The 4 Percent Universe: Dark Matter, Dark Energy, and the Race to Discover the Rest of Reality - Richard Panek (2011)

Part II. Lo and Behold

Chapter 4. Getting in the Game

THE WEIGHT OF the universe. The shape of the universe. The fate of the universe.

They talked about it in those terms. They used this giddy language in proposals to solicit funding. They used it in a brochure to recruit graduate students. They used it with the other members of their collaboration as they all told themselves that they were the ones who were finally going to solve some of the most profound mysteries of cosmology—of civilization itself. They also used this language when they needed to reassure themselves that they weren't rebuilding Babel or emulating Icarus, that their experiment was an exercise not in hubris but in science.

Okay, maybe a little hubris. Saul Perlmutter wasn't a born astronomer. He hadn't collected telescope parts as a child, hadn't sketched the motions of the night sky, hadn't dreamed of solitary vigils on mountaintops, just him and the heavens. Carl Pennypacker wasn't a born astronomer either, though at least his PhD in physics was on a related topic, infrared astronomy. And the other members of their team weren't astronomers. None of them had come to Lawrence Berkeley National Laboratory to do astronomy; astronomy wasn't what LBL usually did. Still, they had reason to think they were in the right place at the right time.

The right place, because LBL and the University of California, Berkeley, had just won a government competition to establish a major new research center. The name on the proposal was the Center for Particle Astrophysics, though because the titular "particle" was dark matter, they could have called it the Dark Matter Center—and, as the first director of the center once said, they probably would have if they'd thought of it.

And the right time, because by the 1980s scientists could proceed under the assumption that they were in possession of the middle and the beginning of the cosmic narrative. They knew that their protagonist—the universe—was expanding. They had a reasonable explanation for how it had gotten to this point in the story—the Big Bang. Now they could ask themselves: What will become of Our Hero?

Did the universe contain enough matter to slow the expansion so much that one day it would stretch as far as it could, stop, and reverse itself, like the trajectory of a tossed ball returning to Earth? In such a universe, space would be finite, curving back on itself, like a globe.

Or did the universe contain so little matter that the expansion would never stop but go on and on, like a rocket leaving Earth's atmosphere? In this kind of universe, space would be infinite, curving away from itself, like a saddle.

Or did the universe contain just enough matter to slow the expansion so that it would eventually come to a virtual halt? In this universe, space would be infinite and flat.

Borrowing from the Big Bang example, astronomers gave the first options the cheerfully inadequate names Big Crunch (too much matter) and Big Chill (too little matter); the third option was the Goldilocks universe (just right). From only one measurement, astronomers could determine the weight, the shape, and the fate of the universe.

Before the 1980s, astronomers had certainly known that the amount of matter in the universe would have an effect on the universe's rate of expansion. What they hadn't known was that they had been missing 90 percent or more of the matter. The possible cosmological implications of this realization had been evident from the start. "Not until we learn the characteristics and the spatial distribution of the dark matter," Vera Rubin had written in Science shortly after the idea gained widespread acceptance, "can we predict whether the universe is of high density, so that the expansion will ultimately be halted and the universe will start to contract, or of low density, and so that the expansion will go on forever."

Now Perlmutter and Pennypacker set out to make that measurement. They recognized that writing the closing chapter in the story of the universe would be challenging in the extreme, and they figured they would be done in, oh, three years.

The question of how the universe will end was as old as civilization, but the difference now was that scientists might be able to go out and make the crucial measurement. Because the discovery of the 3 K temperature had matched a prediction from the Big Bang theory, it had taught astronomers what to think about cosmology: It just might be a science after all. But the 3 K discovery also taught them how to think about cosmology: If you want to understand the history and structure of the universe—if you want to do cosmology—you have to do what Bob Dicke and Jim Peebles had been urging even before the discovery of the cosmic microwave background: think about gravity on the scale of the universe.

Not that astronomers had been altogether ignoring the relationship between gravity and the universe. Much of modern physics and all of modern astronomy had arisen from Newton's epic struggles to derive a law of gravity that was universal. In his Principia, published in 1687, Newton met Plato's challenge to find the calculations on paper that matched the motions in the heavens. The telescope had given astronomers the physical tool to chronicle more and more of those motions. But it was Newton's math that had given them the intellectual tool to make sense of them. The law of universal gravitation was what made cosmology-as-science possible.

Yet it also made cosmology-as-science problematic. A syllogism (of sorts): One, the universe is full of matter; two, matter attracts other matter through gravity; therefore, the universe must be collapsing. So why wasn't it?

