Presented in fulfillment of the requirements of the degree of Doctor of Philosophy. February 2014

Transcription

1 A Multi-Wavelength Study of Grain Growth in Protoplanetary Discs Catarina Coutinho Pedroso Chaves Ubach Presented in fulfillment of the requirements of the degree of Doctor of Philosophy February 2014 Faculty of Science, Engineering and Technology Swinburne University

2

3 Abstract i Protoplanetary disc around young stellar objects contain the building blocks of planets. Observations at millimetre wavelengths are used to directly probe the cooler outer regions and mid-plane of the disc where the bulk of the dust resides. Observations at 1 and 3 mm can provide signatures of growth to mm-sized grains. Signatures of grains up to cm sizes can only be obtained by increasing the observing wavelength to 7 and 15 mm. If thermal dust emission dominates at 7 mm and beyond, the spectral slope should remain constant into the cm bands. However, as the observing wavelength is increased from 3 to 7 and 15 mm, other forms of emission besides thermal dust emission can also be present. The contributions from other processes cause an excess in flux above the expected thermal dust emission, and thus disentangling the emission mechanisms is required before conclusions can be made about the maximum grain size. The aim of this thesis is to study the first stages of planet formation by searching for signatures of grain growth up to cm-sized pebbles and to disentangle the emission mechanisms present in protoplanetary discs at 7 and 15 mm. Our observational results are then combined with radiative transfer modelling to determine disc parameters, such as disc size and geometry, and dust mass and composition, for an interesting source in our sample. We begin by presenting our methodology, which consists of five stages: (1) completing 3 and 7 mm detections of a sample of Chamaeleon and Lupus sources, to complete the long wavelength spectral energy distribution and obtain spectral slopes, which provide information on the emission mechanisms present; (2) producing maps and visibility plots of our sample, to determine if the emission is resolved; (3) calculating the dust opacity index, to constrain the maximum grain size; (4) conducting temporal flux monitoring at 7 and 15 mm and 3+6 cm for a sub-sample of sources to determine the dominant emission mechanism; and (5) running a full parameter sweep with the 3D radiative transfer code mcfost to constrain the disc parameters for an interesting source in our sample, GQ Lup, which has signatures of cm-sized grains and flux stability at 7 and 15 mm. We present the results of our Australia Telescope Compact Array (ATCA) 3 and 7 mm continuum survey of 20 T Tauri stars, which aims to identify protoplanetary discs with signs of grain growth. We detected 90% of the sources at 3 and 7 mm, and determined their spectral slopes, dust opacity indices and dust disc masses. We present temporal monitoring results of a sub-set of sources at 7, 15 mm and 3+6 cm to investigate grain growth up to cm sizes and determine the emission mechanisms present in these sources. Additionally, we investigate the potential correlation between grain growth signatures in the infrared

4 ii (via the 10 µm silicate feature) and millimetre (via the 1 3 mm spectral slope). We find that eleven sources have dominant thermal dust emission up to 7 mm, with seven of these having a 1 3 mm dust opacity index less than unity, suggesting grain growth up to at least mm sizes. The Chamaeleon sources observed at 15 mm and beyond show the presence of excess emission from an ionised wind and/or chromospheric emission. Long timescale monitoring at 7 mm indicates that cm-sized pebbles are present in at least four sources, while short timescale monitoring at 15 mm suggests the excess emission is from thermal free-free emission. Finally, a weak correlation is found between the strength of the 10 µm feature and the mm-spectral slope, suggesting simultaneous dust evolution of the inner and outer parts of the disc. Our survey shows that grain growth up to cm-sized pebbles and the presence of excess emission at 15 mm and beyond are common in these systems, and that temporal monitoring is required to disentangle the emission mechanisms. A source from our sample, GQ Lup, shows both signatures of large cm-sized pebbles and evidence of dominant thermal dust emission up to 15 mm. Additional observations were conducted to ensure flux stability over a timescale of years at 7 and 15 mm, to confirm that thermal dust emission dominates the emission up to 15 mm. High resolution 3 mm observations were also conducted to ensure that the emission was resolved. From these follow-up observations for GQ Lup we find that: (1) the emission is extended at 3 mm, suggesting a disc radius between au; (2) the flux is temporally stable at 3, 7 and 15 mm, suggesting the emission is dominated by thermal dust emission; and (3) large cm-sized pebbles are present in the protoplanetary disc. Our GQ Lup disc size is inconsistent with previous 1.3 mm Submillimetre Array (SMA) observations and modelling by Dai et al. (2010), who suggest a compact disc with an outer radius between au. However, when we push their model to longer wavelengths, the model fails to produce enough flux to match the ATCA data at 3 mm and beyond. The discrepancy between the 1.3 and 3 mm observations could suggest that the marginally resolved 1.3 mm observations did not have sufficient sensitivity to detect the emission from the cooler outer radius of the disc. We use mcfost to model the spectral energy distribution and the 3 mm visibilities of GQ Lup to determine the disc parameters. From our model we were able to constrain the stellar parameters (T eff = 4400 K and R = 1.8 R ) and the disc inclination (i = 30 ). However, we were unable to simultaneously fit the shallow 1.3 mm visibilities and the steep 3 mm visibilities assuming a simple disc structure and one grain composition. Our best fit to the ATCA data between 3 to 15 mm corresponds to a large disc with R out 250 au, M dust = M and a maximum grain size of 1 cm, however this model underestimates the flux from 60 µm to 1 mm. In contrast the 1 mm SMA data is fitted by a dust disc

5 iii mass an order of magnitude lower with R out = 50 au. This model underestimates the flux beyond 3 mm and also does a poor job in fitting the mid-infrared emission. The mismatch between the 1.3 mm and 3 mm visibilities and the modelling leads us to propose a two disc scenario for the GQ Lup system, comprising of a compact circumprimary disc 100 au in size and an extended circumbinary disc with R in < 500 au and R out 700 au, where the companion is in the gap between the circumprimary disc and circumbinary disc. This thesis provides an important dataset of southern hemisphere young stellar objects that can be followed-up by Atacama Large Millimetre Array, which has the resolution to observe the inner 10 au of these discs at sub-mm wavelengths up to 3 mm. Additionally, it provides evidence that grain growth up to mm sizes and the presence of excess emission at 7 mm and beyond are common in classical T Tauri star systems, and to disentangle the thermal dust emission from other contributing emission mechanisms one needs to conduct temporal flux monitoring across a range of timescales (e.g. days, months and years). Finally, it also provides evidence that the large grain population can have a significant contribution to the emission at the longer millimetre wavelengths.

6 iv

7 Acknowledgements v This thesis would have not been completed without the support of many people. I would first like to thank my supervisor Sarah Maddison, for her wonderful guidance, respect, patience and her honesty throughout my PhD. This research could not have been completed without her help. I would also like to thank my other two co-supervisors Christopher Wright and Baerbel Koribalski, for their comments and advise. I thank my collaborators David Wilner, David Lommen and François Menard, for their comments on the paper and on the GQ Lup results and Chris Blake for his amassing explanation of Bayesian statistics. Special thanks to the Australia Telescope Compact Array staff, and to Jamie Stevens and Emil Lenc for helping me with my miriad problems. My PhD is based on ATCA data taken over 4 years, and I am grateful for all the help and advise I received along the way, to get the best results from my data. Thanks also to Philip Edwards for awarding some discretionary time to help complete this project. This research was supported by Swinburne University Postgraduate Research Award and in part by a CSIRO OCE Postgraduate Top Up Scholarship. The Australia Telescope is funded by the Commonwealth of Australia for operation as a National Facility managed by CSIRO. This work was performed on the swinstar supercomputer at Swinburne University of Technology. A huge thanks goes to my local support group. Francesco, the other member of the planets group, with whom I started this four year journey. Thanks for all your support and positive thinking. To my friends Christina, Genevieve, Helga, Sam, Elisa, Guido, Ben and Lina. Thank you so much for all your support, for all the laughs, for the good times and the chocolate. Thanks to all of the other students, postdocs and staff whose paths I crossed during my time at Swinburne University. I had a lot of fun here, thank you, and I sincerely hope that our paths intersect again in the future. Lastly, a big thanks to my family and friends, who have supported me from over 9,490 miles (15,287 km). I know it has been hard, so thanks for all your support and for believing in me. Muito obrigada por tudo.

8 vi

9 Statement of originality vii The work presented in this thesis has been carried out in the Centre for Astrophysics & Supercomputing at the Swinburne University of Technology between 2009 and This thesis contains no material that has been accepted for the award of any other degree or diploma. To the best of my knowledge, this thesis contains no material previously published or written by another author, except where due reference is made in the text of the thesis. The content of the Chapters listed below has appeared in refereed journals. Minor alterations have been made to the published papers in order to maintain argument continuity and consistency of spelling and style. Part of Chapter 2 has been published in Monthly Notices of the Royal Astronomical Society: C. Ubach, S. T. Maddison, C. M.Wright, D. J. Wilner, D. J. P. Lommen, B. Koribalski Grain Growth Signatures of Protoplanetary discs of Chamaeleon and Lupus, 2012, MNRAS, 425, 3137 Catarina Ubach Melbourne, Australia February 2014

10 viii

11 Acronyms ix ALMA Atacama Large Millimetre Array ATCA Australia Telescope Compact Array ATNF Australia Telescope National Facility c2d From Molecular Cores to Planet-Forming Disks CABB Compact Array Broadband Backend CASS CSIRO Astronomy and Space Science CARMA Combined Array for Research in Millimeter-wave Astronomy CPU central processing unit CRIRES CRyogenic high-resolution InfraRed Echelle Spectrograph CSIRO Commonwealth Scientific and Industrial Research Organisation CTIO Cerro Tololo Inter-American Observatory ESO-TIMMI2 Thermal Infrared MultiMode Instrument 2 (on the ESO 3.6-m telescope at La Silla) GAIA Global Astrometric lnterferometer for Astrophysics GI gravitational instability GPU Graphic Processing Unit gstar Supercomputer for Theoretical Astrophysics Research IRAC Infrared Array Camera IRAS Infrared Astronomy Satellite ISM interstellar medium ISO Infrared Space Observatory LABOCA LArge BOlometer CAmera (on the 12-m APEX telescope)

12 x MIPS Multiband Imaging Photometer for Spitzer MMSN minimum mass solar nebula PACS Photodetector Array Camera and Spectrometer PAHs polycyclic aromatic hydrocarbons SED spectral energy distribution SEST Swedish-ESO 15m Submillimeter Telescope SMA Submillimeter Array swinstar Swinburne Supercomputer for Theoretical Academic Research 2MASS Two Micron All Sky Survey USNO U.S. Naval Observatory VLA Very Large Array VLBI Very Long Baseline Interferometry VLT ESO Very Large Telescope WFI Wide Field Imager YSOs Young stellar objects

13 xi

14

15 Contents Abstract Acknowledgments Declaration Acronyms, Abbreviations and Conventions List of Figures List of Tables i iv vi viii xv xviii 1 Introduction Protoplanetary Discs From grains to planets Observations of protoplanetary discs Infrared (Sub)millimetre: 850 µm to 3 mm mm to centimetre Thesis goals Survey sample Methodology Our Approach The instruments For observations: ATCA For modelling Thesis outline Signatures of Grain Growth Published Work Introduction Observations and data reduction Sample Observations and data calibration Analysis and results Which sources are resolved at 3 and 7 mm? xiii

16 xiv Contents Source fluxes Millimetre spectral slopes Emission mechanisms at longer wavelengths Additional Sources Discussion Dust opacity index at 3 mm Dust disc masses Millimetre-sized grains? Correlating grain growth signatures Conclusions A Observing details B Data reduction details C Result details D Monitoring details E Strength and Shape A multi-wavelength study of the GQ Lup system Introduction GQ Lup Observations Results GQ Lup Modelling of GQ Lup Model setup Parameters Results GQ Lup discussion Conclusions A The amplitude uncertainty derivation B uvamp versus uvplt C Fitting results D Model SED results E DK Cha system E.1 Observations and data reduction E.2 Results

17 Contents xv 4 Conclusions and Future Directions Major findings and conclusions of this thesis Future work Bibliography 152 Appendices A Comments on individual sources 165 A.1 Chamaeleon A.2 Lupus

18

19 List of Figures 1.1 Infrared classification of young stellar objects Schematic of a protoplanetary disc Theoretical model of the 10 µm silicate feature from Bouwman et al. (2001) Blackbody spectrum at various temperatures Study approach Decision tree Picture of ATCA in H75 configuration Visibilities for sources observed at 3 mm Visibilities for Chamaeleon and Lupus sources observed at 7 mm Maps for RXJ , IK Lup, ATCA and MY Lup Histogram of the 1-3 mm spectral slopes for the survey sample Millimetre flux versus wavelength for Chamaeleon and Lupus sources Plots of the temporal monitoring of 15 mm band Plots of the temporal monitoring of 3+6 cm band WW Cha results KG 49 results at 44 GHz CU Cha results at 94 and 44 GHz Plots of the 1 3 mm millimetre spectral slope as function of 10 µm strength and shape B.1 Flow chart of calibration method B.2 Bandpass calibration correction B.3 Gain calibration correction C.1 Spectral energy distribution for multi-wavelength survey sample E.1 Plots of the third-order polynomial fit to the infrared continuum Visibilities for GQ Lup at 3 mm using uvamp Visibilities for GQ Lup at 3 mm using uvplt Visibilities for GQ Lup at 7 mm Visibilities for GQ Lup at 15 mm Clean Maps of GQ Lup at 3 mm GQ Lup u-v coverage maps at 3, 7 and 15 mm Clean Maps of GQ Lup at 7 mm Clean Maps of GQ Lup at 15 mm Temporal monitoring of GQ Lup at 3, 7 and 15 mm xvii

20 xviii List of Figures 3.10 Millimetre SED slope for GQ Lup Dai et al. (2010) models for GQ Lup with data from the literature and ATCA 3 and 7 mm data Updated GQ Lup Spectral energy distribution Best fit to the stellar photosphere for GQ Lup Best fit models for GQ Lup (a) Best SED fit. χ 2 SED (b) Best 3 mm visibility fit. χ 2 3mm (c) Best 1 mm visibility fit. χ 2 1mm (d) Model with the minimum χ 2 total Overlaid GQ Lup millimetre visibilities Proposed structure for the GQ Lup system B.1 u-v distance vs time for ant(1)(5) at 3 mm B.2 uvamp plot for baseline (1)(5) B.3 Baseline pairs for GQ Lup at 3 mm in July D.1 Model SED results (a) Best SED fit (b) Best 3 mm visibility fit (c) Best 1 mm visibility fit (d) Model with the lowest χ 2 total E.1 Schematic of the YSO DK Cha E.2 Visibilities for DK Cha at 3 mm E.3 Clean Maps of DK Cha at 3 mm E.4 Visibilities for DK Cha at 7 mm E.5 Clean Maps of DK Cha at 7 mm E.6 Visibilities for DK Cha at 15 mm E.7 Clean Maps of DK Cha at 15 mm E.8 Temporal monitoring of DK Cha at 7 and 15 mm E.9 Millimetre SED of DK Cha

21 List of Tables 1.1 Complete survey sample Temporal monitoring sub-sample Decision table for resolved sources Source list for the multi-wavelength survey Survey results: 3 and 7 mm Survey results: 15 mm and 3+6 cm Spectral slopes for the survey sample Flux values for WW Cha Flux values for CU Cha Dust opacity indices and dust disc masses for the survey sample Signatures of grain growth in the IR and mm for Chamaeleon, Lupus, ρ Ophiucus and Taurus-Auriga sources Correlation results between infrared and millimetre grain growth signatures. 69 2A.1 Observation log for the multi-wavelength survey for Lupus sources A.2 Observation log for the multi-wavelength survey for Chamaeleon sources A.3 Observation log for the additional sources C.1 Complete results summary of 3 and 7 mm flux fittings C.2 Flux values for all Lupus sources C.3 Fluxes for all Chamaeleon sources D.1 Temporal monitoring results for the 7 mm band D.2 Temporal monitoring results for the 15 mm band D.3 Temporal monitoring results for the 3+6 cm bands Basic properties of GQ Lup Observation log for the GQ Lup GQ Lup millimetre results GQ Lup 7 millimetre absolute flux calibrator Average millimetre band fluxes of GQ Lup Photometric data used for our GQ Lup modelling Parameters grid for GQ Lup model Parameter values for the best three models C.1 Summary of 3, 7 and 15 mm flux fitting for GQ Lup E.1 Basic properties of DK Cha E.2 Observation log for the DK Cha xix

22 xx List of Tables 3E.3 Summary of 3, 7 and 15 mm flux fittings for DK Cha E.4 DK Cha millimetre fluxes E.5 DK Cha average millimetre fluxes

23 1 Introduction It is generally accepted that planet formation occurs in protoplanetary discs that surround young stars which result as a byproduct of these processes. The gas and dust contained in these discs are studied through observations and modelling to obtain an understanding of the multi-step grain growth process which results in planetary bodies. Observational signatures of grain growth can be obtained from the infrared and millimetre bands, which probe different regions of the disc and are sensitive to different grain sizes. Since thermal dust emission is dominated by dust grains whose size is similar to the observing wavelength, the 3 15 mm bands are probing larger grain sizes (mm cm) than the infrared bands (subµm µm) (Draine, 2006). However, observations at wavelengths greater than 3 mm may detect more than just thermal dust emission from large grains. The aim of this thesis is to study the first stages of planet formation by disentangling the emission mechanisms present in protoplanetary discs at mm and cm wavelengths, to see if signatures of grain growth to cm-sized pebbles can be detected. 1.1 Protoplanetary Discs Young stellar objects (YSOs) are young stars surrounded by accretion discs. The protoplanetary disc results from the conservation of angular momentum of a rotating collapsing cloud of gas and dust during the star formation process (McKee & Ostriker, 2007). Protoplanetary disc surround two types of YSOs: low mass (M < 2 M ) T Tauri stars and intermediate mass (2 10 M ) Herbig Ae/Be stars. The discs around YSOs have lifetimes from 1 10 Myr, masses ranging between M (Dullemond et al., 2007; Williams & Cieza, 2011) and can extend to several 100 au (Nakamoto & Nakagawa, 1994; Andrews & Williams, 2007a; Akimkin et al., 2012). 1

24 2 Chapter 1. Introduction Lada (1987) proposed 3 classes (Class I III) for YSOs, based on the near-infrared spectral slope between 2 and 20 µm, given by α IR = dlogν F ν dlogν = dlogλ F λ dlogλ, (1.1) where F ν is the flux at frequency ν and F λ is the flux at wavelength λ. Class I objects are defined as having α IR > 0.3, Class II having 1.6 < α IR < 0.3 and Class III objects having very weak infrared excess with α IR < 1.6 (Lada, 1987). The introduction of sub-millimetre telescopes and higher sensitivity infrared telescopes has led to the detection of fainter YSOs, resulting in two additional classes: Class 0 (Andre et al., 1993) and flat spectrum sources (Greene et al., 1994). These larger surveys also led to the classification of transition discs YSOs which lack infrared excess at λ < 25 µm (Strom et al., 1989). Additional physical properties and observational characteristics for each of the classes will now be presented in evolutionary sequence. Class 0 objects are protostars still fully embedded in their parental cloud of gas and dust, with ages < 10 4 years. The emission from these young protostars does not escape the cloud and thus there is no 2 20 µm emission (Andre et al., 1993). These objects were first detected in sub-millimetre surveys using the 15 m James Clerk Maxwell Telescope on Mauna Kea, Hawaii (Andre et al., 1993). Class I objects are still embedded and have ages 10 5 years. The envelope continues to collapse, associated outflow are present and the mid-infrared emission dominates over the blackbody radiation of the protostar see Fig Flat spectrum sources are YSOs with flat spectral slopes: 0.3 < α IR < 0.3 (Greene et al., 1994). Class II objects are T Tauri stars which are no longer embedded. These YSOs are 10 6 years and are generally accreting and have strong H α and UV emission (Barrado y Navascués & Martín, 2003; White & Basri, 2003). Transition discs are YSOs characterised by a dip in their spectral energy distribution (SED)s at mid-infrared wavelengths (< 10 µm) and have emission excess between µm. The deficit is associated with a gap in the disc, while the excess is associated with puffed up rim of the disc at the edge of the gap (e.g. Marsh & Mahoney, 1992; Calvet et al., 2005; Arnold et al., 2012). Class III objects have ages of 10 7 years. They have very weak or no H α emission, indicating that accretion has ended (Barrado y Navascués & Martín, 2003), and thus are considered to host passive discs only reprocessing stellar radiation. For a summary of each of YSO class see Fig This thesis focuses on grain growth signatures in Class II objects. These objects have a central star surrounded by gas+dust protoplanetary disc and are isolated from their parent clouds, unlike Class I objects where disentangling the collapsing cloud core from

25 1.1. Protoplanetary Discs 3 Log(λFλ) Class 0 Cold Blackbody Age < 104 years No optical, 2-20 μm emission Only submm emission Sub-mm Log(λFλ) Class I Infrared excess Blackbody Age 105 years αir > 0.3 Outflow present Infrared excess -0.3 <αir < 0.3 Outflow present Blackbody Disc Log(λFλ) Log(λFλ) Flat spectrum Class II Age 106 years -1.6 < αir < -0.3 Accreting disc (Strong Hα & UV) Blackbody Log(λFλ) Transition disc Disc Lack IR excess at λ < 25 μm Flux excess μm Blackbody Log(λFλ) Class III Disc? Stellar Blackbody Age 107 years αir < -1.6 Passive disc No/weak accretion λ (μm) Figure 1.1: Infrared classification of YSOs. On the left, the characteristic SEDs are shown. In the middle, an overview of the expected near-infrared SED spectral slope, physical properties and observational characteristics are provided. On the right, schematic representations of the geometry of the systems are given. Adapted from Andre & Montmerle (1994).

26 4 Chapter 1. Introduction the disc emission is difficult. Grain growth is also an ongoing process during the Class II phase, which might not the case for Class III objects, where the grains are assumed to have grown to sizes larger than centimetre, though small grains are replenished by collisions between larger bodies. 1.2 From grains to planets A protoplanetary disc is initially composed of gas and dust grains of sub-micron to micron sizes. Planet formation is a multi-stage process by which sub-micron size grains grow to planets (objects 12 orders of magnitude larger in size). Here we will discuss the different mechanisms of growth depending on five object sizes: (1) sub-micron sized dust grains, (2) cm-sized pebbles, (3) km-sized planetesimals, (4) 500 km-sized protoplanets, and (5) 500 km-sized planets. Sub-micron sized grains are coupled to the gas and dominated by Brownian motion, which can lead to low speed collisions and growth through coagulation and Van der Waals electrostatic forces (Safronov & Zvjagina, 1969; Dullemond & Dominik, 2005). Laboratory experiments conducted by Güttler et al. (2010) of particles between µm have shown that there are nine possible collision types between grains and that for realistic collision velocities (< 100 cm s 1 ) growth is possible between compact aggregates and porous compact aggregates. As the grains coagulate to millimetre and cm sizes, they start to decouple from the gas and settle towards the mid-plane due to the aerodynamical drag force acting between the gas and dust. The settling of mm and cm-sized grains leads to grain size sorting, where grain size increases with decreasing height above the disc mid-plane (Weidenschilling, 1977; Dullemond & Dominik, 2005; Laibe et al., 2008). Models have also suggested that grains will experience fast radial migration due to drag (Dullemond & Dominik, 2005; Laibe et al., 2008). This radial migration leads to an accumulation of mm and larger grains in the inner disc (Laibe et al., 2008). Recent observations using the Submillimeter Array (SMA) and Combined Array for Research in Millimeter-wave Astronomy (CARMA) have provided observational evidence of a radial change of the dust properties, suggesting a radial grain size segregation consistent with models (Isella et al., 2010; Pérez et al., 2012). Once grains reach decimetre to kilometre-sizes, collisions between them can lead to fragmentation or bouncing rather than growth (Güttler et al., 2010). Additionally, the radial drift timescales experienced by decimetre to metre-sized objects are expected to be much less than the grain growth timescales to mm-sized objects, and thus grains are expected to be rapidly accreted to the star (Weidenschilling & Cuzzi, 1993). Laibe et al.

