Institute for Computational Astrophysics

Previous Images of the Month - 2011

January February March April May June July August September October November

December 2011

click on image to enlarge

It has been reasonably assumed that the combination of interferometry and asteroseismology might provide sufficient information to produce reasonable models of rapidly rotating stars. The interferometry provides both the oblateness of the star and the inclination angle between observer and the star’s polar axis. Asteroseismology provides the oscillation frequencies which a successful model must match. A good test of this is α Oph, which has been observed with the CHARA interferometric array and asteroseismologically by the MOST satellite. Work by ICA faculty Drs. Deupree and Short, post doctoral fellow Fernando Peña, and graduate student Diego Castañeda has found that even the interferometic and asteroseismological constraints still leave too many degrees of freedom to adequately determine the appropriate stellar model. However, they have found that adding the extra constraint of matching the spectral energy distribution of this star in both the optical and the ultraviolet does constrain the models sufficiently well. The upper figure compares the observed spectral energy distribution and one computed for a 2D stellar model at the observed distance, oblateness, and inclination of α Oph. The solid curve is the computed spectral energy distribution and the dashed curves with circles represent the observed spectral energy distribution. They then computed the oscillation frequencies for this 2D model. The results are shown in the lower figure. The circles represent the observed frequencies and the other symbols the computed frequencies for different numbers of longitudinal modes. Given the number of computed modes, it is perhaps not surprising that all observed modes can be matched, but it is perhaps indicative that seven of the nine highest amplitude modes can be matched to within the observational error by axisymmetric, equatorially symmetric modes.

November 2011

click on image to enlarge

The graph shows the α Oph observed frequencies (circles) and the calculated pulsation frequencies for a 2D stellar model rotating with a surface equatorial velocity of 230 km/s. The rotating models and the pulsation frequency calculations were performed by Dr. Robert Deupree of the ICA. The plot is called an echelle diagram, in which the frequencies, instead of being plotted in one long line, have adjacent segments of the line stacked vertically. This provides good visual resolution of the frequencies while still covering a relatively large frequency range. The individual modes can be identified by the number of nodes they have in the azimuthal direction, m. The computed modes are all the equatorially symmetric modes for |m| ≤ 4. The computed frequency symbols are given by squares (m = 0), diamonds (|m| = 1), inverted triangles (|m| =2), triangles (|m| = 3), and right facing triangles (|m| = 4). Red corresponds to positive values of m, and blue to negative values of m. The observational data have been artificially offset vertically to highlight them. The profusion of computed modes produces so many matches to the observed frequencies that distinguishing between models may be difficult.

October 2011

click on image to enlarge

SMU astronomy graduate student Bobby Sorba and Prof. Marcin Sawicki have been working as part of a group commissioned by the Canadian Space Agency to look into how Canada might contribute to the planned European Euclid space mission. Euclid's goal is to better understand Dark Energy, which is the recently-discovered but mysterious substance that accounts for 70% of "stuff" in the Universe (of the remaining 30%, about 25% is Dark Matter and only 5% is "ordinary" matter). The researchers used the computing power of ACEnet to generate realistic simulations of how adding a Canadian-designed U+G imager to Euclid would affect performance.  The image above, one of the results of this work, shows how adding the extra observations increases the fidelity of photometric redshift distance estimates to galaxies in the distant universe.  The left panel shows the poor accuracy before the U+G imager is added, and the right panel shows the much better accuracy that can be expected with its addition.  Photometric redshifts are a key ingredient in Euclid's planned measurements of Dark Energy.  These results were recently published in the Publications of the Astronomical Society of the Pacific.

September 2011

click on image to enlarge

PHOENIX Non-LTE synthetic spectra of red giants

Simulated spectra for thirteen computer models representing the atmospheres of Red Giant (RGB) stars of "surface temperature" (i.e. effective temperature, Teff) ranging from 4000 to 5500 Kelvin (K), in steps of 125 K (as indicated by colour and on the y-axis of the plot), computed with Version 15 of the PHOENIX atmospheric modelling and spectrum synthesis code (Hauschildt, Allard & Baron 1999).  The z-axis represents the radiation power, or flux (i.e. "brightness"), in each spectral wavelength interval (delta-lambda) of 1 cm that would be measured by an ideal radiation detector of 1 square centimeter at the apparent "surface" of the star.  Note that the spectra become brighter, and the peak flux shifts to shorter wavelength (lambda, x-axis) as Teff increases.  

