Astronomers Announce Discovery of Extremely Distant Quasars
PASADENA—Astronomers have discovered 16 new extremely distant quasars, the result of a search made nearly 40 times more efficient than previously possible by applying artificial intelligence to the new Palomar digital sky survey. This novel technique allows researchers to study more easily the formation of quasars and large-scale structures in the early universe.
"This is one of the first successful major applications of artificial intelligence techniques in astronomy and space science," said Usama Fayyad, a scientist at the Jet Propulsion Laboratory (JPL) in Pasadena, California. "Data mining techniques and automated data analysis are becoming a necessity in this new era of astronomy and space science where instruments can generate tremendous amounts of data. The discovery of these new quasars shows how efficiently scientists can explore vast databases such as the Palomar sky survey, using this novel data-mining technology. And this technique is applicable to many other data-rich fields. It is a truly new way of doing science."
These results are reported in the December issue of the Astronomical Journal in a paper by Julia Kennefick, a postdoctoral researcher at Ohio State University; S. George Djorgovski, an associate professor of astronomy at Caltech; and Reinaldo Ramos de Carvalho, a senior research fellow in astronomy, also at Caltech. Kennefick is a former graduate student of Djorgovski. Some of the technical developments leading to these discoveries have been reported earlier by Djorgovski, Fayyad, and their colleagues.
The astronomers have found 16 new quasars at redshifts greater than 4 (redshift is a measure of distance in cosmology), corresponding to look-back times in excess of 90 percent of the age of the universe. Such objects are exceedingly rare, and finding even a few of them is considered very important by astronomers. The recently discovered quasars are providing a new glimpse of the very early universe.
"We see these quasars at a time when the universe was only a billion years old, when the first structures were just forming," explained Djorgovski. The study in the Astronomical Journal confirms a previous suggestion that the number of quasars diminishes rapidly as one looks back toward earlier epochs in the universe. In other words, astronomers are seeing the appearance of the first quasars, when the universe was only one-tenth of its present age, or possibly even younger.
The scientists, led by Djorgovski, are conducting a systematic search to discover large numbers of extremely distant quasars using a set of sophisticated artificial intelligence (AI) software tools developed for this task in collaboration with Fayyad and his Machine Learning Systems Group at JPL.
The astronomers are applying these AI tools to a new digital survey of the entire northern sky. The digital sky survey is being produced as a collaborative project between Caltech and the Space Telescope Science Institute in Baltimore, Maryland, and is based on the photographic sky survey done with the 48-inch Oschin Telescope, a Schmidt telescope at Caltech's Palomar Observatory in northern San Diego County. When complete, the digital sky survey will contain enough information to fill about 6 million books and will include about 2 billion stars, galaxies, quasars, and other objects.
In order to efficiently process this unprecedented amount of astronomical information, a team of scientists from JPL led by Fayyad, in collaboration with Djorgovski and his former student Nicholas Weir, now at Goldman, Sachs and Company in New York, developed a powerful software system, called the Sky Image Cataloging and Analysis Tool (SKICAT). The SKICAT system incorporates cutting-edge AI technology, including machine learning, machine-assisted discovery, and a high-performance database system to automatically measure and classify the billions of objects in the sky survey images, and to assist astronomers in performing scientific analyses of the resulting catalogs.
The Caltech group used SKICAT to select quasar candidates from catalogs of objects detected in the sky survey, sorting through roughly one million other objects to find each quasar. On photographs, quasars are indistinguishable from ordinary stars in our galaxy.
"This is far more difficult than finding needles in a haystack," said Kennefick. "SKICAT allows us to automatically sort through and pinpoint interesting quasar candidates based on their color, so that we can make the best possible use of the valuable telescope time in checking them out." A previous survey for quasars at comparable distances done at Palomar used about 20 times more nights, with the 200-inch Hale Telescope, and found only nine quasars.
"This great increase in the observing efficiency is due to a combination of the huge amount of data in the sky survey, and the modern software techniques that allow us to explore it," Djorgovski said. "And the more quasars we find, the better we will be able to map these early epochs of the universe."
"Data mining and automated analysis of large databases offer the promise of giving us a handle on the data avalanche generated by NASA instruments on missions to planet earth and elsewhere in the solar system," said Mel Montemerlo, the manager of the Autonomy and Operations Program at NASA headquarters in Washington, DC.
This work was supported by the National Aeronautics and Space Administration, with additional funding from the National Science Foundation.
Contact: at the California Institute of Technology
at the Jet Propulsion Laboratory at Ohio State University Usama Fayyad Julia Kennefick (818) 306-6197 (614) 292-5403 firstname.lastname@example.org email@example.com Fax (818) 306-6912 Fax (614) 292-2928
Written by John Avery