27 1.3. Observations of protoplanetary discs 5 (2012) modelled the dynamics of dust grains in protostellar discs, and found that the right combination of surface density and temperature profiles can be sufficient to halt the inward migration of grains, thereby eliminating this meter-size barrier (or radial-drift barrier ) problem. Laboratory experiments and dust collision models have shown that if the majority of the grains are below cm sizes, the larger grains can sweep up the smaller grains and grow via a combination of the bouncing barrier and fragmentation with mass transfer (Windmark et al., 2012). Once km-sized planetesimals form, there are two models for the transition from protoplanets to planets: the core accretion model and the gravitational instability (GI) model (Cameron, 1978; Armitage, 2009). In the core accretion model the planetesimals grow via direct collisions. This will increase as their mass becomes larger, leading to runaway growth of a few bodies to form the cores of planets. If the cores reach a critical mass 10 M, they start to rapidly accrete gas from the nebula (Mizuno, 1980) to form giant planets, until a gap is opened in the protoplanetary discs which will then slow the growth (Armitage, 2009). The gap also produces pressure bumps where dust can pile up, creating a favourable region for further grain growth (Armitage, 2009). The core accretion model has some limitations. The timescales are to long compared to disc dissipation timescales, thus gas giants would not form and planetesimals might be accreted by the star before growing to protoplanets (Mizuno, 1980). The GI model assumes early in their evolution the protoplanetary disc is very massive and cools rapidly, and disc self-gravity triggers fragmentation leading to direct collapse of nebula gas forming giants planets (Cameron, 1978; Kuiper, 1951). Boss (2011) modelled a range of YSOs and suggested that disc instabilities are capable of forming 1 5 M Jup planets beyond 30 au. The GI model is limited in that is requires a massive, cold disc for fragmentation to occur (Cameron, 1978). This however does not rule out the possibility of the presence of a hybrid mechanism of GI and accretion ( core assisted plus gas capture ) (Boley & Durisen, 2010; Boley et al., 2011), and so the debate continues. This work we will focus on the first stages of the planet formation process, looking specifically at the phase where sub-micron-sized grains grow to cm-sized pebbles. 1.3 Observations of protoplanetary discs: From Infrared to Centimetre The presence of sub-micron grains is indirectly determined through observations of light scattering (e.g. Paresce & Burrows, 1987; Kalas et al., 2006; Pinte et al., 2008) and

28 6 Chapter 1. Introduction observations of thermal dust emission at IR and mm wavebands. The IR regime probes the warm inner surface of the disc, providing information on sub-micron to micron-sized grains, while the mm and cm regimes probe the cooler outer regions as well as the mid-plane of the disc where the bulk of the dust resides see Fig To understand the distribution of the dust phase in protoplanetary discs, multi-wavelength observations are therefore needed. In this section we discuss the grain growth signatures obtained from the detection of thermal dust emission from the IR to cm bands infrared Height [Z/R] millimetre & centimetre R [AU] Figure 1.2: Schematic of a protoplanetary disc. The grey areas show the gas within one (dark) and two (light) pressure scale heights. The black areas show the approximate region of the infrared emission. The millimetre and centimetre emission, tracing the larger dust grains, originates mainly from the mid-plane of the outer disc. Note that the origin of the observed infrared emission strongly depends on the orientation of the disc. Original image adapted from Dullemond & Monnier (2010) and Mulders & Dominik (2012), with further improvements by Dave Lommen. Published in Ubach et al. (2012) Infrared The infrared band is used to probe dust properties such as grain composition and sizes in protoplanetary discs. Grain sizes have been inferred through the analysis of the strength and shape of the 10 µm silicate feature, where the strength can be defined by the peak of the emission and the shape by the flux ratio F 11.3µm /F 9.8µm. Bouwman et al. (2001) modelled 10 µm silicate profile of a sample of Herbig Ae/Be stars using a radiative transfer code with three different components (silica, forsterite and amorphous olivine). They found the change in 10 µm spectral shape was mainly due to grain growth,

29 1.3. Observations of protoplanetary discs 7 where submicron-sized grains have a strong triangular shaped 10 µm feature, while larger micron-sized grains have a weaker and broader 10 µm feature see Fig They also observed a shift in the peak of the 10 µm silicate feature, which could be caused by either grain growth from 0.1 to 2.0 µm, a change in composition from amorphous silicates to a mixture of amorphous silicates and forsterite, or a combination of the two. Crystallinity was also found to play a minor role in the correlation between grain size and peak position of the silicate feature, in agreement with Molster et al. (1999) who suggested that grain growth via coagulation occurs on shorter timescales than the processing of crystalline grains. The distinct shape of the 10 µm feature has also been observed in other surveys of Herbig Ae/Be stars (van Boekel et al., 2003) and in T Tauri stars (Przygodda et al., 2003). Figure 1.3: Theoretical model of the shaped 10 µm silicate feature from Bouwman et al. (2001). Small sub-micron-sized grains have a strong triangular 10 µm spectra, while larger micron-sized grains have a weaker and broader 10 µm feature. One of the main disc studies with the Spitzer Space Telescope was the From Molecular Cores to Planet-forming Discs (c2d) legacy survey, which obtained photometry and spectra from 3.6 to 70 µm for YSOs in nearby star forming regions (Evans et al., 2003; Young et al., 2005; Chapman et al., 2007; Luhman, 2008, 2007). These surveys used the Infrared Array Camera (IRAC) and Multiband Imaging Photometer for Spitzer (MIPS)

30 8 Chapter 1. Introduction instruments to obtain a census of YSOs at different evolutionary stages, and identified stars with infrared excess corresponding to the presence of a protoplanetary disc. Several star forming regions were observed and hundreds of YSOs detected. One of the main proposes of c2d was to study grain growth from subµm µm-sized grains, to provide evidence of the settling of growing grains and to provide the necessary infrared observations to help constrain disc models. Kessler-Silacci et al. (2006) analysed the strength and shape of the 10 µm silicate features of 40 T Tauri stars in Lupus, Chamaeleon, Taurus and ρ Ophiuchus, and found evidence of grain growth from 0.1 to 1.0 µm. They conducted a similar analysis of the 20 µm silicate feature, which was not previously possible due to the confusion between the rising continuum and the emission feature. They found a large 10-to-20 µm flux ratio for a sub-set of the sample, which could only be reproduced by models with large (2 3 µm) grains emitting at 20 µm and smaller (0.1 µm) grains emitting at 10 µm. Because the two silicate features probe different temperatures and hence different depths within the disc, this result provides evidence of settling of large (> 2 µm) µm-sized grains. The results from these surveys would suggest that the strength and shape of the 10 µm silicate feature traces grain growth from µm, while the 20 µm silicate feature traces grain growth from µm. In this work, we use the shape and strength of the 10 µm silicate feature to provide information about grain growth from subµm µm sizes in the surface layers of the protoplanetary discs and compare these to longer wavelength grain growth signatures in the mm bands (Sub)millimetre: 850 µm to 3 mm Sub-mm and mm wavelengths probe the cooler outer regions compared to the IR (Fig. 1.2). This emission is assumed to be dominated by thermal dust emission, which follow the properties of a blackbody an object which absorbs all electromagnetic radiation and is at thermal equilibrium. The spectrum of a blackbody follows Planck s law, B ν (T ) = 2hν 3 /c 2 exp(hν/kt ) 1, (1.2) where h = Js is Planck s constant, c = m s 1 is the speed of light, k = m 2 Kg s 2 deg 1 is Boltzmann s constant, T is the temperature (in K), and ν is the frequency (in Hz). A full derivation of Planck s law and its properties can be found in Kraus (1966) and Rybicki & Lightman (1979). Figure 1.4 presents the blackbody spectrum from 1 K to 10 8 K. When hν << kt, Eq 1.2 simplifies to B ν (T ) = (2ν 2 /c 2 )kt. This is known as Rayleigh-Jeans law and is

31 1.3. Observations of protoplanetary discs 9 represented as the straight line in Fig At (sub)mm wavelengths, observations trace the optically thin emission in the Rayleigh-Jeans regime (hν << kt ). Following Eq. 1.2, the emission in the (sub)mm bands is weaker than that in the infrared band, making large scale surveys at (sub)mm bands time consuming. Sub-mm surveys targeting specific star forming regions have been performed, e.g. Chamaeleon using Swedish-ESO 15m Submillimeter Telescope (SEST) (Henning et al., 1993), Lupus using SMA and Australia Telescope Compact Array (ATCA) (Nuernberger et al., 1997; Lommen et al., 2010) and Taurus-Auriga using SMA and CARMA (Andrews & Williams, 2005). ν (Hz) B ν B ν (erg sec -1 cm -2 Hz -1 ster -1 ) B ν (erg sec -1 cm -2 Hz -1 ster -1 ) B ν ν (Hz) λ (cm) Figure 1.4: Blackbody spectrum from 1 K to 10 8 K (taken from Kraus 1966). To detect signatures of grain growth up to mm sizes, the 1 3 mm slope of the SED α, where F ν α, must be determined. From this, the dust opacity index β (where κ ν ν β ) can be obtained via β α 2, where β < 1 indicates grain growth to mm sizes (Draine, 2006). However, care must be taken when relating α and β, as shallow spectral slopes can result from an extended optically thin disc or a compact optically thick disc (Beckwith & Sargent, 1991). To break this degeneracy the emission has to be spatially resolved at mm wavelengths.

32 10 Chapter 1. Introduction Approximately 30 discs have now been resolved at 3 mm, with 20 having a dust opacity index β < 1 (Natta & Testi, 2004; Lommen et al., 2007, 2009, 2010; Ricci et al., 2010a,b; Guilloteau et al., 2011; Banzatti et al., 2011; Ubach et al., 2012; Pérez et al., 2012; Akimkin et al., 2012), indicating that the dust has reached mm sizes (Draine, 2006). Lommen et al. (2007) obtained 1 mm fluxes with SMA and 3 mm fluxes with ATCA for sources in Chamaeleon and Lupus and found that the majority of the sources have a dust opacity index β consistent with mm-sized grains (β < 1). Lommen et al. (2010) expanded on their 2007 work by obtaining SMA, CARMA, ATCA and Very Large Array (VLA) continuum fluxes at 1, 3 and 7 mm for 31 massive discs around T Tauri stars across five star forming regions, and obtained similar results. Both of the Lommen et al. surveys, however, suffered from a bias due to a small number of bright sources, with detection rates of only 66% and 25% at 3 and 7 mm respectively. Sub-mm and mm observations can also provide estimates of the dust mass at the observing wavelength. The mass is determined via M disc = F ν D 2 κ ν B ν (T dust ), (1.3) where F ν is the observed flux at frequency ν and D is the distance to the source. κ ν is the dust opacity at frequency ν, estimated by Beckwith et al. (1990) to be 0.02 cm 2 g 1 at 1.3 mm assuming a dust opacity index β = 1 and the standard gas-to-dust ratio of the interstellar medium (ISM) of 100. B ν (T dust ) is the brightness for a dust temperature T dust (set to 25 K in Beckwith et al. (1990) and Andre & Montmerle (1994)) as given by the Planck function. The disc mass can then be compared to the minimum mass solar nebula (MMSN) the minimum mass required in a protoplanetary disc to create the planets in our solar system. This value is between 0.01 M (Weidenschilling, 1977) and 0.02 M (Hayashi, 1981). Using Eq. 1.3, Nuernberger et al. (1997) found statistical evidence that M disc /M decreases with increasing stellar mass for a sample of protoplanetary discs in the Lupus 2 and 3 clouds, and determined that the disc masses for that sample were less than the MMSN value of 0.01 M. A similar result was found by Andrews & Williams (2005) for a sample of 24 resolved discs in Taurus-Auriga at 1.3 mm. These results suggest that the star forming regions contain YSOs which might not be capable of forming planets. On the other hand, Henning et al. (1993) found the disc masses at 1.3 mm in Chamaeleon to be on average 0.01 M (assuming κ ν of 0.02 cm 2 g 1 ), which implies some discs have enough mass to create planets. These results, however, assume that the gas-to-dust ratio does not change as the disc

33 1.3. Observations of protoplanetary discs 11 evolves (via κ ν in Eq. 1.3). This has been shown by D Alessio et al. (2006) to not be the case. By modelling dust settling and grain growth, they found that the gas-to-dust ratio needs to be >10% of the standard ISM value for the small grains in the upper layers of the disc, and the ratio decreases for the interior of the disc due to dust settling. Thus a more appropriate measure of κ ν is κ ν = 10(ν/10 12 Hz) β cm 2 g 1, (1.4) which is independent of the gas-to-dust ratio (Beckwith et al., 1990). When substituted into Eq. 1.3, this provides the mass of the dust (M dust ) in the disc rather than the total disc mass. With some knowledge of the gas-to-dust ratio, the total disc mass can then be calculated by multiplying by the gas-to-dust ratio. An analysis of the mm and IR grain growth signatures has suggested a tentative correlation between the 10 µm silicate feature and the 1 3 mm spectral slope, α, for a sample of 30 sources in the Chamaeleon and Lupus star forming regions, which suggest the simultaneous evolution of dust in the inner and outer part of the disc (Lommen et al., 2010). However, Ricci et al. (2010b) found no such correlation for sources in ρ Ophiuchus and Taurus. The 10 µm spectral feature and α come from observations probing different layers of the disc (IR: hot inner upper layer, mm: cold mid-plane), and if confirmed this correlation could imply that grain growth in the inner upper layers of the disc affects grain growth in the mid-plane. This thesis will explore this potential correlation mm to centimetre Signatures of cm-sized pebbles have been observed in protoplanetary discs by extending the dust opacity index relationship (β α 2) to 7 mm and centimetre wavelengths (e.g. Testi et al., 2003; Rodmann et al., 2006; Wilner et al., 2005). However care must be taken as this approach assumes the long wavelength emission comes from thermal dust emission. T Tauri stars are complex systems consisting of an active young star surrounded by a dusty disc and are often accompanied by stellar winds and outflows. Resolving the disc will not necessarily break the degeneracy between large optically thin discs and compact optically thick discs, as compact discs may also have extended ionised winds which emit thermal free-free emission. This means that thermal dust emission is probably not the only emission mechanism detected at longer wavelengths and it may not be the dominant emission mechanism, and in most cases may be too weak to be detected. At 7 mm, both thermal dust emission and thermal free-free emission from an ionised wind have

34 12 Chapter 1. Introduction been detected in YSOs (e.g. Rodmann et al., 2006; Lommen et al., 2009), while at 3 and 6 cm a combination of thermal dust emission, thermal free-free emission and non-thermal chromospheric emission have been detected (e.g. Lommen et al., 2009). Thermal free-free emission (thermal Bremsstrahlung) is essentially unpolarized emission which occurs when two free charged particles interact and are accelerated. For a spherically symmetric, constant velocity wind, there are two possible spectral slopes for the resulting thermal free-free emission depending on the optical depth. For an opaque free-free wind, α ff = 0.6 (Panagia & Felli, 1975), while for an optically thin wind, α ff = 0.1 (Mezger et al., 1967). A full derivation can be found in Panagia & Felli (1975). On the other hand, for a well-collimated ionised wind α ff < 0.6 (Reynolds, 1986). Rodmann et al. (2006) analysed the spectral slope from millimetre through to centimetre wavelengths for a sample of 6 YSOs in Taurus-Auriga and determined that 20% of emission at 7 mm was from thermal free-free emission. Similarly, Lommen et al. (2009) determined the presence of thermal free-free emission in RU Lup at wavelengths longer than 1.2 cm. Another property of thermal free-free emission is the temporal flux variability (e.g. González & Cantó, 2002). Skinner & Brown (1994) monitored the flux of T Tau N using the VLA at 3.6 cm and found that the flux varied by less than 10% on the timescale of years and detected no polarised emission, suggesting the presence of thermal free-free emission. Smith et al. (2003) observed T Tau S and found flux variability on timescales of years by a factor of 10% at 3.6 cm. Likewise, temporal monitoring over a period of two years of WW Cha at 3 and 6 cm has also found flux variability by a factor of 20 50%, suggesting the presence of thermal free-free emission at 3 and 6 cm (Lommen et al., 2009). Chromospheric emission (due to synchrotron or magneto-bremsstrahlung radiation) is polarised radiation caused by particles accelerating in a magnetic field (e.g. Rybicki & Lightman, 1979). In radio astronomy the spectral slope α cm associated with synchrotron radiation is α cm 0.1 (e.g. Anglada et al., 1998). This allows one to differentiate between non-thermal (α cm 0.1) and thermal free-free emission (α ff = 0.6 or α ff = 0.1). Of course, at cm wavelengths, both emission mechanisms can be present in YSOs. It is also possible for non-thermal emission to have a spectral slope α cm = 0.1 (Anglada et al., 1998), and thus additional information is required to differentiate between non-thermal emission with α cm = 0.1 and thermal free-free emission from an optically thin wind. To break this degeneracy one can use two other properties of non-thermal emission: (1) the emission is expected to be polarised, unlike thermal free-free emission (e.g. Rybicki & Lightman, 1979), and (2) non-thermal emission is variable on timescales of hours (Kuijpers & van der Hulst, 1985; Phillips et al., 1991).

35 1.4. Thesis goals 13 These two properties of non-thermal emission have been detected in radio observations of YSOs. The first evidence of non-thermal emission in a classical T Tauri was obtained by Phillips et al. (1993), who conducted simultaneous observations of T Tau N and T Tau S with Very Long Baseline Interferometry (VLBI) and VLA at 18 cm and found the emission to be circularly polarised and variable by a factor of 2 on the timescales of hours, suggesting the presence of non-thermal emission. Additional studies using VLBI and VLA at 2, 3.6 and 6 cm of this system have provided further evidence of polarisation and flux variability by at least a factor of 2 on a timescales of hours to days in T Tau S (Smith et al., 2003; Skinner & Brown, 1994). Similar results for a sample of 9 weak-line T Tauri stars in Taurus-Auriga have been obtained with the VLA at 3.6 cm (Chiang et al., 1996). This means that both spectral slope analysis and temporal flux monitoring are required to disentangle the three potential emission processes present at cm wavelengths, and unfortunately this process is time consuming. Lommen et al. (2009) and Maddison et al. (2010) conducted temporal monitoring of WW Cha and HD using ATCA, and concluded that the stable flux at 16 mm over a period of 2 years for WW Cha and HD was due to the presence of pebble sized grains, indicating that planet formation was well underway in the discs of these two young stars. Prior to these works, the only other source with confirmed cm-sized pebbles in its disc was TW Hya (Wilner et al., 2005). 1.4 Thesis goals The goal of this thesis is to study the first stages of planet formation, by detecting signatures of cm-sized pebbles in the discs surrounding nearby, young, low-mass stars. This requires a study of the signatures of grain growth in a large sample of protoplanetary discs over the millimetre waveband, to determine which emission mechanisms are present, and which discs likely host large pebble-sized grains. Using a multi-wavelength dataset, we ask the following questions designed to probe the nature of the emission from protoplanetary discs: Are other forms of emission present besides thermal dust? If so, what is the dominant emission mechanism? Is the source resolved? Are large grains present? Can a single model simultaneously fit the full SED and mm visibilities? We will build on the Lommen et al. (2007, 2010) 1 and 3 mm dataset and complete the SED from 1 to 7 mm for a sample of T Tauri stars in the Chamaeleon and Lupus southern star forming regions. We will also monitor the longer wavelength emission at 7 and 15 mm and 3+6 cm for variability (or stability) to help determine the contributing The 3 and 6 cm bands are observed simultaneously with ATCA.

36 14 Chapter 1. Introduction emission mechanisms, and resolve sources in the millimetre which show signs of cm-sized pebbles. Finally, we will model the full SED and visibilities for a low-mass T Tauri star which shown signatures of large pebbles sized grains. Our sample includes fainter discs (F 1mm < 100 mjy) than the bright sources studied in Lommen et al. (2007, 2010), which are more representative of the overall disc population. Obtaining fluxes at 3 and 7 mm allows us to evaluate the spectral slope between 1 3 mm, α, and between 3 7 mm, α 3 7, and determine if excess emission above thermal dust emission is present at 7 mm. With the spectral slope we can also calculate the dust opacity, β, which provides information on the maximum grain size, and obtain a more accurate estimate of the dust disc mass (M dust ) of faint discs. Creating a model for a low-mass T Tauri star allows us to determine the physical properties of the protoplanetary disc. 1.5 Survey sample We target 20 T Tauri stars in the Chamaeleon and Lupus southern star forming regions see Table 1.1 for our source list. The sources were selected to overlap with the sample of Lommen et al. (2007, 2010) specifically to complete the detections at 3 and 7 mm, and the Spitzer c2d programme (Young et al., 2005; Alcalá et al., 2008; Merín et al., 2008). Two additional sources were added to the Lommen et al. (2007, 2010) sample. Both GQ Lup and DK Cha, both of which have strong 1.3 mm fluxes (Henning et al., 1993; Dai et al., 2010), which should improve the chance of detection at 3 mm. The Chamaeleon star forming region contains three molecular clouds (Cha I, II and III) located at a distance of pc (Luhman, 2008; Furlan et al., 2009). The Chamaeleon sample includes 9 Cha I cloud sources, the isolated source T Cha, and the Cha II source DK Cha. WW Cha was observed simultaneously with Sz 32, however it was not the primary target of the sample and thus was not included in the analysis (see Lommen et al. (2009) for a detailed study of WW Cha). Lupus is composed of 8 clouds (Lupus 1 8) (Hara et al., 1999). Our sample includes a total of 10 sources from the Lupus 1, 3 and 4 clouds at distances 150±20 pc, 200±20 pc and 165±15 pc respectively (Comerón, 2008), and the isolated source RXJ at a distance of 184 pc (Makarov, 2007). We also present observations of the Herbig Ae/Be star CU Cha and the Cha IR nebula to complete the wavelength coverage from Lommen et al. (2007, 2010). Note that neither CU Cha nor Cha IR nebula were used in the analysis presented in Chapter 2 since neither are T Tauri stars. Further information on individual sources listed in Table 1.1 can be found in Appendix A.

37 1.5. Survey sample 15 Table 1.1: Complete survey sample. The 20 T Tauri stars used in the analysis are numbered from ID Source RA DEC Spectral Cloud Comments References (J2000) (J2000) type Chamaeleon 1 SY Cha M0.5 Cha I 1 2 CR Cha K0 K2 Cha I 1 3 CS Cha K4 Cha I Binary a 1 4 DI Cha G1,G2 Cha I Binary a 1 5 T Cha G2 Isolated 1,2 6 Glass I K4 Cha I Binary a 1 7 SZ Cha K0 G Cha I 1 8 Sz K4.7 Cha I HH jet a 1 9 DK Cha F0 D Cha II Outflow a 3,4 WW Cha K5 D Cha I Observed with Sz 32 1 CU Cha A C Cha I Herbig Ae/Be a 2 KG M5 D Cha I Cha IR nebula a 5,6 Lupus 10 IK Lup K7 D Lupus 1 Binary b 7 11 Sz M2 D Lupus 1 Binary b 7 12 HT Lup K2 Lupus 1 Triple b, HH jet 7 13 GW Lup M2 Lupus GQ Lup K7V D Lupus 1 Planet b 7 15 RY Lup G0V: C Lupus HK Lup M0 Lupus Sz M1.5 Lupus 3 Cold disc b 7 18 EX Lup M0 Lupus MY Lup Lupus RXJ K5 Isolated 9 a See Section A.1 for further details on Chamaeleon sources. b See Section A.2 for further details on Lupus sources. Note that when Chapter 2 results were published, we had not found Hughes et al. (1991) that classified DK Cha as a Herbig Ae star. References: Chamaeleon: (1) Luhman (2004), (2) van den Ancker et al. (1998), (3) Spezzi et al. (2008), (4) van Kempen et al. (2010), (5) Hughes et al. (1991), (6) Feldt et al. (1998). Lupus: (7) Comerón (2008), (8) Mortier et al. (2011), (9) Makarov (2007). (Table from Ubach et al., 2012).

38 16 Chapter 1. Introduction 1.6 Methodology This study consisted of five distinct stages, building at each stage for a more complete understanding of the observed protoplanetary discs. The flow chart in Fig. 1.5 presents the information obtained at each stage of our approach, and Fig. 1.6 presents the questions and possible answers at each stage. Study Approach Stage 1. Obtain fluxes at 3 and 7 mm Obtain mm SED and α & α 3-7 Stage 2. Create maps and visibility plots Obtain disc outer radius limits and determine if mm emission is resolved Stage 3. Calculate β Obtain grain size info Stage 4. Temporal Flux monitoring Constrain emission contribution at 7 & 15 mm Stage 5. Modelling Determine disc and dust properties Figure 1.5: Flowchart of our approach to the study of grain growth in protoplanetary discs. Stage 1: Are other forms of emission present besides thermal dust? We obtain fluxes at 3 and 7 mm for the 20 T Tauri stars in Chamaeleon and Lupus. These fluxes are used to calculate the spectral slopes α and α 3 7, which can then be used to determine if other emission mechanisms are present at 7 mm. Stage 2: Is the source resolved? We create maps and visibility plots at 3 mm of the 20 T Tauri stars to determine if the sources are resolved. Resolved emission allows us to

39 1.6. Methodology 17 assume that the emission comes from a radially extended optically thin disc. Stage 3: Are large grains present? We estimate the dust opacity index using β α 2 for each source in the sample. This allows us to estimate if large grains are present in the resolved emission. Stage 4: What is the dominant emission mechanism? We temporally monitor the flux of a sub-sample of sources. The magnitude and timescale of any flux variability is used to constrain the contributing emission mechanisms. Stage 5: Can a single model simultaneously fit the full SED and mm visibilities? We model a source in our sample with signs of pebble-sized grains using a detailed radiative transfer code, allowing us to further constrain disc and dust properties, such as disc outer radius, grain size ranges, disc surface density and flaring exponents, disc inclination and dust settling Our Approach Measuring the fluxes at 3 and 7 mm for all the sources in our sample provides the millimetre data required to calculate the spectral slopes α and α 3 7, create clean maps and visibility (baseline length vs amplitude) plots, and determine β values thus addressing the first three questions in our methodology (see Fig. 1.6). With these data we are also able to examine other properties of protoplanetary discs: 1. Estimate more accurately the dust mass for faint discs (using Eq. 1.3). 2. Combine the results of these new observations with previous ATCA observations, permitting a temporal monitoring analysis on the timescale of years for sub-sample of sources, which provides information on the emission mechanisms present. 3. Compare the millimetre and infrared grain growth signatures for a larger and more homogeneous sample, to determine if the correlation between the two grain growth signatures exists. A correlation could imply that grain growth in the inner upper layers of the disc affects the grain growth in the mid-plane of the disc. 4. Provide an important dataset of protoplanetary discs for follow-up observations with Atacama Large Millimetre Array (ALMA). ALMA provides high-angular resolution images to investigate the inner disc structure and radially resolve grain properties with multi-wavelength observations. The next step (stage 4) is to conduct temporal flux monitoring at 7 and 15 mm and 3+6 cm to check for flux stability. As discussed in Section 1.3.3, the emission detected

40 18 Chapter 1. Introduction Stage 1. Q: Are other forms of emissions present besides thermal dust? Case A Case B Stage 2. Q: Is the source resolved? I II Stage 3. Q: Are large grains present? Case Ia: Resolved and β < 1 big grains Case Ib: Resolved and β > 1 ISM sized grains Case IIa: Unresolved and β est < 1 potentially big grains Case IIb: Unresolved and β est > 1 ISM sized grains Figure 1.6: Decision tree used in this study. (Continued on the next page.)

41 1.6. Methodology 19 Stage 4. Q: What is the dominant emission mechanism? Stage 5. Q: Can a single model simultaneously fit the full SED and mm visibilities? Constrain disc and dust parameters such as: Inclination (i) Dust disc mass (M dust ) Flaring exponent (β) Maximum grain size (a max ) Surface density exponent (α) Scale height (h) R i n h R out T eff, R Figure 1.6: Continued.

42 20 Chapter 1. Introduction at 7 mm and beyond can result from a range of emission mechanisms: thermal dust emission, thermal free-free emission from an ionised wind, and non-thermal chromospheric emission. In order to disentangle the forms of emission present, multi-wavelength temporal monitoring to determine the level and timescales of flux variability is required. allows us to differentiate between chromospheric emission (variable by 100% or more on hour-to-day timescales, Phillips et al., 1993), free-free emission (variable by 20% on yearly timescales, González & Cantó, 2002), and thermal dust emission from large pebbles (non-varying). Table 1.2: Temporal monitoring sub-sample. Source ID from Table 1.1. The frequency pairs for the observations were as follows: 7 mm band GHz; 15 mm band GHz; 3+6 cm bands GHz. ID Source Millimetre Centimetre bands bands Chamaeleon 2 CR Cha 7,15 3 CS Cha 7,15 4 DI Cha 7,15 5 T Cha 7, SZ Cha 7,15 8 Sz 32 7, DK Cha 7, Lupus 14 GQ Lup 7,15 15 RY Lup 7 17 Sz 111 7,15 19 MY Lup 7,15 20 RXJ This Due to time and source elevation constraints during our allocated observing time, only a subset of our sources were monitored. The temporal monitoring sample is presented in Table 1.2. Priority was given to the brightest sources (requiring the least amount of integration time for a detection) and to sources with previous observations at 7 mm. This allowed us to maximise the number of sources which could be analysed in the available time. The temporal monitoring sample contained both sources with signs of grain growth and sources with signs of excess emission. Using the results from stages 1 4 of our methodology, we can draw conclusions about the grain size and emission mechanisms present for each source. Table 1.3. presents a set of scenarios from which we can make conclusions about each source. First we can check the slope of the SED see stage 1 of Fig 1.6. In case B, a break in the spectral slope suggests the presence of other emission mechanisms besides thermal dust emission. Next

43 1.6. Methodology 21 Table 1.3: Decision table for resolved sources at 3 and 7 mm. See Fig 1.6 for more details on each case. Cases SED Resolved Flux variability A B Ia Ib Conclusion Large grains + thermal dust ISM size grains + thermal dust Large grains + free-free wind Large grains + non-thermal Large grains + multiple emissions ISM size grains + free-free wind ISM size grains + non-thermal ISM size grains + multiple emissions we can check whether the emission is resolved (stage 2 of Fig. 1.6). In case I, we can assume that the emission arises from a radially extended and optically thin disc and hence relate β and α to make inferences about the maximum grain size (stage 3 of Fig. 1.6). Note that unresolved sources at 3 mm are categorised as case IIa or case IIb at stage 2. Finally we check for long wavelength flux variability (stage 4 of Fig. 1.6), allowing us to determine the dominant emission mechanisms. It is important to note that a source can have signatures of large cm-sized pebbles from 1 7 mm and the presence of multiple emission mechanisms in the centimetre wavebands, further demonstrating the need for temporal flux monitoring to disentangle the emission. Thus far we have concentrated on the information which can be obtained from observations, but there are some disc and dust parameters such as surface density exponent, flaring exponent, disc inclination and dust settling exponent that are not constrained through observations. To obtain some information about these parameters, attempts have been made to fit a model to multi-wavelength datasets. Duchêne et al. (2010) attempted to simultaneously fit a radiative transfer model to optical and infrared images, mm visibilities and the full SED of HV Tau C. They found a small total disc mass (10 5 M ) and ISM-sized grains are needed to replicate the scattered light images, while a mass two orders of magnitude higher and grains up to mm sizes are needed to fit the full SED. They attributed the mismatch to the assumption that the dust is perfectly mixed, and proposed that both grain growth and vertical stratification are present in the HV Tau C disc. Olofsson et al. (2011) also simultaneously fitted the mm visibilities and full SED of T Cha, and were able to constrain the inner disc radius and ruled out the presence of a close companion within the a few 10 au of the star. For both of these studies, these

44 22 Chapter 1. Introduction results were only obtained when a model was simultaneously fitted to the multi-wavelength datasets. The final stage of our approach is to use the 3D radiative transfer code mcfost, which is based on the Monte Carlo method (Pinte et al., 2006), to create a model that simultaneously fits the full SED and 1.3 and 3 mm visibilities of a source in our sample with signatures of pebble-sized grains. Radiative transfer codes determine the radiation transfer between particles within a medium via absorption, emission and scattering processes. mcfost randomly generates a number of photon packets which scatter and absorb as they travel through the computational grid (Pinte et al., 2006). By calculating the path of the photons (ray tracing mode), the code calculates the temperature structure of the disc and then determines the flux at each specified wavelength to create the full SED. The temperature structure is determined so that the dust properties can be properly modelled within the disc. From the SED, the visibilities at the desired wavelength can then be calculated (Pinte et al., 2006). mcfost allows the user to model a multiple component system (such as a disc, envelope and cavity) and explore a large parameter space covering the stellar properties, disc geometry and dust properties. It has been successfully used to model a range of systems including debris discs (Schneider et al., 2006; Kalas et al., 2007), discs around brown dwarfs (Guieu et al., 2007), T Tauri stars (Pinte et al., 2008; Duchêne et al., 2010; Olofsson et al., 2011) and Herbig Ae stars (Tatulli et al., 2008; Benisty et al., 2010). Accurate modelling of the temperature distribution and dust disc mass is crucial to our understanding of protoplanetary disc formation. Since fitting the SED alone will not provide a unique disc model, visibilities at multiple wavelengths are needed to further restrict the models. Thus, the final step of our approach is to find a model that can simultaneously fit the full SED and visibilities of sources of interest. There were two good candidates, GQ Lup and DK Cha, from our multi-wavelength survey which show signs of containing large pebble-sized grains. For these sources, additional temporal flux monitoring at 7 mm and 15 mm was conducted to confirm flux stability at these wavelengths. We also obtained high resolution imaging at 3 mm which provided equal baseline 2 spacing and the signal-to-noise required in the visibilities to discriminate between disc models. Radiative transfer modelling is computationally expensive given the large parameter space, and so we focused our modelling on GQ Lup. GQ Lup was first modelled by Dai et al. (2010), who fitted a 50 au radius parametric disc with a power-law grain size 2 A baseline is the distance between two radio antennas.

