FOR years scientists have puzzled over what the universe is made of. Detected matter accounts for only about 10 percent of the universe's mass. But while researchers still can't identify the ``dark matter'' that makes up the remaining 90 percent of the universe, they are starting to eliminate some possibilities.
The dark-matter candidates fall into two broad categories:
* In what John Bahcall, professor of natural science at Princeton University's Institute for Advanced Studies, calls ``the most conservative solution,'' dark matter is like ordinary matter and consists mainly of faint, low-mass stars, such as red dwarfs, which are very difficult to see.
* The more radical solution: that dark matter consists of exotic subatomic particles left over from the Big Bang that formed the universe. These have not been detected either, although physicists hope to do so using powerful new particle accelerators.
In findings released this week in Washington, two researchers appear to have ruled out what for 30 years has been the preferred solution for many astronomers: that low-mass stars held the answer to the dark-matter problem.
``The most fundamental problem in astronomy today, and certainly the most vexing problem we have in particle physics, is the question of what is the dark matter? That problem influences everything we think about the universe,'' Dr. Bahcall says.
``Very reluctantly, I've been dragged into the position of thinking that the most likely solution is going to be a quite radical one, one that will make my physics friends happiest,'' Bahcall adds.
Using the Hubble Space Telescope, a team led by Bahcall and Andrew Gould of Ohio State University surveyed large portions of the halo of the Milky Way, a region thought to host a large number of red dwarfs. At the same time, Francesco Paresce, with the Space Telescope Institute and the European Space Agency, used Hubble to search for faint red dwarfs in a globular cluster, where computer models suggest that he should find an abundance of them.
Both research teams found many fewer red dwarfs than expected. And Dr. Paresce discovered that stars with masses of less than 20 percent of the sun's were very rare, suggesting a far higher mass requirement for star formation than previously thought.
``The universe is such that less than 16 percent of dark matter is made of faint stars,'' Bahcall concludes.
These findings further undercut so-called brown dwarfs as dark-matter candidates, according to Steve Maran, an astronomer with the National Aeronautics and Space Administration's Goddard Space Flight Center in Greenbelt, Md. Brown dwarfs are star-wannabes that never grew massive enough to ignite in a stable fashion.
Although dark matter is difficult to detect directly, its presence has been inferred from its gravitational effects on galaxies, clusters of galaxies, and large clouds of hot gases. The gravity attributed to dark matter helps keep these structures stable.
While the results from the Hubble Space Telescope appear to eliminate low-mass stars as dark-matter candidates, other research suggests that it isn't yet time to write off other forms of ordinary matter as the solution, says Joseph Silk, a professor of astronomy at the University of California at Berkeley.
``Some elegant experiments are being done'' to find massive objects in the galaxy's halo, so-called massive compact halo objects, or MACHOs, Dr. Silk says.
Scientists ``are scanning millions of stars, and they are finding things, such as gravitational lensing, in great numbers,'' Silk says. Gravitational lensing refers to gravity's ability to bend light, and the phenomenon implies the presence of an otherwise unseen object between the observer and the star whose light is being bent.
These experiments, Silk says, are unable to determine the exact mass of the objects doing the lensing. But if the objects have a mass of at least one-tenth of the sun's, they either could be very cold white dwarfs - remnants of burnt-out stars - or black holes. With less than one-tenth of the sun's mass, he adds, they could be brown drawfs.