An Observational Study of Accretion Processes in T Tauri Stars

Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 841

HENRICUS CORNELIS STEMPELS

ACTA UNIVERSITATIS UPSALIENSIS UPPSALA 2003 Dissertation presented at Uppsala University for the Degree of Doctor of Philosophy in As- tronomy, to be publicly examined in Polhemsalen, Angstrû omlaboratoriet,¬ Uppsala, Monday, May 26, 2003 at 13:15.

ABSTRACT

Stempels, H.C.. 2003. An Observational Study of Accretion Processes in T Tauri Stars. Acta Universitatis Upsaliensis. Comprehensive Summaries of Uppsala Dissertations from the Fac- ulty of Science and Technology 841. 33pp. Uppsala. ISBN 91-554-5634-0

This thesis is a detailed observational study of the accretion processes in T Tauri stars (TTS). The interaction between the central star, the circumstellar disk and the magnetic field gives rise to a wide range of features in the spectra of TTS. The current picture of TTS is based on rather simple models assuming that accretion is a homogeneous and axisymmetric process. Although these models have been successful in explaining some observational signatures of TTS such as the shape of emission lines, the static nature of these models makes them unsuitable for describing the strong variability of the veiling spectrum and emission lines of TTS. An improved understanding of this variability is of key importance to study the dynamic processes related to the accretion flow and the winds. This study is based on a set of high-quality spectroscopic observations with the UVES spectrograph at the 8-m VLT in 2000 and 2002. These spectra, with exposure times as short as 10Ð15 minutes, have high spectral resolution and high signal-to-noise ratios and cover a large part of the optical wavelength range. From this dataset we determine the basic physical parameters of several TTS and model their photospheres. These models then serve as a basis for a detailed investigation of variations of the veiling continuum and line emission. We confirm that the level of veiling correlates with some of the strongest emission lines and that coherent changes in accretion occur on a timescale of a few hours, comparable to the free-fall time from the disk to the star. From the properties of the emission lines formed close to the central star and in the stellar wind we derive restrictions on the geometry of the observed systems. Because the intrinsic axial symmetry of a single star makes it almost impossible to disentan- gle rotational modulation from inhomogeneity and axial asymmetry of the accretion flow, we study a series of spectra of a close spectroscopic binary at different orbital phases and derive the 3D structure of flows between the disk and the star. Finally, we calculate the profiles of hydrogen emission lines by iteratively solving 3D NLTE radiative transfer in a state-of-the-art magnetospheric model.

Keywords: Star formation, T Tauri stars, Emission lines, Accretion, NLTE radiative transfer

Henricus Cornelis Stempels. Department of Astronomy & Space Physics. Uppsala Univer- sity. Box 515, SE-751 20 Uppsala, Sweden ([email protected])

c Henricus Cornelis Stempels 2003

ISBN 91-554-5634-0 ISSN 1104-232X Aan mijn ouders

List of Papers

This thesis is based on the following papers :

Paper I Formation of Fe X-Fe XIV coronal lines in the accretion shock of T Tauri stars, Lamzin, S. A., Stempels, H. C., & Piskunov, N. E., 2001, Astronomy & Astrophysics, 369, 965Ð970.

Paper II Spectroscopy of T Tauri stars with UVES – Observations and analysis of RU Lup, Stempels, H. C., & Piskunov, N., 2002, Astronomy & Astrophysics, 391, 595Ð608.

Paper III The photosphere and veiling spectrum of T Tauri stars, Stempels, H. C., & Piskunov, N., 2003, accepted for publication in Astronomy & Astrophysics.

Paper IV NLTE calculations of hydrogen emission lines in T Tauri stars, Stempels, H. C., von Rekowski, B., & Piskunov, N., 2003, to be submitted to Astronomy & Astrophysics.

Paper V The close T Tauri binary V 4046 Sagittarii, Stempels, H. C., & Gahm, G., 2003, to be submitted to Astronomy & Astrophysics.

Contents

1 Introduction ...... 9 2 T Tauri stars ...... 11 2.1 Observational characteristics ...... 11 3 Models of T Tauri stars ...... 15 3.1 Classical accretion models ...... 15 3.2 Magnetospheric accretion models ...... 16 3.3 Time-dependent models ...... 17 3.4 Open questions ...... 18 4 An observational study of T Tauri stars ...... 19 4.1 Observations ...... 19 4.2 Data reduction ...... 20 4.3 Spectral synthesis ...... 20 4.4 Magnetic ﬁelds ...... 22 5 Summary of papers ...... 23 5.1 Measuring accretion rates (Paper I) ...... 23 5.2 The extreme TTS RU Lup (Paper II) ...... 24 5.3 The veiling spectrum (Paper III) ...... 25 5.4 NLTE calculations of hydrogen emission lines in T Tauri stars (Paper IV) ...... 26 5.5 The spectroscopic binary V 4046 Sagittarii (Paper V) ...... 27

Introduction

When I in April 2000 and April 2003 traveled to Chile to observe with one of the largest optical telescopes on Earth, the goal was not to peer in to the far corners of the universe. On the contrary, my main interest was to obtain a glimpse of the dynamical processes around young neighboring stars, only 50 to 100 parsec away, which in astronomical terms is on our doorstep. These stars, called T Tauri Stars, are the younger brothers of the most common stars we know. Even though they are common and in size and luminosity similar to our Sun, they are too faint to be observed with the naked eye. On top of that, the observational features that are typical for young stars are even fainter and highly variable in time, which means that observational studies of T Tauri stars were normally concentrated on just a few aspects of these stars. In the mean time, the distance between the limited observational studies and the theoretical work only became larger, with models that are capable of explaining almost any feature of T Tauri Stars, as long as you put in the correct parameters. The breakthrough came in 2000, when new and larger telescopes opened the door for a ﬁrst comprehensive and simultaneous study of all features present in the optical spectra of T Tauri Stars. When I embarked on this project it was unclear what the results would be, but, in retrospect, the data couldn’t be better. The observations show all processes around T Tauri stars in their full glory, making it possible to improve our understanding of these systems.

