single-temperature LTE model with Tex ≈ 10, 000 K and Nmol ≈ 1 × 1014 cm−2 (fixing rp to 1.3 AU), the emission is very optically thin and cannot reproduce line fluxes with Eu . 6000 K. Despite the success of these two separate LTE solutions for different ranges of Eu (see Figure 2.5.2) we do not conclude that the emission is in LTE. A few transitions lying on a straight line can also be mimicked by a quasi-thermal behavior (GL99), and finding different temperatures when different ranges of Eu are fitted suggests non-LTE excitation. 45


Fig. 2.9 — Rotation diagram of OH lines detected in outburst (see Table 2.2), reported as black squares with error-bars. Linear fits suggest two separate emission components (see Section 2.5.2), and are shown with red/blue dotted lines. Two LTE models can reproduce the line fluxes of each component separately (red/blue crosses), but neither one is able to reproduce all lines together (red crosses fall outside the plot in the high-Eu range). Consideration of the two detections in quiescence (empty circles with error-bars) suggests a lower column density and a temperature consistent with outburst for the component given by high-Eu lines.

In conclusion, this analysis shows that the emission in EX Lupi in outburst can be explained by two LTE components, one warm (T ∼ 600 K) and one hot (T & 4000 K). It is nonetheless likely that a two-temperature LTE description is simply an approximation, and that many factors play a role in producing the scatter/curvature that we see in the rotation diagram: optical depth at low Eu , non-thermal excitation at high Eu (in fact for Tgas & 10, 000 K thermal dissociation would become important), and/or a range of gas temperatures instead of only two. We briefly analyze also the quiescent case. The high-Eu component, with only two detected lines, suggests a comparable temperature and a lower column density as compared to outburst (see Figure 2.5.2). As mentioned above, the low-Eu component is instead not detected. If we assume that the flux from the two dubious detections at ∼ 23.22 and 27.67 µm (Figure 2.1 and Table 2.2) is to be attributed to OH alone and we fit these lines, we find a low-Eu OH column density of ≈ 0.7 × 1017 cm−2 (assuming the same Tex found in outburst, 590 K, and the same emitting area found for water in quiescence, with rp = 0.58 AU). These estimates are summarized in Table 2.5 together with the constraints derived above. 46


2.5.3

C2 H2 , HCN and CO2 rovibrational branches

The Q-branches of C2 H2 , HCN and CO2 produce broad blended features in the Spitzer spectra, contribute to each other’s flux, and are also contaminated by H2 O and OH lines. The shape of the Q-branch, as it is seen with the Spitzer spectral resolution, depends on both the temperature and the column density of the emitting gas. Therefore, our fitting method is not sufficient to constrain the emission from organics, as the line flux alone is very highly degenerate in Tex and Nmol . A careful analysis of the unresolved emission from these molecules requires at least that all other chemical species are well constrained and subtracted from the data, and that the Q-branch profile is accounted for in the model fitting. Two attempts in this direction have been recently shown by Carr & Najita (2011) and Salyk et al. (2011). The constraints on Tex and Nmol they are able to derive are loose such that a characterization of the change in organic emission in the case of EX Lupi between quiescence and outburst would be very hard and uncertain. Therefore in this paper we refrain from any kind of sophisticated analysis of the organic emission, and simply provide a rough estimate of the change observed between quiescence and outburst in terms of a variation in column density. In Figure 2.10 we show the quiescent S05 spectrum and LTE models for C2 H2 , HCN, and CO2 assuming parameters in the ranges derived for other T Tauri disks by Carr & Najita (2008, 2011). Namely, for EX Lupi we assume Tex = 750 K and Nmol ∼ 2–7 ×1015 cm−2 , while we take the emitting area from water emission (rp = 0.58 AU). It can be seen from the figure that the emission observed in EX Lupi does not show any remarkable difference from the models. In Figure 2.11 it can be seen that the models assumed for organics in quiescence are instead not consistent with the outburst A08 spectrum. The emission from C2 H2 at 13.7 µm and HCN at 14 µm is clearly no longer as strong as it was in quiescence. The case of CO2 is more hard to judge. The Q-branch of CO2 at 14.97 µm is close to a strong water line at ∼14.9 µm, and overlaps with one OH cross-ladder doublet (see Figure 2.11 and Table 2.2). It is therefore hard to say if a small contribution from CO2 is still present in outburst, but the observations can be explained by H2 O+OH emission alone. This is consistent with the non-detection of CO2 (and C2 H2 and HCN as well) reported in Pontoppidan et al. (2010a) for EX Lupi in outburst, using the C08 spectrum. We then use our LTE model to add the organic spectral features in the outburst spectrum and derive upper limits. We start with the model parameters assumed in quiescence and decrease Nmol until the line fluxes get below the noise limit in outburst. The result is that all organic molecules have to be reduced in column density by at least a factor ∼ 5 to be undetected in outburst. If the emitting area and/or the temperature increase in outburst then this factor is ∼ 10 or more.

47


Fig. 2.10 — Quiescent spectrum (S05) and LTE models derived for H2 O (blue) and OH (cyan), plotted together with the models assumed for HCN (green), C2 H2 (purple) and CO2 (orange). The sum of all model contributions from different molecules is shown in black and shifted for comparison. A continuum is not added to the model. For the three organic molecules we assume the same emitting area found for water emission (rp = 0.58 AU), and model parameters consistent with the range of values derived for other T Tauri systems by Carr & Najita (2008, 2011), namely Tex = 750 K and Nmol ∼ 2–7 ×1015 cm−2 .

2.5.4

Summary of changes in emission from model results

We report the constraints derived using single-slab LTE models in Table 2.5. In Figures 2.10 and 2.11 we show all together the LTE models derived for water and OH and the models assumed for organics. By comparison of the model results in quiescence and outburst we can summarize the changes in molecular emission as follows. Water becomes more optically thin and the column density decreases (NH2 O from ≈ 3 × 1018 to ≈ 3 × 1017 cm−2 ), but the emission comes from a larger area. Given the degeneracy between parameters, another possible solution is that the temperature increases while the column density remains the same over the same emitting area as in quiescence. However, the larger emitting area of OH in outburst supports the former solution, assuming that the two molecules emit from the same portion of the disk (as found e.g. by Doppmann et al., 2011). So, considering water and OH together we would conclude that the emitting area of warm gas increases 48


Fig. 2.11 — Same as Figure 2.10 but showing the outburst (A08) spectrum. The LTE model derived for low-Eu OH emission is added in grey (two lines can be seen at ∼15 µm). Models for organics are plotted using the same parameters assumed in quiescence. It can be seen that in outburst the Q-branches of C2 H2 at ∼13.7 µm, HCN at ∼14 µm, and CO2 at ∼15 µm are not as strong as in quiescence. Remarkable emission lines from other identified molecules/atoms are labeled. Unidentified species may be responsible for other features, like those at ∼ 13.8 µm.

from ∼ 0.6 to ∼ 1.3 AU in radius12 . Remarkably, while the column density of water - and of C2 H2 , HCN and CO2 - decreases in outburst, the OH column density increases. OH emission can be explained by two temperature components, one characterized by the excitation of very high rotational levels (Eu & 10, 000 K) and one by transitions with Eu lower than ∼ 6000 K. The second component shows a temperature that is consistent with the result found for H2 O in outburst (Tex ∼ 600 K) and increases in column density in outburst for at least a factor 3 (NOH from . 0.7 × 1017 to ≈ 3 × 1017 cm−2 ).

Given the inclination of the disk of ∼ 20◦ (Sipos et al., 2009) these projected equivalent radii may be tracing actual physical disk radii. 12

49


2.6

Implications

Our study demonstrates that accretion outbursts in EX Lupi considerably affect the molecular chemistry in a region that is probably the surface of the disk at radii relevant for planet formation. If confirmed, a change in emitting area of OH and water would show that a larger extent of the disk is heated significantly during outburst. But a greater illumination and heating are not enough to explain the change in emission we observe, and we should consider the consequences of the spectral energy distribution of the emission. UV photodissociation of water is indeed the best candidate we find in the literature to explain the sudden formation of a large amount of OH in outburst. Moreover, if water becomes more optically thin in outburst and its column density decreases, this may allow UV radiation to penetrate deeper in the disk. Organics that might have been shielded by water in quiescence would not be so any more, and could be photodissociated. A thinner layer of water would also probably favor more emission from H2 within disk regions that were previously not heated during quiescence. In the following we briefly discuss the scenario we propose to explain the change in molecular emission observed in EX Lupi between quiescence and outburst, and we compare our results with theoretical predictions from the literature.

2.6.1

H2 O and OH emission between quiescence and outburst

During the 2008 outburst in EX Lupi the mass accretion rate onto the star increased from ∼ 10−10 –10−9 to ∼ 10−7 M yr−1 and the stellar+accretion luminosity L∗ +Lacc to increase by a factor ∼ 4.5 (Sipos et al., 2009; Aspin et al., 2010; Juhász et al., 2011). Whatever be the heating source of the gas in the disk, water formation via gas-phase reactions proceeds vigorously when the gas temperature is above 300 K (e.g. Glassgold et al., 2009; Bethell & Bergin, 2009), and can be expected to increase during an outburst. According to the Glassgold et al. (2009) model, the range of H2 O column densities derived for EX Lupi in the two phases (≈ 1017 –1018 cm−2 ) can be explained with heating by X-rays irradiation alone, provided that H2 formation on grains is efficient enough. However, it should be checked under which conditions formation of water would overcome its destruction under an increased X-ray radiation, and compare with the case of EX Lupi. Unfortunately, X-ray observations were not taken close enough in time to the Spitzer spectra we consider in this work, and comparison of the available data from G¨ udel et al. (2010) (August 2007) and Grosso et al. (2010) (August 2008) does not provide conclusive constraints. Glassgold et al. (2009) also proposed that accretion (mechanical) heating would increase the H2 O column density. This instead can be ruled out in EX Lupi: even if we had the same emitting area in quiescence and outburst the H2 O column density does not increase when the accretion rate is higher, more so if the emitting area is larger in outburst as we assume in our reference case (see Section 2.5.1). 50

2.6. Implications

On the other hand, UV radiation has been proposed to play a key role in the formation/destruction and availability of H2 O, OH, and organic molecules in the planet-forming zones of disks (Bethell & Bergin, 2009). A strong UV radiation can photodissociate water vapor in favor of OH in the upper layers of the warm disk atmosphere, competing with the capability of H2 O and OH to self-shield. The change in emission in EX Lupi is consistent with this scenario: we see NH2 O decreasing in outburst, while NOH increase up to the limit given by the self-shielding mechanism (∼ 2 × 1017 cm−2 , Bethell & Bergin, 2009). Following in this direction, the increase of NOH in outburst suggests that the connection between the strength of OH emission and the strength of the accretion rate found by Carr & Najita (2011) may be attributed to the effect of UV radiation rather than to mechanical heating. Again, the available UV observations do not provide strong constraints but we have at least the evidence that UV radiation was higher at the time of the A08 spectrum than it was in quiescence, as in the U-band (0.36 µm) the observed flux varied by a factor ∼ 50 (Juhász et al., 2011). The strongest evidence in favor of a key role of UV radiation probably comes from OH emission. The coexistence of two temperature components in the MIR spectrum of EX Lupi is an interesting possibility. In previous works using Spitzer spectra of classical T Tauri systems a cold component (T ∼ 500 K) was found by Carr & Najita (2008) fitting OH lines detected in the LH module, while Carr & Najita (2011) proposed a hot component (T ∼ 4000 K) considering lines in the SH module. Based on previous observational evidence, Najita et al. (2010) proposed that UV-irradiated disks may show a twofold OH emission: a prompt non-thermal component from UV photodissociation of H2 O which would eventually relax to a thermal emission at the temperature of the gas. The broad population of OH levels in the range Eu ≈ 400-28,000 K, first observed by Tappe et al. (2008) in the HH211 outflow, was indeed explained with a radiative cascade down the ground-state rotational ladders following UV photodissociation of H2 O, which would preferentially populate OH levels with Eu & 40, 000 K (Tappe et al., 2008; Harich et al., 2000). If our approximation of two temperature components in outburst is confirmed, it would represent the first observational evidence of an ongoing prompt “hot” (in reality nonthermal) emission followed by a thermal emission at the temperature of the gas. In fact, in outburst we find one component that may hardly be explained with thermal excitation, while another component shows the same temperature found for water (see Section 2.5.2). An interesting peculiarity of EX Lupi is the detection of the cross-ladder transitions between ∼ 15 and 19 µm, that were never observed before in any classical T Tauri system. The case of EX Lupi, as well as similar variable sources, might be very useful to clarify the nature of OH emission in the inner regions of circumstellar disks, but more theoretical work is needed. For instance, it would be interesting to know if the proposed excitation mechanism by radiative cascade would produce the curve that we see in the rotation diagram, and, if that is a transitory process, on which timescales/conditions it should be observed as two components rather than a continuum of temperatures.

51


The simultaneous monitoring of UV, X-rays and MIR emission during future outbursts would help in confirming the connection between the variation in accretion rate and changes in H2 O and OH emission. Velocity- and transition-resolved spectral lines might be the key tool to clarify the location of the MIR molecular emission as well as help in understanding the processes responsible for formation/destruction of the molecules we observe (e.g. Pontoppidan et al., 2010b, see also Chapter 3 in this book). More comprehensive chemical disk models accounting for both Xrays and UV variable irradiation would also be very helpful in better constraining water formation and survival in the inner zones of disks based on spectrally resolved synoptic observations of young variable star+disk systems.

2.6.2

The lack of C2 H2 , HCN and CO2 in outburst

The fact that organic emission decreases in outburst is very interesting, and probably the most puzzling of our findings. Production of HCN has been proposed to be driven by strong UV photodissociating N2 in disk atmospheres (Pascucci et al., 2009). In EX Lupi we observe instead a lower HCN column density in outburst, and yet the ratio NHCN /NC2 H2 ∼ 2.7 in quiescence is consistent with the median sun-like star reported in Pascucci et al. (2009). The lack of C2 H2 , HCN and CO2 in outburst seems to preferentially imply the large photodissociation of these molecules by UV, in contrast with the capability of self-shielding of H2 O and OH. If the organic emission probes the same disk radii as H2 O and OH, then their non-detection would show that warm organics are located higher in the disk atmosphere than an optically thick layer that could shield them, whether it is made of dust or H2 O and OH as suggested in Pontoppidan et al. (2010a) and Bethell & Bergin (2009) respectively. The unclear behavior of the emission from these organic molecules in other T Tauri disks (Carr & Najita, 2011) suggests that we still do not fully understand where exactly their emission originates and what are the factors most responsible for their abundance. It would be very interesting to see if organics are detected in EX Lupi again after the outburst. If that was found to be true, than the vertical mixing proposed by Juhász et al. (2011) to be powered-up in outburst may be the mechanism replenishing the disk atmosphere with molecules from closer to the midplane. Monitoring EX Lupi during the current new quiescent phase and future outbursts may be extremely important to check the timescales/processes relevant for organic chemistry in planetforming regions of disks.

2.6.3

Note on H I and H2 in outburst

We do not focus in this paper on the strong lines from H i and H2 that appear in EX Lupi in outburst (see Section 2.3.1). The increase in the accretion rate is certainly a possible explanation for an increase in H i emission (e.g. Pascucci et al., 2007), but photoevaporative winds might also contribute (see e.g. the discussion in Najita et al., 2010). The increase in H2 emission can be explained by emission from a 52

2.7. Concluding remarks

warm disk atmosphere (Najita et al., 2010; Gorti et al., 2011), which in EX Lupi in outburst is probably warmed up in regions that were too cold in quiescence (larger radii and/or deeper layers). It would be interesting to see whether a direct relation between these species and the formation/destruction of H2 O in EX Lupi exists. If warm H2 is indeed available on a larger portion of the disk atmosphere, then water vapor formation might be favored (Glassgold et al., 2009) and increase again the observed column density after the end of the outburst, when UV photodissociation diminishes.

2.7

Concluding remarks

The molecular emission we detect in the quiescent S05 spectrum (from H2 O, OH, HCN, C2 H2 and CO2 ) is comparable to what has been found to be common in several other T Tauri systems (Pascucci et al., 2009; Carr & Najita, 2008, 2011). This is a further confirmation that EX Lupi is not distinguishable from typical T Tauri systems, as already suggested by Sipos et al. (2009) from consideration of the spectral energy distribution in quiescence. Therefore, the remarkable changes in molecular emission observed in the outbursting EX Lupi might be extremely important to constrain chemical-physical processes common to all T Tauri systems. In this paper we have addressed the simple scenario of the increase of one system parameter in EX Lupi: the disk-atmosphere illumination during outburst. To first order this is consistent with an outbursting event that accreted material only from inside an inner hole in the dusty disk (∼ 0.3 AU) and not over global disk scales (Juhász et al., 2011; Goto et al., 2011; Kóspál et al., 2011). Consideration of more sophisticated scenarios needs to be included as soon as high-resolution data will become available. The geometrical structure of the EX Lupi system is probably affected by strong outbursts and may be relevant for the correct interpretation of the changes in emission we observe, including the increase of line fluxes in the LH module. As the stellar+accretion luminosity L∗ + Lacc increased in outburst by a factor ∼ 4.5 (Aspin et al., 2010), the dust-sublimation radius is probably shifted outward by a factor ∼ 2. This can have effects on the inner rim location and the shadowed portion of the disk as well, and/or the location of the snow lines of the different chemical species. The location of the disk we likely probe, the inner few AUs, is indeed where both the shadowing and the snow line are believed to be. In addition to that, UV photodesorption from icy grains is expected to power the production of OH as well as to largely replenish water vapor in the disk atmosphere ¨ (Oberg et al., 2009). Despite the increased UV and the production of OH, we still see a large column density of water vapor in EX Lupi in outburst. If self-shielding of water breaks under particularly harsh conditions (see the case of Herbig AeBe stars, e.g. Pontoppidan et al. (2010a) and Fedele et al. (2011), and transitional disks, e.g. Najita et al. (2010)), then a receding snow line may still provide the ice reservoir needed to sustain a wet disk atmosphere in EX Lupi. Velocity-resolved observations are critical to confirm the location of the different molecular emissions and explore 53


these scenarios. New infrared facilities from the ground (e.g. the soon upgraded VISIR on the Very Large Telescope, and in the future METIS on the E-ELT ) and from space (with MIRI on the James Webb Space Telescope and MIRES on SPICA) will be the key tool in future investigations of the warm molecular emission from the inner regions of disks. The author acknowledges the several colleagues who contributed to this investigation with valuable discussions, in particular Ted Bergin, Klaus Pontoppidan, Inga Kamp, John Carr, and Joan Najita. A.B. thanks all people of the new Star and Planet Formation group at the ETH Zurich for their support during the development of this work, especially Susanne Wampfler and Michiel Cottaar. This work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology.