45 1.7. The instruments 23 distribution with sizes up to 1 mm to their 1.3 mm SMA data. This model, however, does not account for the longer millimetre wavelengths, which provide information on large grains that affects the model temperature distribution and dust disc mass estimates. The aim of stage 5 is to improve the Dai et al. (2010) model of GQ Lup by modelling the full SED up to 15 mm and the visibilities at both 1 and 3 mm. 1.7 The instruments used in this work For observations: ATCA Figure 1.7: ATCA in H75 configuration, taken on 26 September Our multi-wavelength survey used the only interferometer capable of 3 15 mm band observations in the southern hemisphere: the Australia Telescope Compact Array (ATCA). ATCA is located near Narrabri, New South Wales, Australia, at the Paul Wild Observatory and is operated and managed by CSIRO 3 Astronomy and Space Sciences (CASS). Opened on 2 September 1988, ATCA initially consisted of a 3 km east-west track with 5 movable 22-meter diameter antennas (CA01 to CA05) and a sixth fixed 22-meter diameter antenna (CA06) 3 km from the western end of the track, allowing ATCA to have 15 baselines with a maximum baseline of 6 km (Frater et al., 1992). All antennas were fitted with 3, 6, 12 and 21 cm band receivers. It took three major technical upgrades over the course of 9 years to make ATCA the millimetre instrument that it is today 4. The first major upgrade made ATCA the only millimetre wavelength interferometer in the southern hemisphere, outfitting ATCA s five movable antennas with 3 mm receivers and all six antennas with 12 mm receivers (Gough et al., 2004). This upgrade also added a 214 m north-south spur, enabling compact 3 Commonwealth Scientific and Industrial Research Organisation 4 Such at:

46 24 Chapter 1. Introduction hybrid array configurations (H75, H168 and H214) which allow ATCA to obtain full u-v synthesis in 6 hours (versus 12 hours when using the east-west track). The second upgrade, completed in 2008, added the 7 mm receivers to all six antennas (Moorey et al., 2006, 2008), and the latest upgrade replaced the entire backend system with the Compact Array Broadband Backend (CABB). CABB has a bandwidth of MHz (compared to the previous backend which had a bandwidth of MHz), which increased the continuum sensitivity by a factor of four or decreased the observing time for continuum observations by a factor of 16. See Wilson et al. (2011) for a full description of the CABB upgrade. For our multi-wavelength survey, the standard configuration was the most extended hybrid array H214, which allowed us to spatially resolve some of the discs with its beam size of 2.3 at 3 mm. This is 300 au at the distances of the star forming regions of Chamaeleon and Lupus. The hybrid configuration also provided a full u-v synthesis in a shorter integration time (6 hours). Since the sources in our sample all had known sub-mm fluxes, a reliable estimate of the expected 3 mm flux was obtain assuming a typical spectral slope of α = 2.5. The new CABB system allowed us to detect fainter sources at 3 mm than the original sample in Lommen et al. (2007, 2010) and obtain the corresponding 7 mm fluxes. To obtain a 5σ detection for all sources at 3 mm, on-source integration times ranging from hours were estimated (including 100% overhead at 3 mm for calibration). With an average on-source integration time of 2 hours, we were expecting an RMS of 0.61 mjy/beam (assuming H214 configuration, 2GHz CABB continuum, moderate weather and natural weighting), which would allow us to detect all the sources at better than 5σ at 3 mm. At 7 mm with an average integration time of 2 3 hours, we were expecting an RMS of 0.05 mjy/beam, allowing for the detection of all sources at 5σ or better (assuming H214 configuration, 2GHz CABB continuum, moderate weather and natural weighting). For the 15 mm and 3+6 cm observations, to obtain a 5σ detection for all sources, integration times ranging from 2 4 hours at 15 mm, and from 10 minutes to 3 hours at 3+6 cm were estimated assuming thermal dust emission (i.e. extending the spectral slope α 2.5 to cm wavelengths). These integration times resulted in an RMS of 0.05 and mjy/beam for 10 minutes at 15 mm and 2 hours at 3+6 cm (H214 configuration, 2GHz CABB continuum, moderate weather, natural weighting), for which a 70% detection rate was expected at better than 5σ for all 3 wavelengths. For these observations using the standard hybrid configuration we expect a beam size of (without antenna 6) and (with antenna 6) at 15 mm, and (without antenna 6) and 3 2 (with antenna 6) at 3+6 cm. Note that the expected RMS was

47 1.7. The instruments 25 achieved at 3 mm for all sources, and at 7 mm for some sources. However, the expected RMS was not achieved at 15 mm and 3+6 cm due to shorter than required integration times. For the high resolution 3 mm imaging of GQ Lup and DK Cha, the EW352 array configuration was used, as it provides regular baseline spacing needed to generate optimal visibility profiles. With an east-west track, a full 12 hour integration is required to ensure full u-v synthesis and the highest signal to noise. For these observations we expected an RMS of mjy/beam (for EW352, natural weighting, average weather). Both sources were detected at 1 mm (from the literature) and at 3 and 7 mm (from our multi-wavelength survey) with shorter integration times prior to the full synthesis. For the temporal monitoring of DK Cha and GQ Lup, integration times of 2 and 10 minutes respectively were estimated to achieve at least 8σ detections at 7 and 15 mm, resulting in an RMS of mjy/beam (DK Cha) and mjy/beam (GQ Lup) at 7 mm and an RMS of mjy/beam (DK Cha) and mjy/beam (GQ Lup) at 15 mm (for H214, natural weighting, average weather). Note the RMS values were achieve for these observations, further discussion of the RMS values and fluxes will be presented in Chapter For modelling Our mcfost radiative transfer modelling of GQ Lup utilised the Green and Green II supercomputers 5, which offered the capability of running jobs on multiple processors and cores. This code calculates three components: temperature structure of the disc, the full SED and the visibilities at the desired wavelengths. mcfost first calculates the temperature structure and the SED, and then the visibilities are calculated separately. Thus the code needs to be run once for temperature structure and SED, and again for each of the desired wavelengths, in this case 1.3 and 3 mm. This means we need to we run mcfost three times in order to simultaneously fit the full SED and the 1.3 and 3 mm visibilities. The Green machine 6 was installed on May 2007 and comprises 145 nodes each with 8 processors. On this machine, each mcfost model (defined as the temperature structure and SED calculation, and the two visibilities calculations) took approximately 9 to 19 central processing unit (CPU) minutes depending on the model parameters used, therefore requiring a total of 166,250 CPU minutes per source to survey the 9 free parameters. 5 Operated by Centre for Astrophysics and Supercomputing on behalf of Swinburne University of Technology. 6

48 26 Chapter 1. Introduction Since each model is run separately and mcfost is capable of using multiple processors, the actual time required to complete an 9 free parameter sweep is 83,125 (with each model running in sequence using 2 nodes with 4 processors each). The Green II 7 machine is the successor to Green (and currently still under construction), and consists of two facilities: the Graphic Processing Unit (GPU) Supercomputer for Theoretical Astrophysics Research (gstar) and Swinburne Supercomputer for Theoretical Academic Research (swinstar). Green II is a hybrid of traditional X64 processing cores and GPUs with a petascale data storage and QDR infiniband network. swinstar has 86 CPU nodes each with 16 processors. On swinstar, we expect each mcfost model ([temp. + SED] + 2 visibilities) to take approximately 3 minutes to complete, decreasing the total time of a 9 free parameter sweep to 26,244 minutes. Note that in ray tracing mode mcfost performance degrades above 4 processors, thus we run 4 models in parallel on a node rather than one model per node. In addition, we can use a maximum of 256 processors at one time, meaning 64 models can be running at one time, decreasing the total time to 438 hours. Further details of the modelling will be presented in Chapter Thesis Outline The goal of this thesis is to study the first stages of planet formation, to search for signatures of grain growth to cm-sized pebbles and to disentangle the emission mechanisms present in a multi-wavelength survey of the Chamaeleon and Lupus star forming region. To achieve this goal we: 1. Build on the Lommen et al. (2007, 2010) 1 and 3 mm dataset and complete the SED from 1 to 7 mm for a sample of 20 T Tauri stars in the Chamaeleon and Lupus southern star forming regions. 2. Monitor the longer wavelength emission at 7 and 15 mm and 3+6 cm for variability (or stability) to help determine the contributing emission mechanisms. 3. Resolve sources in the millimetre which show signs of cm-sized pebbles. 4. Model the full SED and visibilities for a low-mass T Tauri star which shown signatures of large pebbles sized grains. This study will provide a more accurate estimate of the dust mass for faint discs, a temporal monitoring analysis on the timescale of years for a sub-sample of sources, 7 wiki/

49 1.8. Thesis outline 27 and an investigation of the potential correlation between the infrared (the 10 µm silicate feature) and millimetre (the 1 3 mm spectral slope α) grain growth signatures. We will also be able to determine the emission mechanisms present and some information on the maximum grain size for the sample. Lastly, this study will provide an important dataset of protoplanetary discs for follow-up observations with ALMA. To accomplished this study we used the ATCA to conduct all observations and the Swinburne Supercomputers, Green I and II, to conduct the modelling and ATCA data reduction. In Chapter 2 the results of our 3 and 7 mm continuum survey are presented. We also present our temporal monitoring results and investigation of potential correlation between grain growth signatures in the infrared and millimetre. Chapter 3 presents the results of the high-resolution observations at 3 mm and the longer wavelength temporal flux monitoring of GQ Lup and DK Cha. The results of the detailed radiative transfer modelling of GQ Lup are also presented. In Chapter 4 we conclude with our major findings of the thesis and look to the future.

50

51 2 Signatures of Grain Growth in the Protoplanetary Discs of Chamaeleon and Lupus The dust content of protoplanetary discs is studied at multiple wavelengths to determine the multi-step grain growth process during the first stages of planet formation. Continuum emission from protoplanetary discs at millimetre wavelengths (1 & 3 mm bands) generally arise from optically thin thermal dust emission of grains potentially up to millimetre sizes. Under this assumption, the dust opacity index β can be used as an indicator for grain growth (via β α 2, where α is the spectral slope from 1 3 mm). By extending the dust opacity index relation to 7 mm and beyond, it is possible to determine if grains up to cm sizes are present. However, as the wavelength increases other mechanisms can contribute to the observed emission. If thermal dust emission dominates at 7 mm and beyond, the spectral slope from 7 mm to 6 cm should remain constant. The contributions from other processes cause an excess in flux above the expected thermal dust emission, leading to deviations in the spectral slope. The changes in the spectral slope, combined with temporal flux variability can be used to determine the dominant emission mechanism (e.g., thermal dust emission, free-free emission, chromospheric activity) present in the disc. In this Chapter we present ATCA results of a 3 and 7 mm continuum survey of 20 T Tauri stars in the Chamaeleon and Lupus star forming regions. This survey aims to identify protoplanetary discs with signs of grain growth. We detected 90% of the sources at 3 and 7 mm, and determined the spectral slopes, dust opacity indices and dust disc masses. We also present temporal monitoring results of a small sub-set of sources at 7, 15 mm and 3+6 cm to investigate grain growth to cm sizes and constrain emission mechanisms in these sources. Additionally, we investigated the potential correlation between grain growth signatures in the infrared (10 µm silicate feature) and millimetre (1 3 mm spectral slope, α). 29

52 30 Chapter 2. Signatures of Grain Growth Eleven sources at 3 and 7 mm have dominant thermal dust emission up to 7 mm, with 7 of these having a 1 3 mm dust opacity index less than unity, suggesting grain growth up to at least mm sizes. The Chamaeleon sources observed at 15 mm and beyond show the presence of excess emission from an ionised wind and/or chromospheric emission. Long-timescale monitoring at 7 mm indicated that cm-sized pebbles are present in at least four sources. Short-timescale monitoring at 15 mm suggests the excess emission is from thermal free-free emission. Finally, a weak correlation was found between the strength of the 10 µm feature and α, suggesting simultaneous dust evolution of the inner and outer parts of the disc. This survey shows that grain growth up to cm-sized pebbles and the presence of excess emission at 15 mm and beyond are common in these systems, and that temporal monitoring is required to disentangle these emission mechanisms. 2.1 Published Work This work has been published in the Monthly Notices of the Royal Astronomical Society: C. Ubach, S. T. Maddison, C. M.Wright, D. J. Wilner, D. J. P. Lommen, B. Koribalski Grain Growth Signatures of Protoplanetary discs of Chamaeleon and Lupus, 2012, MNRAS, 425, 3137 The following Tables and Sections were updated: Tables , 2A.3, 2C.1, 2C.2, 2C.3, 2D.1 Sections 2.4.3, 2.4.4, 2.4.4, 2.5.3, 2.6, 2.3.1, The following results were not included in the publication: Data obtained in Tables 2A.1, 2A.2 Sections Appendix 2.A 2.2 Introduction Observational signatures of grain growth in protoplanetary discs can be obtained from both the infrared (IR) and millimetre (mm) bands, which probe different regions of the disc and are sensitive to different grain sizes. IR emission comes from the warm inner ( 1 5 au) and upper layers of the disc, while the mm emission comes from the cooler outer disc regions

53 2.2. Introduction 31 > 10 au and mid-plane where the bulk of the dust resides. The smallest submicron-sized grains are strongly coupled to the gas, but as they grow they begin to decouple from the gas and settle to the mid-plane, resulting in grain size sorting (Chiang et al., 2001; Dullemond & Dominik, 2004). Since thermal dust emission is dominated by dust grains whose size is similar to the observing wavelength (Draine, 2006), multi-wavelength observations are therefore needed to understand the distribution of the dust phase in protoplanetary discs. The strength and shape of the 10 µm silicate feature can be used as a grain growth indicator. Przygodda et al. (2003) found that low mass T Tauri stars with submicron-sized grains typically have a strong triangular shaped 10 µm feature, while discs with micron-sized grains typically have weaker and broader 10 µm features. Similar results were found by van Boekel et al. (2003) for their survey of intermediate mass Herbig Ae/Be stars. An equivalent effect was suggested to be caused by crystallisation (e.g., Honda et al., 2003; Kessler-Silacci et al., 2006). Kessler-Silacci et al. (2006) found an increase in flux near 11.3 µm can be caused by polycyclic aromatic hydrocarbon and/or crystalline forsterite, which can mislead the grain growth interpretation. The 1-3 mm spectral slope, α 1 3mm (hereafter α), where the flux F ν α and ν is the frequency, can be used to estimate the dust opacity index β where dust opacity κ ν ν β. Assuming the emission is optically thin, the dust opacity index can be written as β α 2 and an opacity index β 1 indicates grain growth up to mm sizes (Draine, 2006). This approach has been used to detect mm-sized grains in 50 discs (Natta et al., 2004; Andrews & Williams, 2005; Lommen et al., 2007, 2009, 2010; Ricci et al., 2010a,b). However, care must be taken when relating α and β, as shallow spectral slopes can result from an extended optically thin disc or a compact optically thick disc (Beckwith & Sargent, 1991). To break this degeneracy the discs have to be spatially resolved at mm wavelengths. Recently Ricci et al. (2012) determined that although only a small fraction of optically thick material is required to create a shallow spectral slope in the disc where the maximum grain size 0.1 mm, the required dust overdensities are much larger than the physical processes which concentrate dust (e.g., streaming instabilities, gravitational instabilities) can achieve. This strengthens the notion that shallow spectral slopes at long wavelengths are due to large mm/cm-sized grains in the outer parts of the disc. By extending the dust opacity index relationship to 7 mm and beyond, it is possible to determine if grains up to cm sizes are present in discs (Testi et al., 2003; Wilner et al., 2005; Rodmann et al., 2006; Lommen et al., 2009). However for wavelengths of 7 mm and longer, the emission can result from a range of physical processes in young stellar objects, including thermal emission from dust, free-free emission from an ionised

54 32 Chapter 2. Signatures of Grain Growth wind, and non-thermal chromospheric emission (e.g., Dullemond et al., 2007; Millan-Gabet et al., 2007). While chromospheric emission will be unresolved without the use of very long baseline interferometry, both thermal dust emission from the disc and free-free emission from a wind may be extended, and so simply resolving the emission is not enough to determine the dominant emission process. The spectral slope information from 7 mm wavelengths and beyond can be used to break this degeneracy. Depending on optical thickness, two spectral slopes values are known for emission originating from free-free emission from spherically symmetric, constant velocity, ionised wind. An opaque wind as a free-free spectral slope, α ff = 0.6 (Panagia & Felli, 1975), while for optically thin wind α ff = 0.1 (Mezger et al., 1967). Using these definitions, Rodmann et al. (2006) used the cm spectral slope from 2 to 3.6 cm of four T Tauri stars to determine that 20% of the 7 mm flux was due to free-free emission. Another way to determine the emission mechanisms present in a disc is by temporal monitoring of flux variability at 7 and 15 mm. Thermal dust emission is expected to be constant over time, while thermal free-free emission from an ionised wind is expected to vary by 20 40% on a timescale of years (González & Cantó, 2002; Loinard et al., 2007). Non-thermal emission varies over a timescale of minutes to hours by an order of magnitude or more (Kutner et al., 1986; Chiang et al., 1996). The results from IR and mm surveys have been used to study a potential correlation between the 10 µm silicate feature and the millimetre spectral slope (Lommen et al., 2007, 2010; Ricci et al., 2010b). Lommen et al. (2007, 2010) suggested a tentative correlation between the strength and shape of the 10 µm silicate feature and α for a sample of Chamaeleon, Lupus, and Taurus sources exists. A correlation between these grain growth signatures would suggest simultaneous dust evolution of the inner and outer parts of the disc. The submicron-sized grains would no longer be replenished as grains grow to mm to cm sizes, which flattens the 10 µm silicate feature, while the simultaneous growth of larger particles in the midplane leads to a shallower mm spectral slope (Lommen et al., 2007, 2010). However, Ricci et al. (2010b) found no such correlation in their sample of ρ Ophiuchus and Taurus sources, and suggested that the small sample size and some inconsistency between the spectral slopes of specific sources could have lead to the discrepancy between these two results. In this work we identify discs with signs of grain growth from observations at 3, 7, 15 mm, 3+6 cm of 20 T Tauri stars in the Chamaeleon and Lupus star forming

55 2.3. Observations and data reduction 33 regions conducted using the Australia Telescope Compact Array (ATCA) 1, extending the 3 and 7 mm work of Lommen et al. (2007, 2010). An introduction to the survey and the data reduction process can be found in Section 2.3, with the results of the survey presented in Section 2.4. In Section 2.5 the dust opacity indices, the dust disc masses, and dominant emission mechanism at the longer wavelengths are determined, and the potential correlation between the IR and mm grain growth signatures is explored, with the conclusions presented in Section Observations and data reduction Sample As described in Section 1.5, this survey targets 20 T Tauri stars in the Chamaeleon and Lupus star forming regions at a range of wavelengths presented in Table 2.1. The sources were selected to overlap with the sample observed by Lommen et al. (2007, 2010) and the Spitzer From Molecular Cores to Planet Forming Disks (c2d) programme (Young et al., 2005; Alcalá et al., 2008; Merín et al., 2008). Additional sources with strong 1.3 mm fluxes from Henning et al. (1993) and Dai et al. (2010) were also added to the sample. The sample includes 7 Chamaeleon I cloud sources, the isolated Chamaeleon source T Cha, the Chamaeleon II source DK Cha, 10 sources from the Lupus 1, 3, and 4 clouds, and the isolated Lupus source RXJ The primary aim was to obtain 3 and 7 mm fluxes for all sources. Note that Sz 32 and WW Cha were observed simultaneously at 7 mm and beyond, where Sz 32 was the primary target and WW Cha was in the field of view of the observations. Nine sources in the sample with the highest expected centimetre fluxes were observed at 15 mm and 3 Chamaeleon sources 3+6 cm in order to determine the emission mechanisms see Table 2.1. The expected cm fluxes were estimated by extrapolating the spectral slope from 3 to 7 mm, α 3 7, using the 3 and 7 mm fluxes, where α = d(log F ν )/d(logν). Most Lupus sources were unavailable during the scheduled observing time for long wavelength observations. This work will also present results for WW Cha, Herbig Ae/Be star CU Cha and Cha IR nebula (KG 49) in Section The Australia Telescope Compact Array is operated by the Australia Telescope National Facility (ATNF) which is a division of CSIRO.

56 34 Chapter 2. Signatures of Grain Growth Table 2.1: List of sources observed with ATCA in this survey. Source Distances T eff Wavelengths Comments References (pc) (K) (mm) Chamaeleon 1 SY Cha ,5 2 CR Cha ,15 2,5 3 CS Cha ,7,15 Binary 2,5 4 DI Cha ,7,15 Binary 2,5 5 T Cha ,15,30,60 3,6 6 Glass I ,7 Binary 2,5 7 SZ Cha ,15 2,5 8 Sz ,7,15,30,60 HH jet 2,5 9 DK Cha ,7,15,30,60 Outflow 7,8 WW Cha ,15,30,60 Not primary target 2,5 CU Cha ,7,15,30,60 Herbig Ae/Be 3 KG Cha IR nebula 1,4 Lupus 10 IK Lup Binary 9,11 11 Sz Binary 9,11 12 HT Lup Triple system, HH jet 9,11 13 GW Lup ,11 14 GQ Lup ,7,15 Planet 9,11 15 RY Lup ,11 16 HK Lup ,13 17 Sz ,15 Cold disc 9,11 18 EX Lup ,7 9,11 19 MY Lup ,15 9,11 20 RXJ ,12 See Section A.1 for further details on individual Chamaeleon sources. See Section for full results and analysis. See Section A.2 for further details on individual Lupus sources. References: Chamaeleon: (1) Hughes et al. (1991), (2) Whittet et al. (1997), (3) van den Ancker et al. (1998), (4) Feldt et al. (1998), (5) Luhman (2004), (6) Brown et al. (2007), (7) Spezzi et al. (2008), (8) van Kempen et al. (2010). Lupus: (9) Hughes et al. (1994), (10) Makarov (2007), (11) Comerón (2008), (12) Merín et al. (2010), (13) Mortier et al. (2011). (Table from Ubach et al., 2012).

57 2.3. Observations and data reduction Observations and data calibration From May 2009 to October 2012 we performed continuum observations with ATCA in the mm and cm bands. These observations used ATCA in compact hybrid configurations observations details and specific array configurations can be found in Table 2A.1 and 2A.2. Compact hybrid configurations were chosen to ensure good u-v plane coverage and hence maximum detection sensitivity in relatively short integration times. These observations used the new Compact Array Broadband Backend (CABB) 2 which has a maximum bandwidth of 2 GHz (a factor 16 improvement), higher level of data sampling, and improved continuum sensitivity by a factor of four from the previous backend (Wilson et al., 2011). For these continuum observations we used the 2048 channels of 1 MHz width. The observations were conducted in dual side band mode, with frequency pairs centred at GHz (3 mm band), GHz (7 mm band), GHz (15 mm band), and simultaneous observations at GHz (3 and 6 cm bands respectively). The gain calibrator was observed between each target scan of 10 minutes length at 3 and 7 mm, of 3 or 5 minutes at 15 mm (see Table 2D.2) and of 15 minutes at 3+6 cm. QSO B and QSO B were used as gain calibrators for Chamaeleon and Lupus respectively, Uranus was used as the primary flux calibrator at 3 and 7 mm, and the quasar QSO B was used as the primary flux calibrator from 15 mm to 6 cm. The weather was generally good throughout the observations. We reached a sensitivity of 0.1 mjy/beam required to detect 90% of the sources at better than 5σ for the 3 and 7 mm band observations, and detected 4/6 sources at 15 mm and 1/3 at 3+6 cm. The two 7 mm non-detections had the lowest 1.3 and 3 mm fluxes and the highest RMS during the observations, and the non-detections at longer wavelengths had the lowest extrapolated fluxes of the targeted sample. The data calibration involved 2 main steps (1) the calibration of the bandpass, gain and primary flux calibrators and boot strapping of the fluxes. (2) The calibration of the targeted sources see Figure 2B.1 in Appendix 2.A. Part 1 is composed of 3 main tasks: [1] Bandpass, gain and primary flux calibration. For the May 2010 observations, a second bandpass calibrator was needed to do the bandpass correction of the flux calibrator, since the observations took place 12 hours after the initial bandpass calibration. This required that the bandpass correction for the gain and flux calibrators be performed separately. [2] Flagging. [3] Boot strapping the flux value. This data calibration process was based on the standard CABB procedure described in great detail in the ATCA user guide 3 for all 2 For more information about CABB see: 3 Published in 2010: guide/

58 36 Chapter 2. Signatures of Grain Growth wavelengths using the software package miriad version 1.5 (Sault et al., 1995) provided by Australia Telescope National Facility (ATNF) see Appendix 2.A for additional details on data calibration. This procedure differs from the guide as follows: it is not necessary at millimetre wavelengths to perform a gain calibration on the bandpass, however, due to upgrades in the software to accommodate for the new zoom-mode feature, a gain calibration on the bandpass is required in order to convert the gain tables to a format compatible with the plot routines Analysis and results The flux and RMS values at 3 and 7 mm were extracted from the image plane using imfit and imstat. At the longer wavelengths, values were extracted from the u-v plane using uvfit and uvrms due to the short 3 15 minute scans. A summary of the fluxes are presented in Table 2.2. The resulting source maps were created using natural weighting with the clean algorithm. For sources observed over several epochs, data from each day was calibrated separately and combined with invert Which sources are resolved at 3 and 7 mm? We determined if a source was extended by looking at the visibilities. As the resolution of the array increases, the flux will decrease for a resolved source and remain constant for an unresolved source. Plots of the u-v distance as a function of amplitude created using the uvamp task in miriad are presented in Fig. 2.1 for 3 mm and Fig. 2.2 for 7 mm. These plots suggest DK Cha at 3 and 7 mm and CS Cha at 3 mm have extended emission, while marginally extended emission is observed at 7 mm for CR Cha, CS Cha, SZ Cha and RXJ For these marginally resolved sources, we will take the conservative approach and assume the emission were not resolved. RXJ was resolved at 1.3 mm with the SMA, with an estimated Gaussian size of 1.53 arcsec ( 28 au assuming a distance of 184 pc) (Lommen et al., 2010). Since the H168 array has a resolution of au (beam size of arcsec) and the map in Fig. 2.3a does not appear to have extended emission, it is unlikely the emission was resolved here at 7 mm. The visibilities for Sz 32 and MY Lup presented in Fig. 2.2 suggest the presence of a contributing source near each targeted source location. Following the Fomalont & Wright html/atug.html 4 This problem might be resolved in a new version of miriad.

