Articles by Ulrike Heiter

H.-G. Ludwig: Prospects of Using Simulations to Study the Photospheres of Brown Dwarfs

This talk presented ideas and problems when it comes to simulation of brown dwarf photospheres. A key aspect is the formation and transportation of dust grains. Up to now, all 3D simulations were extrapolations from simulations of hotter stars (M dwarfs), which were dust free. An extension to hot Jupiters (EGPs) is also forseen, which are quite similar to brown dwarfs, hydrodynamically speaking.

Simulating convection is relevant for the energy transport, the thermal structure of the convective layers and mixing of material. Maybe it will also be important for acoustic activity (mechanical heating) and local dynamo action (magnetic activity), we don’t know yet.

In terms of radiation, RHD models will not add substantial information. Modelling radiation is of course important for energy transport and the thermal structure of the radiative layers, while radiation pressure is not important.

Another important ingredient of the simulations will be the formation of liquid and solid condensates (“dust”). This is an opacity source, for which knowing the albedo, the atmospheric chemical composition and the spatial distribution will be necessary.

Further modelling aspects include rotation (advective transport of dust) and external external irradiation by the host star for EGPs (“day/night”), causing a global circulation pattern.

Up to now, toy models without dust have been calculated to derive time scales. The results were presented in a graph showing time scales vs. geometrical height. Convection time scales are 10^2 to 10^3 s, the condensation time scale is about 10^2 s for 100 micron grains. This can be used to derive a grain size limit. The dynamic ranges of the simulation are 10^4 to 10^5 in time and 10^2 to 10^3 per dimension in space. The ratio of radius to convection cell size is about 10^4 (10^3 for EGPs). This shows that the simulations will be unable to encompass all spatial ranges (from global to convection) and one will have to separate global and local models.

References for global EGP models are Cho et al. and Burkert et al. (2005).

As an example for local models, a Teff = 1800 K RHD test model without dust by Ludwig, Allard and Hauschildt was shown, where the convective overshooting turns out to be about 0.35 pressure scale heights. The large range in velocities puts high demands on numerics. Rotation and thermal forcing cause an interplay between global and local models.

In conclusion, we are in the position to obtain realistic models of brown dwarfs including radiation and dust formation. But there will be no unified models in the near future.

The challenges for numerics are the substantial scale ranges (dust grain vs. convective velocities), developing an efficient procedure to treat scattering in the time-dependent multi-D case, the interaction between local and global transport processes, the dust cloud distribution and the local transport of momentum (turbulent velocity).

---

G. Wuchterl: Convection during the formation of gaseous giants and stars

Classical pre-main-sequence evolution starts somewhere in the middle of nowhere in the HR diagram (e.g. at the so-called “birthline”). The only other way is to calculate the full protostellar collapse, which involves challenging physics. To demonstrate this, eight equations for self-gravitating, convective, radiating fluids were shown (see Wuchterl and Tscharnuter 2003, A&A 398, 1081).

Wuchterl and Feuchtinger (1998) is on the supercritical protostellar accretion shock with convection. This requires very high resolution (order of 10^8) and one has to go to very high Courant numbers (10^12).

Improving and testing convection involves a physically plausible time dependent theory. A modified Kuhfuss theory is employed.

The calculations start with an equilibrium Bonnor-Ebert sphere, but the stars do not arrive on the Hayashi line (they are hotter and more luminous after one Myr).

In the HRD, star formation calculations are best displayed with isopleths (lines of constant mass), since the mass is not constant during evolution (accretion!).

Results of a solar mass collapse with convection show that after 1 million years, the cloud memory is lost. The differences to the static picture are that deuterium is burnt (so the birthline seems not to be a useful concept) and the core is radiative.

Two cases of extreme planet formation were presented:

1. GQ Lupi b (ESO VLT NACO observations from June 2004). This is a nearby young star with a faint companion, with a projected separation of 0.7 arcsec (100 AU).
2. HD 149026 b – a transiting Saturn mass planet with a 70 Earth masses core.