2.8

Appendix: a single-slab model of gas in LTE

A single-slab model is used to extract from the spectra the excitation conditions of the emitting gas. We assume the gas to be in local thermodynamical equilibrium (LTE), but account for optical depth effects. We follow a formalism similar to van der Tak et al. (2007), and we ignore the absorption of background radiation by the emitting gas (i.e. the absorption term of the radiative transfer equation). Neglecting mutual overlap between transitions and assuming the line profile function to be Gaussian, the integrated intensity (e.g. erg s−1 cm−2 sr−1 ) of a transition connecting the molecular level u (upper) and l (lower) reads Z ∞ ∆vνul 1 2 Iul = p 1 − exp −τul e−y dy . (2.1) Bν (Tex ) 2 ln(2) c −∞ Here, τul is the opacity at the line center, Bν (Tex ) the Planck function for the excitation temperature Tex , νul the line frequency and ∆v the line width, which we fix to 2 km s−1 . Thep integral in the above expression obtained using a change √ has been −τul in variable y = 2 ln(2)v/∆v and approaches π (1 − e ) for τul < 1. Thus, the usual expression used in escape probability codes (e.g. van der Tak et al., 2007) is recovered for low opacities. The opacity at the line center is given by p ln(2) Aul Nmol c3 gu √ τul = xl − x u , (2.2) 3 gl 4π π ∆vνul with the Einstein-A coefficient of the transition Aul and the total molecular P column density Nmol . The statistical weight and the normalized level population ( i xi = 1) of the upper and lower level are given by gu , gl and xu , xl , respectively. The level population is obtained from the Boltzmann distribution xi = gi exp (−Ei /kTex ) /Q(Tex ), with the partition sum Q(Tex ) and the energy of the level Ei . The relevant molecular parameters are taken from the HITRAN 2008 database (Rothman et al., 2009). To model the Spitzer spectra, i.) the intensity of all transitions of a chosen molecule 54

2.8. Appendix: a single-slab model of gas in LTE

(e.g. H2 O) in the wavelength range of the Spitzer IRS is calculated using equations 2.1 and 2.2, for an assumed Tex and Nmol , ii.) the integrated flux (Ful , in units of erg s−1 cm−2 ) is calculated by multiplying the intensity Iul with the solid angle ∆Ω = A/d2 , where A is the emitting area in the sky, for which we assume a prop jected circle with radius rp = A/π, and d the distance to the source (assumed to be consistent with the distance of the Lupus complex, 150 pc, Lombardi et al., 2008); iii.) a synthetic spectrum (in erg s−1 cm−2 µm−1 ) with spectral resolution R = ∆λ/λ ≈ 600 is generated based on all transitions in the wavelength range considered, with flux given by

F (λ) =

X i

− 1 √ e 2πσλ

λ−λiul √ 2σλ

2

Fuli ,

(2.3)

p where σλ = λ/2R 2 ln(2), and λiul and Fuli are respectively the wavelength and the flux of a single transition.

55


56

“WE CHOOSE TO GO TO THE MOON IN THIS DECADE AND DO THE OTHER THINGS, NOT BECAUSE THEY ARE EASY, BUT BECAUSE THEY ARE HARD.” JOHN F. KENNEDY, MOON SPEECH - RICE STADIUM, SEPTEMBER 12, 1962

3

Exploring the link between water and accretion variability

Based on a paper published in The Astrophysical Journal, 780, 26 (2014), using observations made with ESO telescopes at the Paranal Observatory under programme ID 088.C-0666 (Principal Investigator: A. Banzatti). A. Banzatti1 , M. R. Meyer1 , C. Manara2 , K. M. Pontoppidan3 , L. Testi2

1

ETH Z¨ urich, Institut f¨ ur Astronomie, Wolfgang-Pauli-Strasse 27, CH-8093 Z¨ urich, Switzerland European Southern Observatory, Karl Schwarzschild Str. 2, D-85748 Garching bei M¨ unchen, Germany 3 Space Telescope Science Institute, Baltimore, MD 21218, USA 2

57

CHAPTER 3. Exploring the link between water and accretion variability

Abstract Young stars are known to show variability due to non-steady mass accretion rates from their circumstellar disks. Accretion flares can produce strong energetic irradiation and heating that may affect the disk in the planet formation region, close to the central star. During an extreme accretion outburst in the young star EX Lupi, the prototype of EXor variables, remarkable changes in molecular gas emission from ∼ 1 AU in the disk have recently been observed (Banzatti et al., 2012, see also Chapter 2 in this book). Here, we focus on water vapor and explore how it is affected by variable accretion luminosity in T Tauri stars. We monitored a young highly variable solar-mass star, DR Tau, using simultaneously two high/mediumresolution ESO-VLT spectrographs: VISIR at 12.4 µm to observe water lines from the disk, and X-shooter covering from 0.3 to 2.5 µm to constrain the stellar accretion. Three epochs spanning timescales from several days to several weeks were obtained. Accretion luminosity was estimated to change to within a factor ∼ 2, and no change in water emission was detected at a significant level. In comparison to EX Lupi and EXor outbursts, DR Tau suggests that the less long-lived and weaker variability phenomena typical of T Tauri stars may leave water at planet-forming radii in the disk mostly unaffected. We propose that these systems may provide evidence for two processes that act over different timescales: UV photochemistry in the disk atmosphere (faster) and heating of the disk deeper layers (slower).

3.1

Introduction

The detection of warm gas emission from the inner few AU of young circumstellar disks (Salyk et al., 2008; Carr & Najita, 2008) has opened a new window on the chemical and physical characterization of the environments where rocky planet properties are probably set. Surveys of mid-infrared (MIR) molecular emission from ∼100 circumstellar disks over a range of spectral types and evolutionary stages have already moved important steps forward in this context (Pontoppidan et al., 2010a; Carr & Najita, 2011; Salyk et al., 2011). The observed emission from water, OH, and organic molecules (HCN, C2 H2 , and CO2 ) comes from warm gas (300-1000 K) in the planet-forming zone of the disk, and is generally found to be common in T Tauri disks, while reduced or absent in more massive (Herbig Ae/Be) or more evolved (transitional) disks. Moreover, molecular abundances are found to be not simply inherited from the prestellar phase, but show evidence for an ongoing active chemistry. The infrared molecular gas emission is a privileged tracer of the processing of warm material in evolving protoplanetary disks. Studies of the interplay between accretion histories and the molecular chemistry are essential to our understanding of disk evolution and planet formation, and could perhaps provide hints to help explain the diversity in composition of (exo)planets.

58

3.1. Introduction

An often overlooked element in disk evolution is the effect of non-steady accretion processes during star formation. In this regard, an exceptional experiment was provided by a recent extreme accretion outburst in EX Lupi (the prototype of EXor variables, Herbig, 1989, 2008). This is a classical T Tauri star showing eruptive accretion phenomena that in strength are in between FUor outbursts and the typical variability found in the T Tauri phase of young stars. From comparison of MIR spectra taken during outburst and in a preceding quiescent phase, Banzatti et al. (2012) found that while the molecular spectrum in quiescence showed water, OH, and organic emission typical of T Tauri systems, it changed remarkably in outburst. Water emission increased, suggesting a larger extent of the warm emitting gas caused by the increased disk heating. On the other hand, detection of previously unseen OH lines suggested ongoing OH production via UV photodissociation of water. Strikingly, emission from all organics disappeared. EX Lupi provided the first evidence that the warm molecular gas emission can change dramatically, in the same disk, depending on the accretion phase. If the EXor and classical T Tauri variability are phases of the same accretion history that all forming stars undergo (Hartmann, 2009), effects similar to those observed in EX Lupi might be common in T Tauri stars. To better understand both the properties and the evolution of the molecular gas in the planet formation region of all circumstellar disks we may need to consider the time domain, especially during the non-steadily accreting phases of the star-disk interaction. This work was initiated to probe this idea. In this paper we study the emission observed toward the active T Tauri star DR Tau. Situated in the Taurus-Auriga star-forming region at 140 pc from Earth (Kenyon et al., 1994), DR Tau is one of the most studied classical T Tauri stars. Yet, it is one of the most peculiar as well. It showed a slow brightness increase over ∼ 20 years between 1960 and 1980 (Chavarria-K., 1979; Gotz, 1980), which raised it from being a relatively faint star (. 14 mag in V band) to one of the brightest stars in the Taurus association (∼ 11 mag in V band). Its behavior was attributed to strong non-steady accretion (Bertout et al., 1988), and DR Tau was included in the first list of EXor variables by Herbig (1989) together with EX Lupi. Since 1980, DR Tau has maintained its enhanced brightness while showing large photometric and spectroscopic variability (e.g., Alencar et al., 2001; Grankin et al., 2007). The high and variable veiling produced by accretion has made its spectral classification challenging and dubious. However, recent studies performed in lower veiling phases agree on a K5-K7 spectral type (Mora et al., 2001; Petrov et al., 2011). DR Tau is still a young system (∼ 1–2 Myr) with a relatively massive disk of ∼ 0.01M (Ricci et al., 2010a). It is one of the first disks where warm water vapor emission was detected (Salyk et al., 2008). Together with water, OH and organic molecules have been detected at both mid- and near-infrared (NIR) wavelengths (Pontoppidan et al., 2010a; Mandell et al., 2012; Brown et al., 2013). Given its rich molecular emission and its strong and variable accretion, it is an optimal target for investigating how the former is affected by the latter. In this work, we monitored DR Tau through its MIR water emission lines from the

59


Tab. 3.1 — Three epochs of VISIR observations of DR Tau Epoch

Airmass

Seeing (arcsec)

Doppler (km/s)

Frames

tint (s)

1: 27 Nov 2011 2: 02 Dec 2011 3: 14 Jan 2012

1.38 1.51 1.41

0.4–0.5 0.4–0.7 0.4–0.5

21.15 23.87 44.64

32 40 41

2240 2800 2870

Note. — Details on the values reported in this table are given in Section 3.2.1. Airmasses are averaged per epoch. The seeing is estimated at 12.4 µm using the measured width of the source PSF. Doppler shifts are measured with respect to rest frequencies of the targeted emission lines. The number of frames and the respective on-source integration time of the reduced spectra are given in the last two columns.

inner disk using the ESO Very Large Telescope (VLT) Imager and Spectrometer for the mid-InfraRed (VISIR, Lagage et al., 2004). Simultaneously, the stellar accretion emission was monitored using VLT/X-shooter (Vernet et al., 2011).

3.2

Observations

This monitoring program (088.C-0666, PI: Banzatti) was carried out in service mode using two instruments at the VLT: VISIR (at VLT/UT3) and X-shooter (at VLT/UT2). As part of this program, DR Tau was observed on six nights between November 2011 and January 2012. In three out of these six nights, simultaneous observations with the two VLT instruments were performed, and these are the three epochs studied in this paper4 (see Table 3.1). The first and second epochs were taken 5 days apart; the third was taken 43 days after the second. Such a scheduling spanning different time ranges was proposed in an attempt of observing the star in different phases of its unpredictably variable accretion. However, the exact choice of dates depended on atmospheric conditions, which we required to be photometric5 , and the challenging scheduling of the two VLT instruments, which we required to be used simultaneously. In each epoch, DR Tau was observed for ∼ 2 hours with VISIR, in good atmospheric and airmass conditions (see Table 3.1). Given that DR Tau is a bright star, the X-shooter observations required a much shorter time (∼ 15 min) and were performed during the first quarter of each VISIR epoch. 4

The X-shooter epochs, six in total, will be analyzed together in a future paper. We required clear sky, seeing better than 0.800 , and precipitable water vapor column of less than 2 mm. 5

60

3.2. Observations

Fig. 3.1 — Telluric correction of VISIR spectra. For illustration, we show here the second epoch of observations obtained in this program. DR Tau is shown in black/grey (before/after telluric correction), HD50138 in orange/red (before/after cross-correlation with DR Tau and airmass correction). The two relevant telluric lines used for cross-correlation as explained in the text are at 12.403 and 12.407 µm. The 12.396-µm water line has hardly any telluric counterpart, due to its high upper level energy (∼ 5800 K).

Observing water emission lines from Earth is difficult because every targeted line has its telluric counterpart. The observations were done in November-January to maximize the Doppler shift between the target and Earth, such that the targeted lines were shifted by ∼ 20–45 km/s from their telluric counterparts (Table 3.1). We chose to observe high-energy lines (at 12.396 and 12.407 µm, upper level energy Eu 5000-5800 K), to ensure weak telluric counterparts and minimize absorption effects even in the case of partial overlap of line wings (see Figure 3.1). The line sample was restricted by a limited availability of filters along with a need for long integrations to reach sufficient signal-to-noise ratios (S/N), to be able to probe variations in emission from one epoch to the other. Given these constraints, one VISIR setting per epoch was observed, centered at 12.407 µm. 61


Fig. 3.2 — Three epochs of VISIR data, with the two targeted water lines (rest frequencies are marked with dotted lines). The spectrum in each epoch was reduced and telluric-corrected as described in Section 3.2. Corrections for the heliocentricbaryocentric velocity of the VLT at Paranal and the radial velocity of the star (23 km/s, Appenzeller et al., 1988; Petrov et al., 2011) has been applied to each epoch. The spectra are continuum-normalized to illustrate the relative change in water-tocontinuum emission from epoch to epoch.

3.2.1

VISIR spectra

VISIR was used in the cross-dispersed mode, providing a resolving power R ∼ 20, 000 (∆v ∼15 km/s) at 12.4 µm. The data were taken using standard chopping and nodding techniques for VISIR6 . The telescope nods parallel to the slit between positions on-source (A) and off-source (B) with a sequence ABBA, and a secondary mirror chops parallel to the slit with a sequence ABAB starting in the nod position A, and BABA starting in the nod position B. Each VISIR raw datacube provided by ESO includes the four half-chop-cycle images performed in one nod position (e.g., Ai , Bi , Ai+1 , Bi+1 , where i is the chop cycle number). The slit used in the high resolution mode (width = 0.7500 , length = 4.100 ) does not allow chopping on-slit, as the minimum chopping throw is 800 . Therefore, the B positions do not have the source on the slit and are used only for background subtraction. This effectively reduces the on-source integration time of VISIR in the cross-dispersed mode to half the total time, bringing it to 2940 s in each of our epochs. 6

A detailed description of observing techniques is available in the VISIR User Manual online on www.eso.org.

