They may be halfway across the universe, but distant supernovae have caused quite a ruckus on the earth. Two years ago astronomers discovered that these stellar explosions are dimmer than expected--so much so that the very structure of space seemed to be responsible. Their anomalous dimness intimated that the expansion of the universe may be speeding up rather than slowing down, as cosmologists had always assumed. But now several of the supernova observers worry that their dramatic findings, like some other disputed claims, jumped the gun. Adam G. Riess of the University of California at Berkeley says grimly, "I liken it to the discovery of life in the Mars rock."

 The argument for cosmic acceleration depends on two key measurements for each of several dozen supernovae: the maximum brightness of the explosion, which shows how far away it is and hence when it took place; and the redshift, which records how much the universe has expanded since it occurred. The farthest known supernovae went off 8.4 billion years ago, and since then the universe has doubled in size. Yet at its current expansion rate (as inferred from more recent supernovae) the universe would have tripled in size. Therefore, the expansion rate must have increased. 

All this hinges on differences in brightness of about 25 percent. As it happens, supernovae of the type used by cosmologists naturally vary by about the same amount, probably because of differences in the mass and composition of the stars that give rise to such explosions. This diversity once made supernovae too unpredictable to serve as gauges of distance. But in studies of nearby supernovae, observers noticed that the brighter they are, the slower their glowing embers fade over time. By measuring the rate of fading, observers can now compensate for the variations. Their only assumption is that this procedure also works for faraway supernovae.

 Which is what Riess and his colleagues have recently called into question. The team, a subset of the High-Z Supernova Search, one of the two main groups examining supernovae in a cosmological context, looked at a hitherto neglected attribute of the explosions: the time it takes them to attain their maximum brightness. They spied 10 nearby supernovae and tracked their brightening, in one case using a photograph taken by amateur astronomer Chuck Faranda of Fort Lauderdale, Fla., about a day after the blast began--one of the earliest images ever made of a supernova.

 The researchers determined that the nearby supernovae typically took 20 days to reach peak brilliance. By comparison, a group of far-off supernovae--reported earlier by the other main group, the Supernova Cosmology Project, based at Lawrence Berkeley National Laboratory--took only 17.5 days. Therefore, it seems that stellar explosions unfold differently depending on how long ago they occurred.

 Persis S. Drell and his colleagues at Cornell University have raised another warning flag: different formulas used to compensate for the natural variation in supernovae, each purporting to calculate the true brightness of the explosions, generate slightly different values.

 Many theorists think these discrepancies don't matter. The brightening time depends on the early stages of the explosion, when details make the most difference. But the maximum brightness and fading rate--the only two things cosmologists care about--involve the gradual release of energy by the debris, which should be largely independent of the characteristics of the late star. Other researchers are less sanguine. The disparities might be the dead canary in the coal mine, indicating that the theory is awry even if nobody knows quite why.

 Alternatively, the discrepancies might indicate that acceleration is merely an artifact of observational biases. Andrew Howell and his colleagues at the University of Texas in Austin have pointed out that distant supernovae have been discovered with electronic detectors, whereas nearby ones have mostly been found on photographs. Film tends to overexpose the center of galaxies and so miss any supernovae there, thereby skewing the sample and undermining the procedure that compensates for supernova diversity. But by the same token, selection effects may have led Riess's and Drell's teams to find problems where none really exist. Only new studies, designed specifically to avoid such biases and to check whether near and far supernovae are indeed comparable, will resolve the issue. So was it premature to trumpet the accelerating universe as a revolution in cosmology? Not necessarily. Other pillars of evidence, such as the cosmic microwave background radiation and estimates of the total amount of matter in the universe, remain unshaken. "The supernova evidence was never, in my view, the best evidence for cosmic acceleration," says cosmologist Paul Steinhardt of Princeton University. "The arguments we already had are much more solid." 
 

--George Musser