God In The Equation

by Corey S. Powell

  Just a few years earlier, such a project would have seemed laughable. This time the stumbling block was not telescope power, it was manpower. Nobody could have mustered the time or the staff to comb through all the images on photographic plates looking for a telltale flash, schedule follow-up observations around the world, and analyze with maniacal precision the rising and falling brightness of the supernova. The only reason Perlmutter could now dream of tackling such a challenge was because of some stunning developments in electronics technology, most notably more powerful computers and digital, silicon light detectors called “charge-coupled devices,” or CCDs. “We often treat the history of science as a back-and-forth between theorists and observers, but an equal part of that back-and-forth is the technology. It's more like a three-legged stool, with the technology allowing you to do things you couldn't do before. Einstein gave us the conceptual tools to ask these questions, but we didn't really have the technology tools until five or ten years ago for the kind of stuff I'm doing. Now the experimental results we're getting are going to push cosmology forward again,” he says. Hubble hacked away at the secrets of the universe with the glass and steel of the Hooker telescope. Cosmology had advanced from the industrial era to the information era, and the key tool was the silicon chip.

  CCDs aided the hunt for supernovas in part because they are much more efficient than photographic film, so they can squeeze more performance out of a telescope. More important, they keep an electronic record of every iota of light they receive. This digital tally makes it a snap to single out one exploding star in a field full of galaxies. So Perlmutter could start by creating a reference image of a patch of sky. He would then look again a few weeks later and cross his fingers that bad weather would not ruin the scheduled observing session. Next he would subtract the first image from the second. The digital sky images consist entirely of binary ones and zeroes, so the subtraction should leave nothing but background noise if everything stayed the same. Anything that wasn't there the first time around would pop out immediately. If one star exploded and brightened, it would now be hard to miss.

  It sounds simple enough. In practice, nobody had yet made the system work effectively. Perlmutter found himself spending long hours writing software to combine, clean up, and analyze the CCD images. “A lot of times you think, Boy, you're spending your whole life on this stupid computer,” he says with an enthusiastic laugh. The technology of the late 1980s, before the World Wide Web had even been invented, was barely adequate for the task. An effective supernova search requires a wide-field camera that can look at a lot of galaxies all at once, maximizing the chance that an explosion will appear in any particular image, but early CCDs were quite small. Commonly available computers barely had enough power to process all the data collected by the telescopes. Network links were essential for shuttling information from observatory to observatory, but the Internet was still an exotic and cumbersome device. There's a reason it was used back then primarily by scientists and technofreaks.

  Slowly the situation improved. CCDs got bigger, computers got faster, and Perlmutter began to develop more effective software. “One of the reasons the supernova studies seemed plausible to us was because the technology had advanced so far. We were some of the first people to develop CCD controllers. The computers went through such a change during that same period. When I started, people were using PDP-10s—they didn't have much memory. VAX computers came along and suddenly the programming became much easier. When we started doing the high-Z [large redshift] search, the computers were not quite up to it—but within a year they were. And then networks came up to speed. I was calling people at NASA, asking to borrow lines to Australia by hook and by crook. At every stage of the game, we didn't quite have the technology we needed, but that meant we were ready when the next step came along,” he explains. There were political hurdles as well. Large observatories schedule telescope time only for people who have plausible observing programs, and nobody was sure that the Supernova Cosmology Project could really achieve its ambitious goals. “It was a chicken and egg problem, so you haven't yet found the supernovas, so it's hard to ask for all the time that you want,” Perlmutter says.

  The first big break came in 1992, when the Supernova Cosmology Project bagged its first high-redshift supernova using a new, large CCD detector on the two-meter (six-and-a-half-foot) telescope at the La Palma Observatory in the Canary Islands. Although this single observation was too sketchy to have much scientific merit, it proved that the team was on the right track. Over the next two years, Perlmutter bagged a succession of other supernovas. The technique was working. But amid all the excitement, the sci/religious faithful began to disagree on how to interpret the heavenly messages.