Models give predictions which are in accordance with the observations.

The previous entry was the last one for this week. Unfortunately I did not get past the second day, but at least that one is complete. I have notes for many more talks and will add them later to this blog.

Today, I already attended the very nice concluding talk for the Convection Symposium given by J-P Zahn and will now go to the summary talks of the Binary Stars Symposium by C. Scarfe and V. Trimble.

This talk within Symposium 240 was about how to extract information from spectroscopic binary (SB) spectra by reconstruction and analysis of the component spectra.

At the beginning, the importance of the line broadening function was discussed. For reconstruction, it is assumed that the observed spectrum is composed of a sum of weighted intrinsic component spectra. The intrinsic component spectra are time-independent (the shape of the spectral lines does not depend on orbital phase), but the weights may be time dependent (light and line strength variability). The reconstruction technique is based on velocity differences between the components and their time-variations. In order to apply this technique successfully, one should aim at observations homogeneously distributed in velocity and not concentrated at extreme line separation.

Historically, SB spectra have been analysed by the following techniques:

• Cool giant + hotter, less evolved star (very different stars): subtract template for cool giant (Griffin & Griffin 1986)
• Tomographic separation (Bagnulo & Gies 1991)
• Disentangling in velocity space (Simon & Sturm 1994)
• Disentangling in Fourier space (Hadrava 1995)
• Iteratively correlate – shift – co-add (Marchenko 1998, Gonzalez & Levato 2006)

The application domain of the reconstruction ranges from being a detection tool to detailed analysis of the component spectra. It can be applied over a large range of S/N, from UV to IR, for single lines to large wavelength ranges, and for spectral types from O to G.

As an example, the sharp-lined F8 + close binary (F8 + late-G) system RV Crateris was presented (details in ESO conference proceedings by Hensberge et al.). The disentangling technique revealed 3 components.

The use of the component spectra in astrophysics range from determining fundamental stellar parameters, testing evolutionary models, the spectroscopic detection of eclipses to spectroscopic determination of light ratios. For example, the chemical composition of DG Leo Ab is presented in Fremat, Lampens & Hensberge 2005, MNRAS 356, 545).

One can work in Fourier or velocity space. The options for velocity space are to mask blemishes or unwanted components in the input spectra, weight spectral bins and do an error analysis of the reconstructed components. The options for Fourier space are to analyse the origin and shape of spurious components in the reconstructed component spectra and to weight the Fourier components.

A problem with the technique is that the solutions may not be unique. As a way out one can use external information (photometry, models).

Prospects are:

• Retrieving astrophysical information by modelling the asymmetrical broadening function
• Extending the application domain of the spectral disentangling techniques in order to include better physics

I will not comment on any particular of the hundreds of posters presented during this week (of which the majority are presented at the Binary Stars Symposium). I only took a brief walk through the poster rooms of the Convection and Binary Stars Symposia and took some of the few A4 copies that were left. In general, one could say that the posters got less attention than deserved, at least in the Convection symposium, since there was not a single session dedicated to posters and the poster rooms were located far away from the coffee tables. But pdf files of the posters will be made available on the conference web page. In the Binary Stars Symposium there was one session with poster highlights each day, allowing the poster authors to make short oral presentations. Many of the posters (not presented orally) were about one particular binary system.

I noticed that for my own poster I had supplied too few A4 copies, they were all gone after a short time. You can download an A4 pdf file of my poster, presented at both the Convection and Binary Stars Symposia here.

This review talk opened Session B of the Convection Symposium: “Observational Probes of Convection”. It gave an overview of the classical ways to detect convection in observations.

Convection reaches the photosphere in most stars of Teff 10000), the situation is confusing (see Lyubimkov et al. 2004 and Przybilla et al. 2006).