62

3.2. Observations

The VISIR data were reduced using a suite of self-made IDL routines intended to optimize the standard VISIR data reduction of the ESO Recipe Execution Tool (EsoRex), version 3.9.0. First, small shifts (typically fractions of a pixel) caused by mechanical oscillations of the grating between sequential images needed to be corrected. To do that, in each nod position the individual A and B images were combined, applying the shift estimated by cross-correlation of two telluric lines observed in the region between the targeted water lines. In our data, this procedure worked well in reducing artifacts from sky residuals that would otherwise affect the targeted water emission, if processed only through the standard ESO pipeline (a similar correction was implemented by Carmona et al., 2008). Standard EsoRex procedures were then used to perform image combination, distortion correction, and wavelength calibration as follows. Sequential nod pairs were combined into nod halfcycle images by averaging A and −B. In total, each epoch has 42 such AB frames, whose on-source integration time is 70 s each. We carefully monitored the observing conditions frame by frame, by means of measuring the peak and the width of the gaussian point-spread-function (PSF) fitted to the source trace on the detector, in the spatial direction. An estimate of the atmospheric seeing at 12.4 µm (N -band) was based on the measured width of the PSF, and is reported in Table 3.1. Variable atmospheric conditions can overwhelm the target signal in a given frame, when the sky infrared background varies too much from A to B. The actual number of frames that could be used in each epoch is therefore smaller than the total and differed from epoch to epoch (see Table 3.1). Nod images were then corrected through pixel interpolation for small distortions caused by the instrument, using the EsoRex parameters slit skew = 1◦ and spectrum skew = 0.4◦ , which were found to be appropriate in the 12.4 µm setting. Wavelength calibration of individual spectra was done by cross-correlation with an atmospheric model set on typical Paranal conditions, using the same telluric absorption lines mentioned above. An additional background subtraction was needed to correct for sky background residuals produced by variable atmospheric water vapor within chop cycles. This was performed by a median subtraction row by row in the cross-dispersed direction, as in Pontoppidan et al. (2010b). An individual spectrum was then extracted from each combined (AB) nod frame using optimal extraction to maximize S/N (Horne, 1986). The individual spectra in each epoch were then combined into a weighted average yielding one spectrum per epoch. Flux errors on individual pixels in the averaged spectrum were set to the standard deviation of the weighted mean of pixel values as measured in the total number of frames in each epoch. Telluric absorption had a negligible influence on the two targeted water lines during this monitoring of DR Tau, for the reasons explained at the beginning of this section. Nonetheless, a telluric standard was observed and utilized to correct also for the inhomogeneous detector response (as flat fields are usually not taken for VISIR). We used a Herbig Ae/Be star (HD50138), which was observed after DR Tau each night, selected on the basis of its brightness and absence of emission lines at 12.4 µm. The spectra were normalized to their continua, aligned by cross-correlation,

63


Fig. 3.3 — Relation between airmass and conversion factor Jy/(ADU s−1 ) as derived from photometric standard stars (observed before and after DR Tau in each epoch). Conversion factors as measured from individual standards are shown with crosses (plus error-bars), while the black line is a linear fit to them. The values extrapolated for DR Tau, using the average airmass in each epoch, are shown with diamonds. Epochs are color-coded as in Figure 3.2. and then divided (DRTau/HD50138), after airmass (A) correction of HD50138 with respect to DR Tau. To do that, an exponential correction was applied to the telluric (A /A ) absorption lines in the HD50138 spectrum, as fcorr = fobsDRT au HD50138 . The match between HD50138 and DR Tau after this correction was found to be excellent in all epochs. Figure 3.1 illustrates the telluric correction procedure, while Figure 3.2 shows the three epochs of the final VISIR spectra, reduced and corrected as described in this section. To perform absolute flux calibration of the VISIR epochs, we used photometric standard stars that were observed before and after DR Tau each night. HD22663, HD6805, HD7055, and HD37160 were observed as part of the standard ESO calibrators, and in addition we observed Sirius A. From knowledge of their flux in Jy, a conversion factor Jy/(ADU s−1 ) can be derived using the measured detector counts for each standard, which were observed at different airmasses. Figure 3.3 shows a linear fit to the conversion factors estimated over the three epochs using two standards each night. The model fits the individual epochs reasonably well, supporting the idea of overall stable atmospheric conditions. The continuum in DR Tau was then estimated using the fitted relation between airmass and conversion factor, extrapolated at the average airmass of DR Tau in each epoch (Figure 3.3). The estimated continuum decreases by 25% in the second epoch, and rises by 23% in the third. In the second epoch, the seeing was relatively worse (but still lower 64

3.2. Observations

Fig. 3.4 — Three epochs of X-shooter spectra, with the three arms covering from UVB to NIR. Each spectrum was taken simultaneously to the VISIR spectrum of a given epoch (Figure 3.2). The data were reduced and calibrated as explained in Section 3.2.2. than the slit width, ensuring limited slit losses; see Table 3.1), and the two photometric standards were taken at ∼ 0.4–0.5 lower airmasses with respect to DR Tau. To check whether the decrease in continuum level could be simply due to the general worse seeing conditions, we used the technique described above to estimate the continuum also for the telluric standard HD50138. We found the same trend, a decrease in the second epoch followed by an increase in the third, but confined to within 5%, comparable to the precision obtained on the photometric standards. We therefore concluded that variable atmospheric conditions may have affected by at most 5% the continuum in DR Tau, but that it intrinsically changed by ∼ 20%. This is consistent with the change estimated in the H band from the X-shooter data (see Section 3.4).

3.2.2

X-shooter spectra

The X-shooter echelle spectrograph observes in three spectral arms, providing a simultaneous combined coverage of ∼300–2500 nm: a UVB arm covering ∼300–590 nm, a VIS arm covering ∼530–1020 nm, and a NIR arm covering ∼1000–2480 nm. In each epoch, we asked for two settings: the wide-slit mode (500 ) with 20 s exposure time, and the narrow-slit mode (0.400 in the VIS and NIR arms, 0.500 in the UVB arm) with 120 s exposure time. Four nodding repetitions with the ABBA scheme were done to obtain an optimal background subtraction. The narrow-slit setting 65


provided the higher-resolution spectra (R = 9100, 17400, 10500 in the UVB, VIS, and NIR arms, respectively) used to monitor accretion emission in this work. The wide-slit setting avoids flux losses and was taken to perform the spectro-photometric calibration of the higher-resolution spectrum in each epoch. The data were reduced using EsoRex and the ESO pipeline for X-shooter, version 1.3.7 (Modigliani et al., 2010). The pipeline applies standard reduction procedures to perform background subtraction, spectral extraction, wavelength calibration, and flux calibration in each spectral arm separately. A more accurate absolute flux calibration of the higherresolution spectra was then obtained by scaling each spectral arm to the flux level measured using the wide-slit mode. The overlapping spectral regions between the arms were used to check the results, which were found to agree well in all arms. Finally, the spectroscopic standard stars provided by ESO were used to perform telluric correction through the IRAF task telluric. Figure 3.4 shows the three epochs of the higher-resolution X-shooter spectra, reduced and calibrated.

3.3 3.3.1

Analysis Water line properties

The two targeted water lines are both well detected in all epochs (see Table 3.2), and present a single-peak unresolved profile (Figure 3.2). This is consistent with gas emission from the inner regions of a nearly face-on disk (inclination ∼ 10◦ , where 0◦ is face on), as proposed for DR Tau by Pontoppidan et al. (2011) and Brown et al. (2013) from spectro-astrometry of CO and NIR water emission lines respectively. Assuming Keplerian rotation and a stellar mass of 0.8 M (Ricci et al., 2010a), a lower limit to the emitting radius of the 12.4 µm water vapor in DR Tau is found at & 0.1 AU from our VISIR data. We measured the Doppler shift between the observed line peaks and their rest frequencies (taken from the HITRAN database, Rothman et al., 2009) after correction for the known heliocentric-baryocentric velocity in each epoch provided by ESO for the VLT at Paranal. The heliocentric velocity of both water lines was found to be ∼ 25 km/s in all epochs of our VISIR data. This is consistent with the radial velocity of the star as estimated in 1987 by Appenzeller et al. (1988) and in 2007–2010 by Petrov et al. (2011), who measured an average of 23 ± 2 km/s using stellar photospheric lines. We fit the observed water emission using gaussian functions on top of a linear continuum, estimated over the spectral range covered by the VISIR setting and excluding the region of the targeted emission lines and of telluric absorption lines (12.393–12.409 µm). For line fitting, we use the least-squares fitting routines by Markwardt (2009). We estimate each line flux and FWHM from the peak and the 66

3.3. Analysis

Tab. 3.2 — Properties of water lines observed with VISIR Wavelength (µm)

Transition J Ka Kc

Aul (s−1 )

Eu (K)

Epoch 1

Line flux Epoch 2

Epoch 3

12.39625 12.40708

17 4 13 → 16 3 14 16 3 13 → 15 2 14

7.7 4.2

5781 4945

1.80 ± 0.19 0.91 ± 0.17

1.60 ± 0.18 0.92 ± 0.12

1.46 ± 0.16 0.81 ± 0.13

Note. — Line fluxes are given in 10−14 erg s−1 cm−2 . Line properties are taken from the HITRAN database.

width of the best-fit gaussian function, and the flux error is propagated from the uncertainty on the best-fit parameters. Fitted line widths are consistent with the nominal VISIR resolution of 15 km/s within the errors, and we therefore fix this value in all epochs. Line fluxes are measured in the normalized spectrum of each epoch, and then scaled to the continuum level at 12.4 µm as estimated in Section 3.2.1. The line-to-continuum ratio, or the line flux in the normalized spectrum, increases by ∼ 30% in the second epoch and decreases by roughly the same amount in the third. This contrast “flickering” is observed in both water lines at roughly the same level (Figure 3.2). We therefore estimate the average line flux change between epoch i and epoch i + 1 as linea,i+1 lineb,i+1 + /2 , (3.1) linea,i lineb,i where the two water lines are labeled as a and b. Calibrated line fluxes and their propagated errors are reported in Table 3.2, while their relative changes between epochs are shown in Figure 3.6.

3.3.2

Accretion luminosity

We derive the accretion luminosity (Lacc ) from estimates of the UV excess emission in the Balmer jump region in X-shooter spectra, following previous work (Valenti et al., 1993; Herczeg & Hillenbrand, 2008; Rigliaco et al., 2012). The model and method we adopt are presented in detail in Manara et al. (2013b); here we give a basic description. We fit the data using a model with three components: a photospheric template, a reddening law, and a model for the accretion spectrum. As the photospheric template we use a Class III young stellar object (2MASS J053905400232303) with spectral type K7, similar to DR Tau, from the suite of photospheric templates collected in Manara et al. (2013a). We adopt the reddening law from Cardelli et al. (1989) with RV = 3.1 (appropriate for Taurus, Herczeg & Hillenbrand, 2008) and explore values of AV within 0 and 3 mag, in steps of 0.1 mag. We 67


model the continuum emission of the accretion spectrum as an isothermal slab of atomic hydrogen gas, which is an approximation adequate for young accreting stars (e.g., Valenti et al., 1993; Herczeg & Hillenbrand, 2008). It includes bound-free and free-free emission from both H and H− , assumes local thermodynamic equilibrium (LTE) conditions, and is described by three parameters: the electron temperature (Tslab ), the electron density (ne ), and the optical depth at 300 nm (τ300 ), which is related to the length of the slab. Tslab is explored in the range 5000–11000 K, ne in the range 1011 –1016 cm−3 , and τ300 in the range 0.01–5, which are typical for this kind of description. Additional parameters of the model are two normalization constants: Kphot accounts for the differences in distance and radius between DR Tau (for the stellar luminosity we adopt L? = 0.85 L from Muzerolle et al., 2003) and the photospheric template star, and Kslab accounts for the area of the accretion slab at the stellar surface, as if Lacc was produced by a hot spot. In total, the free model parameters are five: AV , Tslab , ne , τ300 , and Kslab . We fit the data in several spectral ranges in the Balmer and Paschen region (330– 480 nm), including the Balmer jump, the slopes of the Balmer and Paschen continua, and the continuum at ∼710 nm. The best fit is found by minimization of a χ2 -like function defined as the sum of the squared deviations (data – model) divided by the error. The goodness of the best-fit accretion model is checked also visually, by looking at the degree of veiling produced in some photospheric absorption features (namely CaI at 420 and 616 nm, and the TiO lines at 844 nm). More details on this model and the fit methodology can be found in Manara et al. (2013b). The extinction AV is found to be 1.8–1.9 mag in this monitoring of DR Tau and is consistent with being the same in all three epochs. This can be understood in terms of the almost face-on disk geometry (see Section 3.3.1), where changes in accretion columns onto the star would be unlikely to produce large changes in extinction (see also Alencar et al., 2001). We therefore decided to fix AV to a value of 1.8 mag common to all epochs. This simplification is convenient because the largest uncertainty on the derived Lacc is due to the uncertainty in AV , but in a relative sense (comparing epoch to epoch), the uncertainty on the change in Lacc is smaller. Thus we reduce the free parameters to Tslab , ne , τ , Kslab and evaluate the change in the best fit from epoch to epoch as if it were primarily given by accretion. After finding a best-fit slab model in each epoch, Lacc is derived as Lacc = 4πd2 Facc , where d is the distance to the source and Facc is the flux integrated over the 50–2500 nm range in the slab model rescaled with Kslab . Figure 3.5 shows the best-fit models obtained in the three epochs of X-shooter data. The uncertainty on Lacc , as derived from the uncertainty on the slab model parameters, is of the order of 10–20%.

68

3.3. Analysis

Fig. 3.5 — Best fit for the accretion model in each epoch, derived as explained in Section 3.3.2. In each plot, the X-shooter spectrum of DR Tau is shown in black, the photospheric template in dotted line, and the fitted model (slab + photosphere) in red. The photospheric contribution is three orders of magnitude smaller than the accretion slab contribution in all epochs.

69


3.4

Discussion

Figure 3.6 summarizes our results on the relative changes in water emission and stellar accretion, estimated in this monitoring study of DR Tau as described in Section 3.3. The accretion luminosity decreased by ∼ 55% in the second epoch, and increased by ∼ 95% in the third. The continuum in the X-shooter spectra follows this trend (see Figure 3.4), supporting the idea that its variation was related to the change in accretion. For illustration, we show in Figure 3.6 the changes in the Hband continuum level, measured as the average pixel ratio between epoch i + 1 and epoch i in the range 1.52–1.72 µm. Using the absolute continuum estimated from the VISIR data (Section 3.2.1), the 12.4 -µm flux shows the same trend, decreasing in the second epoch and increasing in the third within ∼20–25% (consistent with the variation in the H-band continuum). The change of ∼ 30% in the continuumnormalized water line fluxes (Figure 3.2) can therefore be attributed to the change, similar in fraction but opposite in sign, of the N -band continuum level. In other words, the calibrated water line fluxes are consistent in the three epochs within the errors. In conclusion, while in this monitoring program of DR Tau the accretion luminosity changed within a factor ∼ 2, no change in water emission was detected at a significant level (a 10% decrease in line fluxes is seen by comparison of epoch 3 to 2 but only at a 1σ level, see Figure 3.6). Variable water emission from the disk was previously observed from comparison of a quiescent phase to an accretion outburst in EX Lupi (Banzatti et al., 2012, see also Chapter 2 in this book), using spectra obtained with Spitzer -IRS (Werner et al., 2004; Houck et al., 2004). In outburst, Aspin et al. (2010) estimated an increase in Lacc of a factor ∼ 40, while the flux continuum at 12.4 µm increased by a factor ∼ 5 and water line fluxes in the range 10–35 µm increased by ∼ 50–350% (Banzatti et al., 2012). Specifically, the two water lines at 12.396 and 12.407 µm increased in flux by ∼ 250% 7 (see Figure 3.7). In a thought experiment, if what we observed in DR Tau is a scaled-down version of the effect observed in EX Lupi, for an increase in Lacc of a factor ∼ 2 we should have seen an increase in the 12.4 µm continuum of ∼ 20% and an increase in water line fluxes of ∼ 13% (dividing all values by 20). The results presented in this paper for DR Tau are consistent with this scaled-down prediction from EX Lupi (see Figure 3.6). The change in water emission is consistent only within 2σ and might show an opposite trend, but the small changes are comparable to the precision reached on the line fluxes and are therefore hard to distinguish from the noise. The behavior observed in EX Lupi and DR Tau may be understood in terms of two processes and their different timescales: UV photochemistry and disk heating. Bethell & Bergin (2009) showed that in a typical T Tauri disk model the disk 7

Line fluxes and errors are estimated by fitting the 12.4 µm water emission, which is blended in Spitzer spectra, with a gaussian for each individual line. The basic method is explained in Banzatti et al. (2012), while here we additionally fix the line width to FWHM=λ/750 (Najita et al., 2010, Banzatti et al. submitted, see Chapter 4 in this book).

70

3.4. Discussion

Fig. 3.6 — Relative variations of different tracers (12.4-µm water line fluxes, accretion luminosity, continuum level in the H band and at 12.4 µm), estimated in this monitoring study of DR Tau (see Section 3.3). We compare epoch 2 to 1 and epoch 3 to 2 in a time sequence, taking the ratio of their tracers. Grey stars indicate the changes measured in EX Lupi (Aspin et al., 2010; Banzatti et al., 2012) divided by 20 (see Section 3.4). surface is dominated by photodissociation and fast chemical timescales (. 102 s), while the deeper cold midplane where most of the mass resides in icy solids reaches chemical equilibrium on timescales of > 104 s. Studying how protostellar envelopes react to accretion outbursts, Johnstone et al. (2013) recently showed that thermal equilibrium between gas and dust through collisional heating requires a few weeks in regions with density comparable to the inner disk regions where the 12.4 µm water emission likely comes from (∼ 109 cm−3 , e.g., Banzatti et al., 2012). In EX Lupi the accretion outburst lasted for ∼7 months, long enough to change the thermal disk structure by heating up the disk and shift outward the T & 170 K region (the ´ snowline) from . 0.6 to . 1.2 AU (Abrah´ am et al., 2009), consistent with the change in emitting area of the warm gas estimated by Banzatti et al. (2012). Stronger water lines for higher accretion were observed probably due to new water vapor, evaporated from icy solids beyond the snowline (e.g., Ciesla & Cuzzi, 2006) and/or formed in situ in the gas phase from molecular hydrogen and oxygen (Glassgold et al., 2009; Bethell & Bergin, 2009). The weaker and shorter accretion variability observed in DR Tau (and typical of T Tauri stars) might, instead, have the time to affect only 71


Fig. 3.7 — Comparison of DR Tau (this work) and EX Lupi (Banzatti et al., 2012) in terms of changes in accretion luminosity and water vapor emission at 12.4 µm (black stars). The grey-shaded area shows, for reference, the range in water line flux changes measured by Banzatti et al. (2012) over the entire Spitzer coverage (10–35 µm). While the variation in Lacc observed in DR Tau is typical during the T Tauri phase, and no change in water emission was significantly detected, EX Lupi was studied during a recent extreme accretion outburst, and water emission increased. The breaking point between the two regimes may depend both on the duration and the strength of the accretion variation.

the disk surface layers and its main effect would probably be restricted to a variable UV photochemistry. In this case, weaker water lines might be observed in case of photodissociation. This new monitoring study of DR Tau suggests that during the T Tauri phase, a change in Lacc within a factor ∼ 2 over a few days only is probably not enough to affect significantly the water vapor in the inner disk, at least as probed in the N band. Perhaps the strength and duration necessary to drive stronger water vapor emission (and possibly production) as observed in EX Lupi are provided only during the extreme EXor outbursts (Figure 3.7). 72

3.5. Summary and conclusions

3.5

Summary and conclusions

We investigated the effects of stellar accretion variability on the water vapor at planet-forming radii in circumstellar disks during the T Tauri phase, by means of a new high-resolution UV–to–MIR monitoring of DR Tau. Three epochs of simultaneous VLT/VISIR and VLT/X-shooter spectra were obtained between November 2011 and January 2012. Accretion luminosity was derived from the X-shooter data and found to change within a factor ∼ 2 (decreasing in the second epoch, increasing in the third). No change in water line fluxes at 12.4 µm was measured at a significant level from the VISIR spectra. The enhancement in water emission observed during an extreme accretion outburst in EX Lupi suggests that accretion can drive water vapor production, possibly in connection with a recession of the snowline (Banzatti et al., 2012). While the change in accretion between the epochs in DR Tau was not large enough to distinguish changes in water emission from noise, our monitoring suggests that if the change in Lacc is confined to within a factor ∼ 2, then water vapor emission as probed at 12.4 µm may remain mostly unaffected (within ∼ 10%). It would be interesting to explore until when this is still the case: what conditions are mild and stable enough to leave the inner disk undisturbed, and where the breaking point is (Figure 3.7). This exploration would require constraining two parameters: the strength and the duration of the accretion outburst. While an increase in water line fluxes of & 20-30% is needed to clearly stand out from the noise with the current instrumentation (and EX Lupi suggests that this might happen at an increase in Lacc of a factor & 4), an accretion event duration of at least a few weeks might be necessary to allow for thermal equilibrium at the gas densities probed by the observed water vapor (Johnstone et al., 2013). Much is still to be understood in the production/destruction of water vapor as driven by accretion variability. Future observations promise to shed more light on the role of accretion histories in shaping the molecular environments of the planet-formation region in protoplanetary disks. A.B. is grateful to several colleagues for valuable conversations about the results from this work, many of which happened during the PPVI conference in Heidelberg: Ted Bergin, Simon Bruderer, James Muzerolle, and Ruud Visser. This work is based on observations made with ESO telescopes at the Paranal Observatory under programme ID 088.C-0666.