The models represent spherical stars with the same mass and chemical composition as that of the Sun (Teff=5800 K), but have a surface gravity, g, of 100 grams/(square centimeter) (ie. log(g)=2), about 275 times weaker than that of the Sun, giving them a radius of about eighteen times that of the Sun.  The spectra were sampled with a wavelength interval (Delta-lambda) that varied from 0.01 Angstroms (A) at 3000 A in the ultra-violet (UV) (log(lambda)=3.5 on the x-axis) to 0.037 A at 13000 A in the infrared (IR) (log(lambda)=4.1 on the x-axis), for a total of about half-a-million wavelength points (1 Angstrom (A) = 0.1 nanometers (nm)).  The wavelengths of several spectral absorption ("dark line") features that are important for identifying Red Giant stars, and the chemical species that give rise to them, are indicated.  

The unique feature of these models is that the state of 35 chemical species that are important opacity sources, including Hydrogen (H), Carbon (C), Nitrogen (N), and Oxygen (O), and the first two or three ionization stages of many light "metals", and several iron-peak elements, including iron (Fe), were computed in Non-Local Thermodynamic Equilibrium (Non-LTE).  As a result, thousands of spectral absorption lines of these species are included more realistically than in previous modeling.  Many more chemical species, including diatomic molecules such as Titanium Oxide (TiO) and Carbon Monoxide (CO), were also included, but with the "classical" simplifying approximation of Local Thermodynamic Equilibrium (LTE).  More details can be found in Short & Hauschildt 2010.  All models were computed on ACEnet facilities available to the ICA.

Hauschildt, P., Allard, F., & Baron, E., 1999, Astrophysical Journal, 512, 377

C. Ian Short and Peter H. Hauschildt, "Modeling the Near-UV Band of GK Stars, Paper I: LTE Models", 2010, Astrophysical Journal 718, 1416

August 2011

click on image to enlarge

Question: Why are the oscillations on Procyon so difficult to observe?

Answer: Unlike the sun, the oscillations on Procyon look similar to convective motions at the star’s surface and it is, as a consequence, difficult to distinguish the two.

The Answer explained: Drs. Frank Robertson (Yale), Pierre Demarque (Yale), Thomas Kallinger (Vienna), and David Guenther (SMU) ran a series of 3 D numerical simulations of convection for Procyon. They discovered that most of the energy (or power) associated with granulation (i.e., convective motions at the surface of Procyon) occupied the same frequency range as its pulsations. In the figure above, a power density plot, you can see that the light red band between 900 and 1200 µHz, which corresponds to the frequency range of Proycon’s oscillations overlaps local peak in granulation power. This is not the same as for the Sun, where the granulation power and frequency range of its oscillations (shown as a light blue band) are distinct. The long life-time oscillations are mixed together in the same part of the frequency spectrum as the short life-time granulation motions. The simulations also showed that the region where the oscillations are driven is much closer to the surface, where radiative losses are significant, than for the sun. This explains why the Procyon’s pulsations in luminosity have lower amplitudes than in velocity.

July 2011

click on image to enlarge

The Pulse of a Young Star Cluster

In late 2006/early 2007, Canada's first space telescope, the Microvariability and Oscillations of STars (MOST) telescope, observed the young open cluster NGC 2264 for approximately one month, as requested by the MOST Science Team, of which ICA member David Guenther is a member. Six stars were seen to pulse in multiple frequencies, and here we see the frequencies (in red) with which each of those six stars pulsate (each star's name is in the upper-right-hand corner of each graph).  The blue "bar codes" in each graph represent one of the theoretical stellar-pulsation frequency-spectrum models, computed at the ICA that happens to most closely match the star in question.