T Tauri stars

The evolution of stars occurs on a vast amount of different time-scales. A typical solar-type star spends the first few million years of its life interacting with the gas and dust remaining from the nebula it was born in. Compared with the total lifetime of such stars, over 10 billion years, this is a rather short period. Though short, this period is very interesting from an evolutionary point of view, because this is also the period during which large changes occur in the overall appearance of the stellar system. This is the phase when the circumstellar disk that is so typical for young stars disappears, the star settles into a much more quite and stable state and planetary systems may form. Stars come in all colors and sizes. In terms of size, mass and luminosity, our own sun is a relatively common star. Stars of similar masses as the sun have very similar appearances. However, such uniformity is not present among young solar-type stars. The observational properties of young solar-type stars differ strongly from star to star and there exist several groups of young stars based on observational classification. In my thesis I will focus on one particular group of young solar-type stars. These stars are named T Tauri stars, after their prototype T Tauri. This group of stars was first described by Joy (1945) and is defined by certain observational characteristics.

2.1 Observational characteristics The basic observational characteristics of T Tauri stars (TTS) resemble those of main-sequence stars. Their spectral types range from early G-type (Teff ≈ 6000 K) down to M-type (Teff ≈ 3500 K). TTS have masses of about 0.5 Ð 1.5 M and typical surface gravities are log g ≈ 3.5 Ð 4.0, which is slightly lower than what their main-sequence counterparts have. Typical rotational velocities are small (v sin i ∼< 30 km sec−1). Photometry reveals strong continuous excess emission in both the IR and the UV regions of the spectrum. High-resolution imaging shows large concentrations of dust and gas around TTS. Although direct imaging of circumstellar disks is not yet possible, high- resolution images of TTS and in particular of bipolar outflows (jets) show ob- scurations in the equatorial plane suggesting the presence of a disk. Spectroscopic observations of TTS show many strong emission lines, most

11 14

0 4850 4855 4860 4865 4870 4875 Wavelength

Figure 2.1: Illustration of the strong variability of the Hβ emission line in the spectrum of the very active TTS RU Lup. The figure consists of 35 observations during 4 nights in April 2002. of hydrogen and helium, but also emission lines of metals such as iron and forbidden transitions of oxygen and sulphur are present. The strength of the emission lines is considerable, often more than 10 times stronger than continuous emission. An equivalent width for the Hα line of 10 Ð 100 Aû is not uncommon. Sometimes also the Balmer continuum is seen in emission. The presence of strong emission indicates that large volumes of excited gas are located close to the central star. Emission lines tend to be very wide, with half widths corresponding to several hundreds km sec−1. Emission lines of H, Ca, and Na often exhibit P-Cygni profiles, indicating that mass is ejected from the system through winds. Outflow also manifests itself through the appearance of forbidden emission lines that are typical signatures of excited gas at low-density. The shapes and intensities of forbidden emission lines provide valuable information about mass loss rates and the velocity distribution in the wind. Images of TTS sometimes show strong jets extending to large distances, up to hundreds of AUs, from the star. These jets are often associated with interstellar nebulosity, the so-called Herbig-Haro objects. The absorption line spectrum of TTS is similar to that of main sequence stars, except for the presence of strong absorption lines of lithium. Sometimes a pseudo-continuum is seen on top of the photospheric spectrum, making photospheric absorption lines appear shallow. The presence of an excess continuum and emission lines with high excitation potentials (such as He I 5876) is accounted for by the energy released during mass accretion from the circumstellar disk. Energy balance calculations and measurements of line emission yield typical accretion rates for TTS of 10−8 to 10−9 M per year (Gullbring et al. 1998; Hartigan et al. 1995;

12 Lamzin et al. 2001). Outflow rates are of the same order of magnitude as the accretion rates, depending on which diagnostics are used. Because the T Tauri phase typically lasts 107 Ð 108 years and the estimated mass of circumstellar disks is < 1 M, most of the disk mass will have to be ejected from the system, either through jets in parallel to mass accretion, or directly from the circumstellar disk (a disk wind). Magnetic fields are believed to play an important role in the interaction between the star and the circumstellar disk. Strong magnetic fields have been detected on TTS and there is evidence for solar-like chromospheric activity giving rise to features such as cool and hot spots. Probably the most important feature of T Tauri Stars is their variability. Almost every single quantity that can be measured in TTS is variable on both very short as well as very long time scales. For example, the luminosity of TTS has been noticed to vary on scales of minutes to several years. Also the shape and intensity of emission lines can change considerably, but in most cases no periodic variations were found. Irregular variations indicate that TTS are passing though a highly dynamical phase of their evolution. The observational characteristics of TTS are very distinct and make their discovery rather easy. Since the 1950s large numbers of TTS have been dis- covered with surveys using objective prism-spectroscopy. TTS are numerous, many thousands are know. However, that what makes them so common, their low mass, also makes them rather faint and thus difficult to study. As these stars are associated with star-forming regions they are not spread uniformly in space. The brightest TTS are located in the most nearby star-forming regions, such as the clouds in Taurus-Auriga (140 pc), in Lupus (190 pc), and Chameleon (160 pc) (Wichmann et al. 1998). The fact that TTS usually are found in groups provides valuable information about star formation, in particular that a suitable environment for star birth consists of cold, molecular clouds that are much more massive than a single star. The remains of such clouds make up the circumstellar medium of TTS. Because of their distance, the typical visual magnitude of the brightest TTS is V ≈ 8m. Of course, the brightest stars in a given star forming region are also the most active ones, so the sample of ‘ordinary’ TTS has visual magnitudes of V ∼> 10m. In addition to optical surveys, searches were also performed on the basis of the X-ray activity and UV emission lines present in the spectra of TTS. These searches yielded large numbers of young pre-main-sequence stars with positions and kinematics similar to TTS in the same star formation region, but without strong line emission or IR excesses. Because these objects show Hα in emission and contain traces of Li, a subclass of TTS, the Weak-line T Tauri Stars (WTTS) was introduced. The exact boundary between these two groups is rather arbitrary. Stars with Wλ(Hα) ∼< 5 Aû are classified as WTTS, whereas others are classifies as Classical TTS (CTTS) (Herbig and Bell 1988).