73


74

“AN ADVENTURE IS ONLY AN INCONVENIENCE RIGHTLY CONSIDERED. AN INCONVENIENCE IS ONLY AN ADVENTURE WRONGLY CONSIDERED.” GILBERT K. CHESTERTON, ALL THINGS CONSIDERED, 1908

4

Searching for signatures of icy migrators in protoplanetary disks

Based on a paper submitted to The Astrophysical Journal A. Banzatti1 , K. M. Pontoppidan2 , M. R. Meyer1 , S. Bruderer1

3

1

ETH Z¨ urich, Institut f¨ ur Astronomie, Wolfgang-Pauli-Strasse 27, CH-8093 Z¨ urich, Switzerland Space Telescope Science Institute, Baltimore, MD 21218, USA 3 Max-Planck-Institut f¨ ur Extraterrestrische Physik, Giessenbachstr. 1, D-85748 Garching bei M¨ unchen, Germany 2

75

CHAPTER 4. Searching for signatures of icy migrators in protoplanetary disks

Abstract Steps toward the formation of planets rely on processes that are largely elusive to current observational techniques because they are deeply embedded in the highly optically thick midplanes of protoplanetary disks. One fundamental process is the migration of icy solids from the outer to the inner disk. Here we describe a method to search for observational confirmation of icy migrators from the abundance of water vapor in inner disks. We perform a rotation diagram analysis of infrared water vapor emission, covering a wide range in line opacities and upper level energies. From consideration of three different models, we find a characteristic rotational scatter, whose shape and spread are illustrative of the emitting conditions (temperature, column density, and type of excitation). In particular, we show that its spread is directly related to the range in line opacities and can provide evidence for inner disk water abundances higher than solar oxygen values, likely implying enrichment from inward migration of icy bodies followed by evaporation. We apply this technique to observed data by de-blending unresolved water lines from Spitzer mid-infrared spectra of young water-rich disks. Rotational scatters seem to exceed static solutions and to support water columns at the high end of degenerate fits reported in previous work, favoring the migration scenario if the gas-to-dust ratio in disks is 104 . The Spitzer data provide still tentative evidence, but future confirmation by detection of the weak optically thin lines will finally lift the veil on migration processes occurring inside protoplanetary disks.

4.1

Introduction

Planets are thought to form in circumstellar disks within the first few Myr of evolution, when disks are still rich in dust and gas (Hernández et al., 2007; Ribas et al., 2014) so to provide the material needed to build planetary cores and atmospheres. Most of the processes responsible for disk evolution and planet formation therefore happen embedded in gas- and dust-rich disks, and are hard to observe directly (Armitage, 2011). In particular, solid material coagulating and growing in size in the disk midplane is expected to experience gas drag that causes inward migration (Weidenschilling, 1977), providing regions of over-density in large solids in the inner disk and favoring planetesimal formation (Johansen et al., 2007). A new window for studies of the properties and evolution of planet-forming material at a few AUs in disks was recently opened by disk observations of infrared molecular emission (Carr & Najita, 2008; Salyk et al., 2008). Water vapor, in particular, is expected to be a tracer of both disk evolution and planet formation processes in the inner disk. During star formation, water is delivered to protoplanetary disks from the parent molecular cloud primarily as ice, providing a main constituent of solid mass ¨ (Pontoppidan et al., 2004; Visser, 2009; Oberg et al., 2011b; Bergin & van Dishoeck, 2012). Once in disks, the thermal and density structure of circumstellar matter defines an inner region where water is in the gas phase, and an outer region where it is 76

4.1. Introduction

frozen into ice. The condensation/evaporation front between the two regions, which has both a vertical and a radial structure, is known as the snow line (Pontoppidan et al., 2014). Contrary to the seminal prediction of Stevenson & Lunine (1988), where outward turbulent diffusion and freeze-out of water vapor beyond the snow line leads to dry inner disks, recent infrared observations suggest that some process replenishes these regions with water (Carr & Najita, 2008; Salyk et al., 2008). Two scenarios have been proposed to provide an explanation: evaporation of icy solids migrating inwards from cold outer regions (Cyr et al., 1998; Cuzzi & Zahnle, 2004; Ciesla & Cuzzi, 2006), and, more recently, in situ formation via gas-phase reactions (Glassgold et al., 2009; Bethell & Bergin, 2009; Najita et al., 2011). The former is the natural consequence of viscous drag in disks with a negative radial pressure gradient (Weidenschilling, 1977), while the latter is efficient at temperatures above ∼ 300 K and relies on the available supply of oxygen and hydrogen in the inner disk. While these mechanisms are not mutually exclusive, constraining their relative importance is fundamental to our understanding of how protoplanetary disks evolve and form planets. In fact, the timescale of planet formation is generally thought to be related to the local surface density of solids. The snow line acts as a “cold finger” onto which the inner disk volatiles can condense (Stevenson & Lunine, 1988; Ros & Johansen, 2013), while icy solids pile up as a consequence of inward migration (e.g. Ciesla & Cuzzi, 2006). Both processes increase the local density of solids at the outer edge of the snow line, leading to the efficient formation of planetesimals (Johansen et al., 2007) as well as 5-20 M⊕ giant planet cores (Dodson-Robinson et al., 2009). There is growing evidence that the infrared water vapor emission in inner disks traces directly the snow line (Meijerink et al., 2009; Zhang et al., 2013, Blevins et al., in preparation), but whether it directly traces also dynamical processes linked to planet formation, like migration of solids and enhancements in their local density, is still unknown. This work has been initiated to address this open question: are signatures of icy migrators found in the inner regions of protoplanetary disks? Ultimately, we are interested in finding evidence to confirm whether disks are predominantly static, where planets form locally in a well-mixed chemical environment, or rather dominated by large-scale transport of solids, where the chemistry of the inner regions is connected to the physical processes in the outer regions.

4.1.1

An answer from the water abundance

Static disk models that include gas-phase chemistry assume solar abundance for oxygen (4.2 × 10−4 relative to atomic hydrogen, Grevesse et al., 2010) and predict that most of the oxygen is turned into water producing similar abundances (a few 10−4 relative to hydrogen) of warm water vapor in the inner disk (e.g. Woitke et al., 2009). Water vapor abundances higher than solar have instead been predicted by models that include transport of icy bodies from the outer disk, which evaporate after crossing the snow line (Cyr et al., 1998; Cuzzi & Zahnle, 2004; Ciesla & Cuzzi, 77


Fig. 4.1 — Degeneracy from slab models fits to water emission in Spitzer spectra of young protoplanetary disks (part I). Crosses and diamonds display the best-fit solutions from the disk sample in Salyk et al. (2011) and Carr & Najita (2011) respectively. The superimposed contours illustrate the typical elongated shape of χ2 confidence regions in the T vs N space (Salyk et al., 2011; Banzatti et al., 2012). Observed water columns of NH2 O . 1018 cm−2 have been explained by static disk models with in situ water vapor formation via gas-phase reactions (Glassgold et al., 2009; Bethell & Bergin, 2009; Najita et al., 2011), while higher columns may suggest water enrichment by evaporation of icy migrators from beyond the snow line (e.g. Ciesla & Cuzzi, 2006). 2006). Specifically, Ciesla & Cuzzi (2006) proposed that enhancement of factors ∼ 10–30 as compared to solar can be produced by ongoing migration of icy solids in circumstellar disks during their first Myrs of evolution. Therefore, knowledge of the abundance of water vapor in the inner disk could in principle discriminate between the two origin scenarios and show whether (and in which disks) water vapor is primarily a tracer of chemistry rather than of migration and evaporation. However, constraining the inner disk water abundance has been an observational challenge so far. While previous work could profit from the large coverage in wavelengths (10–37 µm) provided by the Spitzer Space Telescope (Werner et al., 2004), where a forest of strong water emission lines is observed in T Tauri disks, the main limitation was due to its low resolving power (R ∼ 720, see Section 4.3.1). At this resolution, the observed forest of water emission lines is not spectrally resolved, producing confusion between lines from different quantum levels. Slab models have been used so far to estimate physical parameters of the observed water vapor emission: the emitting area, the excitation temperature T , and the column density N (rather than the abundance). These estimates were found to be degenerate, and 78

4.1. Introduction

likely biased, due to the blending of the emission. The range in best-fit model parameters relied on assumptions regarding the emitting areas and on the selection of emission features used in the goodness-of-fit estimates (Salyk et al., 2008, 2011; Carr & Najita, 2008, 2011; Banzatti et al., 2012). Temperatures were found between ∼ 200 and 1000 K with column densities typically 1017 cm−2 , where models with high T and low N are interchangeable with models of low T and high N (see Figures 4.1 and 4.2). Static disk models typically predict observed column densities of warm water vapor of NH2 O . 1018 cm−2 (Glassgold et al., 2009; Bethell & Bergin, 2009; Najita et al., 2011) that can partially explain the results from fits to the emission observed in the inner few AUs of circumstellar disks. High column densities of & 1019 cm−2 that may instead support the migration scenario were found in some disks (Salyk et al., 2011), but it is still unclear whether they were simply tracing the large degeneracy between model parameters (Figures 4.1 and 4.2). For this reason, the migration scenario was never directly addressed in previous observational studies of water spectra in disks. In this work we provide a complementary view to previous studies, tackling the retrieval of water abundances (rather than column densities) so to determine how evidence for an origin from icy migrators can be found. In Section 4.2 we demonstrate how to break the degeneracy between slab model parameters and distinguish high column densities using a rotation diagram analysis. Then we explore how water abundances higher than those produced in the static scenario can be recognized in rotation diagrams, using a two-dimensional disk model. In Section 4.3 we apply this technique to a dozen high signal-to-noise Spitzer spectra of water-emitting protoplanetary disks, to determine whether evidence supporting the high abundances from the migration scenario is found, and we discuss the limitations of the data. In Section 4.4 we discuss the presence of high water abundances in inner disks, the extant challenges to confirm them, and we propose promising directions for future work.

79


Fig. 4.2 — Degeneracy from slab models fits to water emission in Spitzer spectra of young protoplanetary disks (part II). Comparison of degenerate model solutions (marked with red/blue stars in Figure 4.1) on top of representative portions of the Spitzer spectrum of DR Tau. Features that are most noticeably missed by both water models are due to blends with other species like OH (12.65 and 30.3 µm) and HI (11.31 and 12.37 µm).

80

4.2. Rotation diagrams of water vapor emission

4.2

Rotation diagrams of water vapor emission

As a convenient way to illustrate how the abundance of water vapor can be observationally constrained, we use the rotation diagram technique. We start with considering the column density (number of molecules per unit of surface), and will move to the water abundance or concentration (number of molecules relative to hydrogen) in Section 4.2.3. Rotation diagrams are defined such that, in conditions of optically thin emission and local thermal equilibrium (LTE), the observed molecular line fluxes form a straight line, following: NΩ Eu 4πF = ln − , (4.1) ln hνgu Aul Q(T ) T where F is the integrated line flux, Ω is the angular emitting area4 , gu is the statistical weight of the upper level, Q(T ) is the partition function, Aul is the Einstein-A coefficient, and Eu is the energy of the upper level. Rotation diagrams therefore provide a simple visualization of gas properties, as the excitation temperature T and the column density N can be easily derived from the slope and intercept of a linear fit using Equation 4.1. This technique is very convenient also to recognize when the emission departs from the simplest conditions and becomes optically thick and/or the excitation is in non-LTE. In fact, in these cases the observed molecular line fluxes would produce curvatures and/or scatter departing from the straight line given by the LTE optically thin case. The rotation diagram technique has been thoroughly described and discussed by Goldsmith & Langer (1999), but has never been systematically applied to water emission because of lack of measured line fluxes over a significant range in Eu and Aul . While this is not necessary in the case of linear molecules in the simplest conditions (Goldsmith & Langer, 1999), covering a large range in Eu and Aul is essential in the case of the more complex water molecule, especially in circumstellar disks where the emission is not optically thin and non-LTE conditions may sub-thermally excite lines with high Aul (Meijerink et al., 2009; Banzatti et al., 2012). In Sections 4.2.1 and 4.2.2 we use slab models to describe how different conditions (line opacity, LTE/non-LTE excitation) produce distinct features that can be recognized in water rotation diagrams, with particular attention on how to distinguish a high column density. We will then see how that translates into the signatures of a high water abundance using a two-dimensional disk structure in Section 4.2.3.

4.2.1

LTE slab models

The basic properties of rotation diagrams of infrared water emission are displayed in Figure 4.3 using a slab model in LTE (described in Banzatti et al., 2012), where the 4

Equation 4.1 assumes that the angular emitting region Ω equals the telescope beam ΩB so as to cancel out in the left-hand term, which is valid for unresolved point sources observed with the Spitzer -IRS.

81


Fig. 4.3 — Rotation diagrams of infrared water emission lines covered by the Spitzer IRS high-resolution modules (10–37 µm, dynamic range of ∼ 100). Slab models are used to illustrate the basic signatures given by line opacity and by LTE/non-LTE excitation. Left: LTE models. For an optically thin emission all transitions lie on a straight line (upper panel). The more the column density is increased, the more the emission becomes optically thick and produces a scatter of data in the diagram (lower panels). Right: non-LTE models with n(H2 ) = 106 cm−3 . The rotational scatter is increased by non-LTE excitation, but with a distinct shape as compared to the LTE case. Lines at high Eu are not significantly populated (upper two panels) unless high columns are reached (lower panel).

free parameters are the temperature and the column density. In this work we keep the emitting area of the slab fixed, as it produces simply a shift common to all line fluxes along the y-axis in the rotation diagram and is not of interest in the context of this analysis. For convenience, here we keep the notation of LTE even though slab models assume a single temperature, and hence thermal equilibrium is in a global (not local) sense. In the optically thin case (N = 1015 cm−2 in Figure 4.3), water lines form a straight line in the diagram, with slope set by the temperature 82


T = 600 K , N = 1018 cm−2

T = 300 K , N = 1024 cm−2

15

15 8

8

10

10 6

0

4

2

−5

5 Log(τ)

ln(4pF/(hvAulgu))

Log(τ)

ln(4pF/(hvAulgu))

6

5

0

2

−5

0

0

−10 −15 0

4

−10 −2

2000

4000

6000 Eu (K)

8000

10000

−15 0

−2

2000

4000

6000

8000

10000

Eu (K)

Fig. 4.4 — Same as Figure 4.3 (LTE model), but illustrating two additional line properties: the intensity (proportional to dot sizes) and the opacity (shown in bluecolor scale). The spread in the rotational scatter is given by the spread in line opacities. The model to the left assumes a moderate column according to static disk models that produce water by gas-phase chemistry (Glassgold et al., 2009; Bethell & Bergin, 2009; Najita et al., 2011). The model to the right assumes an extreme high column density, to illustrate better the effect of the spread in line opacities. following Equation 4.1. As soon as line opacities increase, by increasing the column density, a scatter is introduced in the diagram following a specific pattern. When a line gets optically thick (τ 1) it “freezes” on the diagram, i.e. its intensity is only weakly dependent on the column density. This happens first to those lines with large Aul (the low-energy corner of the diagram), as the line opacity is proportional to Aul . The optically thin lines, instead, still follow Equation 4.1 and can rise in the diagram together with N . Therefore, unlike the case of linear molecules where a curve would be produced (see e.g. Bruderer et al., 2012), an increase in optical depth for the non-linear water molecule opens a “fan” of data in the diagram. The most optically thin lines (those with the lowest Aul at any given bin in Eu ) define the upper edge of the fan, while the most optically thick ones set the lower edge (Figure 4.4). Therefore, the rotational scatter in LTE is primarily sensitive to the column density. This property can effectively be used to break the degeneracy between model parameters: while the excitation temperature is determined by the slope of the upper edge of the fan, the column density is linked to the spread of the fan, or the maximum distance between the most optically thin and the most optically thick lines. Observationally, this can be difficult to assess without knowing a priori the line opacity, especially from low-resolution spectra: the optically thin lines defining the upper edge of the fan are the weakest and the most difficult to detect, while the strongest and easily detectable lines are optically thick and underestimate the real spread of the scatter (see Figure 4.4 and Section 4.3.1). This introduces uncertainties and biases in gas properties estimates, and has contributed to the degeneracy found in previous work (Salyk et al., 2011; Carr & Najita, 2011; Banzatti et al., 2012, see also Figure 4.1). 83


4.2.2

Non-LTE slab models

What happens to the rotational scatter under non-LTE conditions? To explore this, we simulate the emission from a slab of gas using the RADEX code (van der Tak et al., 2007), which accounts for the density of the main collisional partner that thermalizes the emitting gas. RADEX uses molecular data from the LAMDA database (Schöier et al., 2005) complemented with data from Tennyson et al. (2001) and Faure & Josselin (2008). We explore a wide range in volume densities of the collisional partner H2 where water emission is not thermalized by collisions, down to n(H2 ) = 106 cm−3 that is the case shown in Figure 4.3. Mid-infrared water lines generally require at least n(H2 ) = 108 cm−3 to thermalize, condition that is found in the regions producing the mid-infrared water emission observed in disks (see e.g. Banzatti et al. (2012); Zhang et al. (2013)), but here we explore more extreme conditions to better highlight the effects of non-LTE excitation. Overall, the shape produced in non-LTE is different from the LTE case (see Figure 4.3): it displays curvatures and mimics steeper slopes (i.e. lower temperatures in a “quasi-thermal” behavior, Goldsmith & Langer, 1999). Moreover, lines with Eu higher than a few thousands are significantly populated only when high column densities are reached (N > 1018 cm−2 ), i.e. in a regime already dominated by line opacity. Therefore, different shapes of the rotational scatter provide hints to distinguish whether line opacity rather than non-LTE dominates, provided that the observations cover a large enough range in Eu and Aul . A recent example has been showed by Herczeg et al. (2012), where coverage of a larger portion of the rotation diagram has led to recognition of non-LTE effects and to a substantial revision of the emission observed toward NGC 1333 IRAS 4B. In conclusion, what demonstrated in the previous section for LTE conditions holds even in the non-LTE regime: a large rotational scatter, from low to high values of Eu , is indication of a large column density.