The frequencies in microHertz correspond to periods of several minutes to several hours. Stars that pulsate within this frequency range are known as Delta Scuti stars, named after the prototype star of class, Delta Scuti, the fourth-brightest star in the constellation Scutum (The Shield). The frequencies with which a star can pulsate are determined by the star's internal structure, and the study of these pulsations is known as asteroseismology.  This is where the ICA comes into the picture: in order to understand these pulsations, many stellar evolution and pulsation models must be computed, against which the observations are compared, requiring lots of computer time.  For this particular class of star, stars less than 10 millions years old, known as Pre-Main-Sequence (PMS) stars, approximately one million individual stellar models needed to be computed (and "pulsed") before sufficient computational data existed for meaningful comparisons to be made to the observed pulsation spectra.  It has been the work of graduate student Mike Casey to perform these comparisons for both the stars of NGC 2264, and of any other PMS star for which delta-Scuti-type pulsations have been observed.

Background image was produced by European Southern Observatory. NGC 2264 is a young open cluster in the constellation Monoceros (the Unicorn), a region of active star formation containing very young stars. The region contains the Cone Nebula and the Christmas Tree Cluster.

Mike Casey (PhD Student, ICA)

June 2011

click on image to enlarge

The interaction of radial pulsation and convection has been a long standing problem in stellar astrophysics.  In particular the interaction of convection with radial pulsation can stop the pulsation of a star altogether by transporting heat through the ionization region at the critical phase and removing the driving mechanism for pulsation. This is what happens at the “red edge” of the RR Lyrae instability strip, the cool side of the region in temperature and luminosity space where RR Lyrae stars pulsate.

Purely radiative (heat transported only through radiation) models of radial pulsation in RR Lyrae stars has been unable to match the observed location of the red edge of the RR Lyrae instability strip (the lower effective temperature at which RR Lyrae stars cease pulsating). The ability of convection to terminate pulsational driving is what produces the location of the red edge. The image shows a graphical representation of a radially pulsating star with 2D convection. The interior mass is the independent radial coordinate so that it “flows” with the pulsation. These results are based on a 2D hydrodynamic simulation with the SPHERLS code, which is being developed in the ICA by Ph. D. student Chris Geroux under the guidance of Dr. Robert Deupree, ICA Director. The color scale shows the temperature; with white indicating a temperature of 10,000 Kelvin (the temperature at which hydrogen is ionized). Vectors show the convective velocities of the material. The bottom and left axis show the radial and horizontal extent of the simulated star. Only a small pie slice of the star has been simulated. The white lines on top and bottom show the location of the periodic boundary conditions. Zones outside these lines are copies of zones inside the white lines on the opposite sides.

The small figure shows the growth of pulsation energy with time for a 1D radiative simulation and a 2D simulation including convection. One can see that in the 2D simulation, growth of the pulsation energy is slower than in the 1D simulation, indicating that convection has reduced the driving of pulsation by the hydrogen ionization region. 

See an animation of the pulsating star on youtube.

Chris Geroux (PhD Student, ICA)

May 2011

click on image to enlarge

This is a picture of a galaxy simulation - a model of a galaxy quite similar to our own Milky Way. It probably looks different to most pictures of galaxies you have seen. This is because most of the visible light that comes from a galaxy is from the bright stars, and this is what we see with our telescopes. However, in my research, I am interested in the gas in a galaxy. The light that comes from this gas is much easier to see with an infrared telescope instead.

On the right you can see my whole galaxy, seen from above. It's very clumpy. When we look at our own "real" galaxy, we observe clumps of gas too - we call these Giant Molecular Clouds. I'd like to find out how these clouds form, and what influence they have on the whole galaxy. In particular, these clouds can collide with each other (sequence on left).  Whenever two clouds collide, they lose some of their kinetic energy. In my research I'm calculating how often these clouds collide, and how much energy they lose when they do. This will affect the whole galaxy, as it will slowly be losing kinetic energy. These simulations are difficult to do. This is because we have to simulate little molecular clouds (dozens of light years across), and the entire galaxy (hundreds of thousands of light years across). It is only fairly recently that computers have reached a level where we can simulate this whole range of sizes at the same time. These simulations were performed on Dr. Rob Thacker's "Cerberus" cluster at the ICA, which has over 300 processors. Dr. Rob Thacker was also involved in developing Hydra, the program I use (and modify) to do these simulations, while the cluster is maintained by the multitalented David Lane

David Williamson (PhD Student, ICA)

April 2011

click on image to enlarge

High performance computers can be used to study galaxy collisions and mergers, taking only a few days to simulate events that actually occur over a time span of a few Gyr (10^9 years).  The above 4 images are snapshots taken from such a simulation, where the colours represent different gas densities (with density increasing from pink to red to yellow).  Initially, (0 Gyr) there are two Milky Way-type galaxies on a parabolic orbit around one another.  The effect of gravity causes the galaxies to collide (0.53 Gyr), but there is enough angular momentum in the collision for the two disrupted galaxies to move apart (0.60 Gyr); as can be seen, each galaxy has a tidal tail that follows its original orbital path and/or rotational direction.  The galaxies continue to move apart (0.68 Gyr), with more gas being stripped from the galaxies and going into the tidal tails.  Eventually (not shown), there will be a second collision of the two galaxies, and they will merge to form a new galaxy.