  Perlmutter and his teammates occasionally consulted with other members of the tiny supernova community. One of those who lent a hand was Brian Schmidt, then a graduate student at Harvard. After a while, Schmidt started to feel uneasy. “We were not terribly happy with the way they were analyzing the data at the time—partly it was out of ignorance, partly it was real scientific differences. We thought it would be good for us to do our own, independent analysis of the data the way we wanted to do it,” Schmidt says. He talked to Kirshner and suggested that the time was ripe to launch a separate supernova search. Kirshner was skeptical at first—people had found supernovas before but not been able to squeeze useful information from them. “Yes, we could do it better, but could we do it?” he asked. Schmidt convinced him that they could. Together with a number of other supernova experts in their circle, the twenty-seven-year-old Schmidt began the High-Z Supernova Search and set off on his own road to enlightenment.

  Perlmutter had the advantage of several years of software development. He had also been toughened by his years in the LBL physics group, where people were familiar with fussy, highly technical research projects. But Schmidt had two potent factors acting in his favor: a group of colleagues intimately familiar with hunting for supernovas and a tremendous ability to rise to the challenge. He quickly called his former competitors and organized them into a loose confederation. Then he sat down and hammered away at the same programming problems that had so consumed Perlmutter. Kirshner sounds like a proud parent when he reflects on his disciple's achievements: “Saul's group worked for six years on software, and Brian said, 'I could do that in a month.' And he did.” Perlmutter had laboriously created much of his software from scratch. Schmidt, drawing on his greater familiarity with the astronomical world, swiftly cobbled together existing computer programs into a workable, if less elegant, solution.

  Both teams needed the computer to help them with the insanely complicated task of understanding the supernovas. Silicon processors allowed them to go where Hubble could not. Baade's old headaches still plagued them, however. As difficult as it was to find the Type 1a explosions, it was even harder to understand what they were doing. The scientists wanted to know both how bright each supernova appeared and how bright it truly was, but the cosmos is not a straightforward place. The expansion of the universe reddened the light of each supernova by a slightly different amount. That color change would affect how the CCD detectors measured the light and needed to be accounted for. A scattering of dust between the supernova and us could also change its color and would reduce its apparent brightness; these factors demanded due consideration. Worst of all, preliminary supernova surveys conducted during the late 1980s and early 1990s showed that Type 1a supernovas are not all identical after all. Astronomers' standard candles came in slightly different wattages.

  These problems came to a head in 1991, when astronomers observed two relatively nearby Type 1a supernovas of startlingly different luminosities. The sharp disparity suggested that observers might have to find another way to measure distances. On closer examination, however, the situation was not so grim. Some exploding stars brighten and fade significantly faster than others. Those early surveys uncovered enough Type 1a supernovas in relatively nearby galaxies that a pattern began to emerge:
sluggish supernovas are consistently brighter at their peaks than fleeting ones. In these proximate galaxies, there were other ways to measure distances and evaluate the supernovas. Scientists had no such luxury for remote objects, where supernovas were the only predictable beacons bright enough to be visible. But the correlation between speed and luminosity was so tight that the shape of a supernova's light curve—a plot of its changing brightness over time—could very accurately determine the intrinsic brilliance of the detonated star regardless of its location.

  Enter Adam Riess of the Space Telescope Science Institute, another member of Kirshner's flock. Drawing on this information, he helped devise a statistical technique for eliminating the variations among Type 1a supernovas. A little later, Perlmutter came up with his own, more geometric solution in which he stretched the light curves to correct for the differences. “I drew light curves and they were amazingly close—they all fell on top of each other. It was clear there was some physics making that happen. And the spectra all looked so similar. It looked like we had a tool we could use for distance measurements and the Hubble constant.” Either way, the teams now claim they can figure out the inherent brightness of the explosions to within about 10 percent, an astonishing level of accuracy in the world of cosmology. The supernovas had vindicated themselves. They seem to be almost perfect distance markers. For the first time, scientists had a tape measure large enough for the entire universe—an observational tool that could match the far-reaching spiritual power of Einstein's equations. Now it would be possible to apply the rule of falsification and see what kind of universe God had created. The two teams were off and running.