4.2.3

The inner disk water abundance

So far we have considered slab models, which provide column density estimates that may be representative only of the uppermost disk layers. We now use a realistic disk structure to account for the total abundance of water vapor in the inner disk. For this purpose we adopt the two-dimensional RADLite code and LTE excitation (Pontoppidan et al., 2009). RADLite is a raytracer for infrared molecular emission from circumstellar disks, based on the dust temperature and density structure calculated self-consistently using the RADMC code (Dullemond & Dominik, 2004). RADLite accounts for dust and gas opacities, and here we set the gas temperature equal to the dust temperature, which is thought to be appropriate for the deeper layers giving rise to the mid-infrared water emission. Despite the extant uncertainties in modeling the water emission (Kamp et al., 2013), this approximation is somewhat confirmed by a match of the rotational temperature of the RADLite model to the data (Meijerink et al., 2009). In modeling the water emission, RADLite takes into account the effects of a snow line set by the temperature and density structure of the 84


Fig. 4.5 — RADLite water emission models based on a two-dimensional structure for the disk, showing lines in the Spitzer -IRS high-resolution modules as in Figure 4.3. The water vapor abundance relative to hydrogen X(H2 O) inward of the snow line and the gas-to-dust ratio GTD are varied as shown. For comparison, dotted lines mark the region covered by the static model in Figure 4.4. disk. In static conditions, the water abundance is set to solar oxygen values inward of the snow line (typically where T > 170 K), and a low, constant value of 10−9 per hydrogen outside of it to simulate diffusion and freeze-out onto grains. RADLite has already been used to argue for a “cold-finger” effect depleting water vapor beyond a surface snow line in circumstellar disks (Meijerink et al., 2009). It has also been used to locate the water snow line and measure the depletion of water vapor in TW Hya (Zhang et al., 2013). Here we use it to demonstrate how a water abundance higher than solar values of oxygen can be recognized in rotation diagrams, by artificially increasing the water abundance inward of the snow line to simulate enrichment from evaporation of icy migrators. For illustration, as our reference model we assume a 0.01 M flared disk around a 1 M star and a high gas-to-dust ratio (GTD) of 104 to mimic dust settling towards the midplane, as adopted in Meijerink et al. (2009). RADLite accounts for the water emission coming from a range of disk radii where water is present with different temperatures and densities. This basically adds a curvature to water emission in the rotation diagram as compared to LTE slab models, due to the contribution of different temperature and density components at different radii. Apart from that, models with a water abundance of 10−4 relative 85


to hydrogen (static scenario) cover a similar portion of the rotation diagram as the slab models presented above with NH2 O ∼ 1018 cm−2 . Models with higher water abundances exceed the region of the static scenario, confirming that the spread in the rotational scatter, as given by the spread in line opacities, is a good indicator of the inner disk water abundance (see Figure 4.5). The only other parameter that affects the rotational scatter in a similar way to the water abundance is the gas-todust ratio. Assuming a canonical GTD of 100, typical of the interstellar medium, the scatter is reduced such that a higher abundance is needed to cover the same portion of the diagram. In other words, a high abundance model with GTD = 100 is basically equivalent to a model with solar abundance and GTD = 104 (see Figure 4.5). This degeneracy can be interpreted in a physical sense in terms of how deep in the disk it is possible to observe: if grain growth and settling depletes the atmosphere from small dust grains that normally dominates the opacity (i.e. a larger GTD), water emission comes from a region that extends deeper in the disk and a lower abundance is sufficient to produce a similar scatter in the rotation diagram (see also the discussion in Section 4.4). In any case, the spread of the rotational scatter produced using a two-dimensional disk model is linked to the spread in observed line opacities, similarly to slab models, and indicative of the water abundance. The shape and spread of the rotational scatter of water vapor emission can therefore be used to evaluate the contributions from different effects and in particular the presence of high water abundances in inner disks, as traced by the most optically thin water lines defining the upper edge of the scatter (Figure 4.4).

4.3

Application to Spitzer spectra

The Spitzer Space Telescope has provided, to date, the largest survey of mid-infrared water emission in young circumstellar disks, observing ≈ 100 disks from several starforming regions. The Spitzer -IRS high-resolution modules (Houck et al., 2004) cover the region where water vapor emission is strong in disks (10–37 µm), including thousands of lines that trace a wide range in upper level energies (Eu from 700 to 9000 K) and several orders of magnitude in Einstein-A coefficients Aul (10−6 –102 s−1 ). This is the largest sample of water lines possibly provided by a single instrument today: from the range typically traced by far-infrared observations, e.g. from Herschel (Eu ∼100-1000 K), to the range of near-infrared observations (Eu ∼6000-10000 K). It is therefore the place to start for a first practical demonstration of the technique described above, where a broad range of lines is needed to reveal the shape of rotational scatters. As a first step, in this paper we explore the morphology of rotational scatters of a dozen of water-rich T Tauri disks observed with Spitzer with high signal-to-noise (Table 4.1), whose spectra were published in Pontoppidan et al. (2010a) and Salyk et al. (2011). Fits to individual disks will be performed in combination to additional data from higher-resolution instruments (see e.g. the analysis of RNO90 in Pontoppidan & Blevins, submitted). The Spitzer data were reduced 86

4.3. Application to Spitzer spectra

Tab. 4.1 — Water-emitting disks sample considered in this work Name

ST

M? (M )

Md (M )

Refs.

AA Tau AS 205 CW Tau DR Tau FZ Tau RNO 90 RW Aur TW Cha VZ Cha VW Cha WX Cha

K7 K5 K3 K7 M0 G5 K1 K7 K6 K5 M0

0.8 1.0 1.2 0.8 0.6 1.8 1.3 1.0 0.8 0.6 0.5

0.015-0.15 0.03 0.002-0.004 0.007-0.02 0.002-0.09 0.003-0.007 0.004 ... ... ... ...

1,2 3,7 1,2 1,2 1,2 3,4 1,8 6 5 5 5

S/N

a

155 214 254 291 295 240 230 140 185 307 170

Note. — a Average of the signal-to-noise on the continuum estimated at ∼12.5 and ∼31.5 µm in the Spitzer -IRS spectra utilized in this paper. References. — (1) Andrews & Williams (2005), (2) Ricci et al. (2010a), (3) Andrews & Williams (2007), (4) Ricci et al. (2010b), (5) Hartmann et al. (1998), (6) Feigelson et al. (1993), (7) Andrews et al. (2009), (8) White & Ghez (2001)

using the publicly available Caltech High-Resolution IRS Pipeline (CHIP)5 , optimized to take full advantage of high-quality spectra that use dedicated background observations, nodding on the slit, and cycle redundancy (Pontoppidan et al., 2010a; Carr & Najita, 2011). The reduction procedure is described in Pontoppidan et al. (2010a) and was used for the larger sample of ∼ 75 disks analyzed in Pontoppidan et al. (2010a) and Salyk et al. (2011).

4.3.1

Line flux extraction

The main limitation of Spitzer spectroscopy of water vapor is that even in the IRS high-resolution modules the observed emission lines are not spectrally distinguished due to their vicinity. To perform a rotation diagram analysis, individual line fluxes therefore need to be extracted from observed blended line complexes (a previous example can be found in Herczeg et al. (2012)). In this work we use the following procedure, building on methods used in previous work (the basics were already described in Banzatti et al., 2012, and references therein). We fit the data assuming a 5

www.stsci.edu/∼pontoppi/Pontoppidan web home/Software

87


Fig. 4.6 — FWHM of the instrumental line profile of the Spitzer IRS in highresolution mode, as measured from gaussian fits to single unresolved hydrogen lines in the spectrum of a Herbig Ae/Be star with strong hydrogen emission (HD57150). Linear/quadratic fits to the data are shown with a dotted/dashed line. linear local continuum and a Gaussian line profile for each individual water transition, using the least-squares fitting routines by Markwardt (2009). For each water transition we fix the centroid to its known rest frequency as taken from the HITRAN 2008 database (Rothman et al., 2009). The width of each Gaussian is fixed to the IRS instrumental line width as determined from fits to single unresolved hydrogen lines over the entire spectral coverage (see Figure 4.6). A linear fit gives: FWHM = a + bλ a = −0.0001 ± 0.0008 b = 0.00139 ± 0.00006 ,

(4.2)

equivalent to a spectral resolution R = 720 ± 33, extending up to 35 µm what found by Najita et al. (2010) over the short wavelengths (FWHM ∼ λ/750 over 10–20 µm). A quadratic fit to line widths gives: FWHM = a + bλ + cλ2 a = −0.012 ± 0.003 b = 0.0029 ± 0.0004 c = −4 × 10−5 ± 1 × 10−5 .

(4.3)

The quadratic fit accounts better for the overall bending of FWHM over the range 10-35 µm, and we adopt Equation 4.3. Fixed the line width and the rest frequencies, in our procedure to extract line fluxes from observed line blends we fit only for the 88


peak value of each Gaussian. For n transitions blended in a given complex the number of free parameters is therefore n + 2, and we require at least n + 3 data points for fitting. An example of the fits obtained with this method is shown in Figure 4.7, as applied to a portion of the data and to a model matching the spectral resolution and S/N of the data. Following this procedure, we retrieve ≈ 60–80 individual line fluxes (depending on the object) by extraction of those where the line identification is not degenerate, i.e. where the rest frequencies are separated by at least one pixel (> 0.5 resolution element). Exceptions are ortho-para line pairs with very similar Eu that, assuming an equilibrium ratio 1:3, can be extracted even when their rest frequencies overlap and be considered essentially as a single line (one example is the right-most deblended line in Figure 4.7). As in Banzatti et al. (2012), the error on each line flux is set to the standard deviation of the distribution of 1000 estimates of the same line flux fitting after resampling the line complex with the local noise, and we consider detected only fluxes with confidence higher than 99% (3σ). Additional uncertainty is given by contamination from other species and a complex pseudocontinuum produced by the dense forest of weak emission lines. These introduce unavoidable systematics that we try to mitigate with a careful line inspection. In fitting and extracting water lines, we avoid features contaminated by other species that are known to emit in the mid-infrared, in particular OH, HCN, C2 H2 , CO2 , HI, and [NeII], and we try to select line complexes where at least a few pixels of reliable continuum can be found at each side. Other systematics are more difficult to account for and can only partially be corrected (e.g. imperfect order matches and uncorrected cosmic ray hits, see Pontoppidan et al., 2010a; Carr & Najita, 2011). These errors can generally be mitigated to the point where signal-to-noise ratios (S/N) of & 200 on the continuum are achieved, which is the reason why we limit our analysis to a dozen of the highest S/N spectra only. Nonetheless, it is clear that the Spitzer data have intrinsic limitations that unfortunately do not allow to take full advantage of the rich emission observed. The number of lines that can be extracted still includes only a fraction of those that produce the molecular forest observed towards protoplanetary disks, as it is not possible to de-blend many of the Spitzer line complexes due to dense clustering. The most optically thin lines that are extracted (a few lines with Aul ∼ 10−3 s−1 ) are the weakest and likely the most unreliable, and yet they are needed to trace the upper edge of the rotational scatter. Therefore, while optimal in terms of coverage in Eu and Aul (at least in principle), Spitzer water spectra clearly provide ground to move only a first step until better data become available. The full potential of the rotation diagram analysis presented in the previous sections will be demonstrated when applied to sensitive higher-resolution observations, which for now are still too scarce to allow a comprehensive analysis of this kind (see Figure 4.8).

89


Fig. 4.7 — Top: flux extraction of individual water lines from a blended complex observed in the Spitzer -IRS spectrum (R ∼ 720) of FZ Tau. Individual gaussians are showed in dashed colors, overall fit in black thick line. The centroid of all water lines that could emit in the range considered, as taken from the HITRAN database, is marked with dotted lines. Bottom: the same line complex as fitted in a model of water emission having same sampling and S/N as the Spitzer spectrum (the model is shown in grey at a higher resolution of R ∼ 20, 000 below the fit, for comparison).

90


Fig. 4.8 — Rotation diagrams of water emission observed in the T Tauri disks included in this paper. Water lines extracted from Spitzer -IRS spectra are shown in black. The nominal 1-σ error-bar on the measured line flux is usually smaller than the dot size in this logarithmic plot. Dotted lines mark the region covered by the static models in Figure 4.4 and 4.5, for comparison. In addition, we plot the few water lines published to date from higher-resolution instruments (∆v ∼315 km/s), from: VLT-VISIR in green upward triangles (from Pontoppidan et al. (2010b) and Banzatti et al. (2014)), CRIRES/NIRSPEC in orange downward triangles (from Mandell et al. (2012)), and Herschel -PACS in purple squares (from Riviere-Marichalar et al. (2012)).

4.3.2

Rotation diagrams of Spitzer spectra

Figure 4.8 shows the rotation diagrams of infrared water emission for the disks considered in this work. For the reasons explained in the previous section, the number of line fluxes that can be extracted from Spitzer spectra, while populating quite well the rotational scatter, do not show the full details of its shape, in particular the exact location of the upper edge. However, they allow to make some first interesting considerations. All disks show a rotational scatter with features typically given by LTE emission in combination with a wide range in line optical depths: a fan-like shape that opens towards low Eu . While non-LTE excitation may play a role (see e.g. Banzatti et al., 2012), the detection of lines with high Eu in combination with the shape of the observed scatter confirm that it typically does not dominate the 91


excitation, and that a large range in line opacities is responsible for the emission observed from T Tauri disks. Therefore, the properties revealed by the rotation diagram analysis support what found in previous work: the emission is composed of a mixture of optically thick and thin lines approximately in LTE conditions. Moreover, the spread of the scatter suggests high column densities of NH2 O > 1018 cm−2 , confirming what found by Salyk et al. (2011). If high column densities (columns as high as 1021 cm−2 were found by Salyk et al., 2011) do trace a range in water abundances, RADLite models suggest that these are at least 10−4 and possibly as high as 10−2 relative to hydrogen, especially if GTDs are lower than 104 (see Section 4.2.3 and the discussion in Section 4.4). In fact, in many if not all disks the rotational scatter covers and exceeds the region defined by solar abundances even assuming very high GTD of 104 (compare to Figure 4.5). Water lines that largely exceed the static solution are only a handful in most objects (. 10). They lie in the optically thin part of the rotation diagram (having Aul ∼ 10−2 –10−3 s−1 ), likely close to its upper edge although it is difficult to carefully define it with the present data. These lines are overall weaker and more difficult to detect than the strong optically thick lines at the center/bottom of the scatter, especially in unresolved Spitzer spectra. Their detection is therefore still only tentative, and it is premature to measure individual values of inner disk water abundances without support from higher-resolution data. However, large rotational scatters exceeding the static solution boundaries would already support evidence for water abundances that are probably one (perhaps up to two) order of magnitude higher than solar oxygen values.

4.4 4.4.1

Discussion Origin of large water abundances

Assuming solar elemental abundances, even though chemical disk models predict that most of the oxygen is driven into water (Woitke et al., 2009; Glassgold et al., 2009), water abundances higher than the available oxygen are essentially not possible to produce in a static scenario. How could much higher values, perhaps up to two orders of magnitude higher, be explained? The canonical assumption on the atomic oxygen budget could be wrong: if oxygen were typically much more abundant in protoplanetary disks and in other (young) stars than in the present Sun, this would allow for water abundances higher than those provided by static disk models so far. However, there is evidence that oxygen abundances are close to solar in protoplanetary disks (Tsamis et al., 2011), in young B stars (Przybilla et al., 2008), as well as in FGK stars (Petigura & Marcy, 2011). Therefore, in a static wellmixed disk the oxygen budget would strictly limit the maximum water abundance to ∼ 10−4 relative to hydrogen. A more convincing scenario to explain larger water abundances comes instead from consideration of the migration of solids expected to happen in disks due to gas drag (Weidenschilling, 1977). Accounting for this effect, Ciesla & Cuzzi (2006) demonstrated that water abundances higher than solar 92

4.4. Discussion

oxygen values are produced in inner disks by inward icy migrators evaporating after crossing the snow line, in the first Myrs of disk evolution. Adopting a disk surface density profile as R−p , where R is the disk radius, Ciesla & Cuzzi (2006) estimated the maximum water abundance provided in the migration scenario by adding up the disk mass up to the outer disk radius Rout divided by the mass inside the radius of the midplane snow line Rsnow , assuming that water ice provides most of the solid mass of solid migrators. In this approximation, the maximum enhancement in water vapor concentration possibly produced inside the snow line is ∆X(H2 O)max ∼ (Rout /Rsnow )2−p .

