Bradstreet receives NSF grant for binary star project

Eastern’s physical science professor recently received a substantial National Science Foundation grant to be used in a groundbreaking project involving binary stars.

The grant is an honor, since NSF grants are awarded to only about eight out of 100 applicants per year, according to recipient David Bradstreet.

“If you don’t have a good reputation, you aren’t going to get the grant,” he said. “It’s nice to be recognized and hopefully we’ll be able to do what we said we’d do.”

Bradstreet said the three-year, $400,000 grant, awarded in July, will be used in an attempt to develop sophisticated software capable of automatically analyzing the light curves of binary stars (stars that orbit each other).

Bradstreet applied for the grant in November of 2004 along with four other men: Ed Guinan, from Villanova’s astronomy department and the head of the project, Ed Devinney, a retired industry scientist, Marty DeGeorge, a research scientist with Panasonic, and Joe Giammarco, a post-doctorate mathematical physicist from the University of Pennsylvania.

The team’s proposal was to develop software that would allow a computer to analyze, within seconds, many properties of binary stars including their masses, sizes and temperatures, according to Bradstreet.

These characteristics can be determined from a study of the light curves of the binary systems. Each binary system has one light curve, which is the graph of the brightness of the stars versus the time it takes them to orbit each other, Bradstreet said.

“The light curve is the signature of the stars,” he said. “This is one of the few ways you can directly measure the qualities of stars.”

The present system for studying light curves takes about eight hours per light curve. Thus, fewer than 1,000 of an estimated 10,000 available light curves have been analyzed so far, Bradstreet said.

Satellites being developed now are expected to increase the number of light curves available as they discover more binary systems. They could bring the number of known light curves to 10 million, according to Bradstreet.

This information led to the team’s idea for a grant proposal.

“We’re going to get this firehose of data in the next 10 years, and somebody will have to do it, so why not us?” he said.

Although other groups are also trying to speed up light curve analysis, Bradstreet said his group’s project is unique because they are using neural networks, programs that allow the computer to learn from experience.

“If you give it many, many examples, it learns,” he said. “We’re trying to get the computer to know as much about light curves as Ed [Guinan] and I know.”

They will do so by feeding the computer millions of artificial light curves, giving it the experience necessary for it to accurately analyze real light curves, according to Bradstreet.

“Just like a child. You teach them vocabulary and give them examples, then tell them to write a story,” he said.

The team has already begun the project. DeGeorge has written the neural network program, and Bradstreet has developed an initial sample of 13,320 light curves to be used in a test of the program’s effectiveness, Bradstreet said.

The software, if the team succeeds, will give scientists access to information that can help them better understand how stars form and evolve, he said.

This information also has intrinsic worth, according to Bradstreet.

“Will it make your car cheaper? Probably not,” he said. “But what good is music? It makes life richer. This information is filling out the picture of our knowledge about stars, and from a Christian standpoint it helps us appreciate the Creator more.”