  To be accurate, it was more like the two teams were off and chasing each other through knee-deep molasses. Hunting supernovas calls for a singular mix of frantic activity and almost limitless patience. It begins in a frenzy of administrative activity, securing time on a large telescope just after a new moon, when the sky is dark, and about three weeks later when moonlight again is not a problem. The four-meter (thirteen-foot-wide) telescope at Cerro Tololo Inter-American Observatory, sitting under ink black skies fifteen thousand feet above the Chilean desert, is equipped with a special detector, the Big Throughput Camera, that can see an especially large swath of sky all at once. This camera can snap an image of five thousand galaxies in just ten minutes. Both Perlmutter's and Schmidt's teams depend on this versatile instrument. Once they have secured two images of the same areas of the sky, they must make sure the two views are properly aligned, then account for changes in atmospheric clarity, then eliminate the many flickering objects that are not supernovas, such as variable stars, quasars, and rogue asteroids. During each observing session, they might look at fifty or one hundred parts of the sky, covering hundreds of thousands of galaxies.

  If a blip of light looks promising, a new round of work begins. The scientists make their pilgrimage to the twin huge Keck telescopes atop Mauna Kea, on the Big Island of Hawaii. Thirty-six perfectly polished and aluminized glass hexagons work in unison to form the Keeks' thirty-three-foot-wide cyclops eyes. Each Keck gathers enough light from the supernova suspect to spread the beam into a spectrum, which contains a wealth of data about the star's composition. That information makes it possible to distinguish Type 1a supernovas from the other varieties of exploding stars. Once the astronomers make a positive ID, the real frenzy begins as they scramble to keep the star under nearly constant surveillance in order to produce a sufficiently accurate light curve. Supernovas don't sleep, and neither do the people who study them. For the Supernova Cosmology Project, Perlmutter juggles observing time on telescopes around the globe: the Cerro Tololo Inter-American Observatory 4-meter telescope, William Herschel telescope, Wisconsin-Indiana-Yale-NOAO telescope, European Southern Observatory 3.6-meter telescope, Nordic Optical 2.5-meter telescope, Keck telescopes, and the Hubble Space Telescope. In all, the scientists need to track each supernova for forty to sixty days to get an accurate read. After that comes more analysis to correct for intergalactic dust and other possible sources of error. Final analysis can take a year or more, until the exploded star has faded from view, when it is possible to get a clean view of the galaxy where it lived and died. Meanwhile, each team felt the other breathing down its neck.

  Perlmutter's Supernova Cosmology Group had established an early lead. Despite an abundance of talent, Schmidt's High-Z Supernova Search got off to a rocky start. Just five months after he got the project going, Schmidt made that move to the Mount Stromlo observatory in Australia, about as far as possible from the colleagues whose efforts needed close coordination. “I had just had a child, I had just written software that had never been used before, and I was attempting to look for supernovas and debug the software across twelve time zones between Chile and Australia. It was nearly a disaster,” he says. Then in 1995, shortly after getting started, Schmidt found his first cosmologically significant supernova, and it was clear his efforts were not in vain. The exploding star was dimmer than expected, suggesting that the universe was up to something odd, but Schmidt was too careful to draw any major conclusions from a single data point. The High-Z team hunkered down to gather more data.