This was the question that the cleric Richard Bentley posed to Newton in 1692 while preparing a series of lectures on faith, reason, and the just-published Principia. Newton acknowledged that his argument required "that all the particles in an infinite space should be so accurately poised one among another as to stand still in a perfect equilibrium. For I reckon this as hard as to make not one needle only but an infinite number of them (so many as there are particles in an infinite space) stand accurately poised upon their points." How was such an equilibrium possible? In a later edition of the Principia, Newton appended a General Scholium in which he postulated an answer—the foresight of God: "And so that the systems of the fixed stars will not fall upon one another as a result of their gravity, he has placed them at immense distances from one another."

What made cosmology scientifically suspect for investigators of nature wasn't just this invocation of a supernatural cause—a cause that was, literally, beyond nature. The problem was the effect. Or, more accurately, it was the absence of an effect. Newton's physics was all cause-and-effect, matter-and-motion. Yet what he was proposing in this one instance was a lack of gravitational interaction among the bodies of the cosmos. Having conceived of gravity as action at a distance, Newton was now suggesting the need for inaction at a distance.

Over the following decades and centuries, the more that astronomers discovered about the system of "fixed stars"—that the stars aren't fixed at all but are in motion relative to one another, and that the entire system of unfixed stars, our galaxy, rotates around a common center—the less satisfying was the explanation of inaction at great distances.

Einstein made subtle adjustments to Newton's theory of gravity. And in his 1916 theory of general relativity, he presented calculations on paper that matched the motions in the heavens slightly more accurately than Newton's. Yet he, too, had to account for a universe that, as was evident in "the small velocities of the stars," wasn't collapsing of its own weight. In his 1917 paper "Cosmological Considerations on the General Theory of Relativity," he inserted a fudge factor in his equation—the Greek symbol lambda, "at present unknown"—to represent whatever it was that was keeping the universe from collapsing. Like Newton, he feigned no hypotheses as to what that something might be. It was just ... lambda. But then, little more than a decade later, came Hubble's universe, and with it an elegant and unforeseen solution to the lack-of-collapse conundrum: The reason the universe wasn't collapsing of its own weight was that it was expanding.

Newton hadn't needed God, and Einstein hadn't needed lambda. In 1931 Einstein traveled from Germany to the Mount Wilson Observatory in the mountains northeast of Pasadena and visited Hubble. After reviewing the expansion data for himself, Einstein discarded his fudge factor. In retrospect, physicists of a philosophical bent came to realize, the problem with cosmology hadn't been a supernatural cause (God). And it hadn't been an illogical effect (inaction at a distance). It had been the unthinking assumption behind the syllogism, the premise of the whole cosmology-as-science debate: that the universe was static.

If you took the universe at face value, as even Einstein did, you would have unthinkingly assumed it was, on the whole, unchanging over time. But the universe (yet again) wasn't what it appeared to be. It wasn't static. It was expanding, and that expansion was outracing the effects of gravity—for now, anyway.

But what about over time? A new syllogism presented itself: One, the universe is expanding; two, the universe is full of matter attracting other matter through gravity; therefore, the expansion must be slowing down. The lingering challenge to aspiring cosmologists was no longer Why wasn't the universe collapsing? It was Will it collapse?

Ever since Hubble's discovery of evidence for an expanding universe, astronomers had known how to measure how much the expansion was slowing down, at least in principle. Hubble had used Henrietta Swan Leavitt's period-luminosity relation for Cepheid variable stars to determine distances to nearby galaxies. And he had used the redshifts for those galaxies as equivalent to their velocities as they moved away from us. When he graphed those distances against those velocities, he concluded they were directly proportional to each other: the greater the distance, the greater the velocity. The farther the galaxy, the faster it was receding. This one-to-one relationship showed itself as a straight line on a 45-degree angle. If the universe were expanding at a constant rate, that straight line would continue as far as the telescope could see.

But the universe is full of matter, and matter is tugging gravitationally on other matter, so the expansion can't be uniform. At some far reach of the Hubble relation, the galaxies would have to deviate from the straight line. The graph of their points would begin to gently curve toward brighter luminosity. How much they deviated from the straight line would tell you how much brighter they were at that particular redshift than they would be if the universe were expanding at a constant rate. And how much brighter they were—how much nearer they were—would tell you how much the expansion was slowing.

To make that distinction, astronomers would need to continue to graph distance against velocity. For the velocity axis, they could still use redshift. For distance, however, they had a problem. Cepheid variables are visible only in relatively nearby galaxies. For distant observations astronomers would need another source of light that had a standard luminosity, celestial objects they could plug into Newton's inverse-square law as if they were so many candles at greater and greater distances.