59 2.4. Analysis and results 37 (1974) approach of using the visibility function to determine the distance between double sources, we find that the Sz 32 and MY Lup visibilities are consistent with the location of WW Cha at 35 to the south-west of Sz 32 and an unidentified source at 50 to the north of MY Lup maps are presented in Fig The smaller field of view at 3 mm (30.2 versus 65.4 at 7 mm) did not allow for either WW Cha or the unidentified source to be detected at 3 mm. This is consistent with Lommen et al. (2009), who previously observed and resolved WW Cha at 3 and 7 mm and only detected Sz 32 in the field of their 7 mm observations. Although looking at the visibility data are commonly used in literature to determine source extension (e.g. Andrews & Williams, 2007b), this approach could suffer from phase decorrelation. Phase decorrelation degrades the signal as the baseline increases, causing a similar decrease in amplitude on the longer baselines. Given the generally good weather conditions, the use of compact array configurations, and reasonable RMS values from the ATCA seeing monitor 5, phase decorrelation was not substantial in our observations. To ensure that phase decorrelation was not effecting our visibility amplitudes, we used three other methods to investigate source extension at 3 and 7 mm. If a source is extended, a point source fit will exclude some emission, whilst a Gaussian fit should include most of the emission. A source could be considered extended when (1) the point source fit is less than the Gaussian fit (after accounting for uncertainties), (2) the Gaussian fit is greater than the point source fit plus n times the RMS, where n is a predetermined value (Lommen et al. (2007, 2009) used n = 2 for pre-cabb ATCA data, and assuming a factor of four improvement with CABB, here we use n = 8) and (3) when the synthesised beam is smaller than the Gaussian size obtained from the Gaussian fit. These three alternative approaches were considered (see Table 2C.1) and the results are consistent with the sources considered resolved when using the visibility data (CS Cha and DK Cha), while marginally resolved sources were considered unresolved with these three alternative approaches. Thus from this analysis we have determined that the emission of CS Cha was resolved 6 at 3 mm, and the emission of DK Cha was resolved at 3 and 7 mm. Additionally, Lommen et al. (2007, 2010) resolved SZ Cha, HT Lup, RY Lup and Sz 111 at 3 mm with ATCA. In Table 2.2 sources in boldface were resolved at 3 mm with ATCA in this work or in the literature. Sources resolved at 1 mm with SMA and at 3 mm with ATCA are mark with a G and a Gaussian fit is given. 5 The ATCA seeing monitor is a fixed interferometer tracking the GHz beacon on a geostationary communication satellite at elevation of 60 (Middelberg et al., 2006). 6 This result is not robust. If the error bars are increased from 1σ to 3σ the data can also be fitted by a line (unresolved emission). In this case, only DK Cha would be considered resolved.

60 38 Chapter 2. Signatures of Grain Growth Figure 2.1: Visibility amplitude versus baseline length (or u-v distance) for Chamaeleon and Lupus sources at 3 mm, with the 1σ statistical error bars for each bin.

61 2.4. Analysis and results 39 Figure 2.2: Visibility amplitude versus baseline length (or u-v distance) for Chamaeleon and Lupus sources at 7 mm, with the 1σ statistical error bars for each bin. Continued on the next 2 page.

62 40 Chapter 2. Signatures of Grain Growth Figure 2.2: Continued.

63 2.4. Analysis and results 41 Figure 2.2: Continued Source fluxes A summary of all the continuum fluxes obtained in this survey are presented in Table 2.2 and 2.3. The flux values were obtained by combining both sidebands for all available days. Flux values for each epoch are presented in Tables 2C.2 and 2C.3. Table 2.2 is supplemented with 1.2 mm fluxes for Chamaeleon sources from Henning et al. (1993) with SEST, and for Lupus sources from Lommen et al. (2010) and Dai et al. (2010) with the Submillimetre Array (SMA) 7, as well as some 3 mm data from Lommen et al. (2007, 2010) with ATCA see Table 2.1 for a list of sources observed and wavelengths used in this survey. Four of our Chamaeleon sources have 870 µm LABOCA 8 data (Belloche et al., 2011b), however for consistency we will use the SEST flux data for our analysis of the Chamaeleon sources. All seven sources observed at 3 mm in this survey were detected, and are consistent with the previous upper limits reported in Lommen et al. (2007, 2010). We present for the first time observations of the sources at 7 mm (for exceptions see Table 2D.1) and beyond, 18/20 sources at 7 mm, 6/9 sources at 15 mm and 1/3 sources at 3+6 cm were detected. A 3σ upper limit is provided for all non-detections. If the source was considered resolved in Section or in the literature, the Gaussian fit for the flux is presented in Table 2.2 along with the flux uncertainties in the fits, the RMS values, and the beam size for observations made in this survey. The primary flux calibration uncertainties are not included in Table 2.2. This uncertainty is 20% at 1.2 mm for SEST data and 10% for SMA data. For the ATCA data this uncertainty is 7 The Submillimetre Array is a joint project between the Smithsonian Astrophysical Observatory and the Academia Sinica Institute of Astronomy and Astrophysics and funded by the Smithsonian Institution and the Academia Sinica. 8 LABOCA: LArge BOlometer CAmera service at the 12-m Atacama Pathfinder EXperiment (APEX) telescope.

64 42 Chapter 2. Signatures of Grain Growth 30% at 3 mm and 10% at 7 15 mm and 3+6 cm bands. The beam size of at 7 mm was sufficient to resolve the binary pair IK Lup and Sz 66 with a separation of 6.4 (Reipurth & Zinnecker, 1993). However, Sz 66 was not detected see Fig. 2.3b. Thus the 7 mm fluxes in Tables 2C.1 are most likely only from IK Lup. Similar results were reported by Lommen et al. (2010) for the 1.2 and 3.2 mm observations, with beam sizes of and 2 respectively. In both cases the beam size was sufficient to resolve the binary, suggesting the 2.2 mjy flux reported by Lommen et al. (2010) at 3.2 mm for Sz 66 is likely an over-estimate. If Sz 66 was not detected at 3.2 mm, a 3σ upper limit of 1.2 mjy would be more appropriate. Sz 66 was detected with Spitzer and has a 10 µm silicate feature see Lommen et al. (2010). However, the lack of cold dust from 1.2 to 7 mm would suggest there is no circumbinary disc, and if Sz 66 has a cold dust disc its mass is too low to have been detected by our observations.

65 2.4. Analysis and results 43 Figure 2.3: The cleaned maps using natural weighting for RXJ at 7 mm, IK Lup at 7 mm, ATCA at 3 mm and MY Lup at 7 mm. (A) RXJ at 7 mm with contours at -3,3,5 times the image RMS of 0.2 mjy/beam. (B) IK Lup (Sz 65) at 7 mm with contours at -3,3,6,9,12 times the image RMS of 0.8 mjy/beam. Although the synthesised beam size of was sufficient to resolved the binary with a separation of 6.4, Sz 66 was not detected. (C) Beam flux corrected clean map of MY Lup at 7 mm with ATCA in the north with contours at -3,3,5,11,22,45 times the RMS of 0.3 mjy/beam. (D) ATCA at 3 mm (with a cross in the position of MY Lup) with contours at -3,3,12,21,30,39 times the image RMS of 0.2 mjy/beam.

66 44 Chapter 2. Signatures of Grain Growth Table 2.2: Survey results. (1) Source name. (2) 1.2 mm continuum flux with image RMS in parenthesis. (3) 3 mm flux with image RMS in parenthesis. (4) Beam size at 3 mm. (5) 7 mm flux with image RMS in parenthesis. (6) Beam size at 7 mm. (7) References: For 1.2 mm continuum: 1 Henning et al. (1993) with SEST, 2 Lommen et al. (2010) with SMA, 3 Dai et al. (2010) with SMA. For 3 mm continuum with ATCA: 4 Lommen et al. (2007), 5 Lommen et al. (2010), 6 this work. Source F1.2mm (RMS) F3mm (RMS) Beam Size F7mm (RMS) Beam Size References mjy (mjy/beam) mjy (mjy/beam) arcsec mjy (mjy/beam) arcsec Chamaeleon SY Cha <172.0 <4.8 <0.9 (0.3) 1,4,6 CR Cha (24.2) 6.2 (1.5) (0.1) ,4,6 CS Cha (45.6) 9.4 (0.6) G * (0.1) ,6,6 DI Cha 38.0 (11.4) 2.3 (0.3)* a (0.1) ,6,6 T Cha (17.7) 6.4 (1.0) (0.1) ,4,6,6,6 Glass I 69.9 (22.4) 4.4 (0.1)* (0.1) ,6,6 SZ Cha 77.5 (20.3) 5.8 (0.5) G (0.1) ,5,6 Sz (20.8) 3.1 (0.2)* (0.1) ,6,6 DK Cha (22.0) 49.8 (1.3) G * (0.6) ,6,6 Lupus IK Lup 28.0 (2.8) 3.4 (0.4) (0.1) ,5,6 Sz 66 <8.0 (2.8) 2.2 (0.4) b <0.3 (0.1) ,5,6 HT Lup 73.0 (4.0) 12.0 (1.1) G (0.1) ,4,6 GQ Lup 25.0 (3.0) 3.6 (0.3) (0.1) ,6,6 GW Lup 64.0 (3.7) 8.5 (1.9) (0.1) ,4,6 RY Lup 89.0 (4.9) G 2.8 (0.7) G (0.1) ,5,6 HK Lup (3.9) G 7.3 (2.1) 1.0 (0.1) ,4,6 Sz (4.2) G 5.7 (0.7) G (0.1) ,5,6 EX Lup 21.3 (4.0) G 2.1 (0.3) (0.1) ,6,6 MY Lup 66.1 (3.4) G 8.7 (0.4) (0.1) ,5,6 RXJ (3.9) G 6.7 (0.6) (0.2) ,5,6 A 3σ upper limit is provided for non-detections. Sources in boldface were resolved at 3 mm with ATCA. a Flux at 93 GHz, as 95 GHz was not detected. b Potentially an over estimate see Section mm continuum fluxes from this work. All 7 mm fluxes are from this work. G For a source resolved at 1 mm with SMA and 3 mm with ATCA a Gaussian fit is given, otherwise a point fit is given.

67 2.4. Analysis and results 45 Table 2.3: Survey results. (2) Source name. (3), (4), (5) Fluxes at 15 mm, 3 and 6 cm with u-v plane RMS in parenthesis. Source F 15mm (RMS) F 3cm (RMS) F 6cm (RMS) mjy (mjy/beam) mjy (mjy/beam) mjy (mjy/beam) Chamaeleon 2 CR Cha 0.3 (0.1) 3 CS Cha <0.3 (0.1) 4 DI Cha <0.3 (0.1) 5 T Cha 0.3 (0.1) <0.3 (0.1) 0.3 (0.1) 7 SZ Cha 0.1 (0.02) 8 Sz (0.1) <0.3 (0.1) <0.3 (0.1) 9 DK Cha 1.6 (0.1) 0.8 (0.1) <0.3 (0.1) Lupus 17 Sz (0.02) 19 MY Lup <0.06 (0.02) A 3σ upper limit is provided for non-detections. Sources in boldface were resolved at 3 mm with ATCA Millimetre spectral slopes The mm spectral slopes α and α 3 7 presented in Table 2.4 were determined using the 1, 3 and 7 mm band fluxes from Table 2.2, the uncertainties in the α and α 3 7 were calculated through propagation of the uncertainties of the fluxes presented in Table 2C.1. Histogram of the distribution of α values is presented in Fig The values of α range between 2 to 4. The grain size distribution in disc models evolves and can depart substantially from the n(a) a 3.5 size distribution of the interstellar medium (e.g, Weidenschilling, 1997; Birnstiel et al., 2011). Since fully modelling the grain size distribution for all 20 sources in our survey is beyond the scope of this Chapter, for this analysis we will assume the dust emissivity can be represented by a single power law from mm through to cm wavelengths. Thus if only thermal dust emission is contributing to the flux from 1 to 7 mm the spectral slope would remain constant, whilst an inequality between α and α 3 7 would imply a break in the spectral slope, suggestive of the presence of other emission mechanisms such as thermal free-free and non-thermal emissions contributing to the flux at the longer wavelengths. For this analysis we will consider α and α 3 7 to be consistent (no break at 7 mm) when α 3 7 ± α 3 7 = α± α and inconsistent (a break at 7 mm) when α 3 7 ± α 3 7 < α± α. The uncertainty α is 0.4 for all sources and α 3 7 is given in Table 2.4. The histogram α values show that majority of Lupus sources have a value of α < 3, while majority of Chamaeleon sources have a value α > 3 see Fig From Table 2.4

68 46 Chapter 2. Signatures of Grain Growth a a a α α05. Figure 2.4: Histogram of the spectral slopes α. SY Cha and Sz 66 spectral slopes are unconstrained due to the upper limits at 3.2 and 1.2 mm, thus both are excluded from the analyses. we determined that 11 sources do not have a break in their spectral slope at 7 mm, while 7 sources do have a break and two are undetermined. Note that for the 7 sources with an observed break, the excess flux is well above the expected uncertainty. The presence of the excess emission above thermal dust emission does not imply a lack of grain growth, it indicates the presence of multiple emission mechanisms. Additionally, GW Lup, GQ Lup, Sz 111 and MY Lup appear to have 7 mm fluxes lower than the expected spectral fit from 1 3 mm. Although the 7 mm fluxes for GQ Lup and Sz 111 are below the spectral fit, they are within the expected uncertainty (grey hashed area in Fig. 2.5). However, the 7 mm flux for GW Lup is just below the uncertainty and the 15 mm upper limit for MY Lup is well below the uncertainty. This lack of emission at the longer wavelengths could suggest a binomial distribution of grains (with a lack of cold larger grains in the midplane), or that a single spectral slope is unlikely to fit the emission from 1 mm to 15 mm, or a combination of both. We plot the flux as a function of wavelength in Figs. 2.5, to better study the spectral slope. The solid line represents the spectral slope α presented in Table 2.4 and the dashed-dot lines are the upper and lower limits for α using the flux fit uncertainties and assuming the standard primary flux calibration uncertainties (20% for 1.2 mm, 30% for 3 mm, and 10% for 7, 15 mm and 3, 6 cm). Thus the shaded region represents the flux range expected when thermal dust emission dominates. The dashed line representing free-free emission with α ff = 0.6 (Panagia & Felli, 1975) anchored at the 15 mm data point, is included for sources with cm data. Full SEDs can be found in Fig. 2C.1. CR Cha, DI Cha, T Cha, Sz 32, RY Lup, EX Lup and RXJ all have 7 mm fluxes in excess of purely thermal dust emission consistent with Table 2.4 see Fig. 2.5.

69 2.4. Analysis and results 47 Note that if we were to use the 870 µm fluxes obtained with LABOCA by Belloche et al. (2011a) for the three available Chamaeleon sources instead of the SEST data, the α values would change to 2.3±0.2 for CS Cha, 2.9±0.2 for Glass I and 3.1±0.4 for SZ Cha, which would not change the interpretation of our analysis of these sources. For those sources with observations at 15 mm and longer wavelengths, we can see that all seven Chamaeleon observed at 15 mm have excess emission above that expected from thermal dust, while the two Lupus sources do not show evidence of excess emission above thermal dust at this wavelength. At 3+6 cm, the flux and upper limits of DK Cha, T Cha and Sz 32, suggest the presence of free-free emission from an ionised wind and/or chromospheric emission, which will be discussed further in the next Section. Table 2.4: Spectral slopes from 1-3 mm and 3-7 mm. (1) Source name. (2) Spectral slope from 1-3 mm, α (with an uncertainty of 0.4). (3) Spectral slope from 3-7 mm, α 3 7. (4) Uncertainties in α 3 7. (5) Indication of whether there is a break in the spectral slopes at 7 mm. Sources α α 3 7 α 3 7 Break Chamaeleon SY Cha* >3.9 > CR Cha Y CS Cha N DI Cha Y T Cha Y Glass I N SZ Cha N Sz Y DK Cha N Lupus IK Lup N Sz 66* >1.4 > HT Lup N GQ Lup N GW Lup N RY Lup Y HK Lup N Sz N EX Lup Y MY Lup N RXJ Y * SY Cha and Sz 66 only have upper limits at 1.2 and 3 mm. The sources in bold were resolved at 3 mm.

70 48 Chapter 2. Signatures of Grain Growth Figure 2.5: Millimetre flux versus wavelength for Chamaeleon and Lupus sources. The solid line represents the spectral slope α and the dashed-dot lines are the upper and lower limits for α using the flux fit uncertainties and assuming the standard primary flux calibration uncertainties (20% for 1.2 mm, 30% for 3 mm, and 10% for 7 and 15 mm and 3+6 cm). The dashed line represents the free-free emission with α ff = 0.6 anchored at the 15 mm data point for sources with cm data. Fluxes values from Table 2.2. Note that the error bars include the flux uncertainty from the fitting routine and the primary flux calibration uncertainties. Continued in the next 2 pages.

71 2.4. Analysis and results 49 Figure 2.5: Continued.

72 50 Chapter 2. Signatures of Grain Growth Figure 2.5: Continued.

73 2.4. Analysis and results Emission mechanisms at longer wavelengths Temporal flux monitoring helps to determine the presence of contributing emission mechanisms. For our sample, a subset of 4 Lupus and 2 Chamaeleon sources which have 7 mm fluxes obtained approximately one year prior to our survey see Table 2D.1, allowing us to determine if free-free emission is contributing to the 7 mm fluxes for these sources. We also observed a sub-sample of 6 Chamaeleon sources at 15 mm and 3 of these sources at 3+6 cm for one epoch (see Table 2.2). We can investigate flux variations in these sources by determining a point source fit in the u-v plane for each scan length at these wavelengths allowing us to determine the contributing emission mechanisms 15 mm wavelength and beyond. The results from the long wavelength temporal monitoring are presented in Figs. 2.6 and 2.7, with the maximum and minimum recorded flux values at 15 mm and 3+6 cm presented in Tables 2D.2 and 2D.3 respectively. The literature 7 mm data were obtained from Lommen et al. (2009, 2010). Their observations took place in 2008 with the ATCA using the pre-cabb system. Their observing frequency pairs at different epochs do not always match our central frequencies. However small changes in the observing frequency (42.8 to 43.0 GHz and 45.1 to 45.0 GHz) between different epochs would not result in a change in flux by more than the statistical uncertainty and thus should not affect our conclusions. For our subsample at 7 mm, we can also investigate the potential day-to-day flux variability by obtaining a point source fit flux for the central frequency of the dual sideband for each day the source was observed, resulting in integration times between 10 to 70 minutes depending on the source (see Table 2A.2). In general, sources observed on consecutive days do not have the same integration times, and since Chamaeleon never sets at ATCA, more time was spent on the Chamaeleon than the Lupus sources. For this analysis we define short temporal monitoring as scans separated by less than a day, and long temporal monitoring as scans separated by a day or more. Thermal dust emission is considered dominant when no intra-epoch variability is observed. Thermal free-free emission is likely present when the flux varies during long temporal monitoring by a factor of 20 40%. Non-thermal emission will be considered present when the flux varies by a factor of 2 or more on timescales of minutes to hours. For this analysis we take into account the flux fit uncertainty and the primary flux calibrator uncertainty for CABB data. We only claim variability if the same behaviour is observed in both frequency sidebands.

74 52 Chapter 2. Signatures of Grain Growth Lupus sources RY Lup, MY Lup, RXJ and Sz 111 were observed three times at 7 mm, once in 2008 and twice in 2009 on consecutive days, MY Lup and Sz 111 were also observed twice in 2012 fluxes are presented in Table 2D.1. RY Lup, MY Lup show no evidence of flux variability over daily or yearly timescales, indicating that thermal dust emission is dominant. Sz 111 shows no evidence of flux variability over daily timescales however over a yearly timescales the flux variability is consistent with the presence of thermal free-free emission, while RXJ flux variability is consistent with a presence of thermal free-free emission. The α values for all sources are consistent with the upper limits calculated by Lommen et al. (2010). Chamaeleon sources Temporal flux monitoring at 7 mm is available for CS Cha and Sz 32 see Table 2D.1. The 15 mm temporal monitoring for 6 Chamaeleon sources are presented in Fig. 2.6 with the maximum and minimum recorded flux presented in Table 2D.2. The 3+6 cm temporal monitoring for 3 of those 6 Chamaeleon sources are shown in Fig. 2.7 with the recorded fluxes in Table 2D.3. The temporal monitoring of CR Cha and DI Cha suggest that the excess emission seen in Fig. 2.5 at 15 mm is not from a fast varying chromospheric emission but most likely from thermal free-free emission, consistent with the spectral slope of 0.6 seen in Fig CS Cha was observed at 7 mm three times in 2008 with extended array configurations (1.5 and 6 km), twice in 2009 on consecutive days with a compact array configuration and twice in There is no evidence of flux variability on consecutive days in 2009 (see Table 2D.1), while the three fluxes obtained in 2008 are inconsistent with each other and with the fluxes obtained in This is due to the fact that CS Cha was resolved with the extended arrays in 2008 and unresolved with the compact array in 2009, which could cause an underestimate of the flux when CS Cha was resolved. However, the additional observation obtain in 2012 are consistent with the presence of thermal free-free emission, and this is the likely cause of the excess emission observed in Fig SZ Cha was observed at 7 mm once in 2009 and twice in 2012, the variability observed in 2012 is consistent with thermal free-free emission from an ionised wind. The temporal monitoring of DK Cha at 15 mm is consistent with the presence of some excess emission from thermal free-free emission see Fig. 2.6 and Table 2D.2 and contamination from chromospheric emission was detected at 3+6 cm see Fig. 2.7 and Table 2D.3. Both of these results are consistent with the excess emission at 15 mm and

75 2.4. Analysis and results 53 Flux (mjy) CS Cha 17 GHz Flux (mjy) CR Cha 17 GHz Flux (mjy) GHz Flux (mjy) GHz Time (minutes from first scan) DI Cha GHz Flux (mjy) Time (minutes from first scan) T Cha GHz Flux (mjy) Flux (mjy) GHz Flux (mjy) GHz Time (minutes from first scan) Sz GHz Flux (mjy) GHz Flux (mjy) Time (minutes from first scan) DK Cha GHz 2 Flux (mjy) Flux (mjy) GHz Time (minutes from first scan) Time (minutes from first scan) Figure 2.6: Temporal monitoring of 15 mm band continuum flux at 17 and 19 GHz of six Chamaeleon sources. The scan lengths were between 3 and 5 minutes with total integration times given in Table 2A.2. The fluxes are represented by triangles with corresponding uncertainties, and the red squares represent 3σ detection values. Note that when a source is not detected (fluxes below the red squares), uvfit can sometimes estimates a negative flux. These are included for completeness, however the value itself is incorrect.

76 54 Chapter 2. Signatures of Grain Growth 3+6 cm seen in Fig Flux (mjy) Flux (mjy) T Cha 5.5 GHz 9 GHz Flux (mjy) Flux (mjy) Time (minutes from first scan) Sz 32 2 Flux (mjy) Flux (mjy) DK Cha 5.5 GHz 9 GHz Time (minutes from first scan) 5.5 GHz 9 GHz Time (minutes from first scan) Figure 2.7: Temporal monitoring of 3+6 cm band continuum flux at 9 and 5.5 GHz of three Chamaeleon sources. The scan lengths were 15 minutes with total integration times given in Table 2A.2. The fluxes are represented by triangles with corresponding uncertainties, and the red squares represent 3σ detection values. Note that when a source is not detected (fluxes below the red squares), uvfit can sometimes estimates a negative flux. These are included for completeness, however the value itself is incorrect. Sz 32 was observed once in 2008 and twice in 2009 on consecutive days and once in 2012 at 7 mm. The level of flux variability on consecutive days and between the 2008, 2009 and 2012, indicates thermal free-free emission is contributing to the 7 mm flux. T Cha and Sz 32 have excess emission at 15 mm likely due to contamination from thermal free-free emission from an ionised wind, and excess emission at 6 cm from chromospheric emission see Tables 2D.2, 2D.3, and Figs. 2.6 and 2.7. These results are consistent with the excess emission seen in Fig This is also consistent with Sz 32 being the suggested source that drives the short East-West jet, HH 914 seen at a distance of 0.4 from Sz 32 (Wang & Henning, 2006).

77 2.4. Analysis and results Additional Sources A second source was detected at 7 mm to the northwest of the MY Lup map see Fig. 2.3d for beam flux corrected clean map. This is an unknown source which we identify by ATCA (instrument + J2000 location), first observed in May 2009 at RA 16:00:43, Dec -41:54:42 with an estimated Gaussian flux of 25.0 ± 5.5 mjy after a beam flux correction. Since the source was at the edge of the beam, observations were conducted at 3 mm to confirm the detection and investigate the spectral slope. In August 2010, the source was detected in the same location at 3 mm with a point source fit of 12.0 ± 0.5 mjy and RMS of 0.4 mjy/beam see Fig. 2.3c. This source was also detected (in October 2012) at 15 mm with a point source fit of 13.8 ± 0.2 mjy and RMS of 0.02 mjy/beam. There are no known 1.2 mm or cm band detection at this location in the literature. The mm spectral slope from 3 to 7 mm, α 3 7 = 1.0 ± 0.8, suggests non-thermal emission and the source is likely a background radio galaxy. A range of extra-galactic sources were detected during the 3+6 cm observations, including: SUMS J , SUMS J , SUMS J , G 2MASX J , PMN J , PMN J , SUMSS J , CXO J , SUMS J and SUMS J In addition to these sources, WW Cha, KG 49 and CU Cha were observed during this survey, however they were not included in the analyses since KG 49 and CU Cha are not T Tauri stars and WW Cha was not the primary target. Observation details and specific array configurations can be found in Table 2A.3. WW Cha was detected in the field of view of the Sz 32 observations during 7, 15 mm and 3+6 cm observations see Table 2.5. The continuum fluxes obtained are consistent with Lommen et al. (2009), suggesting there is no variability at 7 and 15 mm. The flux values at 3 cm, decrease by a factor of 2 between 2007 and 2012, which is consistent with Lommen et al. (2009) findings of variability and suggesting thermal free-free emission, while the upper limit obtained at 6 cm is consistent with Lommen et al. (2009). We also found that when the continuum flux at 3+6 cm are broken into the individual scans, the fluxes are varying by a factor of 8 Fig 2.8, suggesting the presence of non-thermal chromospheric emission at 3+6 cm, which is consistent with the visibility plot presented in Fig KG 49 was observed at 7 mm with ATCA. The disc was not resolved, however, a significant drop in flux is observed after the first visibility data point and then the flux plateaus with a slight decrease suggesting that both the disc and the known nebula surrounding KG 49 are being detected see Fig We obtained a 7 mm point source

78 56 Chapter 2. Signatures of Grain Growth Table 2.5: Flux values for WW Cha from the literature and this work. Source denoted as Sz 32 and WW Cha when WW Cha was not the primary target. (1) Date. (2) Source name. (3) Total integration time used to determine flux. (4) Frequency (combined frequency pairs). (5) Point flux. (6) RMS. (7) Beam size. Date Source T int Freq Flux RMS Beam size (minutes) (GHz) (mjy) (mjy/beam) (arcsecs) 5 Oct 2007 WW Cha ± Mar 2008 WW Cha ± Oct 2007 WW Cha ± Mar 2008 WW Cha ± May-09 a Sz 32 and WW Cha ± Aug-12 WW Cha ± Jul-11 c Sz 32 and WW Cha ± May 2006 WW Cha ± Oct WW Cha ± Oct WW Cha ± Oct WW Cha ± Nov WW Cha ±0.13 c Mar WW Cha ± May 2006 WW Cha ± Oct WW Cha < Oct WW Cha ± Oct WW Cha ± Nov WW Cha ±0.20 c Mar WW Cha ± Jul-11 c Sz 32 and WW Cha < Jul-12 Sz 32 and WW Cha Oct WW Cha < June 2007 WW Cha ± Jul-11 c Sz 32 and WW Cha < Jul-12 Sz 32 and WW Cha < Oct WW Cha < June 2007 WW Cha < From Lommen et al. (2009) ( c the 16 mm values from the 2 November 2007 data were obtained with antenna 6 included, causing the significantly lower point-source flux. ) Three antennas only.