13 Although there are indications that WTTS lack the dense circumstellar disks of CTTS, the evolutionary relation between CTTS and WTTS is unclear. Other subclasses of T Tauri stars I would like to mention brieﬂy are the FU Ori and YY Ori stars, named after their prototypes. These subclasses are based on very strong activity, and should be considered extreme objects within the group of TTS as a whole. YY Ori stars exhibit signs of very strong mass accretion, and FU Ori stars show exceptionally strong variations in luminosity. It is unclear if these subclasses represent a short stage in the evolution of TTS or that these stars represent unusual physical conditions. In this thesis I will concentrate only on CTTS and WTTS. More extensive overviews of TTS are presented in the review on TTS by Bertout (1989) and in the book of Hartmann (1998).

14 Models of T Tauri stars

As was shown in the previous chapter, the observational characteristics of TTS differ in several respects from the spectra of main-sequence stars. These differences provide important clues about what is happening behind the scenes. The common explanation for these observational characteristics is interaction of the central star with the circumstellar disk through mass accretion. In this chapter I present an overview of accretion models that describe T Tauri stars and their characteristics.

3.1 Classical accretion models The first attempt to explain the observational characteristics of T Tauri stars with a theoretical model for interaction of the central star with a circumstellar disk was made by Lynden-Bell and Pringle (1974). Simple dynamical ar- guments (conservation of angular momentum combined with a central force field, in this case gravity) will make gas and dust around a young star settle in a flattened circumstellar disk. In this model the inner parts of the disk rotates more quickly than the outer parts in order to maintain disk matter in stable or- bits. Such rotation is also called Keplerian rotation. The energy balance in the disk is disturbed by shear between disk annuli that carries angular momentum to the outer regions of the disk. The argument of shear between disk annuli is rather mechanical, and was later replaced by the more physical process of kinematic viscosity due to turbulent mixing of gas between different radii. Be- cause kinematic viscosity is difficult to measure directly, Shakura and Sunyaev (1973) introduced the free parameter α into accretion disk modelling. The α- parameter relates the local viscosity v, the sound speed cs and the disk height H by v = αcsH. Typical α values for circumstellar disks are found to be ∼ 0.001 or smaller. A comprehensive overview of classical accretion disks is given by Pringle (1981). The result of the outward momentum transport is an inward flow of matter which causes a dynamically complex equatorial boundary layer near the surface of the star where matter decelerates to the stellar rotational velocity. This model has successfully been used to explain the continuous IR and UV excess emission (Adams and Shu 1986; Bertout et al. 1988, for example). On the other hand, the assumption that the central star freely accretes matter

15 from its circumstellar disks implies that the rotational velocity of the central star steadily increases towards break-up velocities, typically several hundred km/s for TTS. This is contradicted by surveys of rotational velocities in TTS, showing that typical values are far from this value. (Vogel and Kuhi 1981; Bouvier et al. 1986). TTS with higher rotational velocities are known, but the statistical distribution of rotational velocities cannot support the assumption of unrestrained accretion.

3.2 Magnetospheric accretion models A large step forward was taken when magnetospheric accretion models (Ghosh and Lamb 1979) were successfully applied to TTS (Camenzind 1990; Koenigl 1991). In these models the inward flow of matter couples to a dipolar magnetic field, creating a funnel-shaped accretion flow. The main advantage of apply- ing magnetic fields to the accretion process is that it eliminates the equatorial boundary region and provides for a magnetospheric coupling between the central star and regions of the accretion disk that do not rotate at very high velocities. It also gives a theoretical basis for disk winds and collimated outflows (Shibata and Uchida 1985), as well as for spots on TTS (Bouvier and Bertout 1989). The requirement that TTS exhibit magnetic fields with strengths of ∼> 1 kG was later confirmed observationally (Basri et al. 1992; Guenther et al. 1999; Johns-Krull and Valenti 2000). The success of magnetospheric accretion has led to the development of a number of models that mainly differ in how the magnetic field couples to the disk. The first attempt for a comprehensive magnetospheric accretion model with an extensive theoretical basis for the large range of observed properties of TTS is what now is commonly known as the Shu model (Shu et al. 1994, and subse- quent papers). This model is an adaptation of the model of Ghosh and Lamb. The weak point in the model of Ghosh and Lamb is that the magnetic field lines thread the disk at many different radii, which constantly will wrap up the field lines. Shu et al. resolve this problem by imposing that the magnetic field configuration rotates synchronously with the star. Under the assumption that the disk is conductive, the magnetic field will then only penetrate those regions in the circumstellar disk that are (almost) in co-rotation with the star. As a consequence the inward flow of hot and ionized matter will, upon arrival near the co-rotation radius, be forced to follow the magnetic field lines. De- pending on whether the field line to which the matter attaches penetrates just inside or outside the co-rotation radius, the matter will flow either onto the star or be ejected as a disk wind. The Shu model was mainly motivated by the need for a general explanation of the overall appearance of TTS based on a magneto-hydrodynamical theory with a strong physical framework, but it is not straightforward to verify this

16 model with direct observations. The main reason is that the Shu model does not give direct predictions of the thermodynamical properties of the in- and outflows. There have also been more phenomenological attempts to model the observations. Hartmann et al. (1994) constructed magnetospheric models with the ambition to reproduce the specific shapes of emission lines in the spectra of TTS. However, the fixed magnetosphere in their models is purely dipolar and matter is assumed to accrete along the magnetic field lines in free-fall. The model provides no mechanism for coupling of the magnetic field to the disk and the temperature-pressure structure is prescribed. Therefore it is not sur- prising that the line profiles they obtained agree qualitatively with observations (Muzerolle et al. 1998a,b, 2001).