Bisector curvature (line profile asymmetry) is another way to detect convection. Gray and Nagel (1989) showed that bisectors are reversed for cool (K) vs. hotter stars (F), with the reversion taking place at about G0. A stars have reversed bisectors, late B stars have no curvature at all.

The use of 3D models is limited – if one disagrees with observation, testing changes is time consuming. On the other hand, MLT and other convection models can be used for testing.

Conclusions:

• Stellar atmosphere velocity fields are clearly detectable in the spectrum.
• The behaviour over the HR diagram is varied, the largest velocities are found in supergiants
• Modelling is making progress at connecting convection theory with observations.

H. Kjeldsen: What Can We Learn About Convection From Asteroseismology?

This talk presented work done together with Tim Bedding and focussed on solar type stars. First, schematic movies of non-radial oscillation modes were shown and the classification of modes by degrees l and radial order n reviewed. The construction of power spectra and echelle diagrams was described. From the echelle diagram, the density and age of a star can be inferred.

For the solar like stars alpha Cen A and B, data have been obtained with UCLES at AAT and UVES at VLT. The precision is 50-70 cm/s with UVES for alf Cen A, i.e. almost as high as with a solar satellite. The echelle diagrams show l = 2, 0, 3, 1 modes, with a much smaller oscillation amplitude for alf Cen B. alf Cen B has a much higher density than alf Cen A and Sun. alf Cen A models do not fit well. alf Cen B models fit better, but small deviations remain.

beta Hydri is a G2 subgiant (an “old Sun”). For this star, CORALIE, UCLES and HARPS data have been obtained (one can “see the star oscillating on screen”). The echelle diagram shows normal l = 2, 0 modes and a crazy l = 1 mode, which could actually be mixed modes (due to so-called avoided crossing). Models by Di Mauro et al. (2003) seem to show indication for avoided crossing. Mixed modes are extremly sensitive to convection in the cores of these stars.

What we can learn about convection from asteroseismology: Mixed modes tell us about core convection (bet Hyrdi), structure the in echelle diagram gives information about the outer convection zone, and the surface can be studied with p-mode lifetimes and the “noise” background.

---

G. Cauzzi: Solar High Resolution Spectral Observations Compared with Numerical Simulations

In this talk, spatially resolved spectral observations were presented, which can be used to examine the models presented before. A movie of 80 arcsec of the solar surface at 7200Å during 1 hr with a time step of 20 s was shown. The diffraction limit is 0.24 arcsec = 160 km.

The observations are imaging spectroscopy (rapidly tunable narrow band filters, mostly IR with high transparency of 15-20%) using short exposure times (20-50 ms) and a large 2D field of view 60 arcsec squared. Narrow passbands (R > 200000) and sequential spectral sampling (10-20 points per line) is used. The instrument is called IBIS and the system works with adaptive optics. As an example, quiet Sun data from June 2004 were shown – the Fe I line at 709 nm at 16 spectral positions with R = 250000.

Simulations and LTE spectral synthesis for 6 snapshots were provided by M. Asplund, with a horizontal extent of 6×6 Mm (120 km step), matching the observations, and 1 Mm vertical extent (+800 to -200 km, step 10 km). The synthetic data were smeared with the telescope and atmospheric PSF and the instrumental resolution.

The spatial power spectra agree between observations and simulations. The line vs. continuum intensities at +-60 mÅ agree. The intensity distribution with wavelength and the velocity distribution were found to fit as well.

Reverse granulation is seen in the simulation and the observations. When plotting equivalent widths vs. continuum intensity, the simulation lies a little bit higher. Average profiles are fit by the simulations when using the higher Fe abundance (7.67).

Conclusions:

• Excellent agreement of high resolution observations with simulations
• Validates both realism of simulations and reliability of instrument
• Useful tool additional to abundance analysis

A.G. Kosovichev: Helioseismic inferences on subsurface solar convection

This talk was given at quite high speed, so it was difficult to take notes. Here is an attempt of a summary.