79 2.4. Analysis and results 57 ( m ( m 9.Wa 9 5.ia 5.a 5.Wa 5 9.Wa 9 5.ia 5.a 5.Wa 5 9i 9n W55 C55 h55 a55 ( m ( m C W1T W 9 W1CT W ( m 91CT 9 T1T t 9 T99 W999 WT99 C999 CT99 ( m Figure 2.8: WW Cha results. Top left: Visibility amplitude versus baseline length (or u-v distance) at 44 GHz. Top right: Millimetre flux versus wavelength. Thermal dust emission appears to be dominant to 15 mm. Excess emission is observed at the cm wavebands. Bottom: Temporal monitoring of the continuum flux at the 15 mm band (17+19 GHz), and at the 3+6 cm bands (9+5.5 GHz). The scan lengths were 3 minutes in length for all wavelengths with total integration times given in Table 2.5. The fluxes are represented by triangles with corresponding uncertainties, and the red squares represent 3σ detection values. Note that when a source is not detected (fluxes below the red squares), uvfit can sometimes estimates a negative flux. These are included for completeness, however the value itself is incorrect.

80 58 Chapter 2. Signatures of Grain Growth fit of 6.38±1.08 mjy and a RMS of 0.86 mjy/beam see Table 2C.1. The millimetre flux versus wavelength plot presented in Fig. 2.9 is consistent with the thermal dust emission and with a β = 0.44 suggesting the presence of grain growth. Figure 2.9: KG 49 results at 44 GHz. Left: Visibility amplitude versus baseline length (or u-v distance). Right: Millimetre flux versus wavelength. CU Cha was observed at 3, 7 and 15 mm and 3+6 cm see Table 2.6. Following the procedure outlined in Section 2.4.1, we determine CU Cha was resolved at 3 and 7 mm with a β = 0.44 and no excess emission being detected up to 15 mm, suggesting grain growth up to cm-sizes Fig Similarly to DK Cha, T Cha and Sz 32, CU Cha also shows signs of the presence of non-thermal emission mechanisms at 6 cm both in the visibility plot and short timescale temporal monitoring Fig Table 2.6: Flux values for CU Cha. (1) Date. (2) Source name. (3) Total integration time used to determine flux. (4) Frequency (combined frequency pairs). (5) Point flux. (6) RMS. (7) Beam size. Date Source T int Freq Flux RMS Beam size (minutes) (GHz) (mjy) (mjy/beam) (arcsecs) Jul-10 CU Cha ± 1.0 G Jul-10 CU Cha ± 0.8 G Jul-11 CU Cha ± Jul-11 CU Cha ± Jul-12 CU Cha 20 9 < Jul-11 CU Cha ± Jul-12 CU Cha < G Gaussian fit. 2.5 Discussion In this section we aim to identify the protoplanetary discs with signs of mm and cm-sized grains. Thus far we have determined the sources with extended emission (i.e. which we

81 2.5. Discussion 59 ( m ( m 9.Ca 9 5.ia 5.a 5.Ca 5 9.Ca 9 5.ia 5.a 5.Ca 5 9i 9n C55 U55 h55 a55 ( m ( m ( m h 62C h 62U 62C U2U T C666 U666 ( m Figure 2.10: CU Cha results at 94 and 44 GHz. Top: Visibility amplitude versus baseline length (or u-v distance) at 94 and 44 GHz. Middle: Millimetre flux versus wavelength. Thermal dust emission appears to be dominant to 15 mm. Excess emission is observed at the cm wavebands.bottom: Temporal monitoring of the continuum flux at the 15 mm band (17+19 GHz), and at the 3+6 cm bands (9+5.5 GHz). The scan lengths were 3 minutes in length for all wavelengths with total integration times given in Table 2.5. The fluxes are represented by triangles with corresponding uncertainties, and the red squares represent 3σ detection values. Note that when a source is not detected (fluxes below the red squares), uvfit can sometimes estimates a negative flux. These are included for completeness, however the value itself is incorrect.

82 60 Chapter 2. Signatures of Grain Growth assume to be from resolved discs) in the 3 mm band, breaking the degeneracy between radially extended optically thin discs and compact optically thick discs with shallow spectral slopes. Analysis of the continuum fluxes in the mm and cm bands and the resulting spectral slopes allow us to identify sources with a break in spectral slope at 7 mm, indicating contributions to the emission at 7 mm and beyond from other emission mechanisms other than thermal dust. Here we calculate the dust opacity index at 3 mm in order to estimate the maximum grain sizes at this wavelength, and then estimate the dust disc masses at 3 mm. Finally, we explore a previously claimed tentative correlation between grain growth signatures in the IR and mm regimes using this data set Dust opacity index at 3 mm The 1 and 3 mm band emission are in the Rayleigh-Jeans limit and is expected to be mostly optically thin. The spectral slope α can be used to estimate the dust opacity index β. Assuming the emission is optically thin, the dust opacity index can be written as β α 2 and an opacity index β 1 indicates grain growth up to mm sizes (Draine, 2006). Table 2.7 presents the β values determined assuming optically thin emission for all the sources in our sample. We found that 2 of the 6 resolved sources and 5 of the 12 unresolved sources have β < 1, suggestive of grains up to mm sizes see columns two and three of Table 2.7. Due to the uncertainties in the β values determined through the propagation of the uncertainties in the α values, the symbol? will be used when 0.7 < β < 1.4.

83 2.5. Discussion 61 Table 2.7: Dust opacity indices and dust disc masses. (1) Source name. (2) Dust opacity index (uncertainty is 0.4). (3) Large grains? Y when maximum grain size up to mm-sizes (assumed when β 1), N when β > 1,? when 0.7 < β < 1.4. (4) Temperature power-law exponent, q, using the IRAS 60 and 100 µm fluxes. (5) Product of dust disc mass and dust opacity at 3 mm. (7) The dust opacity evaluated using Eq (8) The dust disc mass evaluated using the dust opacity κ3mm. Sources β Large qiras Mdust κ3mm κ3mm Mdust Mtotal β α 2 grains? q = 2/(3 αir) 10 4 M cm 2 g 1 cm 2 g M (g/d = 100) Chamaeleon SY Cha* CR Cha 1.3? E-02 CS Cha 0.9? E-02 DI Cha 1.1? E-03 T Cha 1.1? E-03 Glass I 1.2? E-02 SZ Cha 0.9? E-03 Sz N E-02 DK Cha 0.9? E-01 Lupus IK Lup 0.4 Y E-03 Sz 66* 0.1 HT Lup 0.0 Y E-03 GQ Lup 0.2 Y E-04 GW Lup 0.3 Y E-03 RY Lup 1.9 N E-02 HK Lup 0.9? E-02 Sz Y E-03 EX Lup 0.6 Y E-03 MY Lup 0.3 N E-03 RXJ N E-02 * SY Cha and Sz 66 only have upper limits at 1.2 and 3 mm. Boldface sources were resolved at 3 mm. Sources with for qiras have no available IRAS 60 and/or 100 µm fluxes.

84 62 Chapter 2. Signatures of Grain Growth This approximation of β does not account for the contribution from optically thick emission. Taking this into account, Beckwith et al. (1990) showed that β (α 2)(1 + ), (2.1) where is the ratio of optically thin to thick emission given by p [(2 q) ln(1 p/2) τ] 1, (2.2) where q and p are the power-law exponents of the disc temperature and surface density profiles respectively, and τ is the average disc opacity at the specific frequency. logarithmic dependence on is only valid when 2 p q = 0 see Beckwith et al. (1990) for details. We determined the temperature profile exponent q for our sources using IRAS 60 and 100 µm fluxes obtained from the Infrared Science Archive 9 using q = 2/(3 α IR ) 10, and found 0.2 < q < 0.7 with an uncertainty of 0.4, with the exception of HT Lup and GQ Lup with q > 2 see Table 2.7. For a flat spectrum disc a value of q = 0.5 and p = 1.5 is expected (Beckwith et al., 1990), and high angular resolution observations suggest p 1.5 (Wilner et al., 2000; Testi et al., 2003). If we adopt p = 1.5, a τ = 0.02 at 3.3 mm (Lommen et al., 2007) and use the q values in Column 4 of Table 2.7, with uncertainties such that the q values are consistent with 0.5, then 0.2 (with exception of HT Lup and GQ Lup where approximation in Eq. 2.2 is invalid). There are three sources (HT Lup, HK Lup, and MY Lup) for which Spitzer 24, 60, 70, and 100 µm fluxes exist. Using the MIPS 11 fluxes at 24 and 70 µm obtained by Merín et al. (2008), we re-calculated q and found values of 0.4, 0.9, and 0.6 for HT Lup, HK Lup, and MY Lup respectively. The significant decrease in q from 2.9 to 0.4 for HT Lup is well above the 0.4 uncertainty level, while the increase in q for HK Lup and MY Lup is within the uncertainty. The increase in q for HK Lup and MY Lup leads to no significant change in the and β values, while the decrease in q to 0.4 for HT Lup allows for the use of the approximation (obtaining a = 0.2). Note that the majority of the IR spectrum for HT Lup at these wavelengths are upper limits (Merín et al., 2008). If the detected 9 Housed at the Infrared Science Archive (IRSA) at the Infrared Processing and Analysis Centre (IPAC), California Institute of Technology. 10 Where α IR = d(log F ν )/d(log ν) is the IR spectral slope for λ 100 µm (Beckwith et al., 1990; Rodmann et al., 2006) 11 Multiband Imaging Photometer for Spitzer is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. The

85 2.5. Discussion 63 fluxes were determined by integrating between both detections and upper limits, it could artificially increase the overall fluxes. The other sources in the sample were detected at these wavelengths and thus were not affected. Knowledge of the inner and outer radii of the disc is needed to determine an exact value of (see equation 20 of Beckwith et al. 1990). Taking = 0.2 changes β 0.3, which does not change the interpretation of our analysis (with exception of CR Cha which increases to 1.6 and thus would be considered to have ISM-sized grains). For further analysis we will use the values for β calculated in Table Dust disc masses The dust disc mass can be determined via M dust = F ν D 2 κ ν B ν (T dust ), (2.3) using the distances D presented in Table 2.1, the 3 mm fluxes F ν from Table 2.2, the brightness, B ν (T dust ), for a dust temperature T dust (assumed to be 25 K) given by the Planck function, and the dust opacity, κ ν at ν = 94 GHz, evaluated using the β values in Table 2.7 via the Beckwith et al. (1990) equation κ ν = 10(ν/10 12 Hz) β cm 2 g 1. (2.4) The resulting κ 3mm and M dust values are presented in Table 2.7. Note that κ 3mm varies from the canonical 0.9 cm 2 g 1 used in the past works (e.g., Beckwith et al., 1990; Andrews & Williams, 2007a; Ricci et al., 2010b). The dust disc masses which range from M are similar to those found in other star forming regions (e.g., Ricci et al., 2010a). Assuming a gas-to-dust ratio of 100, we find eight sources have a total disc mass (gas+dust) greater than 0.01 M, the minimum mass solar nebula according to Weidenschilling (1977); and six have a total disc mass (gas+dust) greater than 0.02 M, the minimum mass solar nebula according to Hayashi (1981). Note that there is a lot of uncertainty in the dust mass calculations. One of the main sources of uncertainty comes from the dust opacity, which is dependent on the chemical composition, size, and shape of the grain (e.g., Draine, 2006). A second source of uncertainty comes from the assumed dust temperature, where a change of ±5 K can cause a 20% change in the dust mass. The uncertainty in the primary flux calibration and the uncertainty in the source distances of ±50 pc (Luhman, 2007; Hughes et al., 1994;

86 64 Chapter 2. Signatures of Grain Growth Comerón, 2008) also contributes to the total uncertainty of the dust masses. That said, the method employed here is that generally used in the literature (e.g. Andrews & Williams, 2005; Lommen et al., 2009, 2010; Ricci et al., 2010a) and one can expect the disc dust masses to be constrained within a factor of 2 to 10 due to these systematic uncertainties Millimetre-sized grains? Our 3 and 7 mm fluxes presented in Table 2.4 and in Figs. 2.5, indicate that 11 sources do not have a break in their spectral slopes at 7 mm, suggesting that thermal dust emission is dominant up to 7 mm. From the derived dust opacity indices (assuming β α 2) in Table 2.7 where the second column presents the β values and third column indicates the existence of grains up to mm sizes we find that 6/11 sources, all in Lupus, have β < 1, suggesting grain growth up to mm sizes, while 3 sources have ISM sized grains with β > 1. The majority of the Chamaeleon sources (7 sources) have 0.7 < β < 1.4, making it difficult to analyse the results given that the uncertainty in β is 0.4. Assuming the emission from 1-7 mm is optically thin, our results would suggest disc s dust grains in Lupus sources are generally larger than Chamaeleon sources, and the lack of excess emission observed at 7 mm for Lupus sources would suggest less stellar activity chromospheric emission and potentially more evolved systems than Chamaeleon sources and hence potentially older. However, compared to published ages of the different star forming regions, it is unclear if this is indeed the case. Lupus 3 and Chamaeleon I have similar ages (3 6 Myr), while Lupus 1 is thought to be younger than Lupus 3 (Hughes et al., 1994; Luhman, 2007; Merín et al., 2008). Individual sources with signs of large grains have ages ranging from < 1 14 Myrs (Hughes et al., 1994; Lawson et al., 1996). The lack of a break at 7 mm, is an indication of no excess emission which could be a signature of age. However, no distinction is found in the ages of individual sources with a break (3 10 Myrs) and sources without a break ( Myrs) (Hughes et al., 1994; Lawson et al., 1996; Hussain et al., 2009; Schisano et al., 2009). The temporal monitoring results for both mm and cm wavelengths presented in Section provide further clues, allowing us to rule out other sources of emission besides thermal dust at 7 mm and beyond. Note of the 7 sources with β < 1 only Sz 111 and MY Lup were monitored for flux variability at 7 mm. We found no flux variability at 7 mm for MY Lup, Sz 111, and conclude that thermal dust emission dominates in these sources and hence they have grains up to 1 cm in size. Our results also suggest all seven Chamaeleon sources observed at 15 mm have excess emission above thermal dust, with temporal monitoring suggesting the emission comes

87 2.5. Discussion 65 from thermal free-free emission. On the other hand, MY Lup and Sz 111 do not show evidence of excess emission above thermal dust at this wavelength, suggesting cm-sized pebbles are present in the disc. At 3+6 cm all three sources were found to have some excess emission. Thus it is difficult to obtain grain size information for these source beyond 7 mm Correlating grain growth signatures Thus far we have determined signatures of mm grain growth in our survey sample by evaluating β α 2, and found 35% of the discs have grains up to mm sizes. Another grain growth signature is the 10 µm silicate feature, which probes the warm inner ( 1 5 au) and upper layers of the disc, while the mm band probes the the cooler outer disc regions (> 10 au) and mid-plane. Lommen et al. (2007, 2010) suggested that a tentative correlation exists between the strength and shape of the 10 µm feature and mm spectral slope for a sample of discs in Chamaeleon, Lupus and Taurus. However, Ricci et al. (2010b) found no such correlation in their Taurus sample. As noted by Lommen et al. (2010), a correlation between these two regions of the disc is unexpected, and if confirmed, it could imply that grain growth in the inner upper layers of the disc and in the mid-plane occur simultaneously. Here we use our sample to further investigate this potential correlation. For this analysis, Spitzer Infrared Spectrograph 12 (IRS) spectrum data from µm was obtained from the Spitzer Heritage Archive 13 for all our sources and an additional seven ρ Ophiucus sources (to coincide with the Ricci et al. 2010b sample), as well as 11 Taurus-Auriga and two additional Chamaeleon and Lupus sources (to coincide with the Lommen et al sample) see Table 2.8. All the infrared data presented in Table 2.8 were processed with the S18.18 pipeline. Following Kessler-Silacci et al. (2005); Furlan et al. (2006), we define the shape of the 10 µm silicate feature as the flux ratio at 11.3 µm and 9.8 µm, F 11.3µm F 9.8µm, where F 11.3µm and F 9.8µm are the integrated flux over a 0.4 µm band centred at 11.3 and 9.8 µm, and define the strength of the 10 µm silicate feature as: strength = (F 10µm F cont ) F cont, (2.5) where F 10µm is the integrated flux from µm and F cont is a third-order polynomial 12 Spitzer Space Telescope is operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. For more information on IRS see Houck et al. (2004) 13 Housed at the Infrared Science Archive (IRSA) at the Infrared Processing and Analysis Centre (IPAC), California Institute of Technology

88 66 Chapter 2. Signatures of Grain Growth fit to the continuum from µm and µm. For each source presented in Table 2.8 the strength and shape of the 10 µm silicate feature was calculated by using the three IRS spectrum tables containing the wavelength ranges µm, 9 19 µm or µm, and 7 14 µm (corresponds to modules SL, SH, LL and SL respectively), with the exception of Taurus-Auriga sources which used two tables (9 19 µm and 7 14 µm). Note that the method of finding the 10 µm strength and shape is sensitive to the wavelength range chosen. Here we used the ranges from µm, µm, µm to determine the 10 µm strength, and 0.4 µm bandwidth to determine the 10 µm shape, which are consistent with the ranges used by Furlan et al. (2006) see Figure 2E.1 for the third-order polynomial fits to the infrared continuum. No IRS spectrum was available for SY Cha, DK Cha and Sz 111, and only the 9 19 µm table was available for GQ Lup and RU Lup and hence these sources are not listed in Table 2.8. T Cha and RXJ are isolated sources and were not included in the statistics of the Chamaeleon and Lupus sample. SZ Cha has a strong Polycyclic Aromatic Hydrocarbons emission band at 11.3 µm and which was not included in the calculation of F 11.3µm. The 1 3 mm spectral slopes α are from Table 2.4 for Chamaeleon and Lupus, from Ricci et al. (2010b) for the ρ Ophiucus sources and from Rodmann et al. (2006); Andrews & Williams (2007b) for Taurus-Auriga see Table 2.8. The millimetre spectral slope α as function of 10 µm strength and shape where plotted in Fig To test these two correlations we used the statistical package R (R Development Core Team, 2011) to determine the Spearman rank correlation coefficient, where the rank coefficient, r, is a measure of the how two values are related (r = 0 weak correlation, r = ±1 strong correlation), p-value is a measure of the probability that the obtained r value happens by chance. With a p-value of 0.17 and 0.39, we found the rank coefficients for the 15 sources in Chameleon and Lupus to be r = 0.37 and r = 0.24 for strength versus α and shape versus α respectively see Table 2.9. Thus we conclude that the correlations between 10 µm strength versus α and the 10 µm shape versus α cannot be confirmed with this sample. Spearman statistics was then performed on the sample of 36 sources from the four star forming regions and on each individual star forming region listed in Table 2.8. The results suggest no correlation is present for each individual star forming regions (Lupus 1 and 3 only have three sources, thus were not included in the individual analysis) see Table 2.9. A weak correlation between the strength of the 10 µm silicate feature and α is found for the whole sample, while no correlation appears present between the shape

89 2.5. Discussion 67 Table 2.8: Strength and shape of the 10 µm silicate feature for all Chamaeleon and Lupus sources from this work, and supplementary ρ Ophiucus and Taurus-Auriga sources. (1) Source name. (2) 10 µm strength values. (3) 10 µm shape. (4) 1 3 mm spectral slope. Source Strength Shape α Chamaeleon CR Cha CS Cha DI Cha Glass I SZ Cha* Sz WW Cha a Lupus IK Lup HT Lup GW Lup RY Lup HK Lup EX Lup MY Lup IM Lup a Isolated T Cha RXJ ρ Ophiucus SR b IRS b WSB b WSB b DoAr b RNO b Wa Oph b Taurus-Auriga DG Tau c DO Tau c AA Tau d CI Tau d DL Tau d DM Tau d DN Tau d DR Tau d FT Tau d GM Aur d AS d * SZ Cha has a strong Polycyclic Aromatic Hydrocarbons emission band at 11.3 µm. a Additional Chamaeleon and Lupus sources from Lommen et al. (2010). b α obtained from Ricci et al. (2010b). c α obtained from Rodmann et al. (2006). d α obtained from Andrews & Williams (2007b).

90 68 Chapter 2. Signatures of Grain Growth 7e 7 e he h ge == =ge g= = =geg hg = ge g ge ge ge h he he == = 6 = = = Figure 2.11: The 1 3 mm spectral slope α as a function of the 10 µm strength (top) and shape (bottom) using data from Tables 2.8. Black circles are the Cham I sources, red ties are the Lupus sources, blue triangles are the ρ Ophiucus, open diamonds are the isolated sources and pink diamonds are the Taurus-Auriga sources.

91 2.5. Discussion 69 of the 10 µm silicate feature and α. There are multiple processes affecting the shape of the 10 µm silicate feature (grain growth, crystallisation and opacity) that would make the detection and interpretation of a possible correlation difficult. For example both RXJ and WSB 52 have a 10 µm shape of 1.02 and 1.01 respectively, indicating a flat shape structure for both sources. However, this is not the case, RXJ has a strong, boxy feature, while WSB 52 has a very flat structure see Fig. 2E.1. It was only when Spitzer allowed for sufficient statistics that such ambiguities were cancelled out and the correlation between the strength and the shape of the 10 µm feature was confirmed (e.g. Kessler-Silacci et al., 2006). Compared to previous analysis of this correlation, our sample is the largest (36 sources compared to 7 sources in Lommen et al. (2007), 30 sources in Lommen et al. (2010) and 21 sources in Ricci et al. (2010b)), excludes sources for which only upper limits on α are known, contains no upper limits for 1 3 mm spectral slope α, and all the IR data were process and analysed consistently. However, the sources used may have several intrinsic differences. In particular, Cha I and ρ Ophiucus have very different environments, with Cha I sources being fairly isolated (e.g. Luhman, 2007) while ρ Ophiucus sources are deeply embedded (e.g. McClure et al., 2010). Although the ρ Ophiucus values in Table 2.8 were not corrected for extinction, similar results were obtained by Ricci et al. (2010b) using extinction corrected values for the 10 µm shape. Ideally we would like the same sample number per star forming region for this type of analysis. A larger sample may also mitigate the intrinsic differences in the 10 µm silicate feature, allowing for a more definite conclusion on the presence or absence of a correlation with α. Table 2.9: Spearman rank correlation coefficient values and percent confidence levels (p-values) for each star forming region in Table 2.8 used in the correlation analysis for the 10 µm silicate feature strength and shape with the 1-3 mm spectral slope α. Strength vs. α Shape vs. α # of sources rank coeff. p-value rank coeff. p-value Cham & Lupus All Cha I ρ Ophiucus Taurus

92 70 Chapter 2. Signatures of Grain Growth 2.6 Conclusions Continuum observations were carried out with ATCA at 3, 7, 15 mm, and 3+6 cm for 20 T Tauri stars located in Chamaeleon and Lupus. We analysed the mm fluxes in order to determine the spectral slopes, maximum grain sizes, and dust masses in these discs. Using supplementary data from the literature we conducted temporal monitoring of the fluxes of a subsample of our sources over short (less than one day) and long (months to years) timescales to help constrained the emission mechanisms present in these discs. We also analysed the potential correlation between the millimetre spectral slope and the strength and shape of the 10 µm silicate feature suggested by Lommen et al. (2007, 2010). Our 3 and 7 mm continuum fluxes indicate that 11 sources do not have a break in the spectral slope at 7 mm, suggesting thermal dust emission is dominant to wavelengths as long as 7 mm. We found that 6 of those sources (all in Lupus) have a dust opacity index less than unity (assuming β α 2), suggesting grain growth up to at least mm sizes. All seven Chamaeleon sources observed at 15 mm, have excess emission above thermal dust, and the two Lupus sources do not show evidence of excess emission above thermal dust at this wavelength, suggesting the presence of cm-sized pebbles. At 3+6 cm, DK Cha, T Cha and Sz 32 have flux upper limits which suggest the emission at 3+6 cm is due to an ionised wind and/or chromospheric emission. We obtained dust disc mass estimates ranging from M. Assuming a gas-to-dust ratio of 100, we find eight sources have a total disc mass (gas+dust) greater than 0.01 M, the minimum mass solar nebula according to Weidenschilling (1977); and six have a total disc mass (gas+dust) greater than 0.02 M, the minimum mass solar nebula according to Hayashi (1981). The monthly and yearly temporal flux monitoring revealed no excess emission at 7 mm for GQ Lup, MY Lup and CS Cha, indicating thermal dust emission dominates in these sources and hence that they have grains up to 1 cm in size. Sz 111 shows no evidence of flux variability over daily timescales however over a yearly timescales the flux variability is consistent with the presence of thermal free-free emission. At 15 mm the excess emission for CR Cha, CS Cha, DI Cha, T Cha, SZ Cha, Sz 32 and DK Cha is not from a fast varying chromospheric emission, but most likely from thermal free-free emission consistent with the spectral slope of 0.6.

93 2.6. Conclusions 71 Supplementing our sample with 16 other sources from Taurus-Auriga and ρ Ophiucus, no correlation between the shape of the 10 µm silicate feature and the 1-3 mm spectral slope α was found, while the strength of the 10 µm feature appears to correlate weakly with α. Atacama Large Millimeter Array will be vital in obtaining high resolution maps at high sensitivities of protoplanetary discs in the southern hemisphere at (sub)mm wavelengths, providing a better understanding of disc sizes and further analysis of mm grain growth signatures. However, at 7 and 15 mm, the ATCA and VLA will continue to provide invaluable information on other emission mechanisms present in protoplanetary discs. Acknowledgements Special thanks the ATCA staff, the Spitzer Science Center Helpdesk and Leonardo Testi, Lucca Ricci and François Menard for useful discussions. Thanks also to Philip Edwards for awarding some discretionary time to help complete this project. This research was supported in part by a CSIRO OCE Postgraduate Top Up Scholarship. CMW acknowledges support from the Australian Research Council through Discovery Grant DP and Future Fellowship FT This research has made use of the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration, and data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.

94 72 Chapter 2. Signatures of Grain Growth 2.A Observing details ATCA observing logs for the 11 sources in Lupus (Table 2A.1) and the 9 sources in Chamaeleon (Table 2A.2) at 3, 7, 15 mm and 3+6 cm. Table 2A.1: ATCA Observing log for Lupus sources. Phase calibrator QSO B (1) Observation date. (2) Sources observed. (3) Total integration time used in analysis in this Chapter. (4) Central frequency pair of observations. (5) Array configuration. Observation dates Sources T int Frequency pairs Array (min) (GHz) Config. May-09 Sz 65-Sz , 45 H214 May-09 HT Lup 40 43, 45 H214 Aug-10 GQ Lup 44 93, 95 H168 Aug-10 GQ Lup 80 43, 45 H168 Aug-12 GQ Lup 55 43, 45 H75 May-09 GW Lup 60 43, 45 H214 May-09 RY Lup 60 43, 45 H214 May-09 HK Lup 50 43, 45 H214 May-09 Sz , 45 H214 Aug-12 Sz , 45 H75 Oct-12 Sz , 45 H214 Oct-12 Sz , 19 H214 Aug-10 EX Lup 70 93, 95 H168 May-09 EX Lup 60 43, 45 H214 May-09 MY Lup 50 43, 45 H214 Aug-12 MY Lup 50 43, 45 H75 Oct-12 MY Lup 80 43, 45 H214 Oct-12 MY Lup 40 17, 19 H214 May-09 RXJ , 45 H214 Aug-10 RXJ , 45 H168 Aug-10 ATCA , 95 H168 ATCA array configuration information: configurations/upcoming configs.html Primary flux calibrator Uranus. Primary flux calibrator QSO B Note 1: Antenna 2 offline. Note 2: Antenna 6 offline.