  By the end of 1996, Perlmutter had collected twenty-three distant supernovas and completed analyzing seven of them. These explosions lay four billion to seven billion light-years from the earth, or as much as halfway to the visible edge of the cosmos. Now came the moment of truth. Nobody had ever tested the big bang this way before. All those elaborate theories about dark matter and Lambda were about to meet the sharp sword of observational falsification. Soon, it seemed, humans would know the fate of their universe. If the cosmic expansion were slowing down, as everyone expected, then the light from the very distant supernovas should be a little brighter relative to their redshifts than it would be if there were no deceleration, because their light would have been more severely stretched by the faster expansion in the past. This was, in fact, what his preliminary results seemed to indicate. The Supernova Cosmology Project also had some bad news for fans of Lambda. Those first seven supernovas showed no clear sign of any hidden springiness in space. Lambda was probably small, if it existed at all.

  Within a year, however, the picture reversed. As Perlmutter worked through more of the observations, the average explosion was looking fainter and fainter. In other words, the universe was not slowing down as much as his team had thought, which meant that the amount of matter in the universe had to be quite modest. Thinking back, Perlmutter now brushes aside his original conclusions as the product of rough, preliminary data. Kirshner recalls, however, that Perlmutter stood by those findings until additional observations forced him to change his tune. At any rate, the real answer emerged over the course of 1997. A follower of Hubble's puritanical style of sci/religious faith, Perlmutter believes that cosmic truth will reveal itself if you just pursue it hard enough. “If the data are telling you something you're not happy with, you have to go with the data, but if the data are telling you something surprising, you have to assume there's something wrong with it. We all didn't believe it at first—assumed it would go away. Little by little we all got convinced. We never sat back and changed our minds. We'd just say, 'I don't believe it, do you believe it?' Well, must be this thing wrong,” he explains. But with each new observation, the data were growing more insistent, yet the apparent rate of deceleration—and hence the inferred cosmic density—was looking smaller and smaller.

  “It was getting ridiculously low,” Perlmutter says. He started saying that if the numbers fell any further, he'd have to conclude that the density of the universe is zero. “I guess we're not here,” he joked nervously. Then the numbers fell even more, and Perlmutter found himself looking at minus signs. The supernovas all looked too dim. They were farther away than their redshifts seemed to suggest, evidence that the space between them and us has been expanding at an accelerating rate. In other words, the universe is not slowing down at all. It is speeding up.

  Even a completely empty univers
e could not speed up unless some outward pressure were acting on it. Reluctantly, Perlmutter and his teammates turned to Einstein's old mystical hedge, Lambda, to make sense of the universe. “I told my review committees that we could go after the vacuum energy of the universe [that is, Lambda], which is clearly a fundamental physics question. But I didn't seriously think I was going to find it,” Perlmutter says. Like almost all of his colleagues, he assumed Lambda had vanished after the era of inflation. Wanted or not, the intangible found its way into his research. An accelerating universe means that “empty” space must be full of energy.

  Late in the year, members of the Supernova Cosmology Project started showing their startling results at scientific gatherings. But the real bombshell came on January 8, 1998, when Perlmutter presented his analysis of forty distant supernovas at a press conference at the American Astronomical Society meeting—a high-profile summit of professional astronomers, which is always extremely well attended by the media. The next day, headlines like SCIENTISTS SEE COSMIC GROWTH SPURT popped up in the papers, Saul Perlmutter was a science celebrity at the ripe age of thirty-eight, and cosmologists were another step closer to reconstructing the history and fate of the universe.

  Schmidt's team members had also been making significant progress. During 1997, they picked apart the light from more supernovas and kept finding that, as with the 1995 explosion, the stars were all peculiarly dim. In the autumn, Adam Riess had finished studying the light curves and sent his findings to Schmidt, who was stunned by what he saw. It was obvious to him that the big bang was accelerating. At the time, though, Perlmutter's only public results stated the exact opposite. “I was very concerned. Why were we getting completely different numbers from Saul's?” Schmidt wondered. He kept the findings quiet while he checked and rechecked the results. Little did he know that Perlmutter's latest supernovas were now telling the exact same tale.