May 20, 2009 | A new technique developed at NCAR will help asteroseismologists learn about stars from their oscillations, or “starquakes.” These variations in the brightness of stars reveal information about their internal structures, in much the same way that earthquakes are used for studying Earth’s inner structure.
NASA’s Kepler satellite, launched in March, will retrieve a flood of data on stellar oscillations over the next several years. The satellite is using a space telescope to continuously monitor the brightness of more than 100,000 stars in its field of view as it orbits around the Sun.
Travis Metcalfe (ESSL/HAO and CISL) has developed a computational method for deriving reliable asteroseismic information from Kepler’s observations. “The amount of data will be so immense that we can’t possibly keep up without automating the procedure,” he says.
The cornerstone of the method is a stellar modeling pipeline that is driven by a parallel genetic algorithm, a low-level form of artificial intelligence. The algorithm takes data from Kepler and converts it into information about a star’s age, temperature, surface brightness, and radius. The process, which is very computationally intensive, is being run on CISL’s bluegene supercomputer.
Travis has tested the algorithm against data on the oscillations of our nearest star, the Sun, to see if it accurately captures the Sun’s properties. “We have the advantage with the Sun that we know those properties extremely well, since they’ve all been measured,” he explains. The results were very successful, with the algorithm capturing the Sun’s observed oscillations and other properties with high precision.
Travis is collaborating with Matthew Woitaszek (CISL) to make the pipeline available to the community as a TeraGrid Science Gateway beginning next October, so that researchers without access to supercomputers can upload data sets from Kepler and run the algorithm. For more information, visit the Asteroseismic Modeling Portal.