3.3 Time-dependent models The need for more sophisticated models that allow time-dependent accretion has led to the development of numerical models that simulate the interaction of the circumstellar environment and the magnetic field by solving magnetohydrodynamic equations (MHD models) and following the evolution of the medium and the magnetic field. The calculation of such models is computa- tionally very expensive and therefore the main differences between different flavors of MHD models are the different computational simplifications, the as- sumptions on the initial configuration of the magnetic field and the choice of boundary conditions. Also, the dynamical range of the interaction between a TTS, its disk and a magnetosphere is enormous, so MHD models tend to concentrate on describing a small number of TTS properties. Although MHD modelling is still challenging in many aspects, these models have been successful in improving our understanding that magnetic fields play an important role in the accretion process. For example, Miller and Stone (1997) use dipolar field models to show that, if the magnetic pressure exceeds the pressure from the accretion flow, matter will be loaded onto the magnetic field lines and accretion will be highly time-dependent. In a similar way Good- son et al. (1997) show that dipolar fields threading the circumstellar disk will wrap up and curl outwards, and can drive both collimated polar outflows as well as disk winds. Accretion along such extended magnetic field lines will then be strongly time-dependent. Hirose et al. (1997) show that reconnection events may be responsible for jet outflows and provides a mechanism for slow- ing the rotation of the central star. on RekowskiVON REKOWSKI et al. (2003) show that the origin of the magnetic field must not be of pure stellar origin, and can as well be through a dynamo active accretion disk. Although such models provide tantalizing self-consistent solutions for accretion processes and are in qualitative agreement with indirect observables

17 such as accretion rates and wind densities, such models have not yet been val- idated through a direct and independent comparison with observations. In con- trast to older models, these models provide full information about the temperature- pressure structure of the interstellar medium and can be used for radiative transfer calculations of the emerging spectrum (paper IV).

3.4 Open questions Accretion models have provided useful insights in the physics of the accretion process in TTS. The idea that accretion is controlled by the magnetosphere is now commonly accepted. Still, there are several open questions related to the dynamics and evolution of TTS, and many of these are related to the interaction of the magnetic field with the circumstellar disk. It is not yet known what the origin is of the magnetic field that plays such an important role in the accretion process. It is also uncertain how the magnetic field couples to the disk, what its geometry is, and how it relates to the disappearing of the disk and the formation of planetary systems. Apart from questions about the magnetic field configuration, research on the evolution of TTS is far from its concluding remarks. The evolutionary relation between CTTS and WTTS is not solved and little is known about TTS evolution in general. Of course, the fact that TTS are located in well- confined groups and may be prone to local differences, makes answering such questions difficult. However, with the arrival of new high-quality observations of ‘ordinary’ TTS, there is a new window opening on the research of T Tauri stars.

18 An observational study of T Tauri stars

It is evident that the TTS phenomenon as such is very complex and that there are many different ways to investigate TTS. Up to now, studies of variability in TTS were limited to very few diagnostics. However, recent progress in optical engineering has provided the astronomical community with a new generation of telescopes and instruments that are much more sensitive than the previous generation. This makes it possible to simultaneously study a large number of diagnostics, and in that way take on an entirely new challenge, investigating variability of ‘ordinary’ TTS on time scales of minutes to hours, the dynamic time scales that are typical for these small-scale systems. In this thesis I have used this opportunity in order to obtain a more complete picture of the processes surrounding TTS and their evolution. I also investigate the validity of an existing MHD model.

4.1 Observations During 2 nights in April 2000 and 4 nights in April 2002 I observed a number of TTS with the UVES spectrograph on the 8-m Very Large Telescope (VLT) in Chile. The powerful UVES/VLT combination can obtain high quality spectra of TTS with very short exposure times (∼ 10 minutes, depending on luminosity). Another advantage is that VLT/UVES can produce spectra throughout a large part of the optical wavelength range in a single exposure. This allows to simultaneously study diagnostics that are related to different processes in TTS. The observations consist of a time series of high-resolution (R ≈ 60000) and high signal-to-noise (∼ 200 Ð 300) spectra of CTTS and WTTS throughout a large part of the optical wavelength range (3500 <λ<6700 A).û The targets we observed are all known to exhibit variations in luminosity and spectral line shapes. The typical exposure time for the targets, all in the range 8m

19 As a consequence, it is possible to obtain a more or less complete image of almost all visible time-dependent processes that occur on short time scales in TTS. This fact, combined with the excellent quality of the spectra in terms of signal, spectral resolution and time resolution, make the spectra one of the best sets of optical spectra of TTS obtained up to date. The dataset obtained in 2002 is similar to the dataset of 2000. The main difference is that I used a different set of targets and a longer time base to investigate spectral variability. In the second run I included the short-period spectroscopic binary V 4046 Sgr, which is the topic of paper V. As this dataset is extensive and relatively new, the ﬁnal analysis of the major part of this dataset is not yet completed.

4.2 Data reduction The basic reduction of the data obtained from ESO was done with help of the IDL-based package REDUCE by Piskunov and Valenti (2002). This package uses 2-dimensional modelling of the echelle orders to perform optimal extrac- tion of the spectra. ESO does provide pre-reduced spectra from its pipeline reduction software, but this software appears to the end-user as a black box. Because we were one of the first users of UVES, we opted to use REDUCE as an independent verification of the quality of the pipeline-reduced spectra. Piskunov and Valenti show in a direct comparison that REDUCE produces spectra of similar or better quality than the UVES pipeline. Because REDUCE was in its final phase of development, our reduction of UVES data with REDUCE served as a final test for the package.