(4.4)

Ciesla & Cuzzi (2006) assumed a snow line located at Rsnow ∼ 5 AU, an outer disk radius of Rout = 100 AU, and p = 1, producing a ∆X(H2 O)max of ∼ 20 which would give a water abundance of a few ×10−3 relative to hydrogen. Water-rich disks around T Tauri stars observed by Spitzer generally comply with the Ciesla & Cuzzi (2006) model, in being typically massive and young (see Table 4.1 and references therein). However, the water snow line is more likely close to ∼1 AU in T Tauri disks (Carr & Najita, 2008, 2011; Meijerink et al., 2009; Salyk et al., 2011), which would allow for water abundances up to 10−2 in inner disks according to the approximation in Equation 4.4. This is still assuming that all water ice migrates from the outer disk to inside the snow line, which is an extreme and unlikely condition. However, if typical disk radii extend further than 100 AU and/or p is less than 1, as found in millimeter studies (e.g. Andrews & Williams, 2007), then inner disk water abundances of 10−3 – 10−2 could in principle be easily reached in the migration scenario even retaining enough mass to form planets beyond the snow line. Together with large differences in volatile composition in the asteroid belt (Walsh et al., 2011) and observations of significant variations in hydrogen isotopic composition of chondrites (Jacquet & Robert, 2013), also the water vapor abundance observed in inner young disks may therefore provide evidence that planet formation does not happen in static well-mixed environments, but rather in dynamically active conditions where radial migration of solid bodies is efficient.

4.4.2

Confirmation of large water abundances

If possible and expected, why has the ice migration and evaporation scenario never been observationally confirmed yet? Most studies of molecular emission from disks so far ignored the possibility of enriched water abundances as due to migration processes and assumed solar values for the oxygen budget, assuming that only chemistry would be responsible for the observed emission. Moreover, infrared spectra of water emission have been typically interpreted using simple slab models, which provide only column densities of the observed emitting layer and do not constrain the abundance of water in the inner disk. Static disk models produce observable columns of NH2 O . 1018 cm−2 (Glassgold et al., 2009; Bethell & Bergin, 2009; Najita et al., 2011) and could at least partially explain the values found in previous analyses of 93


Spitzer spectra (see Figure 4.1). However, the depth of the layer of observable water vapor depends on the conditions posed in the model, like the assumption on the GTD ratio. Total column densities up to NH2 O . 1022 cm−2 (Walsh et al., 2012) could in principle be observed if disks were transparent all the way down to the disk midplane. There is evidence that upper layers in disk atmospheres are more transparent than lower layers toward the midplane, likely as a result of dust grain growth and settling (Brittain et al., 2005; Rettig et al., 2006; Horne, 1986). A larger portion of the disk would be visible above the layer where dust becomes optically thick, but there is no way that the inner disk is entirely transparent in the vertical direction, also because the gas emission itself becomes optically thick at high column densities (see Section 4.2). Observed column densities are therefore inadequate to solve the problem of the total amount of water vapor in inner disks, as they simply provide an unknown fraction of it. What is really needed is to constrain the absolute abundance of water vapor, and compare it to solar oxygen values to find evidence for enrichment from migration. In this work we use a two-dimensional disk model, RADLite, which accounts for the total water abundance in the inner disk and allows comparison with solar oxygen. Our explorations of RADLite models still show a dependency on the assumed GTD, as shown in Section 4.5: a canonical GTD = 100 with enriched water abundance is basically equivalent to a model with GTD = 104 and solar abundance. We explore high values of GTD = 104 as they were proposed by Meijerink et al. (2009) in attempting to match the water emission from T Tauri disks, especially at wavelengths shorter than ∼ 20 µm. However, observations provide evidence for GTDs typically lower by at least one order of magnitude than that in young disks: in the range 200-1000 (Brittain et al., 2005; Rettig et al., 2006; Horne, 1986) or extending even lower (0.1-1000, Meeus et al., 2010; Thi & Herschel GASPS Team, 2011). In this work we demonstrate that high GTDs are not the only way to increase the water vapor emission and that another (possibly more likely) solution is that the water abundance itself is higher than solar (see Figure 4.9), as proposed by disk evolution models that include solids migration. While GTDs lower than 104 already suggest that the migration scenario would be preferred over the static one in some disks (see Section 4.2.3 and Figure 4.9), additional constraints are provided by combination of other molecular gas tracers. From self-consistent simultaneous fitting of highresolution water and CO ro-vibrational lines, Pontoppidan & Blevins (submitted) find GTD of ∼ 500 and a water abundance higher than solar oxygen by one order of magnitude (10−3 relative to hydrogen) in one of the water-emitting T Tauri disks, RNO90. This result suggests that enrichment of water vapor happens in the inner disk of RNO90, likely due to inward migration of icy solids from the outer disk.

94

Fig. 4.9 — Model and observed spectra of water vapor in inner regions of protoplanetary disks. RADLite models with a GTD of 100 and X(H2 O) = 10−2 (migration scenario, top) or X(H2 O) = 10−4 (static scenario, bottom) are compared with Spitzer spectra of CW Tau (GTD ∼ 300, Horne, 1986) and TW Hya (GTD = 100 and X(H2 O) = 10−4 , Zhang et al., 2013). Prominent emission features produced by species other than water in the observed spectra (OH, HCN, CO2 , HCO+ , HI, and [NeII]) are marked with grey crosses.

4.4. Discussion

95


4.4.3

Future observations

Future sensitive high-resolution surveys will provide good ground for measuring the spread in rotational scatters of water emission from young disks, possibly probing the strength of the migration process and the accumulation of planet-building solids beyond the snow line. While some of the strongest optically thick lines could already be measured in many T Tauri disks using Spitzer observations, the specific goal of future surveys shall be to target some of the most optically thin lines tracing the upper edge of the rotational scatter. As the line opacity is not known a priori without making assumptions on the emitting conditions, low Aul values shall be used as a proxy to search for the most optically thin lines. Moreover, it is important to trace the upper edge of the scatter over a large range in Eu , so to distinguish opacity effects from non-LTE excitation and constrain the (range in) emitting gas temperature(s). Lines with Aul lower than 10−2 s−1 and with Eu from several down to a few 1000 K are scattered all over the 5-30 µm spectrum of water, making the task challenging for spectrographs providing narrow spectral windows like most ground-based instruments available today. The MIRI spectrograph on the James Webb Space Telescope will instead provide a wide continuous coverage in wavelength, from 5 to 28 µm, and be the optimal instrument to bring the findings pioneered by Spitzer to the next step, providing a moderate resolution of R ∼ 3000 that is enough to at least partially resolve line blends (see Figure 5 in Pontoppidan et al., 2010a). Another task for future sensitive high-resolution surveys will be to determine whether (and where) rarer water isotopes are populated and detected in disks, supporting the presence of high water abundances. H18 2 O has not been detected in Spitzer spectra so far, but the blending with the main isotope does not allow to derive strong constraints as its lines are fewer and weaker than those 18 of H16 2 O. We modeled H2 O with RADLite finding that a non-detection in Spitzer spectra is consistent with a solar 16 O/18 O ratio of ∼ 500 (Wilson & Rood, 1994) even in the case of H16 2 O abundances higher than solar. If lower isotope ratios are present in some disks, a few H18 2 O lines may become strong enough to be detected at wavelengths longer than ∼ 30 µm, which is where sensitive higher-resolution instruments shall target in the future. If the migration of icy solids from the outer disk is responsible for producing the water vapor emission observed inward of the snow line in protoplanetary disks, trends are expected to be observed between the water abundance and the evolutionary stage of the disk. According to Ciesla & Cuzzi (2006), the number of icy migrators decreases with time, as due to the depletion of solid material from the outer disk and the formation of larger bodies accreting migrators within the disk before they reach the snow line. Disks in different phases of evolution will therefore show large water abundances in earlier phases, and lower or absent water emission in later phases. An example of this might already be seen in the difference between the water spectrum of the evolved disk in TW Hya (Calvet et al., 2002; Najita et al., 2010; Zhang et al., 2013) and that of water-rich young T Tauri disks (see Figure 4.9). The large suppression of water emission in some T Tauri disks (Pontoppidan 96

4.5. Conclusions

et al., 2010a) may therefore be evidence for a range in water abundances depending on the disk evolutionary phase. This is consistent at least qualitatively also with the difference observed between T Tauri and Herbig disks: while the former typically have strong water emission, the latter do not show any water at all (Pontoppidan et al., 2010a; Fedele et al., 2011). While harsh UV photodissociation is usually invoked to explain the non detection of water, another interesting possibility comes from considering that disk evolution proceeds faster around the more massive disks around Herbig Ae/Be stars (Carpenter et al., 2006; Meyer, 2009), which are therefore more difficult to catch in their early phase of icy migrators crossing the snow line, especially if protoplanets form fast and accrete them within the disk. The C/O ratio of protoplanets in the inner disk would also vary depending on the location ¨ of formation and the strength of the migration process (Oberg et al., 2011a). Protoplanetary disks with low or no water vapor inward of the snow line could imply that water is locked in non-migrating planetesimals outside the snow line (Ciesla & Cuzzi, 2006). Initial evidence for this scenario has been very recently proposed by Najita et al. (2013) and Pascucci et al. (2013), who found a range in C/O ratios in inner disks around T Tauris and Brown Dwarfs, arguing in favor of the molecular emission observed at mid-infrared wavelengths being linked to migration processes of icy bodies.

4.5

Conclusions

Building up on the large degeneracies found in previous work and the opportunity to find observable signatures of the migration of icy solids in protoplanetary disks, we have performed a rotation diagram analysis of infrared water emission to identify the effects of large water abundances. By comparison of three different models that account for line opacity, LTE, non-LTE, and/or a two-dimensional disk structure, we find that large rotational scatters do provide a way to obtain observational evidence for water abundances larger than solar oxygen values, supporting the ice migration scenario. It is still unclear whether the migration of icy solids crossing the snow line would be the rule rather than the exception in producing water abundances in inner disks, but if GTDs are lower than 104 enhanced abundances are very likely required and the migration scenario would be the best available explanation of the strong water emission observed towards many T Tauri disks. Future analyses of sensitive higher-resolution spectra will detect and measure the most optically thin lines and constrain rotational scatters, possibly revealing trends expected with disk evolution. This requires the availability of infrared spectrographs with resolving power higher than at least a few thousands, so to resolve most of the water line blends, and providing large spectral coverage in the 5-30 µm window. SOFIA and JWST are therefore optimal candidates to move steps forward in lifting the veil on elusive processes of planet formation, confirming whether water vapor in inner disks is sensitive to migration processes and tracing the accumulation of planet-building solids beyond the snow line in protoplanetary disks. 97


A.B. is thankful to Jarron Leisenring and Nikoletta Sipos for helpful contributions concerning Spitzer spectra and RADMC models. This work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology.

98

“IT IS THE LARGE THING THAT IS SECRET AND INVISIBLE; IT IS THE SMALL THING THAT IS EVIDENT AND ENORMOUS.”

GILBERT K. CHESTERTON, THE EVERLASTING MAN, 1925

5

Conclusions & Outlook

It is hard to draw conclusions from a work in progress, especially as this PhD research has provided a starting point more than an end to the topics discussed. Together with summarizing the main results, the goal of this short concluding chapter is therefore also to highlight the extant questions, opening perspectives for future investigations. 1. Water has a simple molecular structure that produces a complex energy structure. Its energy structure, in addition to the high energy barrier of neutralneutral reactions, makes water vapor an optimal tracer of warm-hot (300 . T . 5000 K) and dense (n & 108 cm−3 ) environments (Chapter 1). These conditions are typically found at planet-forming radii in young, gas- and dustrich circumstellar disks. Infrared water vapor emission is therefore a privileged probe of these regions, among all other astronomical sources, and can be used as a tracer of the physical and chemical conditions found in inner disks. Being it linked to the processes and evolution of regions where planet properties are set, the water vapor observed in protoplanetary disks will increasingly attract attention in the coming years, especially in connection to the search for an Earth twin and other possibly habitable planets. 2. Mass accretion enhancements producing outbursts in young stars are able to affect the conditions of the molecular environments at planet-forming radii in disks (Chapters 2 and 3). The extent to which the inner disk is affected depends on the duration and on the strength of a given accretion flare. In fact, the strong (a factor 40 increase in the accretion luminosity) and long (7 months) accretion outburst monitored in EX Lupi in 2008 produced a dramatic change in the molecular emission. Water vapor emission increased possibly in connection to a recession of the snowline in outburst, an idea that we are currently investigating in detail in a dedicated modeling study. OH emission increased providing strong evidence for ongoing UV photodissociation of water. 99

CHAPTER 5. Conclusions & Outlook

The fate of organic molecules, which disappear from the outburst spectrum, is still not clear and is the object of an ongoing project to use new highresolution post-outburst data in 2014 to investigate the long-term evolution of the chemistry triggered during outburst. Whether accretion outbursts leave a permanent imprint in the disk chemistry or not is still a matter of investigation. While EX Lupi provided the opportunity to notice something that may be relevant to understand the evolution of the molecular gas in all young disks, our monitoring of DR Tau showed a lower limit to these effects. In fact, the water emission observed in DR Tau remained mostly stable during the moderate (a factor of 2 in accretion luminosity) and short (a few days) accretion variability monitored in our 2011-2012 campaign. This accretion behavior can be considered representative of T Tauri stars during most of their evolution, but strong accretion flares are thought to be ubiquitous during star formation. The relative effect of UV photochemistry and accretional heating, acting on different timescales, deserves future investigation by monitoring accreting stars in a range of flaring events. A detailed investigation over a range of timescales, including not only water but also OH and organics, has never been performed so far and may open completely new perspectives in disk chemistry studies in the future. 3. Focusing on water vapor, it is clear that we are just at the beginning of understanding the details of the emission observed toward young protoplanetary disks. This is due to different reasons: the intrinsic complexity of the emission and the limitations of the extant data and models. On the one hand, the complexity of the water spectrum makes its interpretation very challenging, especially if the emission is not produced in the ideal conditions of low optical depth and local thermal equilibrium (LTE). The emission observed from disks can in general be roughly approximated by an optically thick gas close to LTE, but non-thermal excitation processes have been suggested to play a role at least in some cases (see e.g. Chapter 2 in this book). As soon as high optical depth and non-LTE effects come into play, large ranges of rotational levels are needed in order to constrain the emission (see Chapter 4). Uncertainties in the molecular data (e.g. the collisional rates), however, still limit our understanding and modeling of the water emission regardless of the observations. Concerning the data, obtaining reliable and sensitive, high-spectral-resolution data is vital for the development of this field of research. Consideration of large samples of molecular transitions is clearly the way forward to a more comprehensive understanding of inner disks properties and evolution. Water lines de-blended from unresolved Spitzer spectra can be used to complement the (still) sporadic higher-resolution data until new large surveys of molecular emission become available. Growing evidence that water vapor in inner disks traces the snowline and possibly the migration of icy solids is being provided by analyses of Spitzer spectra (see Chapter 4), although it still needs to be confirmed by future observations. If it is connected to icy solids migrating in100

ward of the snowline, water vapor in disks could be used in the future to trace the strength of migration processes in different phases of disk evolution as well as the amount of water (and oxygen) that is held in the gas rather than in the solid phase, thus probing the molecular budget available to forming planets. In connection to a snowline recession, this will also allow to estimate the amount of icy solids (and their sizes?) that are evaporated during accretion outbursts and other energetic events, possibly providing a link to some observed properties of chondrites in our Solar System. Given the properties of water emission and the conditions in which it is produced, my opinion is that a large spectral coverage (5–40 µm) and a moderate spectral resolution (R & 3000) are an optimal compromise and in general preferable over a much higher resolution obtained on a few water lines only. These capabilities will be provided by the James Webb Space Telescope, which promises to bring this field of research to the next level of accuracy and depth. 4. As a final personal conclusion, I shall add that the data “deserve to be studied with the most careful attention to exactness, with the attention to understand things as they are, to listen to their delicate truth, to be every day ready to revise the interpretation in order to decipher, with a greater loyalty, their speechless language” (cf. the Preface). Even the failures I had sometimes to face during my PhD could not convince me of the opposite. I had many times the joy of the gift of understanding from the data some hidden information (for instance when attempting the absolute flux calibration of the VISIR data in Chapter 3), also thanks to my intense (often stubborn) dedication. After the joy of discovery, of a new direction that nobody ever took before, there is no higher joy in research than achieving to understand the sense of something we see. This in itself has always the renewed flavor of a discovery of something new. It is the re-discovery of a piece of truth. I can confirm this even from the most unsuspected and least productive (in terms of papers and career) thing I have deeply dedicated myself to during my PhD: an old-fashioned 1929 experiment of physics, part of the advanced lab of the Bachelor at the ETH, that I supervised for three semesters. The suffering and the joy that I experienced there, while messing up with the experiment to understand what was going on and what the data were telling us, were the same I had in my own work of research. There too, following the data even when they show something unexpected proved to be worth more than any other approach.