95 2.A. Observing details 73 Table 2A.2: ATCA observation log for Chamaeleon sources. Phase calibrator QSO B (1) Observation date. (2) Sources observed. (3) Total integration time used in analysis in this Chapter. (4) Central frequency pair of observations. (5) Array configuration. Observation dates Sources T int Frequency pairs Array (min) (GHz) Config. Aug-10 SY Cha 30 43, 45 H168 May-09 CR Cha 80 43, 45 H214 Aug-12 CR Cha 10 43, 45 H75 Jul-11 CR Cha 42 17, 19 H214 Jul-10 CS Cha 60 93, 95 EW352 May-09 CS Cha 80 43, 45 H214 Aug-12 CS Cha 10 43, 45 H75 Oct-12 CS Cha 80 43, 45 H214 Jul-11 CS Cha , 19 H214 Oct-12 CS Cha , 19 H214 May-10 DI Cha , 95 H214 Jul-10 DI Cha 80 93, 95 EW352 May-09 DI Cha , 45 H214 Aug-12 DI Cha 10 43, 45 H75 Jul-11 DI Cha , 19 H214 May-09 T Cha 60 43, 45 H214 Aug-12 T Cha 10 43, 45 H75 Jul-11 T Cha 39 17, 19 H214 Jul-11 T Cha , 9 EW352 Jul-12 T Cha , 9 H168 May-10 Glass I 50 93, 95 H214 May-09 Glass I , 45 H214 Aug-12 Glass I 10 43, 45 H75 May-09 SZ Cha , 45 H214 Aug-12 SZ Cha 20 43, 45 H75 Oct-12 SZ Cha , 45 H214 Oct-12 SZ Cha , 19 H214 May-10 Sz , 95 H214 May-09 Sz 32 and WW Cha , 45 H214 Aug-12 Sz 32 and WW Cha 10 43, 45 H75 Jul-11 Sz 32 and WW Cha 42 17, 19 H214 Jul-11 Sz 32 and WW Cha , 9 EW352 Jul-12 Sz 32 and WW Cha , 9 H168 Jul-10 DK Cha 43 93, 95 EW352 Jul-10 DK Cha 54 43, 45 EW352 Aug-12 DK Cha 10 43, 45 H75 Jul-11 DK Cha 18 17, 19 H214 Jul-11 DK Cha , 9 EW352 Jul-12 DK Cha , 9 H168 Primary flux calibrator Uranus. Primary flux calibrator QSO B Note 1: Antenna 6 offline. Note 2: Errors with antenna 6 polarisation. Note 3: July 28 antenna 5 offline; July 29 antenna 3 offline for last hour; Flux calibrator bootstrapped from gain calibrator.

96 74 Chapter 2. Signatures of Grain Growth Table 2A.3: ATCA observation log for additional sources. Phase calibrator QSO B (1) Observation date. (2) Sources observed. (3) Total integration time used in analysis in this Chapter. (4) Central frequency pair of observations. (5) Array configuration. Observation dates Sources T int Frequency pairs Array (min) (GHz) Config. Aug-12 WW Cha 10 43, 45 H75 Jul-10 KG , 45 EW352 Jul-10 CU Cha 60 93, 95 EW352 Jul-10 CU Cha 70 43, 45 EW352 Jul-11 CU Cha 60 17, 19 EW352 Jul-11 CU Cha , 9 EW352 Jul-12 CU Cha , 9 H168 Primary flux calibrator Uranus. Primary flux calibrator QSO B B Data reduction details The data calibration procedure presented here was based on what is presented in the ATCA user guide 14 (last updated on 27 April 2011). Further information on all the miriad tasks mentioned below (and throughout this document) can be found in the user guide 15. The data calibration was broken up into two main parts consisting of (1) the calibration of the bandpass, gain and primary flux calibrators and boot strapping of the fluxes. (2) The calibration of the targeted sources see Figure 2B.1. Part 1 is composed of 3 main tasks: 1. Bandpass, gain and primary flux calibration One of the most important steps in setting up the array before commencing observations is to calibrate the delays. Due to the separation of each antenna, the signal reaches an antenna at different times, to correct for this, the signal is delayed before reaching the correlator. The delay calibration ensures that the signals reaches the correlator at the same time independent of distance. Though we set-up the system to nano-second accuracy, nothing is perfect, and small changes in the gain with frequency are possible (Fomalont & Perley, 1999). To correct for this a strong bandpass calibrator is observed at the beginning of each observation and every 12 hours after that see Figure 2B.2 for an example of a bandpass before and after calibration. mfcal is the task used to determine the bandpass correction. The bandpass table is then copied to the gain calibrator using gpcopy. The gain calibrator is observed in between target source observations, the propose of the gain calibrator is to correct for 14 The ATCA user guide: 15 observing/users guide/html/atug.html miriad user guide:

97 2.B. Data reduction details 75 MFCAL GPCOPY Flagged=yes GPCAL Part 1 GPCOPY BLFLAG Function applied to: Bandpass Gain Calibrator Primary Flux calibrator All Calibrators Primary Flux & Gain Calibrator Target sources GPBOOT MFBOOT Flagged=no Check calibration by determining the flux of gain and primary flux calibrator. GPCOPY Flagged=yes BLFLAG Part 2 Flagged=no DONE Figure 2B.1: Flow chart of calibration method. This was broken into two main parts: (1) The calibration of all the calibrators. (2) The calibration of the targeted sources. Before applying the calibration to the targeted sources, a check of the calibration was done by determining the flux of the gain and primary flux calibrators and comparing them to the ATCA calibrator database. The colours represent the function that was applied to the source, e.g. mfcal (red) was applied to the bandpass, while gpcopy and gpcal (dark yellow) were applied to the gain calibrator.

98 76 Chapter 2. Signatures of Grain Growth Figure 2B.2: An example of phase versus channel number for QSO B before and after a bandpass calibration for all baselines at 15 mm band. Top: Before applying the bandpass calibration. Notice the phases are not flat and have a large spread between -50 and 50. Bottom: After applying the bandpass calibration. Here the phases are flat and the spread is gone (which is what we are looking for). NOTE that each baseline pair is being plotted and only baselines pairs with the reference antenna (in this particular case antenna 3) will have a phase value of zero. rapidly varying atmosphere changes and slow varying instrument changes (Cornwell & Fomalont, 1999). A good gain calibrator is a point source close (< 5 ) to the target source see Figure 2B.3 for an example of a gain calibrator before and after calibration. The final step is to copy both the bandpass table and the gains table to the primary flux calibrator using gpcopy. 2. Flagging The flagging of known RFI and bad channels are performed prior to calibration, however, all other flagging should be performed after calibration since disentangling between a phase decorrelation and noise is difficult. Thus, after the initial full calibration, we view all the visibilities with blflag, to remove any unknown bad channels, bad weather intervals and/or bad section of data. If flagging was performed, then task [1] must be re-done, since both the phase and gains solutions will change after flagging. NOTE that if a major problem occurred during observations, the time and antenna(s) affected were recorded and the data were removed prior to calibration. 3. Boot strapping One of the most important calibration procedure for this research

99 2.B. Data reduction details 77 ft u ft u 1 5y3 5.y3. 0y3 0 5y3 1 5.y3. 0y3 0.h.s ft u Figure 2B.3: A sample plot of amplitude (Jy) versus time (hours) for QSO B before and after a gain calibration for all baselines at 15 mm band. Top: Before gain calibration was applied, some spread in the amplitude over the baselines is observed over time. Bottom: After gain calibration, the spread in amplitude over time is significantly decreased. was to ensure that our flux values were within the expected uncertainties. Two miriad tasks are used in sequence to ensure this: gpboot which corrects for the gain tables, by assuming that the gains are differ by a factor; and mfboot which corrects for the flux scale of the visibility dataset, also assuming that it is off by a factor. To obtain the factor, the measured visibilities are compared to models of the known primary flux calibrators. Before applying the calibration to the targeted sources, the calibration was tested by determining the flux of the gain and primary flux calibrators and comparing the obtained values to the ATCA calibrator database. If the obtained values were within the expected uncertainties at the observed wavelengths (30% for 3 mm and 10% for 7, 15 mm and 3, 6 cm) and within the expected variability of the source, the calibration was considered good, and proceed to part 2 of the data calibration process. Otherwise, the calibration was checked for inconsistencies. Part 2 is compose of 2 main tasks: [1] to copy the phase and gain tables and flux correction to the target sources using the task gpcopy. [2] Flagging using blflag. If no flagging is required then the target sources are calibrated see Figure 2B.1.

100 78 Chapter 2. Signatures of Grain Growth 2.C Result details The complete results used in Section analysis to determine if a source was extended are given in Table 2C.1, the full spectral energy distribution plots are given in Figure 2C.1, and the flux values for the sample at all the observed wavelengths, at the multiple epochs, are given in Tables 2C.2 and 2C.3. Figure 2C.1: Full SEDs for Chamaeleon and Lupus sources. Data obtained from the literature and this work. The stellar photosphere was obtained using the method of Furlan et al. (2006) normalising sources of spectral type later than G by their J-band flux, and earlier spectral types by their V-band flux. The photometry data were obtained from published catalogs. Legend: * WFI (Wide Field Imager); + 2MASS (Two Micron All Sky Survey); CTIO (Cerro Tololo Inter-American Observatory); USNO (U.S. Naval Observatory); IRAC (Infrared Array Camera on Spitzer Space Telescope); IRAS (Infrared Astronomy Satellite); MIPS (Multiband Imaging Photometer for Spitzer); ESO-TIMMI2 (Thermal Infrared MultiMode Instrument 2) on the ESO 3.6-m telescope at La Silla; LABOCA (LArge BOlometer CAmera) on the 12-m APEX telescope, (Belloche et al., 2011b); SEST (Swedish-ESO 15m Submillimeter Telescope), Henning et al. (1993); SMA (Submillimeter Array), Lommen et al. (2010); ATCA (Australia Telescope Compact Array), Lommen et al. (2007, 2010) and this work. Continued in the next 3 pages.

101 2.C. Result details 79 Figure 2C.1: Continued.

102 80 Chapter 2. Signatures of Grain Growth Figure 2C.1: Continued.

103 2.C. Result details 81 Figure 2C.1: Continued. NOTE the stellar photosphere of MY Lup was normalised to the J-band flux since its spectral type is unknown.

104 82 Chapter 2. Signatures of Grain Growth Table 2C.1: Summary of 3 and 7 mm flux fittings. (1) Source name. (2), (3) Point source and Gaussian fit fluxes. (4) Gaussian size obtained from the Gaussian fit. (5) Synthesised beam size using natural weighting. Note this value is dependent on u-v coverage and array configuration used. (6) Image RMS (natural weighting). (7) Factor of σ the point source fit is below the Gaussian fit (Fg=Fp+nσ). (8) Description of the amplitude as a function of u-v distance plots (uvamp) presented in Figs. 2.1 and 2.2. (9) If the emission was considered resolved or not. b Sources FP FG Gaussian size Beam size RMS n UVAMP a Resolved b (mjy) (mjy) (arcsec) (arcsec) (mjy/beam) Fg=Fp+nσ F,D,S Y,N 3 mm CS Cha 7.8 ± ± ± ± D Y DI Cha 2.3 ± ± ± ± F N Glass I 4.1 ± ± ± ± F N Sz ± ± ± ± F N DK Cha 33.0 ± ± ± ± D Y GQ Lup 3.6 ± ± ± ± F N EX Lup 1.8 ± ± ± ± F N 7 mm Chamaeleon SY Cha <1.0 c CR Cha 1.7 ± ± ± ± D N CS Cha 1.5 ± ± ± ± D N DI Cha 0.9 ± ± ± ± F N T Cha 3.0 ± ± ± ± F N Glass I 0.7 ± ± ± ± F N SZ Cha 0.7 ± ± ± ± F N Sz ± ± ± ± S N DK Cha 6.6 ± ± ± ± D Y Lupus Sz ± ± ± ± F N HT Lup 3.4 ± ± ± ± F N GQ Lup 0.6 ± ± ± ± F N GW Lup 0.6 ± ± ± ± F N RY Lup 1.0 ± ± ± ± F N HK Lup 1.0 ± ± ± ± F N Sz ± ± ± ± F N EX Lup 1.4 ± ± ± ± F N MY Lup 1.1 ± ± ± ± S N RXJ ± ± ± ± D N KG 49 a 6.38 ± ± ± ± F N a UVAMP: F flat, D drop, S sinusoidal. Resolved: Y yes, N no. c A 3σ upper limit is given for non-detections.

105 2.C. Result details 83 Table 2C.2: Flux values for all Lupus sources at all observed wavelengths at multiple epochs. (1) Date. (2) Source name. (3) Total integration time used to determine flux. (4) Frequency (combined frequency pairs). (5) Point flux. (6) RMS. (7) Beam size. Date Source T int Freq Flux RMS Beam size (minutes) (GHz) (mjy) (mjy/beam) (arcsecs) May-09 b Sz 65-Sz ± May-09 b HT Lup ± Aug-10 b GQ Lup ± Aug-10 b GQ Lup ± Aug-12 GQ Lup ± May-09 b GW Lup ± May-09 b RY Lup ± May-09 b HK Lup ± May-09 b Sz ± Aug-12 Sz ± Oct-12 Sz ± Oct-12 Sz ± Aug-10 b EX Lup ± May-09 b EX Lup ± May-09 b MY Lup ± Aug-12 MY Lup ± Oct-12 MY Lup ± Oct-12 MY Lup < May-09 b RXJ Aug-10 b RXJ ± Aug-10 b ATCA ± ATCA array configuration information: configurations/upcoming configs.html b Phase calibrator QSO B , primary flux calibrator Uranus.

106 84 Chapter 2. Signatures of Grain Growth Table 2C.3: Fluxes for all Chamaeleon sources at all observed wavelengths at multiple epochs. (1) Date. (2) Source name. (3) Total integration time used to determine flux. (4) Frequency (combined frequency pairs). (5) Point flux (unless otherwise noted). (6) RMS. (7) Beam size. Date Source T int Freq Flux RMS Beam size (minutes) (GHz) (mjy) (mjy/beam) (arcsecs) Aug-10 b SY Cha < May-09 a CR Cha ± Aug-12 CR Cha ± Jul-11 c CR Cha ± Jul-10 a CS Cha ± May-09 a CS Cha ± Aug-12 CS Cha ± Oct-12 CS Cha ± Jul-11 c CS Cha < Oct-12 CS Cha ± May-10 a DI Cha ± Jul-10 a DI Cha < May-09 a DI Cha ± Aug-12 DI Cha ± Jul-11 c DI Cha < May-09 a T Cha ± Aug-12 T Cha ± Jul-11 c T Cha ± Jul-11 c T Cha < Jul-12 T Cha 20 9 < Jul-11 c T Cha ± Jul-12 T Cha < May-10 a Glass I ± May-09 a Glass I ± Aug-12 Glass I < May-09 a SZ Cha ± Aug-12 SZ Cha ± Oct-12 SZ Cha ± Oct-12 SZ Cha ± May-10 a Sz ± May-09 a Sz 32 and WW Cha ± Aug-12 Sz 32 and WW Cha ± Jul-11 c Sz 32 and WW Cha ± Jul-11 c Sz 32 and WW Cha < Jul-12 Sz 32 and WW Cha 20 9 < Jul-11 c Sz 32 and WW Cha < Jul-12 Sz 32 and WW Cha < Jul-10 a DK Cha ± 1.3 G Jul-10 a DK Cha ± 0.2 G Aug-12 DK Cha ± 0.7 G Jul-11 c DK Cha ± Jul-11 c DK Cha ± Jul-12 DK Cha 20 9 < Jul-11 c DK Cha < Jul-12 DK Cha < G Gaussian fit. ATCA array configuration information: configurations/upcoming configs.html a Phase calibrator QSO B , primary flux calibrator Uranus. c Phase calibrator QSO B , primary flux calibrator QSO B

107 2.D. Monitoring details 85 2.D Monitoring details The complete results used for the monitoring analysis presented in Section mm fluxes are given in Table 2D.1, 15 mm in Table 2D.2 and 3+6 cm in Table 2D.3. Table 2D.1: Results of 7 mm flux monitoring of RY Lup, Sz 111, MY Lup, RXJ , CS Cha and Sz 32 from this work and the literature. (1) Source name and dust opacity index β (see Section 2.5.1). (2) Date of observation in YYYYMMDD format. (3) ATCA array configuration. (4) References: pre-cabb data from Lommen et al. (2009) and Lommen et al. (2010), and CABB data from this work. (5), (6) continuum fluxes from point source fits obtained at approximately 6.7 and 7.0 mm respectively with the RMS given in parenthesis in units of mjy/beam. Source Date Array Reference F(6.7 mm) b F(7 mm) b (mjy) (mjy) RY Lup H214 Lommen et al. (2010) < 0.6(0.2) β = H214 This work < 1.2(0.4) < 0.9(0.3) H214 This work 1.1 ± 0.1(0.1) 0.8 ± 0.1(0.1) MY Lup H214 Lommen et al. (2010) 1.3(0.1) β = H214 This work < 1.2(0.4) < 1.2(0.4) H214 This work 1.0 ± 0.1(0.1) 1.0 ± 0.1(0.1) H75 This work 1.0 ± 0.2(0.07) 1.0 ± 0.2(0.09) H214 This work 1.1 ± 0.2(0.04) 1.4 ± 0.2(0.04) RXJ H214 Lommen et al. (2010) < 0.5(0.2) β = H168 This work 1.1 ± 0.2(0.2) 1.1 ± 0.2(0.2) Sz H214 Lommen et al. (2010) < 0.6(0.2) β = H214 This work < 0.9(0.3) < 0.9(0.3) H214 This work 0.3 ± 0.1(0.1) 0.3 ± 0.1(0.1) H75 This work 0.4 ± 0.1(0.06) 0.9 ± 0.2(0.08) H214 This work 0.4 ± 0.1(0.05) 0.4 ± 0.1(0.04) CS Cha A Lommen et al. (2009) 1.0±0.3(0.1) β = B Lommen et al. (2009) < 0.8(0.3) < 0.7(0.2) B Lommen et al. (2009) < 1.1(0.4) 1.4±0.3(0.2) H214 This work 1.7±0.4(0.3) 1.2±0.3(0.3) H214 This work 2.0±0.3(0.3) 1.9±0.2(0.2) H75 This work 1.3±0.3(0.2) 1.0±0.3(0.2) H214 This work 0.52±0.03(0.04) 0.64±0.04 (0.04) Sz c H168 Lommen et al. (2009) 0.8±0.1(0.2) β = H214 This work < 0.5(0.2) < 0.5(0.16) H214 This work 1.5±0.5(0.3) 1.0±0.4(0.2) H75 This work 0.7±0.2(0.2) 0.6±0.2(0.2) b A 3σ upper limit is given for non-detections. c Sz 32 was detected in the field of view of WW Cha observations. The

108 86 Chapter 2. Signatures of Grain Growth Table 2D.2: Results of the 15 mm flux monitoring of 6 Chamaeleon sources, listing the highest and lowest fluxes obtained from a point source fit in the u-v plane. A 3σ value is given for non-detections. RMS = 0.1 mjy/beam. 17 GHz 19 GHz Sources Time F high F low F high F low (minutes) (mjy) (mjy) (mjy) (mjy) CR Cha ± 0.1 < ± 0.2 < 0.2 CS Cha ± 0.1 < ± 0.1 < 0.2 DI Cha 5 < 0.3 < ± 0.1 < 0.3 T Cha ± 0.2 < ± 0.2 < 0.3 Sz ± 0.1 < ± 0.1 < 0.2 DK Cha ± ± ± ± 0.2 Table 2D.3: Results of the 3+6 cm flux monitoring of 3 Chamaeleon sources, listing the highest and lowest fluxes obtained from a point source fit in the u-v plane. A 3σ value is given for non-detections. RMS 5.5 GHz = 0.1 mjy/beam and RMS 9 GHz 0.2 mjy/beam. All sources had a scan length of 15 minutes. 5.5 GHz 9 GHz Sources F high F low F high F low (mjy) (mjy) (mjy) (mjy) T Cha 0.9 ± 0.1 < ± 0.1 < 0.5 Sz ± 0.1 < ± 0.1 < 0.5 DK Cha 0.7 ± 0.1 < ± 0.1 < 0.4

109 2.E. Strength and Shape 87 2.E Strength and Shape The third-order polynomial fits to the infrared continuum obtained from the Heritage Infrared Archive of each source are presented. Spitzer data has been published in previous works (Furlan et al., 2006; Kessler-Silacci et al., 2006; Lommen et al., 2010). Figure 2E.1: The third-order polynomial fit to the infrared continuum used to obtain the strength of the 10µm silicate feature. The Spitzer infrared data were obtained from the Heritage Archive. Continued on next page.

110 88 Chapter 2. Signatures of Grain Growth Figure 2E.1: Continued.

111 3 A multi-wavelength study of the GQ Lup system In the previous Chapter we presented the results of our mm survey of 20 T Tauri stars in Chamaeleon and Lupus. From our sample, we found two sources GQ Lup and DK Cha which show signs of hosting large mm-sized grains and have their emission dominated by thermal dust emission. In this Chapter we present ATCA follow-up observations at 3, 7 and 15 mm of GQ Lup (for DK Cha see Appendix 3.E). These observations extend the temporal monitoring to timescales of years, allowing us to determine which emission mechanisms are present, and provide higher-resolution imaging at 3 mm to ensure that if the emission is extended. We also model GQ Lup to obtain further constraints on the disc and dust properties. 3.1 Introduction Using the β approximation (β α 2), for our sample of T Tauri stars in Chapter 2, we determined that 50% our of protoplanetary discs in the Lupus clouds have grains of at least mm sizes. Ideally one would like the relation between α and β to be extended to 7 mm and beyond, to determine if grains up to cm sizes are present in the disc (Testi et al., 2003; Wilner et al., 2005; Rodmann et al., 2006; Lommen et al., 2009). However, as discussed in Chapter 2, at longer wavelengths the emission may no longer be purely thermal dust emission. There are two other physical processes in YSOs whose emission is commonly detected at cm wavelengths: thermal free-free emission from an ionised wind and non-thermal emission from chromospheric activity, or indeed a combination of both (e.g. Lommen et al., 2009). An indication of these emission mechanisms can sometimes manifest as a break in the spectral slope from 3 to 7 mm or 7 to 15 mm or from 15 mm to 6 cm (see Section and Rodmann et al. (2006); Lommen et al. (2010); Carrasco-González et al. (2012)). Excess emission is a clear indication that additional 89

112 90 Chapter 3. A multi-wavelength study of the GQ Lup system processes other than thermal dust emission are contributing. The spectral slope from 7 mm and beyond can be used to distinguish between these emission mechanisms. For thermal free-free emission from a spherically symmetric, constant velocity wind, the spectral slope depends on the optical depth. For an opaque free-free wind α ff = 0.6 (Panagia & Felli, 1975), while for an optically thin wind α ff = 0.1 (Mezger et al., 1967). For non-thermal emission from chromospheric activity, the spectral slope α nt 0.1 (e.g. Anglada et al., 1998). Using these definitions, Rodmann et al. (2006) used the cm spectral slope from 2 to 3.6 cm to determine that 20% of the 7 mm flux in a sample of four T Tauri stars was due to free-free emission. Another way to determine the emission mechanisms present in a disc is by temporal monitoring the flux at 7 and 15 mm. Thermal dust emission is constant over time, while thermal free-free emission may vary by 20 40% on a timescale of years (e.g. González & Cantó, 2002; Smith et al., 2003; Carrasco-González et al., 2012), and non-thermal emission can vary over a timescale of minutes to hours by an order of magnitude or more (Kutner et al., 1986; Chiang et al., 1996). Because temporal flux monitoring surveys are time consuming, only three sources with confirmed cm-sized pebbles have been monitored at the 7 and 15 mm and 3 and 6 cm wavebands: TW Hya (Wilner et al., 2000), WW Cha (Lommen et al., 2009) and HD (Maddison et al., 2010). Multi-wavelength datasets are also used to constrain YSO models created using radiative transfer codes. Since fitting the SED alone does not provide a unique model, visibilities at multiple wavelengths are generally used to further constrain the fit. This method has been used to model a few sources such as IM Lup (Pinte et al., 2008), T Cha (Olofsson et al., 2011) and GQ Lup (Dai et al., 2010). These models, however, do not account for the longer mm and cm wavelengths, which provide the information on large grains that affect the model s temperature distribution and dust disc mass estimates. In Chapter 2, we found two sources which are good candidates for hosting large cm-sized grains and appear to have flux stability at 7 mm over the timescale of months: DK Cha and GQ Lup. Both sources have no break in their spectral slope at 7 mm, suggesting thermal dust emission is the dominant emission mechanisms to 7 mm. In the case of DK Cha, no excess emission was observed to 15 mm. The flux remains constant at 7 and 15 mm in the temporal monitoring conducted over short timescales, suggesting again that thermal dust emission dominates. In addition, both sources have β < 1, suggesting the presence of cm-sized pebbles. DK Cha was resolved at both 3 and 7 mm, while GQ Lup was only marginally resolved at 3 mm. In this Chapter we present ATCA follow-up observations at 3, 7 and 15 mm of GQ Lup (for DK Cha see Appendix 3.E).