4.3 Spectral synthesis In order to study and quantify spectral features that are caused by the accretion process, a proper understanding of the underlying photospheric spectrum of the central star is essential. Fortunately, it is relatively straightforward to model the photosphere of the central star with synthetic spectra. For TTS one- dimensional radiative transfer calculations of late-type stellar spectra are fully adequate, because only a small fraction of the stellar surface is affected by accretion. The accuracy of these radiative transfer calculations depends mainly on the quality of the model atmosphere and the atomic line data used to calculate the line opacities. In addition a proper knowledge of the basic stellar parameters (Teff , log g, [Fe/H], v sin i, vrad,vmic and vmac) is needed to produce the final profiles. For these radiative transfer calculations I have extensively made use of the SME package (Valenti and Piskunov 1996). The advantage of this package is that it can find a consistent set of basic stellar parameters us-

20 1.5

1.0

0.5

0.0

6110 6120 6130 6140

Figure 4.1: An example of the close agreement between synthetic and observed spectra of the TTS CT Cha. The difference between the observations and the synthetic spectra is the thin line around the horizontal axis. The dashed horizontal line is the contribution of the veiling continuum. The dark regions were used for determination of the continuum level, and the white regions were excluded from the synthetic spectrum modelling because of spectral lines with poor parameters (near 6137 A)û or lines that were not present in the atomic line list (for example Li I at 6104 A).û ing the method of least-squares. The model atmospheres I use throughout this thesis come from the MARCS consortium (Gustafsson et al. 2002) and atomic line data from the VALD database (Piskunov et al. 1995; Kupka et al. 1999; Ryabchikova et al. 1999). Modelling of TTS spectra is only marginally more complicated than in the case of main-sequence stars. In TTS non-stellar contributors to the stellar flux, are the accretion disk, the accretion shock and the circumstellar medium. Be- cause of its low temperatures, continuum radiation from the accretion disk is hardly visible in our UVES observations. Continuum radiation of the accretion shock makes stellar absorption lines appear shallow and is visible in almost the entire observed wavelength range, hence the name veiling continuum. As the intensity of the veiling continuum only weakly depends on wavelength, one can for short wavelength intervals (up to 50 Ð 100 A)û model TTS spectra by simply adding a constant contribution to the synthetic spectra. In fact, in this way one can subtract any continuous and featureless contribution to the stellar spectrum. In order to determine the contribution of the veiling continuum to an observed stellar spectrum, I used the following model based on the definition of Hartigan et al. (1989). In this model, the veiling factor V is defined as :

V = Ipc/Isc (4.1) where Ipc is the intensity of the pseudo-continuum and Isc the intensity of the stellar continuum. Under the assumption that the veiling factor changes only throughout a short wavelength interval, the observed line depth d∗ of the stellar

21 absorption lines in this interval will be : d d∗ = (4.2) 1+V where d is the unveiled absorption line depth. More details on the procedure to determine the level of veiling is presented in paper II. A large advantage of using synthetic spectra as templates for the stellar spectrum is that the radiative transfer calculations provide the absolute flux level of the continuum. Therefore, it is possible to determine the level of veiling, and thus model the stellar spectrum without any a priori knowledge of the continuum level from the observed spectra. Especially in the bluest spectral regions it is difficult (and often impossible) to determine the continuum level in observed spectra. As long as there are lines with different central depths present in the synthetic spectra, this method is not degenerate. A comparison between synthetic and observed line profiles is shown in figure 4.1. The agreement between the synthetic and observed spectra is very good, even though MARCS model atmospheres are 1-dimensional and plane- parallel hydrostatic LTE models. Modelling spectra of TTS with the method described above is an excellent tool to study the properties of the photosphere of the central star. Any radiation that is not accounted for in the stellar spectrum represents the contribution of the accretion region and the circumstellar environment. These residual spectra are the key to investigating the accretion process.

4.4 Magnetic fields Magnetic fields do not only play an important role in the process of accretion. The presence of magnetic fields also generates magnetic activity in the stellar atmosphere. A high level of magnetic activity tends to generate active cool and hot regions (plages), which will influence the temperature structure of the atmosphere producing spectra that differ from the quiescent photosphere and from the shock regions. In principle, the effect of magnetic fields should be taken into account when modelling the stellar photosphere or the contribution of the veiling continuum.

22 Summary of papers

The numbering of the papers in this thesis is chronological and reﬂects the transition of purely observational work towards interpretation of accretion processes in TTS though comparison of observations and existing models.

5.1 Measuring accretion rates (Paper I) In the accretion shock on the surface of TTS, temperatures can rise as high as 106 K, far higher than ordinary temperatures of stellar atmospheres. The continuous radiation from these hot accretion shock regions has been used to estimate the accretion rates of TTS (Hartigan et al. 1995). At such high temperature, one also expects the formation of coronal emission lines (Gahm et al. 1979; Imhoff and Giampapa 1980). In this paper we use energy balance calculations to demonstrate that the observed equivalent widths of Fe XÐXIV coronal emission lines can be used to obtain independent estimates and upper limits of TTS accretion rates. The advantage of using coronal emission lines as estimators for accretion rates is that these emission lines are formed under optically thin conditions, and therefore the accretion rate is directly proportional to the emission line intensity. There is no need to perform accurate modelling of the physical conditions in the accretion shock. Due to the complex structure of the Fe atom, coronal lines of Fe appear at many different places in the stellar spectrum, from the UV to the near IR. Even though temperatures in the accretion shock region are high and Fe is a relatively abundant species in stellar atmospheres, the predicted speciﬁc intensities of coronal lines from a TTS is of the order of 103 erg cm−2 s−1 sr−1, which is not particularly high compared to the radiation from the central star (∼ 4 · 105 erg cm−2 s−1 sr−1 Aû −1 at λ = 6000 A).û For example, even in a combined high-quality VLT/UVES spectrum of RU Lup with a S/N ratio of ∼ 1000 it is impossible to detect these lines. In the UV, the stellar contribution is not as high and (weak) Fe coronal emission lines have been detected.