101

CHAPTER 5. Conclusions & Outlook

102

CV Andrea Banzatti born October 27, 1984 in Milan, Italy Member of the American Astronomical Society Education Postdoctoral Fellow at the Space Telescope Science Institute, Balti- 2013more MD, USA Ph.D. at the Institute for Astronomy, ETH Zurich Thesis title: Water vapor in protoplanetary disks: a probe of physics and chemistry of planet formation conditions Advisor: Prof. Dr. M. R. Meyer (ETH) Co-advisor: Dr. K. M. Pontoppidan (STScI)

2009-2013

Master’s thesis project, ESO Headquarters, Garching, Germany Thesis title: Evolution of protoplanetary disks: first steps toward planet formation Advisor: Dr. L. Testi (ESO) Co-advisor: Prof. Dr. M. Bersanelli (INAF)

2008-2009

Master’s Degree in Physics (cum laude), Universitá degli Studi di Milano, Italy

2006-2009

Bachelor’s Degree in Physics, Universitá degli Studi di Milano, Italy Thesis title: The Sunyaev-Zeldovich Effect Advisor: Prof. Dr. G. Bertin (INAF)

2003-2006

103

CV

Additional Training School From Planets to Life, Villars-sur-Ollon, Switzerland

Apr. 2011

7th IRAM Interferometry School, Grenoble, France

Oct. 2010

Teaching Experience ETH VP Lab assistant: Thermionic Emission

3 semesters

ETH VP Lab assistant: Astrowoche

3 semesters

ETH VP Lab assistant: Ultrasound Waves ETH AP Lab assistant Co-supervisor of a semester project, Spectral energy distributions of circumstellar disks, student Patrick Joos

1 semester 2 semesters 1 semester

Observing & Instrumentation Experience Co-I of an ESO large program (189.C-0313, VLT-VISIR, 24 nights) to survey infrared molecular emission in 65 protoplanetary disks (time awarded but not performed for instrument failure)

2012

Production of a new pipeline for reduction of VLT-VISIR data

2012

PI of an ESO program (088.C-0666, VLT-VISIR and VLT-Xshooter, 11.3 hours) to monitor accretion variability and water emission in DR Tau

2011/2012

Observer (4 nights) at the Mt. Bigelow 61” Telescope of the Steward Observatory, the University of Arizona

Oct. 2011

Service Activities Initiator and organizer, the PhD work-in-progress lunch, Institute for Astronomy, ETH Zurich

2012-2013

SOC contributor and speaker: the “Planet-Z” initiative

2012-2013

LOC member, the Swiss Astronomical Society meeting in Zurich

Oct. 2012

104

Publications Banzatti, A., Pontoppidan, K. M., Meyer, M. R., Bruderer, S., ApJ, submitted Searching for signatures of icy migrators in protoplanetary disks Banzatti, A., Meyer, M. R., Manara, C. F., Pontoppidan, K. M., Testi, L., ApJ, 780, 26 A UV-to-MIR monitoring of DR Tau: exploring how water vapor in the planet formation region of the disk is affected by stellar accretion variability

2014

Banzatti, A., Meyer, M. R., Pontoppidan, K. M., Bruderer, S., Proto- 2013 stars and Planets VI, Heidelberg, Poster 2S034 Observations of warm water in protoplanetary disks and its connection to disk evolution and planet formation Banzatti, A., Meyer, M. R., Bruderer, S., et al., ApJ, 745, 90 2012 EX Lupi from quiescence to outburst: exploring the LTE approach in modeling blended H2 O and OH mid-infrared emission Banzatti, A., Testi, L., Isella, A., et al., A&A, 525, A12 New constraints on dust grain size and distribution in CQ Tauri

2011

Various Unrefereed Publications Wasser als Geburtshelfer, Neue Zuercher Zeitung

2013

Astronomy course of ETH Zurich at Diavolezza, C. Monstein, A. Ban- 2012 zatti, L. Dedes, Journal of the Society of Amateur Radio Astronomers SARA Bestimmung des Sonnendurchmessers mittels Interferometrie, C. Monstein, A. Banzatti, L. Dedes, Zeitschrift der Schweizerischen Astronomischen Gesellschaft ORION

2011

105

CV

106

Acknowledgements If while writing this thesis I had to struggle many times to find the right words, to convey the right meaning or simply to (try and) be correct, this is most true here. Not just because I am not a native English speaker, with great despair of my supervisor (whose “warm” recommendation after our very first meeting was to read “Elements of style”, by Strunk and White I think - needless to say that I never did it). Here, it is hard to find the right words simply because there is no way to express exhaustively the gratitude I owe to so many people. But, of course, I will try anyway. Let me start there, then, from my supervisor, Michael R. Meyer. I do not think that any acknowledgments would be taken seriously and considered honest, especially in a PhD thesis, if they are free from any (even little) critics. Therefore, for the sake of making the reader believe what I will write in positive, I will now start from the negative. Moreover, the deeper the negative, the brighter the positive will look like, as detector images teach (but long before them the good Shakespeare - Hamlet, Act II, Scene II). Therefore, here I will not spare any of my critics. I have never seen anybody getting so irritatingly cantankerous when under stress, apart probably myself. Michael can let you repeat something you said, if it does not comply with his standards of precision, with such an intimidating attitude to disconcert (if not silence) anybody. Even senior scientists. The reader might now guess how I, a mere PhD student, felt during our private meetings, especially early in the PhD. It took almost four years until I learned to ignore the attitude and focus on the content. After all, it is a good thing to learn. So here comes my first acknowledgment. I thank you, Michael, for giving me such a hard time. It made me stronger, like a sword is made stronger by putting it into the flames and then hammering it with generosity. But at the same time, I have never seen anybody changing so readily from one attitude to its opposite. You could think that Michael hates you, but the day after you find yourself into his consideration at the same level as before, if not higher. What he really hates is confusion, unfair behavior, imprecision. He likes to understand, to see, being this English word really 107

ACKNOWLEDGMENTS

a wonderful combination of two things that should always go together. Michael is a versatile astronomer, a thoughtful scientist, a great man. He has a sincere passion for education, for understanding the truth of things in depth, for the origin and fate of stars, for telescopes and detectors (and photons above all), for a correct estimate of the uncertainties, and for “interesting conversations with interesting people”. I cannot list all the little and great things I have learned from him (although I never attended entirely any of his university courses..). I shall only mention the patience, and the most delicate respect, he had when I had a hard time during the first year of my PhD, actually the hardest time I have ever had in my life. But I shall say no more words on that here. He understood and allowed me to keep my artistic heart, while growing as a scientist. He taught me that there is nothing wrong in making mistakes, as we learn by doing (“provando e riprovando”). There is greater value in being willing to change and be corrected, rather than in just being correct. I really do not envy those who are always correct (or at least think so). They miss something essential in this world: realize that we are nothing, and that we need help from outside from the very first instants of life (as my dear Tommaso made very clear). Just like a seed is nothing, without water and light. Water, and light. This is my PhD in short. And I should move on to the other people I need to thank, before it gets too late and the survivors of my twenty-five readers, the few who made it thus far, leave this reading for better things. My dear Marianne Chiesi deserves the second place in this acknowledgments, if not a tied first. She is way more than what she looks like from outside, and much more than simply a secretary in Michael’s group. She is a woman, like few are left in this world. For instance, she loves wine. I am not saying that she drinks it (this is obvious). She loves it, as she loves all the beautiful things of this world. She loves Beauty, and this is something unfortunately getting rarer today. I will always remember the sparkling eyes she had after looking at the floor (the floor!) of the great cathedral we have in Milano (il Duomo). I think that this is why we found ourselves so tight in an uncommon friendship, as happens to people that (truly) look in the same direction (and here I do not simply mean the floor). She has a strong maternal attitude that is so comforting, and that only women can (so irreplaceably) bring in this world. She was so essential for me and the success of my PhD, from all points of view. Particularly during the hard times. I would have left early if it was not for her, missing all the great things that happened later. For the same reason I should right away thank also Mr. Toivanen, for his assistance, for his calm and relaxed positivity, and for his long strip beard. I owe great gratitude to Klaus Pontoppidan, who was supposed to join Michael’s group here in Zurich and be one of my mentors over all the PhD. In the end, he was one of my main collaborators anyway, but from the distance. A distance that will get reduced by a lot, very soon. I learned from him many technical things that Michael never had the time to address with me. He is a bright scientist, who is more disposed than Michael towards a good compromise between precision and production. He has a good sense for the opportunities, and for what is most important to pursue in our field. I still clearly remember his talk at the Herschel conference in Grenoble, which 108

gave me some fundamental ideas for my PhD. Moreover, he has all the knowledge of chemistry that I still don’t have. We are quite complementary, and this is something potentially very powerful. We will see what comes out of that in the next few years. I am looking forward to that. I owe an intense gratitude to Simon Bruderer, who rescued me from the chasm in which I ended up at the end of my first year of PhD. He gave me a rope and all the necessary tools to rise again, and be able to walk the distances that I walked in the following three years. Looking back, I would have never remotely even guessed where I am now. I owe to him much of all this thesis. He is a good and fun guy, serious and very knowledgeable. He is a crazy chemistry modeler, and he generally has an opinion on any chemical matters. Often it is a cautionary opinion, and his typical tendency is to curb. In other words, he uses water to extinguish any sprinkle of fire, as he did many times with me. And not without reasons... But fortunately he could not extinguish the fire that burns inside me. He is another one with whom I had to learn, sometimes, to ignore the attitude and focus on the content. As I had to change my own attitude, and this is another important thing to realize. Someone who truly always has an opinion, or better, an answer to any question is my gigantic officemate Michiel Cottaar. And the problem is that, most times, he is right. I once started keeping track of the number of questions he answered me, compared to vice versa. I readily stopped, as it was an unfair count. There were not even the conditions for statistics. And if there is one field where you cannot trick Michiel, that is statistics. Michiel is a big Dutch man with a big spontaneous generosity. So big that he lost track of its origins. He happily shares his immense knowledge with anybody, and thanks to me he now has a permanent queue of PhD students (sometimes even some postdoc) at the door. One day or the other I am going to sell tickets and make a living out of it. (By the way, he can also think in the Fourier space in his head..). I could write pages and pages about him and his pleasant, ironic (sometimes sarcastic), and fun presence, but I need to move on. Next, I want to thank all PhD students who followed me on the much-envied (by postdocs) idea of the PhD work-in-progress lunch, that I started with Antonio Garufi, Sandro Tacchella, and Andreas Faisst. Their support and participation has been vital not only for the lunch itself, but for myself as well. Especially Antonio has had an affection, and has supported me, in a way that only a true friend can have. Every Hamlet needs his Horatio, or will fall in the abyss following a ghost. Sandro has greatly contributed with his unique passion (and the skills of a leader), and Andreas with his crazy photographic reports from all over the world (and his unique observational attitude). Elena Borisova participated with a delicacy and affection that only women can have in this world, not to talk of her marvelous cakes. If I am the captain (as Antonio keeps saying), Thomas Bschorr is for sure the mother of all PhDs. Not only for his looong experience as a PhD, nor simply because his PhD (in terms of pain) is giving him the opportunity of a similar experience to being in labor. He has an acute and careful attention to everybody and each one, like nobody else in the institute. Sandro, Thomas, and Elena were also those who followed me in my crazy idea of the IfA Christmas QUIZ (slides are available for 109

ACKNOWLEDGMENTS

whom is interested), an adventure that I will remember forever. Neven Caplar arrived later, but his eastern-European spirit (and cursing, and sneezing) brought an irreplaceable contribution to the PhD zoo. Not to talk of her marvelous cakes (still Elena’s, as Neven’s were terrific). The list of PhDs I met during my stay would be much longer, and I regret not having had more time to know better each of them, nor to write a few words here. A special acknowledgment is fully deserved by the group of Swiss people that I first met when I arrived, Susanne Wampfler, Esther Buenzli, and the much intimidating Lucia Kleint. Wonderful girls with a strange vocation for little lights in the sky. Susanne shared with me more than all the others, about suffering and hopes and chemistry of water. To be complete, this list should include Andreas Bazzon, Christian Monstein, and Hans-Martin Schmid. Andreas has a great spirit, Christian a great stomach, and Hans-Martin is simply the greatest Swiss guy I ever met (in terms of fun). If I consider the great experience I had with each one of them, I cannot really understand the general belief of Swiss-German people as unfriendly and unsocial. I really had a lot of fun with them, from the basements of the ETH to the top of Diavolezza. To the Star and Planet Formation Group, now! Vincent Geers has been one of the most influential postdocs for me, and by the time he left I can really say that we were good friends, and not simply colleagues. He is another one from whom I could learn a love for precision and details above all. He is the guy that should be deeply taken into consideration when writing a new data reduction pipeline for an infrared spectrograph, as I hope it is happening now with his new MIRI position in Dublin. He has a great heart and a profound curiosity, and as all the most sensitive men he has to fight some bad past experiences to be able to cross some doors. Carolin & Leonidas are an incredible mix of opposite things, if you think, just as an example, of how well Germans and Greeks go along with each other nowadays. Two great guys with whom I shared not only the love for good food, particularly the awesome Greek food that was completely new to me, but also the love for fantasy stories. In many D&D nights I learned, thanks to them, that I am not only a Paladin, but also a Barbarian. And that the two coexist in me, and there is nothing wrong with that (unless the Barbarian rages.. then better keep away). Richard Parker would probably never suspect that, but he played an essential role in shaping my attitude towards our weekly scientific meetings. It is from him, from looking at how he behaved, that I slowly started to engage more personally, bringing my own plots and questions. To some extent, I should admit that the idea of the PhD work-in-progress lunch got some of its origins from his behavior. It is such a relief that men can be good examples to each others even without having the declared purpose. As it is such a relief that good men and women, serious with what they do, still exist. Farzana Meru is another bright example of that. But also a bright example of a calm happiness that can compete with any sudden problem. I once sent her an angry email for something she organized sloppily (at least so I thought), and after that we became truly good friends. (We shall admit that this is not common.) The problem is that she might not remember it, as she is involved 110

in so many things at one time that she does not have the time to think while she does them. Sometimes not even the time to realize what is going on around her. Nikoletta Sipos is another one who I will not be able to thank enough here. She helped me a lot with SED modeling issues, and it is most unfortunate that nothing of that could be included in one of my papers. The problem is that I am still not sure if the benefits can at all compete with the complications and degeneracies that SED modeling forces you to go through. But Niki is way more than SED modeling, even if she dedicated her scientific life to it. She is most sensitive, very fun (and unfortunate, unfortunately), and very happy. It’s good to have her around, if you can cut on the SED model details to some extent. Sascha Quanz is a real man of success, and he deserves some good position of management. Hopefully that would not happen at the expenses of his scientific contribution, as he is one of the most clever and bright observational astronomers around. In addition, he never gave up on his family for the sake of his career. And I should not forget to say one more essential thing about him: that he always has a kind, warmly reassuring, good word for PhD students when they need it. In his own way, at least... Jarron Leisenring shared with me an anxious passion for T Tauri stars and their variability, IDL issues and careful data handling, and more than anything else its cats. But now, time and space are running out, and I have to be brief. At the unfair expenses of the higher consideration that these next people would fully deserve. (Not that I did a much better job with the people above...) I thank Kevin Heng for his great passion and for his high consideration of observations (despite being a hard-core theoretician), Henning Avenhaus for his endless questions and his nice photographs, Maddalena Reggiani for helping me find a good apartment for my family and for her generous cakes, Katarina Kovac, Javiera Guedes Selame, and Antonio Pipino for their warm friendship, Andrina Nicola for her smiles and her care of me (thanks for the coccinelle!), Silvia Garbari for her gentle presence in our office and for her love for beauty and photography. Last, Fiorella Cagnetta gave me the great honor of sharing something that only good friends share, and we now walk together on the same road. A quite intense collaboration with Carlo Manara allowed me (and him as well) to dig deeper into the many facets and caveats of studying accretion in T Tauri stars. I should also acknowledge at least some of the senior colleagues whom I had the pleasure and opportunity to talk to: Didier Queloz was most influential on me, I greatly admire Simon Lilly even though I did not have the pleasure to work with him, Inga Kamp has a unique passion for chemistry and inexplicably spent with me 3 hours talking about my research, Ilaria Pascucci helped me to move my first steps in the analysis of EX Lupi and to “long for the endless immensity of the sea”, Peter Abraham is a really serious astronomer, while Avi Mandell took me and my work very seriously. Leonardo Testi was always supportive and available, and much I still owe him from the time spent at ESO for my Master’s thesis project. And now that we are at the end of it, this is the right place for what is most important. If during my PhD I received water way more than I would have ever dreamt of, my dear Giulia (and now Tommaso too) gave me the light, without 111

ACKNOWLEDGMENTS

which a seed would only rot away. The best moments of my time, recently, were all those evenings when going home after a hard day of work I found them happy, waiting for me. This is perhaps the most important thing that we need to find in this life, someone who is (happily) waiting for us, whatever our conditions are. In saying so, my thoughts immediately go also to the family where I come from and to all my friends in Italy (and now abroad) whose names I cannot list here. All this would not have been (and be) possible without them.