The idea of helioseismology is to measure travel times of resonant frequencies. Global helioseismology estimates frequencies of normal modes from oscillation power spectra. Time-distance helioseismology measures travel times of acoustic or surface gravity waves.

The depth of the solar convection zone can be measured with helioseismology (about 0.29 solar radii). It is more shallow in polar regions.

Differential rotation produces a “tachocline” – a rotational shear layer mostly below convection zone (see Kosovichev 1996, ApJ 469, L61 – analysis of BBSO data).

There are differences between the standard solar model and seismic models, which result in a difference in the solar radius of about 300 km. This could be caused by convective overshoot at the top of the convection zone.

New local helioseismology provides maps of the solar surface (synoptic maps of subsurface flows). Supergranulation can be observed by time-distance observations, vertical flows are difficult to measure. The observations show that the supergranulation pattern moves faster than the plasma.

Magnetoconvection in sunspots as well as solar cycle variations because of meridional circulations were also discussed.

Conclusions:

• Local and global helioseismology provide important constraints for convection in the Sun.
• Large scale numerical simulations are needed to interpret the data.
• Helioseismology can be used to verify simulations.

---

M. Asplund: Convection and the solar elemental abundances – does the Sun have a sub-solar metallicity?

Solar system abundances can be measured in meteorites (very high accuracy, but depleted in volatile elements) or the solar atmosphere (modelling dependent, very little depletion). The solar atmosphere is dynamic and three dimensional (3D) due to convection.

1D solar atmosphere models make various simplifying assumptions, but have the advantage that radiative transfer can be treated in detail. 3D solar atmosphere models are more realistic but have simplified radiative transfer. They are essentially parameter free.

The temperature structure in 3D is very different from 1D – it is very steep in upflows and there are significant inhomogeneities. The mean 3D structure is similar to 1D MARCS models but cooler than the semi-empirical Holweger-Müller model.

3D line profiles differ from 1D profiles. The spectrum formation is highly non-local and non-linear and strongly biased towards upflows. Profiles of an observed solar Fe line were shown and compared with 1D and averaged 3D line profiles. The 3D profile agrees without the need for micro- and macroturbulence. Line asymmetries and shifts are very well reproduced.

Solar C, N, O abundances have been derived using a 3D solar atmosphere model, non-LTE line formation when necessary, and atomic and molecular lines with improved data. Details are in Asplund et al. (2000-2006).

Oxygen diagnostics: In 1D, atomic and molecular lines give discordant results (log O = 8.6-8.9). In 3D, there is excellent agreement. The [O I] 630 nm line is blended with Ni, which gives a correction of -0.13 dex (not noticed in 1D), actually larger than the difference 3D minus 1D (-0.08 dex). The O I 777 nm feature has been calculated with full 3D non-LTE line formation, and the main difference is non-LTE minus LTE (-0.2 dex). This is most significant at the limb. OH vibration-rotation lines in the infrared have been calculated with 3D LTE line formation. Molecular lines are extremely sensitive to temperature.

Carbon diagnostics: Again, 1D there are discordant results (8.4-8.7), while in 3D there is excellent agreement. CO vibration-rotation lines as well as weak CO lines give an abundance of 8.39+-0.05. There are still problems with the strongest CO lines. C and O isotopic ratios can be derived, which agree with terrestial values.

There are also preliminary 3D results for all elements from Li to Ni.

The implications are a siginificantly lower solar metal mass fraction Z – from 0.0194 (Anders & Grevesse 1989) to 0.0122 (Asplund et al. 2005). This alters the cosmic yardstick, and makes the Sun normal compared with its surroundings (e.g. OB stars). The problem is that solar interior models with the new abundances are in conflict with helioseismology (there is no solution yet).

Conclusions:

• 3D + non-LTE + new atomic/molecular data give significantly revised solar CNO abundances
• The new abundances solve a lot of problems but are terrible for solar modelling
• Coming soon: New 3D models with improved radiative transfer treatment