23 5.2 The extreme TTS RU Lup (Paper II) The primary target of the observing runs in 2000 and 2002 was the star RU Lup. This star is one of the most active TTS known. RU Lup has high levels of veiling and strong emission lines. It also exhibits strong variability in photometry, in the shape of its spectral lines and in most other diagnostics and was the first object to be fully reduced. It was also the first object for which we used synthetic spectra to model the veiling contribution. We investigated the simultaneous evolution of a range of diagnostics in order to put constraints on the accretion process and its geometry. Veiling in RU Lup is very strong and variable. Because we find that changes in veiling occur on shorter timescales than other diagnostics, we used the level of veiling as the prime indicator of activity against which we compared changes in other diagnostics. We find that the veiling continuum of RU Lup may change considerably (up to about 20 %) during a single night. We also find that the changes in the veiling factor are coherent on time scales on the order of one hour. This time scale is comparable to the dynamic time scale of free-fall from the inner edge of the disk to the star. Such strong variation in the veiling factor over a period of only a few hours cannot be explained by rotational modulation. We also investigated variations in the spectral distribution of the veiling continuum. Such changes can be caused in two ways. Temperature changes in the accretion region will modify the spectral slope of the veiling continuum, while changes in the projected size of the accretion region will affect the overall intensity of the veiling spectrum. In order to investigate the relation between changes in temperature and size of the accretion region, we used the following model describing the contribution of the veiling continuum as a scaled blackbody spectrum. This modifies the stellar flux as follows :

Ftot =(1− Sff ) · Fstar + Sff · πBλ(Tpc) (5.1) where Tpc is the temperature of the pseudo-continuum (veiling continuum) and Sff the size of the formation region relative to the total stellar surface (the so-called filling factor). Although these two parameters are not independent, we could, using the method of least-squares and error propagation derive that significant changes in Tpc and Sff occur within a single night. The strongest emission lines with high excitation potentials such as the Balmer lines, He I and II,NaII D and Ca II H & K, evolve mostly on much longer timescales than the veiling factor. Therefore we concluded that these lines are formed in a large physical volume. The shapes of most of these lines show short-time variability due to mass in- and outflows. Metallic emission lines with low excitation potentials such as Fe I,FeII and Ca II are abundant in the spectrum of RU Lup. We found a strong similarity

24 4

Veiling factor 1

0 52380.0 52381.0 52382.0 52383.0 52384.0 Modified Julian Day

Figure 5.1: This figure illustrates the large variability in the contribution of the veiling continuum in RU Lup. These measurements are from the 2002 observing run, and up to now unpublished. between Fe lines of different multiplets, indicating that these lines form under optically thin conditions and that the shape of these lines is dominated by the velocity field of the circumstellar medium. Forbidden emission lines of [O I] −1 and [S II] show very high outflow velocities (∼> 200 km s ). The low-velocity component in the [O I] line exhibits changes in shape during both nights. Variability in veiling seems to precede variability in the emission lines. However, the short time base of this study (2 nights) prevents us from drawing any decisive conclusions on the sequence of events. Absorption features in the blue wings of Ca II K and Na I D show that strong mass outflow occurs from RU Lup. The shape of the outflow features is similar in different emission lines. From the equivalent width of these absorption features we derive that the column density changes considerably between the two nights. Because we see individual absorption components that change independently, outflow must be structured. Just like the accretion flow, outflow is not homogeneous. The very low v sin i of RU Lup (9.1 ± 0.9 km s−1) combined with very low levels of interstellar extinction and the indication that strong outflow occurs along the line of sight leads us to believe that RU Lup is observed close to pole-on. This may explain the strong activity of RU Lup compared to other TTS. From the interstellar extinction we derive an upper limit of 200 pc on the distance to RU Lup.

5.3 The veiling spectrum (Paper III) The previous study clearly demonstrated the value of the veiling factor to the analysis of TTS. However, there are only few studies of the spectrum of the veiling continuum. Because the accuracy of veiling measurements strongly depends on the quality of the observations and the agreement of the template spectrum with the photospheric spectrum of the central star, error bars on veil-

25 ing measurements have generally been large, often up to ±1. In this study we make extensive use of synthetic spectra as templates for the stellar component (see also section 4.3), demonstrating that it is possible to accurately determine the level of veiling using synthetic spectra. In addition, we have been able to obtain new and more precise measurements on the basic stellar parameters of the observed targets, as well as measurements of the contribution of the veiling continuum throughout the optical range. We successfully checked our method against two non-veiled stars, the Sun and 61 Cyg A. When we determined the veiling spectra of 5 TTS we unexpectedly found that the spectral distribution of the level of veiling we recovered is not purely continuous. For the most strongly veiled stars we found a region with high veiling values in the region 5000 Ð 5700. We carefully checked our results against other possible sources of continuum emission such as instrumental artifacts, excess line emission and sys- tematic errors in our method of veiling measurement, but could not relate the excess emission to any of these sources. We found that our result is not incom- patible with previous investigations of the veiling spectrum. This region with high veiling values requires further investigation.

5.4 NLTE calculations of hydrogen emission lines in T Tauri stars (Paper IV) As was shown in chapter 3, models of TTS play an important role in improving our understanding of accretion processes. The fact that magnetospheric accretion is the main process responsible for producing the majority of observational characteristics in TTS is now generally accepted. While the initial models only provided a solution for the configuration of magnetic field, newer models based on solving MHD equations follow the evolution of the entire medium and take into account the interactions between different physical processes. These models produce self-consistent results and are in agreement with the global properties of TTS, such as accretion rates and wind densities. Still, these properties are only indirect confirmations that the resulting physical conditions of theses models are correct. Up to date, there has not been any detailed comparison of MHD models with high-quality observations of TTS. With high-resolution spectroscopic spectra of TTS being available, it is now possible to do such a direct comparison. In this paper we show that it is possible to synthesize hydrogen line profiles from the physical conditions predicted by the MHD model of von Rekowski et al. (2003). The magnetic field in this model is generated by a disk dynamo and drives disk outflow and polar accretion. From simulations of the evolution of the model a self-consistent solution for the configuration of the magnetic

26 field and the physical conditions was obtained. We use this model as input for computing emission lines through 3-dimensional radiative transfer. Solving the problem of radiative transfer in the complex and dynamic circumstellar environment of TTS is not straightforward, because the medium is strongly inhomogeneous and the radiation field strongly anisotropic. This creates a situation where the spectral distribution of radiation is dramatically inconsistent with the local kinetic temperature, leading to non-equilibrium level populations, commonly known as non-local thermodynamic equilibrium (NLTE). In this paper we describe a new 3D NLTE radiative transfer code for a 5- level atom with continuum capable of computing hydrogen line profiles in a CTTS environment. We solve the problem of radiative transfer on a equispaced cubic grid with 26 lines of sight passing though each point. Schematically, the process of solving radiative transfer we employ can be represented by : new ni → Sν → Jν → J¯ν → ni (5.2) with the common representation of the radiative transfer quantities, see Can- non (1985) for more details. In our formultation we make the additional as- sumptions of complete redistribution, an optically thin continuum, and that the central star and the accretion shock are the only energy source in the system. The first results of our 3D NLTE calculations of hydrogen line profiles show that we are able to produce emission line profiles, but that there still are strong discrepancies between the observed and calculated profiles. We also derive a list of improvements that need to be implemented in the code before we will be able to constrain the input parameters of MHD models and obtain a better understanding of the processes occuring around TTS.