113 3.2. GQ Lup 91 In this work we obtain follow-up high-resolution 3 mm maps of GQ Lup, and we extend the temporal flux monitoring at 7 and 15 mm to a period of a year. We also use the 3D radiative transfer code mcfost to model the GQ Lup system to help constrain disc parameters. The aims of these follow-up observations and the modelling are to: 1. Confirm that the emission is resolved at 3 mm. This required observations with an extended array configuration that provided equal sampling of the u-v plane to the longer baselines. 2. Determine flux stability at 7 and 15 mm. This required temporal monitoring in both the 7 and 15 mm bands on timescales of days, months and years, to determine if the presence of thermal free-free emission and/or non-thermal emission can be ruled out. 3. Determine if larger cm-sized pebbles are present in the disc. This required an analyses of the long wavelength emission to determine the dust opacity index β out to 15 mm, in conjunction with the results of temporal monitoring. GQ Lup introduced in Section 3.2, with the observations and data reduction process discussed in Section 3.3. The results are presented Section 3.4 and Section 3.5 presents the modelling of GQ Lup. A discussion of the results is provided in Section 3.6, with the conclusions in Section GQ Lup GQ Lup is a classical T Tauri star of spectral type K7V D (Neuhäuser et al., 2005, 2008b), located in the Lupus 1 cloud at a distance of 156 ± 50 pc (Crawford, 2000; Franco, 2002; Neuhäuser et al., 2008b) see Table 3.1 for basic properties. Both H α and H β lines have been detected in emission (Seperuelo Duarte et al., 2008; Donati et al., 2012), indicating ongoing accretion, while mid-infrared and far-infrared excess is assumed to result from the disc (Hughes et al., 1994). Soft and hard X-ray emission, which is not typical for a classical T Tauri star (Neuhäuser et al., 1995), has also been detected and are most likely from coronal emission (Krautter et al., 1997) GQ Lup has a companion denoted GQ Lup b, situated 0.7 ( 100 au) west (Neuhäuser et al., 2005). More recent spectro-astrometric observations determined a physical separation of 240 au (Pontoppidan et al., 2011). It is not yet known whether GQ Lup b is a brown dwarf (Guenther et al., 2005; Neuhäuser et al., 2008a; Lavigne et al., 2009) or a planet (Neuhäuser et al., 2008b). Neuhäuser et al. (2008b) detected no significant change

114 92 Chapter 3. A multi-wavelength study of the GQ Lup system in the position angle, with an orbital motion of 2-3 mas/yr assuming a distance of 156 ± 50 pc. Lavigne et al. (2009) used GAIA 1 synthetic spectra to estimate an effective temperature of 2400±100 K and a mass between 8-60 M Jup, putting the companion in the brown dwarf regime. SMA and ATCA observations found no thermal emission at the location of GQ Lup b assuming a separation of 0.7 (Dai et al., 2010; Ubach et al., 2012). Recently, Donati et al. (2012) suggested the radial velocity change observed between the two spectro-polarimetric epochs may be due to an additional companion orbiting GQ Lup at a distance of a few au. Table 3.1: Basic properties of GQ Lup. Parameter GQ Lup References RA (J2000) 15 h 49 m 12 s.10 1 DEC (J2000) Sp. Type K7V D 2,4 Cloud Lupus 1 3 Distance (pc) 156 ± 50 4 Comments Companion at 240 au. 2 References: (1) Teixeira et al. (2000); (2) Pontoppidan et al. (2011); (3) Carballo et al. (1992); (4) Neuhäuser et al. (2008b) Stellar properties The effective temperature of GQ Lup was first estimated by Hughes et al. (1994) to be 3890 K assuming a K7 spectral type and using a dereddened SED. A similar technique was used by Neuhäuser et al. (2005) who determined T eff = 4060 ± 300 K assuming a spectral type of K7eV. McElwain et al. (2007) fitted multi-colour optical photometry with a NextGen stellar atmosphere model with T eff = 4200 K and log(g) = 4.0. Recently, Donati et al. (2012) used the high-resolution spectro-polarimeter ESPaSOn at the CFWT 2 and deduced T eff = 4300 ± 50 K. Donati et al. (2012) also determined the stellar radius and the stellar mass to be R = 1.7 ± 0.2 R and M = 1.05 ± 0.07 M respectively. Both values are consistent with the previously published lower resolution values of R = 2.55 ± 0.41 and 1.8 ± 0.3 R by Broeg et al. (2007) and Seperuelo Duarte et al. (2008) respectively, and M = 0.8±0.2 M by Seperuelo Duarte et al. (2008). 8.4 ± 0.3 days (Donati et al., 2012). 1 Global Astrometric lnterferometer for Astrophysics 2 Canada-France-Hawaii Telescope The stellar rotation period was determined to be

115 3.3. Observations 93 Disc properties The inner disc radius (R in ) has been constrained to < 0.1 au through modelling to fit the high-resolution VLT 3 CRIRES 4 spectra (Hügelmeyer et al., 2009) and infrared (λ 1 µm) spectro-astrometric observations (Pontoppidan et al., 2011). SMA observations at 225 GHz (beam size of ) did not resolve the disc (Dai et al., 2010). Modelling to fit the 225 GHz data using a 2D radiative transfer code and assuming a face-on orientation suggests R out = 75 au, with a dust disc mass of M, suggesting that the disc is truncated by the companion (Dai et al., 2010). Our ATCA observations at 94 and 44 GHz (beam size of and respectively), suggested the disc was resolved at 3 mm and unresolved at 7 mm (Section 2.4.1). A dust disc mass of M was estimated at 3 mm. No evidence of excess emission above thermal dust from 1 to 7 mm was found and the dust opacity index, assuming β = α 2, was found to be β = 0.2 ± 0.4, which suggests the presence of large mm-sized pebbles (Section 2.5.1). These observations constrain the outer disc radius between 100 and 700 au (assuming a distance of 156 pc). There is a discrepancy in the predicted value of the disc inclination, with current estimates between 30 and 60. Pontoppidan et al. (2011) obtained a disc inclination of 65 ± 10 from spectro-astrometric observations, consistent the earlier measurement of 51 ±13 from B-band photometry (Seperuelo Duarte et al., 2008). However, a 1D radiative transfer model of the inner disc (between and 0.5 au) found that an inclination of 51 results in emission lines broader than the observations (Hügelmeyer et al., 2009). The best fit model to the line widths was obtained when a disc inclination of 22 was used for the inner warm disc with R in = au. This is consistent with the value obtained from photometric monitoring of 27 ± 5 by Broeg et al. (2007) and the value deduced from spectropolarimetric observations of 30 by Donati et al. (2012). 3.3 Observations Here we present our follow-up observations of GQ Lup. (For completeness we will also include GQ Lup results presented in Chapter 2 in the following tables and figures.) Continuum observations were conducted with ATCA in 2011 and 2012, in dual side band mode with frequency pairs centred at GHz (3 mm band), GHz (7 mm band) and GHz (15 mm band). Note that antenna 6 was offline for observations conducted on 8 and 10 July 2011 see Table 3.2 for the observing log. CABB was set to 3 ESO Very Large Telescope (VLT) 4 CRyogenic high-resolution InfraRed Echelle Spectrograph (CRIRES)

116 94 Chapter 3. A multi-wavelength study of the GQ Lup system 2048 channels of 1 MHz width and ATCA was in a compact hybrid configuration during the flux temporal monitoring observations (mainly H214, but also H75 and H168). For the high-resolution imaging at 3 mm, we used the east-west configuration EW352, which offers equal sampling of the u-v plane to the longer baselines for optimal amplitude versus baseline profiles.

117 3.3. Observations 95 Table 3.2: ATCA observation log for GQ Lup. Frequency pairs Observation Tint Array Flux Notes (GHz) date (min) configuration calibrator(s) 93,95 21 August H168 Uranus mediocre weather, antenna 2 offline 93,95 31 July EW352 Uranus 43,45 21 August H168 Uranus 43,45 10 July H Antenna 6 offline 43,45 4 August H , Neptune 43,45 22 Oct H Antenna 6 offline 17,19 8 July H Antenna 6 offline 17,19 18 July H ,19 22 Oct H Antenna 6 offline Data originally presented in Ubach et al. (2012). Uranus was observed on the 30 July 2011.

118 96 Chapter 3. A multi-wavelength study of the GQ Lup system The 3 mm observations on 31 July 2011 used a bootstrapped value of Uranus from 30 July 2011 for the absolute flux calibrator. It should be noted that using a bootstrapped value for the primary flux calibration at 3 mm can increase the absolute flux calibration uncertainty to 40% 50% (instead of the usual expected 30%). The absolute flux calibration was performed using QSO B for all 7 and 15 mm data. In August 2012, Neptune was also observed and could be used as an absolute flux calibration see Table 3.2. On 31 July 2011, the weather conditions were mediocre for the first hour of the 3 mm observations, with significant improvement for the rest of the observing run, with a seeing monitor RMS of 100 µm. The 7 and 15 mm observations from 2011 and 2012 had average weather conditions. Due to poor u-v coverage at the longer baselines, and since antenna 6 data suffered from a polarisation error at 7 and 15 mm in 2011 and a cryo-failure in October 2012, all data from antenna 6 was removed from the analysis. Generally good weather was observed for in 2010 see Section for further details. The phase calibrators QSO B was observed between target observations of GQ Lup, to calibrate the phases. The data calibration followed the same process as described in Appendix 2.A, which involved obtaining the bandpass, phase and flux information from the calibrators to calibrate the targeted sources. An absolute flux uncertainty of 30, 15 and 10% is expected at 3, 7 and 15 mm respectively, unless otherwise noted. 3.4 Results The flux and RMS values for all observations were extracted from the visibilities using uvfit and uvrms. For frequencies observed over several epochs, data from each day was calibrated separately and then combined with invert to create the clean maps. For completeness, the following Figures and Tables are supplemented with the GQ Lup (3 and 7 mm) data presented in Chapter 2. The fluxes were obtained by fitting either a point source or a Gaussian to the visibilities. In order to use the appropriate fit, we need to determine if the emission was resolved. Note, the total uncertainty is defined as the flux uncertainty from the fit plus the absolute flux uncertainty at each wavelength as defined in Section 3.3. Unless otherwise noted, only the flux uncertainty from the fit is included in the Figures and Tables. The resulting flux from a Gaussian fit will be used for resolved emission, while a point source fit will be used for unresolved emission. The standard method of determining if the emission is resolved is by plotting the visibility amplitudes as a function of u-v distance. These plots are created using the uvamp task in miriad, which obtains the flux by taking the vector average of the binned

119 3.4. Results 97 u-v dataset in annuli according to u-v distance. This task also calculates the statistical error bars and the expected amplitude for zero signal 5. The expected amplitudes for zero signal are depicted as a histogram, and if the statistical error bars are within with these expected amplitudes, the flux at that corresponding u-v distance is considered a non-detection. The bins were chosen as follows: the maximum number of bins cannot exceed the number of baseline pairs available, and the length of the first bin needs to be greater than the length of the shortest baseline pair (in kλ), and finally we try to obtain equal uncertainty for all bins. Similar plots can also be created through the real and imaginary parts of the visibilities, which can be obtained by taking the scalar average over the integration time of the flux per baseline pair using the uvplt task. This method provides one data point per baseline pair, avoiding the bin dependence. The error bars are ±1 standard deviation of the scalar average. The amplitudes can then be evaluated in R 2 + I 2 and the amplitude uncertainties via σ amp = ((R σ R ) 2 + (I σ I ) 2 )/(R 2 + I 2 ), where R and I are the real and imaginary parts of the visibility amplitudes respectively and σ R and σ I are the uncertainties in the real and imaginary parts of the visibility amplitudes respectively see Appendix 3.A for derivation of the uncertainty. For the most part, both methods give consistent results. However, note since uvamp determines an average of the amplitudes for a fixed u-v distance over time, while uvplt determines the average for fixed baseline length amplitude over time, the flux for the last bin could be different if the structure differs from a uniform symmetric disc see Appendix 3.B for an example. both methods to determine if the emission was extended. In Sections and 3.E.2, we will use a combination of GQ Lup GQ Lup was detected at all epochs and all wavebands. To ensure that the position of GQ Lup was consistent throughout the different epochs, the RA and DEC were set to 15 h 49 m 12 s.10 and respectively using uvedit after applying the bandpass and gain corrections tables. In Chapter 2 the GQ Lup was considered to be marginally resolved at 3 mm, the flux at 7 mm was found to be constant over a timescale of a month, and all the evidence from observations point to the presence of pebbles. Was GQ Lup resolved? At 3 mm, we used uvamp to create the visibility plot for both epochs see Fig 3.1. In Chapter 2, we suggested that the August mm emission was marginally resolved. 5 See miriad task documentation:

120 98 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.1: GQ Lup 3 mm uvamp plots of the visibility amplitude versus u-v distance for both epochs, with 1σ error bars for each bin. The expected amplitude for zero signal is represented as a histogram. On the left, August 2010 data and on the right the July 2011 data. The July 2011 data was fitted by Gaussian 6 with a FWHM= /kλ. In July 2011, the higher-resolution 3 mm observations show a clear decrease in flux with u-v distance, but a surprising increase in flux for u-v distances > 100 kλ. The observed falloff is suggestive of extended emission and can be fitted by a Gaussian 6 with a FWHM= /kλ, this is an emission size of 1.8 at FWHM, which corresponds to 270 au assuming a distance of 156 pc. However, it is unclear what is causing the flux increase at the longer baselines 7, we will discuss this further in Section 3.6. We also plot the July 2011 visibilities using uvplt and found the flux also decreases with baseline length and can be fitted by the same Gaussian as uvamp plot, however, the increase on longer baselines was not observed see Fig 3.2. Note that this is not necessarily an inconsistency between the methods. uvplt and uvamp average over different quantities (fixed baselines lengths vs fixed u-v distance) slightly differently, and thus use different parts of the dataset to determine the flux see Appendix 3.B. The mismatch between the visibility plots on the longer baselines could be due to the structure of the emission see Section 3.6 for a more detailed discussion. For a consistent comparison between the two epochs, we recreated the visibility plot for August 2010 using uvplt, and overlaid the July 2011 visibilities see Fig Note that antenna 2 was offline in August 2010, bringing the number of baselines down to 6 and we ( ) 6 F(ρ) = exp (πfwhmx) 2, where x is the u-v distance, FWHM is the full width at half-maximum 4ln2 intensity. 7 It should be noted that we checked both the phases and amplitudes versus channels for both the bandpass and phase calibrators and found both calibrations were performed correctly.

121 3.4. Results 99 did not achieve the RMS required to detect emission below 3σ (where σ = 0.3 mjy). We see that both epochs are consistent and the data can be well fitted by the same Gaussian with a FWHM= /kλ (or 1.8). We conclude that the 3 mm emission is resolved. Figure 3.2: GQ Lup 3 mm uvplt plots of the visibility amplitude versus baseline length (u-v distance) for each epoch. The error bars are ±1 standard deviation of the scalar average. The data was fitted by Gaussian with a FWHM= /kλ (or 1.8). At 7 mm, we used uvamp to create the visibility plots for all four epochs which are presented in Fig The emission is unresolved at all epochs, which is consistent with our previous August 2010 findings. Note that in July 2011 the two longer baseline amplitudes are within the uncertainty and thus are considered non-detections. A factor of more than two increase in the total flux was observed in the August 2012 data, suggesting the presence of other emission mechanisms. At 15 mm we used uvamp to create the visibility plots presented in Fig. 3.4 and from that the emission was unresolved. Besides looking at the visibilities, we explored three other methods to determine if the sources were extended. The source could be considered resolved if: (1) the point source fit was less than the Gaussian fit of the flux (taking into account absolute flux uncertainties), (2) the point source fit was n times the RMS smaller than the Gaussian fit (F g > F p nσ), where n is a pre-determined value (n = 2 for pre-cabb ATCA data (Lommen et al., 2007, 2009) and n = 8 for CABB data, assuming a factor of 4 improvement in the continuum sensitivity), and (3) the Gaussian size obtained from the Gaussian fit was greater than the natural weighted synthesised beam (see Section of Chapter 2). We explored these alternative approaches in Table 3C.1 and found them to be consistent with the visibilities

122 100 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.3: GQ Lup 7 mm uvamp plots of visibility amplitude versus u-v distance for each epoch, with the 1σ error bars for each bin. The expected amplitude for zero signal is represented as a histogram. Figure 3.4: GQ Lup 15 mm uvamp plots of visibility amplitude versus u-v distance for each epoch, with the 1σ error bars for each bin. The expected amplitude for zero signal is represented as a histogram.

123 3.4. Results 101 results. We conclude that the 3 mm emission was resolved, while the 7 and 15 mm emission was not. Fluxes and flux variability A summary of the continuum fluxes for each epoch are presented in Table 3.3. The fluxes for each epoch were obtained by combining the frequency pairs. A point source fit was used if the emission was considered unresolved in Table 3C.1, otherwise a Gaussian fit was used. Table 3.3: GQ Lup millimetre results. Chapter 2 are also included. For completeness the 2010 results obtained in Date T int Flux RMS Beam Size Array (min) (mjy) (mjy/beam) (arcsec) conf. 3 mm band (94 GHz) 21 Aug ± H July ± EW352 7 mm band (44 GHz) 21 August ± H July ± H214 4 August ± H75 22 Oct ± H mm band (18 GHz) 8 July ± H July ± H Oct ± H214 Data originally presented in Ubach et al. (2012). a Gaussian fit to the flux, as emission was resolved. The July mm observations used a more extended array, EW352, compared with the 2010 observation. Although this only improves the beam size in one direction, the increased integration time allowed us to obtain an RMS= 0.1 mjy, and allowed us to determine that the emission is resolved. The July 2011 point source fit underestimates the flux compared with the 2010 flux calculation (Table 3C.1), while the Gaussian fit is consistent with the previous observation. Looking at the 3 mm maps in Fig. 3.5, we see the emission was only resolved in one direction in 2011 and the emission is elongated with the beam. For reference the combined u-v coverage for both epochs is also presented in Figure 3.6, with the evenly spaced baselines from EW352 and the scattered tracks from H168. At 7 mm the point source fits are consistent, with exception of the 4 August 2012

124 102 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.5: GQ Lup 3 mm maps created using natural weighting with the clean algorithm. Contours are at -3, 3, 6, 9, 12, 15 times the image RMS. In August 2010: RMS is Jy/beam, beam size of arcsec. July 2011: RMS is Jy/beam, beam size of arcsec. Combined: RMS is Jy/beam; beam size of arcsecs. The red cross marks the location of GQ Lup as defined in Table 3.1.

125 3.4. Results 103 Figure 3.6: GQ Lup u-v coverage maps at 3, 7 and 15 mm (Top left 3 mm, top right 7 mm and bottom 15 mm).

126 104 Chapter 3. A multi-wavelength study of the GQ Lup system observation where the flux increased by a factor of 2, suggesting a contribution from other emission mechanisms besides thermal dust emission see Table To ensure that this flux variability was not caused from an absolute flux calibration error, a point source fit for the absolute flux calibrator QSO B was obtained for each epoch 8 see Table 3.4. The results from the fits are consistent with the flux uncertainty of 0.1 Jy and a variability of Jy observed in the ATCA calibrator catalogue 9 and the known absolute flux uncertainty of 15% at 7 mm. The calibration of the August 2012 data was also completed with the alternate absolute flux calibrator Neptune. Taking into account the known absolute flux uncertainty, we found both QSO B and Neptune provided the same point source fit value for GQ Lup in August In addition, two other sources (MY Lup and Sz 111) were observed during the same hour as GQ Lup on 4 August 2012 and no flux variability was observed for either of those sources see Table 2D.1 in Chapter 2. This suggests the observed increase for GQ Lup was not caused by a calibration error. Taking a closer look at the 7 mm maps without and with August 2012 observations (Fig. 3.7), there is no significant change in the morphology of the emission. Table 3.4: The point source fit values for the absolute flux calibrator QSO B for the frequency pairs in the 7 mm band. QSO B was self-calibrated unless otherwise noted. Date Flux (43 GHz) Flux (45 GHz) (Jy) (Jy) 10 July ± ± August ± ± August ± ± Oct ± ± Using Neptune as the absolute flux calibrator. At 15 mm, the point source fits for all three epochs are consistent (see Table 3.3), and GQ Lup was unresolved at this waveband see Fig In Fig. 3.9 we plot the results of our temporal flux monitoring at 3, 7 and 15 mm, including the four 7 mm observations over 2 years and the average fluxes at all three wavelengths presented in Table 3.5. For this analysis, we adopt the definitions introduced in Chapter 2 where short temporal monitoring is defined as epochs separated by less than a day and long temporal monitoring as epochs separated by a day or more. When no flux variability between epochs is present, the emission was considered to originate from thermal dust emission. A factor of 20-40% variability in the flux over long temporal 8 QSO B was self-calibrated. 9

127 3.4. Results 105 Figure 3.7: GQ Lup 7 mm maps created using natural weighting with the clean algorithm. Contours are at -3, 3, 6, 9, 12, 15 times the image RMS. For August 2010: RMS is Jy/beam, beam size of arcsecs. July 2011: RMS is Jy/beam, beam size of arcsecs. August 2012: RMS is Jy/beam, beam size of arcsecs. October 2012: RMS is Jy/beam, beam size of arcsecs. For the combined (thermal): RMS is Jy/beam, beam size of arcsecs. Combined (all): RMS is Jy/beam, beam size of arcsecs. The red cross marks the location of GQ Lup as defined in Table 3.1.

128 106 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.8: GQ Lup 15 mm combined map created using natural weighting with the clean algorithm. Contours are at -3, 3, 6 times the image RMS of Jy/beam and a beam size of arcsecs. The red cross marks the location of GQ Lup as defined in Table 3.1. The short integration times did not allow a map for each individual epoch to be created. monitoring can generally be considered to be from thermal free-free emission, and flux variability by a factor of 100% or more over short temporal monitoring can be considered to be from non-thermal emission. Table 3.5: Average millimetre band fluxes of GQ Lup from all epochs of the observations presented in Table 3.3. At 7 mm the emission with (all) and without (thermal) the August 2012 data is included. The total uncertainty was taken into account for each epoch to determine the average. Wavelength Flux Total uncertainty Notes (mm) (mjy) (mjy) From Dai et al. (2010) thermal all Before including the August 2012 data, the 7 mm fluxes showed no inter-epoch variability over timescales of months and years, suggesting the dominant emission mechanism is thermal dust emission. However, in August 2012 the 7 mm flux increased by more than a factor of two, which suggests the presence of other emission mechanisms. Since the next

129 3.4. Results 107 available observation was 6 weeks later, there is no way of knowing if this excess lasted for more than one day. To further test the 7 mm variability, we tried to look for intra-epoch variability at 7 mm (the phase calibrator is observed for 2 minutes between each 10 minute target scan). In August 2012, we observed multiple sources in the Lupus clouds along with GQ Lup. Since we needed more than 20 minutes on source to obtain a > 3σ detection, we could only divide the total integration time of 53 minutes into two intra-epochs (a 30 minute and a 23 minute scan), which were separated by 2 hours. We obtained point source fits of 1.3 ± 0.1 mjy and 1.2 ± 0.1 mjy respectively, suggesting that the flux remain fairly constant over 2 hours. A constant flux over such a short timescale might suggest that the excess emission at 7 mm was from thermal free-free emission rather than non-thermal emission. If we assume the other three 7 mm epochs are dominated by thermal dust, then the average thermal dust emission over these three epochs is 0.57 ± 0.16 mjy (see Table 3.5), and we estimate that > 50% of the emission observed on 4 August 2012 was excess emission. This value is higher than the 20% excess emission estimated in the Taurus-Auriga sample of Rodmann et al. (2006) and on the high-end of the expected flux variability of thermal free-free emission. Long temporal monitoring at 15 mm shows that the flux is constant and suggests that the dominant source of emission is thermal dust emission see Fig Note we were unable to determine short-term intra-epoch variability at 15 mm, since the source was not detected in a single 10 minute scan. Millimetre spectral slope Figure 3.10 shows the mm SED for GQ Lup. The data includes the fluxes from each epoch presented in Table 3.3, along with the 1.3 mm SMA flux from Dai et al. (2010). We assume a single power law can be used to represent the dust emission from mm through to cm wavelengths. For this analysis, we use the least-squared method to fit two lines to the average fluxes (Table 3.5) to determine the millimetre spectral slope α. To obtain the average flux at each wavelength, the total uncertainty for each epoch was taken into account. This allows us to determine the uncertainties of α, In Fig. 3.10, the dashed blue line is the fit the thermal dust emission (i.e., excluding the 4 August 2012 data), while the solid green line is a fit to all the data. Error bars include the total uncertainty. We find that α therm = 2.31 ± 0.36 and α all = 2.28 ± The value for α therm is consistent with the previous results of α therm = 2.2 ± 0.4 from Chapter 2. No break in the slope from

130 108 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.9: Temporal flux monitoring of GQ Lup at 3, 7 and 15 mm using values from Table 3.3. The green side triangles correspond to 3 mm, solid black circles to 7 mm and solid blue triangles to 15 mm. The total uncertainty is included in the error bars. The averages represent the average fluxes at each waveband (as presented in Table 3.5). The green dash-dot-dot line corresponds to the average 3 mm flux; the black dash-dot-dash line corresponds to the average 7 mm flux from thermal dust emission only; the red dash-dash line the average 7 mm flux for all the emission; and the blue solid line the average 15 mm flux.

131 3.4. Results mm is observed. Averaged flux values all thermal Flux Jy Wavelength metres Figure 3.10: Millimetre SED for GQ Lup. Plotted are the fluxes from each epoch presented in Table 3.3 and the 1.3 mm flux value obtained by Dai et al. (2010) with SMA. The average fluxes presented in Table 3.5 were used to determine the least-squared fits to the millimetre slope. Dust opacity index Our multi-epoch observations of GQ Lup show that the fluxes at 3 and 15 mm were constant over a timescale of a year, and that at 7 mm the flux could be variable on a timescale of months. The emission was found to be extended at 3 mm. We will assume therefore that the 3 mm emission comes from an extended optically thin region, and thus we can use the spectral slope to estimate the dust opacity index. This in turn can be used to estimate the maximum grain size. We assume that the dust opacity index β can be estimated from the millimetre spectral slope α therm via β α therm 2. For GQ Lup we found the α therm = 2.31 ± 0.36 from 1 15 mm, which gives β 0.31 ± 0.36, suggesting grain growth to cm-sized pebbles. This result is consistent with β estimates from 1 to 3 mm obtained by Ubach et al. (2012).

132 110 Chapter 3. A multi-wavelength study of the GQ Lup system Summary of GQ Lup observations We have detected GQ Lup at 3, 7 and 15 mm in multiple epochs over a period of 2 years. The emission was resolved at 3 mm, but not at 7 or 15 mm. The fluxes at 3 and 15 mm were found to be constant on timescales of a year. We did however, find evidence of excess emission at 7 mm for one of the epochs. This excess was estimated to be double the average thermal flux emission expected at 7 mm and no intra-epoch variability was found over a 2 hour period. This could suggest the excess emission is from thermal free-free emission, although non-thermal dust emission can not be ruled out given the large change in flux. Finally, we determined a dust opacity index β 0.3, suggesting the disc contains pebble-sized grains. 3.5 Modelling of GQ Lup In the previous sections we presented multi-wavelength observations of GQ Lup, and found the system could be hosting large pebble-sized grains. This result, along with the large microwave dataset and additional optical and infrared data from the literature, makes GQ Lup a good candidate for radiative transfer modelling. From observations we can obtained estimates of dust masses, disc outer radii and maximum grain sizes at a particular wavelength. However, this is not the complete picture of the system. There are some disc parameters (such as flaring exponent, disc inclination, dust settling exponent and the surface density exponent) which are not constrained through observations. Modelling is a way to combine the known information from multi-wavelength observations to gain additional insights into unconstrained disc and dust parameters and obtain an overall picture of the system. Moreover, modelling can be used to explore suggested inconsistencies between datasets. In the following sections we will use the available GQ Lup multi-wavelength data along with theoretical information to model the GQ Lup system. Attempting to fit multi-wavelength datasets with one model to explore dust and disc properties is not new. Pinte et al. (2008) constrained the inner and outer disc radii of IM Lup using infrared imaging, millimetre visibilities, the full SED and radiative transfer modelling. Their modelling also demonstrates that the scale height varies with the flaring index, suggesting the outer parts of the disc are in hydrostatic equilibrium (Pinte et al., 2008). These results were only obtained through successfully fitting a model to the multi-wavelength dataset. As discussed in Section 3.2, GQ Lup was modelled by Dai et al. (2010) using the radiative transfer code radmc (Dullemond & Dominik, 2004). They considered a parametric

133 3.5. Modelling of GQ Lup 111 disc with a sharp outer edge to mimic tidal truncation by the companion, assuming a surface density power law Σ(r) = Σ o (r/r o ) α d truncated at an outer radius, R out, where Σ o is the reference surface density at r o and α d is the surface density exponent. They also assumed a disc scale height of the form h(r) = h o (r/r o ) β h, where h o is the reference scale height at r o and 1 + Ψ is the flaring angle. For the star, they took T eff = 4060 K and L = 1.5L. The inner disc radius was set to 0.09 au (the dust sublimation radius), the maximum grain size to 1 mm and the gas-to-dust ratio to 100 (Dai et al., 2010). They focused their modelling on two parameters: R out and the surface density power-law index p. R out was set to 25, 50 and 75 au and α d to 1 and They compared their models to the full SED and their 1.3 mm SMA visibilities, and found R out = 50 au to be a good fit to the SED and the visibilities. However, this model does not account for the longer mm wavelengths, which provide information on the large grains that affect the model s temperature distribution and dust disc mass estimates present in the disc. For our modelling we used the 3D radiative transfer code mcfost rather than radmc. Although both codes are similar, they determine the flux and dust disc masses differently. radmc calculates the flux at a given wavelength from the dust disc mass, where the dust disc mass is determined by integrating the surface density over the inner and outer disc radii(dullemond & Dominik, 2004). mcfost calculates the flux for each cell grid using the temperature profile and the dust mass of each cell. The temperature profile is determined by evaluating the radiative equilibrium equation assuming the dust grains are at radiative equilibrium within the entire volume. The dust opacities are independent of temperature and two possible assumptions are made about the gas and dust interactions: either the gas and dust are in local thermodynamic equilibrium throughout the disc or there is no thermal coupling between the gas and dust (Pinte et al., 2006; Pinte, 2006). Similar to radmc, the dust mass is then determined by integrating the surface density and outer disc radii (Pinte et al., 2006; Pinte, 2006). We first reproduce the Dai et al. (2010) models using mcfost, to see if their models are consistent with our ATCA 3 and 7 mm fluxes. We found that the mcfost models are a good fit to the SED (up to 1 mm) and match well the 1.3 mm SMA visibilities, consistent with Dai et al. (2010) see Fig However we find that their models fail to produce enough flux to match our 2010 ATCA data at 3 and 7 mm, and are very poor fits to the visibility data at 3 mm. The aim this section is to improve the GQ Lup model by using mcfost to model the thermal dust emission and simultaneously reproduce the full SED up to 15 mm and the visibility data at 1 and 3 mm. We use the photometric data from the literature along

134 112 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.11: Dai et al. (2010) models for GQ Lup with data from the literature and ATCA 3 and 7 mm data. Top: SED of GQ Lup with 3 and 7 mm ATCA fluxes from Chapter 2. Bottom left: 1.3 mm SMA visibilities, right: 3 mm ATCA visibilities. The solid curves mimic the models of Dai et al. (2010) with the different R out and the squares correspond to the observed data. The models are clearly a poor fit to the 3 and 7 mm ATCA data. Figures created by Francois Menard using mcfost originally published in our ATCA 2011 April proposal.