This simulation was performed by James Wurster using the Hydra SPH-AP3M code on Dr. Rob Thacker's computer cluster at the ICA.

March 2011

click on image to enlarge

Numerous scales of a stellar jet being launched from a Keplerian (rotating) disc, maintained as boundary conditions on the left-hand side, as computed by AZEuS. Colour contours represent the Alfvénic Mach number, solid black lines are lines of magnetic induction, and dashed lines indicate nested grids where each grid has twice the resolution of the grid containing it. Grids are fixed, although AZEuS is capable of adaptively adding, modifying, and eliminating grids as needed. In this image, the jet is launched along the bottom quarter of the left boundary of the bottom panel. A pseudo-steady-state production of "knots" at the launch site is established from very early on that persists during the entire simulation (~ 100 yr). These knots merge and form larger clumps before blending into what becomes a magnetically-confined `"nose-cone" (e.g., Clarke, Norman, & Burns, 1986, ApJL, 311, L63), giving the jet its familiar bow-shock enshrouded appearance at the observational scale (top panel). The jet is followed to about 2,500 AU.

Credits: J. P. Ramsey & D. A. Clarke

February 2011

click on image to enlarge

Quakes on gaseous planets

Oscillations (pulsations) in stars and planets can be used to probe their internal structure.

Just as the scientific field of seismology tries to understand the internal structure of the Earth, asterosesimology (an area of the research interest in the ICA) tries to do the same for stars and planets.

The 'beach-ball-like' figure shows  (in fake colors) surface motions on a Jupiter-like gaseous planet; green patches show motions of matter going outwards from the planet, while white patches show inward motions. The pattern smoothly reverses  itself roughly every half rotation.  The Coriolis force is behind these motions, and the oscillating pattern is called an inertial mode.

In order to produce this figure Dr. Fernando Pena computed solutions to the appropriate equations using high performance computers and then used the 3D graphic workstations and the Data Cave available at the ICA to create a 3D animation of a pulsating planet, here seen as a 2D-projected snapshot at an instant of time.

Astronomers have observed pulsating stars for hundreds of years. However, pulsations have yet to be observed in planets others than the Earth.  Theory predicts the light variations associated with the pulsations would be very small, and therefore difficult to detect. As technology improves we hope that in the next decade we will be able to detect them.

Take a deeper look into the pulsations in Dr. Pena's youtube

January 2011

click on image to enlarge

Modern astrophysics relies on high quality observations and up-to-date theoretical models. To bridge the gap between the two, statistics and probability theory provide us with the tools to evaluate which models fit best to our observations. As our models become more complex due to increasing numbers of parameters, and as telescopes deliver more and more data, the need for sophisticated statistical methods and computer resources increases dramatically. Only with high-performance computation facilities, such as ACEnet, can this challenge be met.

At the ICA, Michael Gruberbauer uses state-of-the-art methods for solving statistical inference problems in order to study the stars. The figure shows a so-called "power spectrum" of a pulsating star. The star was observed for more than 150 days by the french CoRoT satellite (artists impression: (c) CNES 2005). The power spectrum reveals the presence of a large number of pulsation modes (indicated by the various peaks in the spectrum). Each mode lets the star "ring" like a wobbling soap bubble but always in a slightly different way, most importantly at different frequencies. By studying these modes, details of the stellar interior can be inferred. The red line on top of the spectrum shows a statistical model that was used to detect the exact frequency values and the shape of the modes by employing a sophisticated statistical method called "nested sampling". The model consists of 61 free parameters, and it would take weeks to months to evaluate it using a modern Desktop computer. It only takes hours to days on the ACEnet clusters. This allows Michael to compare a number of different models to see which one fits best.