5.5 The spectroscopic binary V 4046 Sagittarii (Paper V) V 4046 Sgr is a very close spectroscopic binary consisting of two K stars with T Tauri signatures. As the period of the system is short (P =2.42 days), there is strong variability in most observables, especially in the intensity and shape of the emission lines. V 4046 Sgr has not been studied with very high resolution spectroscopy before. The spectrum of V 4046 Sgr shows all the typical characteristics of T Tauri stars and is surrounded by a circumbinary disk. We detect strong Li absorption, but no shell-absorption components (enhanced absorption due to circumstellar matter). Despite the low temperature of the components we cannot detect any TiO absorption lines. Though weak, veiling is present in the spectra. We ﬁnd that interstellar extinction is virtually absent. We extended our techniques for calculating synthetic spectra to spectroscopic binaries, and the resulting proﬁles agree very well with observations.

27 In this study, we also performed for the first time emission line analysis by adapting synthetic spectra without any a priori determination of the level of the continuum (as described in section 4.3). Instead we used the continuum level provided by the synthetic spectra calculations. Especially in the blue parts of the spectrum this technique was very successful in subtracting photospheric components from the compound stellar specta, yielding much cleaner emission line profiles. We find that equivalent widths of the Ca II H & K lines do not change through the orbital period, while the individual components follow the stellar velocities closely. Also, the widths of these lines are larger than the rotational velocity of the components, indicating that these lines form in global chromospheric networks on each star. Previous studies of the Hα line in V 4046 Sgr showed that there exist a double period in the peak emission as well as the line width. From our observations we can deduce that this behavior is due to narrow line emission coupled to the orbital motion of the components. Long- term photometric monitoring of V 4046 Sgr has shown that the system exhibits small-scale photometric variability in Johnson B and Stromgren¬ y and b, and that these variations peak when the binary is near quadrature. However, we cannot bring variability in emission line strength or veiling in agreement with photometric variability. The shape of Balmer emission lines is very similar, especially the higher members of the series. Such similarity indicates that these lines have a common velocity structure of their formation regions. Therefore we developed two models to explain the shapes and evolution of the lines H8 Ð H10. The first model assumes that all emission is coupled to the stellar components. The solution we obtain consists of asymmetric profiles for each star, which can be interpreted as in- and outflow coupled to the polar regions. In the second model we reproduce the profiles with four Gaussian components, of which two components couple to the stars, and two to concentrations of gas with orbital velocities that are considerably higher than those of the stars. Such a configuration may be compatible with common scenarios for binary accretion from a circumbinary disk.

28 Acknowledgments

During the last years I have really enjoyed working on this thesis. Therefore it is a pleasure for me to write these senctences to thank all of you who made this project possible.

First of all, I would like to thank my supervisor, Nikolai Piskunov, a very good teacher and accessible at (almost) all times. For fruitful collaboration I would like to thank Sergei Lamzin, Gosta¬ Gahm and Brigitta von Rekowski. It has also been a pleasure to discuss selected parts of my work with Kjell Eriksson. I am also grateful for the help of Bertil Petterson, Paul Barklem and Bernd Freytag, who in some way or another contributed this thesis.

There are also many persons who contributed without really knowing it. Mak- ing the department a stimulating and scientiﬁcally interesting environment to work is their achievement. They are all too many, so therefore : ”Thank you all!”

Finally, I would like to express my gratitude to Uppsala University, and in particular the Faculty of Science and Technology, for recognizing my engagement for my fellow students by extending my position with one year.

Nunc hilares, si quando mihi, nunc ludite, Musae!

29 References

Adams, F. C. & Shu, F. H.: ‘Infrared spectra of rotating protostars’, 1986, ApJ, 308, 836

Basri, G., Marcy, G. W., & Valenti, J. A.: ‘Limits on the magnetic ﬂux of pre-main-sequence stars’, 1992, ApJ, 390, 622

Bertout, C.: ‘T Tauri stars - Wild as dust’, 1989, ARA&A, 27, 351

Bertout, C., Basri, G., & Bouvier, J.: ‘Accretion disks around T Tauri stars’, 1988, ApJ, 330, 350

Bouvier, J. & Bertout, C.: ‘Spots on T Tauri stars’, 1989, A&A, 211, 99

Bouvier, J., Bertout, C., Benz, W., & Mayor, M.: ‘Rotation in T Tauri stars. I - Observations and immediate analysis’, 1986, A&A, 165, 110

Camenzind, M.: ‘Magnetized Disk-Winds and the Origin of Bipolar Out- ﬂows.’, 1990, Reviews of Modern Astronomy,3,234

Cannon, C. J. ‘The transfer of spectral line radiation ‘Cambridge University Press, 1985

Gahm, G. F., Fredga, K., Liseau, R., & Dravins, D.: ‘The far-UV spectrum of the T Tauri star RU Lupi’, 1979, A&A, 73, L4

Ghosh, P. & Lamb, F. K.: ‘Accretion by rotating magnetic neutron stars. III - Accretion torques and period changes in pulsating X-ray sources’, 1979, ApJ, 234, 296

Goodson, A. P., Winglee, R. M., & Boehm, K.: ‘Time-dependent Accretion by Magnetic Young Stellar Objects as a Launching Mechanism for Stellar Jets’, 1997, ApJ, 489, 199