135 3.5. Modelling of GQ Lup 113 Table 3.6: Photometric data used for our of GQ Lup modelling. Wavelength Flux Uncertainty Reference (µm) (W/m 2 ) (W/m 2 ) WFI WFI WFI WFI WFI MASS MASS MASS Spitzer/IRAC AKARI Spitzer/IRAC Spitzer/IRAC IRAS Spitzer/IRAC Spitzer/IRAC IRAS AKARI AKARI AKARI Dai et al. (2010) This work This work This work

136 114 Chapter 3. A multi-wavelength study of the GQ Lup system Figure 3.12: GQ Lup spectral energy distribution. This is an updated version of the SED presented in Chapter 2, which incorporates the IRAS, Akari and 15 mm ATCA data. The stellar photosphere (dashed line) was obtained using the method of Furlan et al. (2006), which assumes a blackbody and normalises to the J-band flux. Photometry from Table 3.6. Legend: * WFI; + 2MASS; IRAS; IRAC; Akari; SMA, Dai et al. (2010); ATCA, this work. with the SMA and ATCA millimetre observations see Table 3.6 for the photometry and Fig for the full SED of GQ Lup. For this analysis we exclude our August 2012 ATCA data at 7 mm, which shows signatures of excess emission. We use the Dai et al. (2010) 1.3 mm visibilities and our July mm visibilities (Fig. 3.1) Model setup mcfost is a 3D radiative transfer code based on the Monte Carlo method. The computation uses monochromatic photon packets which are propagated in a spatial grid in cylindrical coordinates, where light scattering properties and re-emission are also taken into account (Pinte et al., 2006). Our model consists of a central star surrounding by a disc and does not include any contribution from the companion. The disc s geometry is defined by inner and outer disc radii (R in, R out ), a parametric surface density Σ(r) = Σ o (r/r o ) α d and scale height h(r) = h o (r/r o ) β h, where Σ o is the surface density and h o the scale height at r o = 100 au, α d is the surface density exponent and β h is the disc flaring exponent. The dust grains are assumed to be spherical and homogeneous and the grain size distribution follows the differential power law dn(a) a p da, between the maximum and minimum grains sizes (a max and a min respectively), where we adopt the Mathis et al. (1977) size distribution with p = 3.5.

137 3.5. Modelling of GQ Lup Parameters The parameters in the model can be divided into two groups: stellar parameters and disc parameters (which include the dust parameters). To minimise the number of free parameters we started by fitting the stellar parameters T eff and R to the stellar photosphere assuming a stellar mass of M = 1.05 M (Donati et al., 2012), a NextGen stellar atmosphere model with log(g) = 3.5 (Donati et al., 2012) and a distance to GQ Lup of 156 pc (Neuhäuser et al., 2008b). The parameter range for T eff and R cover the values presented in the literature, with T eff values of 4000, 4200 and 4400 K (Neuhäuser et al., 2005; McElwain et al., 2007; Donati et al., 2012) and R values of 1.6, 1.7, 1.8 and 1.9 R (Broeg et al., 2007; Seperuelo Duarte et al., 2008; Donati et al., 2012). From 1000 models, we found a Bayesian probability of occurrence of 0.8 that T eff = 4400 K and R = 1.8 R best represent the stellar photosphere. λ*f λ (W/m 2 ) Model Obs Wavelength (microns) Figure 3.13: Best fit to the stellar photosphere for GQ Lup. Blue curves correspond to the model while green points to the observed fluxes. For the disc parameters we set the gas-to-dust ratio equal to the ISM value of 100. The minimum grain size is fixed at a min = 0.03 µm, which is small enough that the exact value has no effect on the model (Pinte et al., 2008). From the Spitzer data between 5 35 µm (inner 10 au of the disc), Olofsson et al. (2010) found grains with sizes 0.1, 1.5 and 6 µm to be composed mainly of amorphous silicates and some crystalline silicates. Thus we set the grain composition to the standard Draine & Lee (1984) astronomical silicates with the ultraviolet bump smoothed. We explored 9 disc parameters: disc inclination i, R in, R out, flaring exponent β h, a max, M dust, surface density exponent α d, scale height h o and

138 116 Chapter 3. A multi-wavelength study of the GQ Lup system dust settling exponent ξ. We allow i to take values between 0 60 to cover the three values presented in the literature: 0, 30 and 50 (Broeg et al., 2007; Seperuelo Duarte et al., 2008; Hügelmeyer et al., 2009; Pontoppidan et al., 2011; Donati et al., 2012). R in ranges between 0.07 and 0.09 au (Hügelmeyer et al., 2009; Dai et al., 2010; Pontoppidan et al., 2011) which also includes the value used in Dai et al. (2010) model of 0.09 au. The 1.3 and 3 mm data were used to set the range for R out, where the minimum value of 50 au was obtained from the 1.3 mm SMA data (beam size of ) and the maximum value of 500 au was obtained from the 3 mm 2010 ATCA data (beam size of ). M dust values are set by the 3 mm dust mass estimate of M and the uncertainty of that value (an order of magnitude). From the mm spectral slope, we know there are grains up to mm sizes, thus a max values range between µm. The last four free parameters, β h, α d, h o and ξ, were not constrained by observations, thus we used the ranges presented in Pinte et al. (2008) and Dai et al. (2010). Table 3.7: Parameters grid for GQ Lup model. Parameters Range N sample T eff (K) R (R ) a min (µm) gas-to-dust i (degrees) R in (au) 0.07, 0.08, R out (au) 50, 100, 250, M dust (10 3 M ) 0.1, 0.5, 1 3 β h 1.0, 1.1, a max (µm) ,1 10 3, α d -1.5, -1.0, h 1 (au) 11.0, 12.0, ξ 0.0, 0.1, Scale height of the disc at r o = 100 au. A summary of each parameter range is presented in Table 3.7. This results in a total of 87,480 models. Note mcfost determines the flux for each inclination automatically, thus the parameter sweep produces 8,748 models each with a SED and visibilities at 1.3 and 3 mm for each of the 10 inclination angles Results The SED and visibility fits are performed by minimising the reduce χ 2 value given by

139 3.6. GQ Lup discussion 117 n χ 2 (F obsλ F modelλ ) 2 =, (3.1) λ=0 F 2 λ χ 2 red = χ2 /(N 1 ϕ), (3.2) where F obsλ and F λ are the values for the observed flux and total uncertainty in the observed flux at wavelength λ respectively. F modelλ is the modelled flux at wavelength λ as determined by mcfost. N is the number of observed data points and ϕ is the degrees of freedom in the model. We evaluated the χ 2 red for the SED and the 3 mm visibilities, and χ2 for the 1 mm visibilities for all the models. Note that there are only three fluxes for the 1.3 mm visibilities, thus determining the χ 2 red would be inappropriate given that we have 9 free parameters. In Fig we present four best fit models: (a) the minimum χ 2 red for the SED χ2 SED ; (b) the minimum χ 2 red for the July mm visibilities χ2 3mm ; (c) the minimum χ2 for the Dai et al. (2010) 1.3 mm visibilities χ 2 1mm and (d) the model with the minimum total χ 2 red, which we defined as χ2 total = χ 2 SED + χ2 3mm, the minimum of the sum of the χ2 red. Since we cannot estimate a meaningful χ 2 1mm, this value is not included in χ2 total. Although these four models do a better job at fitting the SED from 3 7 mm than the Dai et al. (2010) models, we are still underestimating the infrared emission from µm, and all our models are a poor fit to the 1 mm visibilities. Table 3.8 presents the parameters for each of the four models presented in Fig Models with R out = 250 au and M dust = M consistently do a better job at fitting the longer wavelengths in the SED, while R out = 50 au is needed for the 1 mm visibilities, yet is still a relatively poor fit! Even though our parameter grid was not overly extensive, the ranges of values investigated were driven by observations and theoretical constraints, and yet we were unable to find a model that fits all 3 datasets the SED, 1 and 3 mm visibilities. 3.6 GQ Lup discussion From both the multi-wavelength data and the 3D radiative transfer modelling, it appears that some key information about the GQ Lup system is missing. Our model results suggests that to fit the longer mm emission in the SED and the 3 mm visibility structure we need a massive extended disc with an i 30, R out = 250 au and grains up to cm sizes.

140 118 Chapter 3. A multi-wavelength study of the GQ Lup system λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) λ*f λ (W/m 2 ) 1e χ 2 = red mm ATCA Baseline (meters) λ*f λ (W/m 2 ) 1e χ 2 = mm SMA Baseline (meters) (a) Best SED fit. χ 2 SED. λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) λ*f λ (W/m 2 ) 1e χ 2 = 6.55 red mm ATCA Baseline (meters) λ*f λ (W/m 2 ) 1e χ 2 = mm SMA Baseline (meters) (b) Best 3 mm visibility fit. χ 2 3mm λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) λ*f λ (W/m 2 ) 1e χ 2 = red mm ATCA Baseline (meters) λ*f λ (W/m 2 ) 1e χ 2 = mm SMA Baseline (meters) (c) Best 1 mm visibility fit. χ 2 1mm λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) λ*f λ (W/m 2 ) 1e χ 2 = 9.63 red mm ATCA Baseline (meters) λ*f λ (W/m 2 ) 1e χ 2 = mm SMA Baseline (meters) (d) Model with the minimum χ 2 total. Figure 3.14: Best fit models for GQ Lup. Left: the modelled SED curve overlaid with the photometric measurements; middle: July 2011 visibilities at 3 mm; right: 1.3 mm visibilities. Blue curves correspond to the model while green points to the observed fluxes. For larger versions of the SEDs see Appendix 3D.1.

141 3.6. GQ Lup discussion 119 Table 3.8: Parameter values for the four best fit models. (a) best χ 2 SED, (b) best χ2 3mm, χ 2 1mm, (c) best χ2 total = χ2 SED + χ2 3mm Model i M dust ξ α d β h R in R out a max h (degrees) (10 3 M ) (au) (au) 10 4 (µm) (au) χ 2 total a b c d These values are dramatically different from low mass compact disc with R out = 50 au obtained by Dai et al. (2010), which are more consistent with the best fit model we obtained for the 1 mm visibility structure. On the other hand, the values for R in = 0.09 au 10 from our best fit SED model is consistent with the fixed values set by Dai et al. (2010). Figure 3.15: Overlaid GQ Lup millimetre visibilities. Both sources scaled such that peak= 1 mjy. We note a potentially significant difference between Dai et al. (2010) and our models are the stellar parameters. For our models we used T eff = 4400 K and R = 1.8 R, while Dai et al. (2010) used T eff = 4060 K and R = 2.2 R. Although their values are within the early estimates by Neuhäuser et al. (2005) of T eff = 4060 ± 300 K and estimates by Broeg et al. (2007) and Seperuelo Duarte et al. (2008) of R = 2.55±0.41 and 1.8±0.3 R respectively and give a good fit to the stellar photosphere, we did not find the Dai et al. (2010) values for T eff and R to be a good fit for the stellar photosphere data when other 10 Note our sublimation radius is slightly different from Dai et al. (2010) due to the different T eff.

142 120 Chapter 3. A multi-wavelength study of the GQ Lup system disc properties are taken into account. It should also be noted that neither the Dai et al. (2010) and our models fit the mid-infrared emission, and as we have shown, we were not able to find a model which fits both the shallow 1.3 mm and the steep 3 mm visibilities simultaneously. This mismatch in slopes is clearly seen when we scale the visibilities at 1.3 and 3 mm and overlay them on one plot see Fig Assuming both the 1.3 and 3 mm wavelengths probe the same disc, the same grain sizes and grain compositions, it is unclear what physical mechanisms could cause a difference in the grain distribution and/or grain composition between the two wavelengths which should be probing the same region of the disc. One possibility is that the 1.3 mm band is probing a compact circumprimary disc, while the 3 mm band is probing an extended circumbinary disc and the (unresolved) circumprimary disc. Such a scenario could potentially explain the inability to fit both datasets.however, very little is known about either the orbit of GQ Lup b or its mass. If the companion is a brown dwarf, we can expect the disc to be truncated. Following Artymowicz & Lubow (1994), and assuming an extreme mass ratio between GQ Lup and GQ Lup b of 2:1 assuming that the companion is in the disc plane, we would expect the inner portion of the circumbinary disc to be truncated at 2.5a, where a is the separation of the binary which we assume to be 240 au. This gives R in 600 au for the circumbinary disc. For the outer edge of the circumprimary disc, Artymowicz & Lubow (1994) found r p 0.37a, which gives us an r p 89 au for GQ Lup. If the separation is instead 0.7 ( 100 au) (Neuhäuser et al., 2005), then R in for the circumbinary disc would be 250 au and r p 37 au. Although these estimates look promising. GQ Lup b mass is poorly constrained and is certainty less than 0.5 M. This would result in a reduced disc truncation. Furthermore, we need to check that these disc sizes are consistent with both the 1.3 and 3 mm observations. The ATCA July mm observations had resolution of 500 au, and thus we are not able to resolved any structure < 500 au, including the companion or any gap created by the companion. Taking a conservative look at the 1.3 mm SMA data, which has synthesis beam of , we suggest the data has a resolution of 100 au, and since the primary beam is 50 at 1.3 mm, the interferometer should have detected the circumbinary disc emission greater than 3σ, where the RMS is 2.8 mjy. The 1.3 mm clean map of Dai et al. (2010), shows some 2σ structures detected beyond 0.7 from the source, which is encouraging. Note short baselines (< 70kλ) where not used in the observations. Additionally, M dust at 1.3 mm less than M dust at 3 mm, and the model results suggest a more massive disc ( M ) is needed to replicate

143 3.6. GQ Lup discussion 121 <700 au circumbinary <500 au ~100 au GQ Lup b circumprimary Figure 3.16: Proposed structure for the GQ Lup system. Image not to scale. the 3 mm visibilities, while an order of magnitude smaller mass is needed to match the 1.3 mm. These results are consistent with the suggested scenario of the 1.3 mm emission coming from a compact circumprimary disc while the 3 mm emission comes from both discs. Finally, using our 7 mm ATCA observations we can set an upper limit for the circumbinary disc outer radius of 700 au, since our unresolved October mm observations had a resolution of Figure 3.16 presents a schematic of the proposed two disc scenario. We propose the GQ Lup system might have a compact circumprimary disc 100 au in size and a circumbinary disc with R in < 500 au and R out 700 au, with the companion orbiting in the gap between the circumprimary and circumbinary discs 11. For grain growth to occur in the outer regions of the disc, a grain trapping mechanisms must be present. One possible mechanism is large-scale spiral density waves caused by disc-planet interactions (Tanaka et al., 2002). Spiral density waves have been detected in the infrared images of the transition disc around HD135344B (Garufi et al., 2013; Muto et al., 2012) and in the near-infrared images of the transition disc around HD (Boccaletti et al., 2013). In both causes the pile-up regions are suggested to be sites of grain growth. GQ Lup s companion located at 240 au (Pontoppidan et al., 2011) places it well within the outer disc region and given it large (though uncertain) mass, it is likely to be interacting with the disc. This will likely result in a gap, as proposed by our 11 Note the gap might be smaller, and possibly not completely devoid of grains.

144 122 Chapter 3. A multi-wavelength study of the GQ Lup system model. Additionally, the companion could generate large-scale spiral waves in the outer disc, which would trap grains at the pressure maxima and result in grain growth. Note that the 7 and 15 mm emission is unresolved and thus we cannot pinpoint the location of the pebble size grains within the disc. Is the 1 mm emission coming from a circumprimary disc while part of the 3 mm emission from a circumbinary disc? Does this system have two discs? How would this scenario change if a GQ Lup b is planet instead of a brown dwarf? In order to answer these questions we would need to obtain shorter spacing data at 1.3 mm to determine if circumbinary disc emission is present and a high resolution image at 3 mm to resolve the inner 70 au of the disc to determine if a gap exists. More detailed modelling is also required to test if our two disc model can simultaneously fit the 1.3 and 3 mm visibilities and if a two disc scenario would also fit the mid-infrared emission, which currently is being missed by both our best models and the Dai et al. (2010) models. If a model is found, we would not only be able to constrain their different disc sizes but also the masses and the grain properties in both discs. 3.7 Conclusions Follow-up observations of GQ Lup were carried out with ATCA at 3, 7 and 15 mm. We analysed the 3 mm high-resolution data to determine if the emission was extended. We conducted temporal flux monitoring at 7 and 15 mm to determine if the flux was stable over a range of timescales and if the dominant emission mechanisms is thermal dust emission. From these results, we were able to analyse the mm spectral slope from 1 15 mm to determine if cm-sized pebbles were present in these protoplanetary discs. We detected GQ Lup at 3, 7 and 15 mm in multiple epochs. The emission was resolved at 3 mm and the flux was found to be generally stable at 3, 7 and 15 mm, suggesting the emission is dominated by thermal dust emission. There was one epoch at 7 mm were evidence of excess emission was detected, possibly from an outburst. Lastly, we are able to confirm the presence of large pebble-sized grains in GQ Lup protoplanetary disc. In addition to these observations, we conducted 3D radiative transfer modelling of the GQ Lup system using mcfost. From our models we were able to constrain the stellar parameters (T eff = 4400 K and R = 1.8 R ) and the disc inclination (i = 30. However, we were unable to simultaneously fit the shallow 1.3 mm visibilities and the steep 3 mm visibilities assuming a simple disc structure and one grain composition. Our best fit to the ATCA data between 3 to 15 mm corresponds to a large disc with R out 250 au, M dust = M and a maximum grain size of 1 cm, while the best fit to the SMA

145 3.7. Conclusions mm data corresponds to a small disc with R out 50 au and M dust = M, with both models underestimating the flux from µm to 1.3 mm. This mismatch between the 1.3 mm and 3 mm visibilities and the modelling leads us to propose a two disc scenario for the GQ Lup system, comprising of a compact circumprimary disc 100 au in size and an extended circumbinary disc with R in < 500 au and R out 700 au, where the companion is in the gap between the circumprimary disc and circumbinary disc. Future work will involved a more detailed investigation in the millimetre wavebands to see if a two disc system is detected and determine if a two disc system model can simultaneously fit the full SED and both the 1.3 and 3 mm visibilities.

146 124 Chapter 3. A multi-wavelength study of the GQ Lup system 3.A The amplitude uncertainty derivation The amplitude uncertainty was obtained by taking the full derivative (total differential) of R 2 + I 2, where R and I are the real and imaginary parts of the visibility amplitudes respectively, and σ R and σ I are the uncertainties in the real and imaginary parts of the visibility amplitudes respectively. amplitude = a = R 2 + I 2 σ amp = [ ( a R σ R ) 2 ( ) ] a 2 + I σ I ( = [ R 2 ( ) ] R) R 2 + I σ I R 2 + I σ 2 I = [(R σr ) 2 R 2 + I 2 + (I σ I) 2 ] R 2 + I 2 = (R σ R ) 2 + (I σ I ) 2 (R 2 + I 2 )

147 3.B. uvamp versus uvplt B uvamp versus uvplt There are two methods that we use to create the visibility plots: uvamp and uvplt. uvamp obtains the flux by taking the average amplitude within a u-v distance over time. The u-v dataset is binned in annuli according to u-v distance. This task also determines the statistical error bars and the expected amplitude for zero signal. uvplt uses the real and imaginary parts of the visibilities, calculates the amplitude by taking the average over each baseline over time. This methods provides one data point per baseline pair, avoiding the bin dependence. The error bars are ±1 standard deviation of the scalar average. The amplitude values can then be evaluated using R 2 + I 2 and the amplitude uncertainties via σ amp = ((R σ R ) 2 + (I σ I ) 2 )/(R 2 + I 2 ) see Appendix 3.A for derivation of the uncertainty. If the target is a point source, averaging the amplitude over baselines over time or within a u-v distance over time will not change the amplitude vs u-v distance plot. But if the source is elliptical (extended emission) or a two disc system, then averaging over baseline over time can provide a different result from averaging within u-v distance over time. Take for example the flux coming from baseline (1)(5) (ant(1)(5) is the longest baseline for the EW352 array configuration) for GQ Lup at 3 mm in July If the amplitude is averaged over that baseline pair over time, the result is one data point at 96 kλ with flux= 0.73 mjy see Fig. 3B.1. Figure 3B.1: Left: u-v distance vs time for ant(1)(5) at 3 mm; right: average u-v distance over time for ant(1)(5). In order to obtain the same dataset with uvamp, the select function needs to be set to select = ant(1)(5) and the bins to bin=4,200,klamb (bin size of 200 kλ). The resulting

148 126 Chapter 3. A multi-wavelength study of the GQ Lup system flux= 0.73 mjy at 100 kλ see Fig 3B.2. This provides consistent results between both methods of creating the visibility plots. Figure 3B.2: uvamp plot at 3 mm for ant(1)(5). However, in general this is not how uvamp plots are generated. Fig 3B.3 is a plot of u-v distance as a function of time, and if we select a region between kλ (equivalent to the last bin in the uvamp plots of the 3 mm GQ Lup observations in July 2011 see Fig. 3.1), there are two baseline pairs (ant(1)(5) and ant(2)(5)) which cross that position of the u-v distance during the time interval 07: The baselines from kλ correspond to au assuming a source distance of 156 pc. If the system is a compact circumprimary disc with an extended circumbinary disc, where the circumprimary disc is truncated at < 500 au, then the average flux from a u-v distance between kλ would be mainly (if not all) from the circumprimary disc. On the other hand, uvplt averages the amplitudes over the entire ant(1)(5) baseline pair to obtain a single point at 96 kλ, which covers from kλ over the time period of :00. This includes an additional 6 hours on the shorter baselines ( au, assuming 156 pc), weighting the average towards the extended emission from the circumbinary disc. In this case, this affect diminishes with decreasing baseline pair length, since the shorter baselines cover a smaller range in u-v distance. Therefore, plotting the visibilities with uvamp and uvplt will be consistent to a u-v distance of 60 kλ. We can use the visibilities to directly obtain information about the disc extent. Information about the disc geometry, however, can be obtained indirectly by deprojecting the visibilities. This is generally achieved by following the method of Lay et al. (1997), where the average visibilities are determined within concentric annuli created assuming the disc is circularly

149 3.B. uvamp versus uvplt 127 Figure 3B.3: Baseline pairs for GQ Lup at 3 mm in July Left: all the baseline pairs and both frequency pairs are plotted as a function of u-v distance, with ant(2)(5) and ant(1)(5) on the far right. Right: ant(2)(5) (the two curves on the far left) and ant(1)(5) and frequency pairs at u-v distance between kλ. symmetric and flat, defined as R u-v = (d 2 a + d 2 b )1/2, where the major axis d 2 a = (u 2 + v 2 ) 1/2 and the minor axis d 2 b = (u2 + v 2 ) 1/2 cos ϕ cos i, i is the inclination angle between line of sight and disc axis, ϕ = arctan(v/u) P.A. and P.A. is the major axis position angle. The scatter in the imaginary part of the visibilities will be minimal (close to zero) when the correct disc position and inclination angles are used in the deprojection (Lay et al., 1997). Note that an uncertainty in the star s position can also create some scatter in the imaginary part of the visibilities (Lay et al., 1997; Hughes et al., 2007). If the deprojected visibilities pass through a null, it suggests a sharp edge in the emission which can indicative the presence of a gap. Hughes et al. (2007) used this method to determine that an inner gap is present in TW Hya disc. Note that this method is highly dependent on knowing the accurate location of the central star (to within a few hundredths of an arcsecond), which is generally not achieved (Hughes et al., 2007). Alternatively, information about the disc geometry can be obtained by comparing observational results to models generated by radiative transfer modelling of the disc system (Pinte et al., 2008; Andrews et al., 2012).

150 128 Chapter 3. A multi-wavelength study of the GQ Lup system 3.C Fitting results Table 3C.1: Summary of 3, 7 and 15 mm flux fittings for GQ Lup. (1) Date of observation. (2) Total on-source integration time used in the flux fitting. (3) & (4) Point source fit and Gaussian fit. (5) Gaussian size obtained from the Gaussian fit and error in the fit. (6) Synthesised beam size using natural weighting. (7) Visibilities RMS. (8) Factor of σ the point source fit is below the Gaussian fit (Fg=Fp+nσ). (9) Description of the amplitude as a function of u-v distance plots (uvamp) presented in Figs. 3.1, 3.3 and 3.4. (10) If the emission was considered resolved or not. Date Tinit FP FG Gaussian size Beam size RMS n uvamp Resolved (min) (mjy) (mjy) (arcsec) (arcsec) (mjy/beam) F, D Y, N 3 mm band (94 GHz) Aug ± ± ± ± ?? July ± ± ± ± D Y Comb ± ± ± ± D Y 7 mm band (44 GHz) August ± ± ± ± F N July ± ± ± ± F N August ± ± ± ± F N Oct ± ± ± ± F N Comb.(thermal) ± ± ± ± F N 15 mm band (18 GHz) July ± ± ± ± F N July ± 0.04 N/A F N Oct ± ± ± ± F N Comb ± ± ± ± F N Values originally presented in Chapter 2. uvamp: F-flat, D-drop. Resolved: Y-yes, N-no. Unable to determine a reliable Gaussian fit.

151 3.D. Model SED results D Model SED results λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) (a) Best SED fit. λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) (b) Best 3 mm visibility fit. λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) (c) Best 1 mm visibility fit. λ*f λ (W/m 2 ) χ 2 red = Obs Wavelength (microns) (d) Model with the lowest χ 2 total. Figure 3D.1: The modelled SED fit overlaid with the photometric measurements.

152 130 Chapter 3. A multi-wavelength study of the GQ Lup system 3.E DK Cha system DK Cha is an isolated YSO (Hughes et al., 1989) in Cha II (Olnon et al., 1986) at a distance of 178 ± 18 pc (Whittet et al., 1997) see Table 3E.1. It is a near face-on (i 18 ) system (van Kempen et al., 2009; Garcia Lopez et al., 2011) which is surrounded by an envelope and molecular outflow (Hughes et al., 1991; van Kempen et al., 2006, 2009). It is thought to be transitioning from an embedded disc to a protoplanetary disc (van Kempen et al., 2010; Garcia Lopez et al., 2011). Table 3E.1: Basic properties of DK Cha. Parameter DK Cha References RA (J2000) 12 h 53 m 17 s.2 1 DEC (J2000) Sp. Type F0 2 Cloud Cha II 3 Distance (pc) 178 ± 18 4 Comments Transitioning between Class I and II. 5 (1) van Kempen et al. (2006), (2) Spezzi et al. (2008), (3) Olnon et al. (1986), (4) Whittet et al. (1997), (5) van Kempen et al. (2006, 2009) Stellar properties DK Cha was first classified a Herbig Ae star by Hughes et al. (1991). It is the brightest infrared (Olnon et al., 1986; Porras et al., 2007) and sub-mm source (Henning et al., 1993) with known H α emission in Cha II (Hughes et al., 1991; Spezzi et al., 2008; Garcia Lopez et al., 2011). It is variable in the near-infrared (Hughes et al., 1991; Hughes et al., 1989) with variability on the timescale of days and years (Hughes et al., 1991). Hughes et al. (1991) detected a prominent H α P-Cygni profile and blue-shifted [OI] and [SII] forbidden lines in the optical spectra, suggesting the presence of an outflowing wind. This outflow has also been suggested to be driving the HH52-54 group of Herbig-Haro objects 14 to the north-east (Hughes et al., 1989; Hughes et al., 1991), however other possible driving sources have also been suggested (Bjerkeli et al., 2011). Spezzi et al. (2008) estimated the effective temperature to be 7200 ± 170 K, a stellar radius of 2.77 ± 0.14 R and a stellar mass of 2 M.

153 3.E. DK Cha system 131 Line of sight for DK Cha Figure 3E.1: Schematic of the YSO DK Cha, with the different components indicated. IWS is the internal working surface. On this scale the protoplanetary disc is not visible. Figure taken from fig. 5 of van Dishoeck et al. (2011), with a similar figure presented in Visser et al. (2012).