A.B. Uitikon Waldegg, conceived and written mostly in the night of June 25th, 2013

112

Bibliography ´ Abrah´ am, P., et al. 2009, Nature, 459, 224 Alencar, S. H. P., Johns-Krull, C. M., & Basri, G. 2001, AJ, 122, 3335 Andrews, S. M., & Williams, J. P. 2005, ApJ, 631, 1134 Andrews, S. M., & Williams, J. P. 2007, ApJ, 671, 1800 Andrews, S. M., Wilner, D. J., Hughes, A. M., Qi, C., & Dullemond, C. P. 2009, ApJ, 700, 1502 Appenzeller, I., Reitermann, A., & Stahl, O. 1988, PASP, 100, 815 Armitage, P. J. 2011, ARA&A, 49, 195 Aspin, C., Reipurth, B., Herczeg, G. J., & Capak, P. 2010, ApJ, 719, L50 Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481 Banzatti, A., Testi, L., Isella, A., et al. 2011, A&A, 525, A12 Banzatti, A., Meyer, M. R., Bruderer, S., et al. 2012, ApJ, 745, 90 Banzatti, A., Meyer, M., Pontoppidan, K., & Bruderer, S. 2013, Protostars and Planets VI, Heidelberg, July 15-20, 2013. Poster #2S034, 34 Banzatti, A., Meyer, M. R., Manara, C. F., Pontoppidan, K. M., & Testi, L. 2014, ApJ, 780, 26 Beckwith, S. V. W., Sargent, A. I., Chini, R. S., & Guesten, R. 1990, AJ, 99, 924 Bergin, E. A., Aikawa, Y., Blake, G. A., & van Dishoeck, E. F. 2007, Protostars and Planets V, 751 113

BIBLIOGRAPHY

Bergin, E. A., & van Dishoeck, E. F. 2012, Royal Society of London Philosophical Transactions Series A, 370, 2778 Bertout, C., Basri, G., & Bouvier, J. 1988, ApJ, 330, 350 Bethell, T., & Bergin, E. 2009, Science, 326, 1675 Borucki, W. J., Koch, D., Basri, G., et al. 2010, Science, 327, 977 Borucki, W. J., Koch, D. G., Basri, G., et al. 2011, ApJ, 736, 19 Brittain, S. D., Rettig, T. W., Simon, T., & Kulesa, C. 2005, ApJ, 626, 283 Brown, L. R., Troutman, M. R., & Gibb, E. L. 2013, ApJl, 770, L14 Bruderer, S., van Dishoeck, E. F., Doty, S. D., & Herczeg, G. J. 2012, A&A, 541, A91 Calvet, N., D’Alessio, P., Hartmann, L., et al. 2002, ApJ, 568, 1008 Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 Carmona, A., van den Ancker, M. E., Henning, T., et al. 2008, A&A, 477, 839 Carpenter, J. M., Mamajek, E. E., Hillenbrand, L. A., & Meyer, M. R. 2006, ApJL, 651, L49 Carr, J. S., & Najita, J. R. 2008, Science, 319, 1504 Carr, J. S., & Najita, J. R. 2011, ApJ, 733, 102 Chavarria-K., C. 1979, A&A, 79, L18 Chiang, E., & Youdin, A. N. 2010, Annual Review of Earth and Planetary Sciences, 38, 493 Chiang, E., & Laughlin, G. 2013, MNRAS, 431, 3444 Ciesla, F. J., & Cuzzi, J. N. 2006, Icarus, 181, 178 Cohen, M., & Witteborn, F. C. 1985, ApJ, 294, 345 Cyr, K. E., Sears, W. D., & Lunine, J. I. 1998, Icarus, 135, 537 Cuzzi, J. N., & Zahnle, K. J. 2004, ApJ, 614, 490 D’Alessio, P., Calvet, N., & Hartmann, L. 2001, ApJ, 553, 321 Dodson-Robinson, S. E., Willacy, K., Bodenheimer, P., Turner, N. J., & Beichman, C. A. 2009, Icarus, 200, 672 Doppmann, G. W., Najita, J. R., Carr, J. S., & Graham, J. R. 2011, ApJ, 738, 112 114

BIBLIOGRAPHY

Dullemond, C. P., & Dominik, C. 2004, A&A, 417, 159 Dullemond, C. P., Hollenbach, D., Kamp, I., & D’Alessio, P. 2007, Protostars and Planets V, 555 Eisner, J. A. 2012, ApJ, 755, 23 Faure, A., & Josselin, E. 2008, A&A, 492, 257 Fedele, D., Pascucci, I., Brittain, S., Kamp, I., Woitke, P., Williams, J. P., Dent, W. R. F., & Thi, W.-F. 2011, ApJ, 732, 106 Feigelson, E. D., Casanova, S., Montmerle, T., & Guibert, J. 1993, ApJ, 416, 623 Gibb, E. L., Whittet, D. C. B., Boogert, A. C. A., & Tielens, A. G. G. M. 2004, ApJs, 151, 35 Glassgold, A. E., Meijerink, R., & Najita, J. R. 2009, ApJ, 701, 142 Goldsmith, P. F., & Langer, W. D. 1999, ApJ, 517, 209 Gorti, U., Hollenbach, D., Najita, J., & Pascucci, I. 2011, ApJ, 735, 90 Goto, M., et al. 2011, ApJ, 728, 5 Gotz, W. 1980, Information Bulletin on Variable Stars, 1747, 1 Grankin, K. N., Melnikov, S. Y., Bouvier, J., Herbst, W., & Shevchenko, V. S. 2007, A&A, 461, 183 Grevesse, N., Asplund, M., Sauval, A. J., & Scott, P. 2010, Ap&SS, 328, 179 Grosso, N., Hamaguchi, K., Kastner, J. H., Richmond, M. W., & Weintraub, D. A. 2010, A&A, 522, A56 G¨ udel, M., et al. 2010, A&A, 519, A113 Harich, S. A., Hwang, D. W. H., Yang, X., Lin, J. J., Yang, X., & Dixon, R. N. 2000, J. Chem. Phys., 113, 10073 Hartigan, P., Hartmann, L., Kenyon, S. J., Strom, S. E., & Skrutskie, M. F. 1990, ApJ, 354, L25 Hartmann, L., & Kenyon, S. J. 1996, ARA&A, 34, 207 Hartmann, L., Calvet, N., Gullbring, E., & D’Alessio, P. 1998, ApJ, 495, 385 Hartmann, L. 2009, Accretion Processes in Star Formation: Second Edition, by Lee Hartmann. ISBN 978-0-521-53199-3. Published by Cambridge University Press, Cambridge, UK, 2009 115

BIBLIOGRAPHY

Herczeg, G. J., Karska, A., Bruderer, S., et al. 2012, A&A, 540, A84 Herbig, G. H. 1970, Memoires of the Societe Royale des Sciences de Liege, 19, 13 Herbig, G. H. 1977, ApJ, 217, 693 Herbig, G. H. 1989, European Southern Observatory Astrophysics Symposia, 33, 233 Herbig, G. H. 2007, AJ, 133, 2679 Herbig, G. H. 2008, AJ, 135, 637 Herbig, G. H., Aspin, C., Gilmore, A. C., Imhoff, C. L., & Jones, A. F. 2001, PASP, 113, 1547 Herbst, W., Herbst, D. K., Grossman, E. J., & Weinstein, D. 1994, AJ, 108, 1906 Herczeg, G. J., & Hillenbrand, L. A. 2008, ApJ, 681, 594 Hernández, J., Calvet, N., Hartmann, L., et al. 2005, AJ, 129, 856 Hernández, J., Hartmann, L., Megeath, T., et al. 2007, ApJ, 662, 1067 Hogerheijde, M. R., Bergin, E. A., Brinch, C., et al. 2011, Science, 334, 338 Horne, K. 1986, PASP, 98, 609 Houck, J. R., et al. 2004, ApJs, 154, 18 Ida, S., & Lin, D. N. C. 2008, ApJ, 673, 487 Kenyon, S. J., Dobrzycka, D., & Hartmann, L. 1994, AJ, 108, 1872 ´ et al. 2011, ApJ, 736, 72 Kóspál, A., Jacquet, E., & Robert, F. 2013, Icarus, 223, 722 Johansen, A., Oishi, J. S., Mac Low, M.-M., et al. 2007, Nature, 448, 1022 Johnstone, D., Hendricks, B., Herczeg, G. J., & Bruderer, S. 2013, ApJ, 765, 133 Joy, A. H. 1945, ApJ, 102, 168 Juhász, A., Dullemond, C., van Boekel, R., et al. 2011, arXiv:1110.3754 Kamp, I., Thi, W.-F., Meeus, G., et al. 2013, A&A, 559, A24 Lagage, P. O., Pel, J. W., Authier, M., et al. 2004, The Messenger, 117, 12 Lahuis, F., et al. 2006, c2d Spectroscopy Explanatory Supplement (Pasadena: Spitzer Science Center) 116

BIBLIOGRAPHY

Larsson, B., Liseau, R., & Men’shchikov, A. B. 2002, A&A, 386, 1055 Lehmann, T., Reipurth, B., Brandner, W. 1995, A&A, 300, L9 Lissauer, J. J. 1993, ARA&A, 31, 129 Lombardi, M., Lada, C. J., & Alves, J. 2008, A&A, 480, 785 Lynden-Bell, D., & Pringle, J. E. 1974, MNRAS, 168, 603 Natta, A., Grinin, V., & Mannings, V. 2000, Protostars and Planets IV, 559 Manara, C. F., Testi, L., Rigliaco, E., et al. 2013, A&A, 551, A107 Manara, C. F., Beccari, G., Da Rio, N., et al. 2013, A&A, 558, A114 Mandell, A. M., Mumma, M. J., Blake, G. A., et al. 2008, ApJl, 681, L25 Mandell, A. M., Bast, J., van Dishoeck, E. F., et al. 2012, ApJ, 747, 92 Mannings, V., & Emerson, J. P. 1994, MNRAS, 267, 361 Markwardt, C. B. 2009, Astronomical Data Analysis Software and Systems XVIII, 411, 251 Mayor, M., & Queloz, D. 1995, Nature, 378, 355 Meeus, G., Pinte, C., Woitke, P., et al. 2010, A&A, 518, L124 Meijerink, R., Pontoppidan, K. M., Blake, G. A., Poelman, D. R., & Dullemond, C. P. 2009, ApJ, 704, 1471 Mendoza V., E. E. 1966, ApJ, 143, 1010 Meyer, M. R., Backman, D. E., Weinberger, A. J., & Wyatt, M. C. 2007, Protostars and Planets V, 573 Meyer, M. R. 2009, IAU Symposium, 258, 111 McLaughlin, D. B. 1946, AJ, 52, 109 Modigliani, A., Goldoni, P., Royer, F., et al. 2010, Proc. SPIE, 7737, Mora, A., Mer´ın, B., Solano, E., et al. 2001, A&A, 378, 116 Muzerolle, J., Calvet, N., Hartmann, L., & D’Alessio, P. 2003, ApJL, 597, L149 Najita, J. R., Carr, J. S., Strom, S. E., Watson, D. M., Pascucci, I., Hollenbach, D., Gorti, U., & Keller, L. 2010, ApJ, 712, 274 ´ amkovics, M., & Glassgold, A. E. 2011, ApJ, 743, 147 Najita, J. R., Ad´ 117

BIBLIOGRAPHY

Najita, J. R., Carr, J. S., Pontoppidan, K. M., et al. 2013, ApJ, 766, 134 ¨ Oberg, K. I., Linnartz, H., Visser, R., & van Dishoeck, E. F. 2009, ApJ, 693, 1209 ¨ Oberg, K. I., Murray-Clay, R., & Bergin, E. A. 2011, ApJL, 743, L16 ¨ Oberg, K. I., Boogert, A. C. A., Pontoppidan, K. M., et al. 2011, ApJ, 740, 109 Pascucci, I., et al. 2007, ApJ, 663, 383 Pascucci, I., Apai, D., Hardegree-Ullman, E. E., Kim, J. S., Meyer, M. R., & Bouwman, J. 2008, ApJ, 673, 477 Pascucci, I., Apai, D., Luhman, K., Henning, T., Bouwman, J., Meyer, M. R., Lahuis, F., & Natta, A. 2009, ApJ, 696, 143 Pascucci, I., Herczeg, G., Carr, J. S., & Bruderer, S. 2013, ApJ, 779, 178 Pavlyuchenkov, Y., Semenov, D., Henning, T., Guilloteau, S., Piétu, V., Launhardt, R., & Dutrey, A. 2007, ApJ, 669, 1262 Petigura, E. A., & Marcy, G. W. 2011, ApJ, 735, 41 Petrov, P. P., Gahm, G. F., Stempels, H. C., Walter, F. M., & Artemenko, S. A. 2011, A&A, 535, A6 Pinilla, P., Birnstiel, T., Ricci, L., et al. 2012, A&A, 538, A114 Polyansky, O. L. 1985, Journal of Molecular Spectroscopy, 112, 79 Pontoppidan, K. M., van Dishoeck, E. F., & Dartois, E. 2004, A&A, 426, 925 Pontoppidan, K. M., Meijerink, R., Dullemond, C. P., & Blake, G. A. 2009, ApJ, 704, 1482 Pontoppidan, K. M., Salyk, C., Blake, G. A., Meijerink, R., Carr, J. S., & Najita, J. 2010a, ApJ, 720, 887 Pontoppidan, K. M., Salyk, C., Blake, G. A., Käufl, H. U. 2010b, ApJ, 722, L173 Pontoppidan, K. M., Blake, G. A., & Smette, A. 2011, ApJ, 733, 84 Pontoppidan, K. M., Salyk, C., Bergin, E. A., et al. 2014, arXiv:1401.2423 Przybilla, N., Nieva, M.-F., & Butler, K. 2008, ApJL, 688, L103 Quanz, S. P., Amara, A., Meyer, M. R., et al. 2013, ApJ, 766, L1 Quanz, S. P., Avenhaus, H., Buenzli, E., et al. 2013, ApJ, 766, L2 Rettig, T., Brittain, S., Simon, T., et al. 2006, ApJ, 646, 342 118

BIBLIOGRAPHY

´ Mer´ın, B., Bouy, H., & Maud, L. T. 2014, A&A, 561, A54 Ribas, A., Ricci, L., Testi, L., Natta, A., et al. 2010, A&A, 512, A15 Ricci, L., Testi, L., Natta, A., & Brooks, K. J. 2010, A&A, 521, A66 Rigliaco, E., Natta, A., Testi, L., et al. 2012, A&A, 548, A56 Riviere-Marichalar, P., Ménard, F., Thi, W. F., et al. 2012, A&A, 538, L3 Ros, K., & Johansen, A. 2013, A&A, 552, A137 Rothman, L. S., et al. 2009, , J. Quant. Spectrosc. Radiat. Transfer, 110, 533 Salyk, C., Pontoppidan, K. M., Blake, G. A., Lahuis, F., van Dishoeck, E. F., Evans, N. J. 2008, ApJ, 676, L49 Salyk, C., Pontoppidan, K. M., Blake, G. A., Najita, J. R., & Carr, J. S. 2011, ApJ, 731, 130 Shapiro, S. S., Wilk, M. B. 1965, Biometrika, 52, 591 Schöier, F. L., van der Tak, F. F. S., van Dishoeck, E. F., & Black, J. H. 2005, A&A, 432, 369 ´ ´ Kun, M., Moór, Sipos, N., Abrah´ am, P., Acosta-Pulido, J., Juhász, A., Kóspál, A., A., & Setiawan, J. 2009, A&A, 507, 881 Sohl, F., Wagner, F. W., & Rauer, H. 2012, arXiv:1211.3331 Sonnentrucker, P., Neufeld, D. A., Gerakines, P. A., et al. 2008, ApJ, 672, 361 Stahler, S. W., & Palla, F. 2005, The Formation of Stars, by Steven W. Stahler, Francesco Palla, pp. 865. ISBN 3-527-40559-3. Wiley-VCH , January 2005 Stevenson, D. J., & Lunine, J. I. 1988, Icarus, 75, 146 Strom, S. E. 1972, PASP, 84, 745 Tappe, A., Lada, C. J., Black, J. H., Muench, A. A., 2008, ApJ, 680, L117 Tennyson, J., Zobov, N. F., Williamson, R., Polyansky, O. L., & Bernath, P. F. 2001, Journal of Physical and Chemical Reference Data, 30, 735 Teodorani, M., Errico, L., & Vittone, A. A. 1999, Mem. Soc. Astron. Italiana, 70, 417 Thi, W. F., & Herschel GASPS Team 2011, IAU Symposium, 280, 356P Tsamis, Y. G., Walsh, J. R., V´ılchez, J. M., & Péquignot, D. 2011, MNRAS, 412, 1367 119

BIBLIOGRAPHY

Valenti, J. A., Basri, G., & Johns, C. M. 1993, AJ, 106, 2024 van der Tak, F. F. S., Black, J. H., Schöier, F. L., Jansen, D. J., & van Dishoeck, E. F. 2007, A&A, 468, 627 van der Tak, F. 2011, arXiv:1107.3368 Vasyunin, A. I., Wiebe, D. S., Birnstiel, T., Zhukovska, S., Henning, T., & Dullemond, C. P. 2011, ApJ, 727, 76 Vernet, J., Dekker, H., D’Odorico, S., et al. 2011, A&A, 536, A105 Visser, R. 2009, Ph.D. Thesis Wallace, L., Bernath, P., Livingston, W., et al. 1995, Science, 268, 1155 Walsh, K. J., Morbidelli, A., Raymond, S. N., O’Brien, D. P., & Mandell, A. M. 2011, Nature, 475, 206 Walsh, C., Nomura, H., Millar, T. J., & Aikawa, Y. 2012, ApJ, 747, 114 Weidenschilling, S. J. 1977, MNRAS, 180, 57 Werner, M. W., et al. 2004, ApJs, 154, 1 White, R. J., & Ghez, A. M. 2001, ApJ, 556, 265 Wilson, T. L., & Rood, R. 1994, ARA&A, 32, 191 Woitke, P., Thi, W.-F., Kamp, I., & Hogerheijde, M. R. 2009, A&A, 501, L5 Zhang, K., Pontoppidan, K. M., Salyk, C., & Blake, G. A. 2013, ApJ, 766, 82

120