Institute for Computational Astrophysics
click on image to enlarge
Many stars are (literally) vibrating. Because the vibrations are sound waves trapped inside the star they yield a rich and complex harmonic structure (the oscillation spectra) that when interpreted through computer modeling gives us a picture of the interior of the star. The oscillations are visible as minute (1 part in 1 million) variations in the star’s luminosity and can be seen from satellites (such as Kepler) that are sensitive to small variations in the light output from stars.
To interpret the oscillation spectra astronomers compare numerically computed oscillation spectra of stellar models to the observed star’s oscillation spectrum. Unfortunately, the complicated physics of the surface layers, which are highly turbulent, dominates the oscillation spectra and hinders efforts to explore the physics of the deeper interior.
In Fig. A we show an echelle diagram for a model star. The diagram plots the frequency of each oscillation mode in a manner that reveals its harmonic structure. We have used a simple heuristic model of the surface physics to show its potential impact on the frequencies. This can be seen as an increasing size of the deviation of the red and blue curves from the black, reference curve).
Unfortunately, the observed oscillations rarely extend to the lower frequencies where the surface effect is minimal. Consequently, any model fit to the observations could be significantly skewed by the unknown magnitude of the surface effect. We get around this issue by using Bayesian statistics and marginalizing out all surface effects. Effectively, when determining the probability of how well a given set of models fit the observed oscillation spectrum, we cancel out all possible surface effects. By doing so in all our models, we can then evaluate the other model physics.
In Fig. B we show an example of how well this works. A test model was constructed with some interesting interior physics; specifically, it was given a 0.2 pressure scale height overshoot beyond the outer convective core boundary. In addition we contaminated the oscillation spectra of the models with a Gaussian uncertainty of 0.5 µHz and added a surface effect perturbation. We then used our Bayesian codes to marginalize the (unknown) surface effect (effectively cancel it out of the comparisons) and compare our standard models with varying amounts of overshoot to the three test models. The Bayesian analysis determines which models have the highest probability (in figure B this is represented by the Evidence) of being a fit to the test model. Models with the highest evidence are the most probable. In all three test models, the code correctly identified the 0.2 core overshoot models as the most likely best model fits; even when there was significant surface effect contamination.
Dr. David Guenther - ICA & SMU Astronomy & Physics Professor
click on image to enlarge
A balance of pressure and gravitational forces produces "hydrostatic equilibrium" in astrophysical systems. Departures from this equilibrium in stars often cause "waves" that force surface layers to expand and contract. These pulsation events produce variations in luminosity that -under the right conditions- can become periodic and measureable from Earth. Similar to how a string can show different "modes" of vibration, stars undergoing pulsations can show several different "modes" of pulsation. The very special conditions required for these modes to occur mean that if they are successfully modelled, it is possible to learn much about the internal structure of the star.
One of the most important processes that can change the stellar pulsation properties of a star is rapid rotation. A rapidly rotating star is no longer spherical and this break in symmetry makes modelling pulsations more complex. Understanding how rotation affects individual modes of pulsation is an important step towards better understanding some of the pulsations shown by rapidly rotating stars.
PhD student Diego Castaneda working together with former ICA director Dr. Robert Deupree, have computed several multidimensional rotating models to study their pulsational properties. By applying linear adiabatic perturbation codes to these models, several thousands of modes have been calculated in an effort to expand the understanding of oscillation modes for rapidly rotating stars. The figure shown (left) presents a sample of how the calculated oscillation frequencies of two modes change at different rotation rates for a star with a mass of 2.5 times the mass of the Sun. One interesting aspect of this frequency "evolution" is that there are rotational velocities at which "different" modes interact with each other mixing their own properties, as is the case inside box shown in the figure. The 2D contour plots on the right show the before and after of the meridional cross section of the pressure perturbations of the interacting modes enclosed in the box. In the case of the pressure perturbations, the lower frequency mode is on the left. This is one of the reasons why the mode identification can be much more challenging compared to non-rotating stars.
click on image to enlarge
The most reliable way to determine the redshift of a galaxy is to use spectroscopic observations to measure the wavelength shift of its spectral lines. When spectroscopy is not available, we compare photometric observations of the galaxy to a set of model galaxies. We can then select the best fitting model to determine a ‘photometric redshift’. In this case, the choice of an appropriate set of models is crucial. The model space must encompass all observations, but if it’s too large, searching the space can become computationally infeasible. This figure shows the projections of a set of model galaxies onto colour-colour planes of a set of observations. PhD candidate Anneya Golob, working with ICA faculty member Dr. Marcin Sawicki, is working on statistical methods to design optimal model spaces for finding photometric redshifts. Using ACENET’s computing resources, these techniques will be used to efficiently analyze observations of 3 million galaxies detected in the CLAUDS survey.
click on image to enlarge
Globular Clusters (GCs) are the largest and brightest groupings of stars within galaxies, easily observed in the Milky Way, and in distant galaxies where individual stars cannot be resolved. In GCs where the individual stars cannot be resolved, information about the stellar population is not readily observable. However, this information may still be extracted from the Integrated Light (IL) of a cluster. Doctoral candidate Mitchell Young, working with ICA faculty member Dr. Ian Short are developing a library of stellar atmospheric models and spectra for use in studying the IL of GCs. Each stellar model/spectrum pair in the library takes between 1-3 days to compute using the atmospheric modelling and spectrum synthesis code PHOENIX, running in parallel on 4 or 8 processors on the ACENET clusters Fundy and Mahone. By interpolating between the library models to create spectra for individual stars in a GC and combining them into a synthetic IL spectrum, populations of stars in GCs may be studied.
The white circles and black dots represent stellar models in the library, produced by assuming either Local Thermodynamic Equilibrium (LTE), a simplifying assumption that is not realistic, or Non-Local Thermodynamic Equilibrium (NLTE), the realistic treatment but also more computationally demanding. The three coloured tracks are a GC population, represented by the theoretical Teramo Isochrones of the BaSTI group, sampled at different ages but with all other parameters held constant. The blue track is 10 billion years old, the average young age of GCs in the Milky Way. The green track is 14 billion years old, roughly the age of the universe. The red track is 15 billion years old, and shows what the oldest GCs may look like in another billion years.