Guenther, E. W., Lehmann, H., Emerson, J. P., & Staude, J.: ‘Measurements of magnetic ﬁeld strength on T Tauri stars’, 1999, A&A, 341, 768

Gullbring, E., Hartmann, L., Briceno, C., & Calvet, N.: ‘Disk Accretion Rates for T Tauri Stars’, 1998, ApJ, 492, 323

30 Gustafsson, B., Edvardsson, B., Eriksson, K., Grae-J¿rgensen,û U., Mizuno- Wiedner, M., & Plez, B.: ‘A Grid of Model Atmospheres for Cool Stars’, 2002, In ASP Conf. Ser. 288: Stellar Atmosphere Modeling

Hartigan, P., Edwards, S., & Ghandour, L.: ‘Disk Accretion and Mass Loss from Young Stars’, 1995, ApJ, 452, 736

Hartigan, P., Hartmann, L., Kenyon, S., Hewett, R., & Stauffer, J.: ‘How to unveilaTTauri star’, 1989, ApJS, 70, 899

Hartmann, L. ‘Accretion processes in star formation ‘Cambridge University Press (Cambridge astrophysics series; 32) ISBN 0521435072, 1998

Hartmann, L., Hewett, R., & Calvet, N.: ‘Magnetospheric accretion models for T Tauri stars. 1: Balmer line proﬁles without rotation’, 1994, ApJ, 426, 669

Herbig, G. H. & Bell, K. R. ‘Catalog of emission line stars of the orion popu- lation ‘Lick Observatory Bulletin no. 1111, Santa Cruz, Lick Observatory, 1988

Hirose, S., Uchida, Y., Shibata, K., & Matsumoto, R.: ‘Disk Accretion onto a Magnetized Young Star and Associated Jet Formation’, 1997, PASJ, 49, 193

Imhoff, C. L. & Giampapa, M. S.: ‘The ultraviolet spectrum of the T Tauri star RW Aurigae’, 1980, ApJ, 239, L115

Johns-Krull, C. M. & Valenti, J. A.: ‘Measurements of stellar magnetic ﬁelds’, 2000, In ASP Conf. Ser. 198: Stellar Clusters and Associations: Convection, Rotation, and Dynamos, 371

Joy, A. H.: ‘T Tauri Variable Stars.’, 1945, ApJ, 102, 168

Koenigl, A.: ‘Disk accretion onto magnetic T Tauri stars’, 1991, ApJ, 370, L39

Kupka, F., Piskunov, N., Ryabchikova, T. A., Stempels, H. C., & Weiss, W. W.: ‘VALD-2: Progress of the Vienna Atomic Line Data Base’, 1999, A&AS, 138, 119

Lamzin, S. A., Stempels, H. C., & Piskunov, N. E.: ‘Formation of Fe X-Fe XIV coronal lines in the accretion shock of T Tauri stars’, 2001, A&A, 369, 965

Lynden-Bell, D. & Pringle, J. E.: ‘The evolution of viscous discs and the origin of the nebular variables.’, 1974, MNRAS, 168, 603

31 Miller, K. A. & Stone, J. M.: ‘Magnetohydrodynamic Simulations of Stellar MagnetosphereÐAccretion Disk Interaction’, 1997, ApJ, 489, 890

Muzerolle, J., Calvet, N., & Hartmann, L.: ‘Magnetospheric Accretion Models for the Hydrogen Emission Lines of T Tauri Stars’, 1998a, ApJ, 492, 743

Muzerolle, J., Calvet, N., & Hartmann, L.: ‘Emission-Line Diagnostics of T Tauri Magnetospheric Accretion. II. Improved Model Tests and Insights into Accretion Physics’, 2001, ApJ, 550, 944

Muzerolle, J., Hartmann, L., & Calvet, N.: ‘Emission-Line Diagnostics of T Tauri Magnetospheric Accretion. I. Line Proﬁle Observations’, 1998b, AJ, 116, 455

Piskunov, N. E., Kupka, F., Ryabchikova, T. A., Weiss, W. W., & Jeffery, C. S.: ‘VALD: The Vienna Atomic Line Data Base.’, 1995, A&AS, 112, 525

Piskunov, N. E. & Valenti, J. A.: ‘New algorithms for reducing cross-dispersed echelle spectra’, 2002, A&A, 385, 1095

Pringle, J. E.: ‘Accretion discs in astrophysics’, 1981, ARA&A, 19, 137

Ryabchikova, T. A., Piskunov, N., Stempels, H. C., Kupka, F., & Weiss, W. W.: ‘The Vienna Atomic Line Database — a Status Report’, 1999, In Proceed- ings of the 6th International Colloquium on Atomic Spectra and Oscillator Strengths (ASOS 6). Ed. W. L. Wiese. & D. C. Morton 176 pages., 162

Shakura, N. I. & Sunyaev, R. A.: ‘Black holes in binary systems. Observa- tional appearance.’, 1973, A&A, 24, 337

Shibata, K. & Uchida, Y.: ‘A magnetodynamic mechanism for the formation of astrophysical jets. I - Dynamical effects of the relaxation of nonlinear magnetic twists’, 1985, PASJ, 37, 31

Shu, F., Najita, J., Ostriker, E., Wilkin, F., Ruden, S., & Lizano, S.: ‘Mag- netocentrifugally driven ﬂows from young stars and disks. 1: A generalized model’, 1994, ApJ, 429, 781

Valenti, J. A. & Piskunov, N.: ‘Spectroscopy made easy: A new tool for ﬁtting observations with synthetic spectra.’, 1996, A&AS, 118, 595

Vogel, S. N. & Kuhi, L. V.: ‘Rotational velocities of pre-main-sequence stars’, 1981, ApJ, 245, 960 von Rekowski, B., Brandenburg, A., Dobler, W., & Shukurov, A.: ‘Structured outﬂow from a dynamo active accretion disc’, 2003, A&A, 398, 825

32 Wichmann, R., Bastian, U., Krautter, J., Jankovics, I., & Rucinski, S. M.: ‘HIPPARCOS observations of pre-main-sequence stars’, 1998, MNRAS, 301, L39