And ever since Hubble's discovery of evidence for an expanding universe, astronomers had also known about the standard-candle candidate that the Berkeley team would decide to pursue. In 1932, the British physicist James Chadwick discovered the neutron, an uncharged particle that complements the positively charged proton and the negatively charged electron. In 1934, the same Caltech astrophysicist who had recently suggested that galaxy clusters might be full of dark matter, Fritz Zwicky, collaborated with the Mount Wilson astronomer Walter Baade on a calculation showing that, under certain conditions, the core of a star could undergo a chain of nuclear reactions and collapse. The implosion would race inward at 40,000 miles per second, creating an enormous shock wave and blowing off the outer layers of the star. Baade and Zwicky found that the surviving ultracompact core of the star would consist of Chadwick's neutrons, weighing 6 million tons to the cubic inch and measuring no more than 60 miles in diameter.

Astronomers had already identified a class of stars that suddenly flared brighter, then dimmed, and they had named this phenomenon "nova," for "new star" (because its sudden brightening might suggest it was "new" to us). Baade and Zwicky decided that their exploding star deserved a classification all its own: "super-nova."

Almost at once, Zwicky initiated a search for supernovae. He helped design an 18-inch telescope that became the first astronomical instrument in use on Mount Palomar, and soon newspapers and magazines across the country were keeping a running tab of how many "star suicides" his survey had discovered. Baade, meanwhile, suggested that because supernovae seemed to arise from the same class of objects, they might serve as standard candles, though he cautioned in a 1938 paper that "probably a number of years must elapse before better data will be at our disposal."

The wait turned out to be half a century. In 1988, the National Science Foundation awarded the University of California, Berkeley, six million dollars over five years to establish the Center for Particle Astrophysics. The center would take multiple approaches to the mystery of dark matter. One was to try to detect particles of dark matter in the laboratory. Another looked for signs of dark matter in the cosmic microwave background. A third approach explored dark matter through theory. And another group would try to determine how much matter was out there, dark or otherwise, by using supernovae as standard candles.

The skepticism was sure to be high: physicists doing astronomy? Pennypacker and Perlmutter knew they would eventually have to convince the astronomy community, as insular and guarded as any in science, that physicists working at a particle physics institution could be capable in their line of work. But first they were going to have to convince themselves.

Lawrence Berkeley National Laboratory had invented accelerator particle physics as the world had come to know it. In the late 1920s the physicist who would become the lab's namesake, Ernest Lawrence, conceived of an accelerator that shot particles not in straight lines, as linear accelerators did, but in circles. Strategically placed magnets would deflect the particles just enough to prod them to follow the closed curve, around and around, faster and faster, to higher and higher levels of energy. Lawrence's first "proton merry-go-round"—or cyclotron—was five inches in diameter, small enough to fit inside any room bigger than a broom closet in the physics building on campus. In 1931 he'd moved his operations into an abandoned building, the former Civil Engineering Testing Laboratory—the first official site of the Berkeley Lab "Radiation Laboratory." By 1940, a version of the cyclotron had reached a diameter of 184 inches, and the experiment had outgrown the Rad Lab. Lawrence secured from the university a promontory above the campus. But his legacy wasn't just the complex of buildings that over the decades would come to line Cyclotron Road. It was all the particle accelerators in various places around the world that circle underground, miles-long snakes devouring their tails.

What fun was that? You could be a cog in the biggest wheel in the history of the planet, you could even be the most important cog, but you'd still be a cog—assuming you lived long enough for the next generation of accelerators to come online. Luis Alvarez had been a key contributor to the construction of the Bevatron, a proton accelerator with a 400-foot circumference that opened at LBL in 1956. But his was not the kind of mind that aspired to cogdom a couple of decades hence. On his own he used principles of physics in a frame-by-frame analysis of the Zapruder film of the assassination of President Kennedy to see whether the "one bullet, one gunman" hypothesis was tenable. (It was.) He probed an Egyptian pyramid with cosmic rays to learn whether it held secret passages. (It didn't.) He wanted to mount a "High Altitude Particle Physics Experiment," but the LBL director at the time refused to give him access to lab funds. Not because of what the experiment would be doing—particle physics was what LBL did. Rather, the problem was how HAPPE would be doing it: aboard a balloon. "We are an accelerator lab," the lab director Edwin McMillan, a Nobel laureate in Physics, told Alvarez. "If we stop doing accelerator physics, our funding will disappear."

For Alvarez, turning your back on a particle physics experiment because it went up in the air instead of around and around betrayed a lack of imagination. Alvarez quit the leadership of his own LBL physics group in protest and got funding elsewhere, including from NASA, to see if HAPPE could fly. (It crashed.)

By the mid-1970s, Alvarez had his own Nobel Prize in Physics, and the lab had a new director. One day Alvarez was reading a magazine article about an experiment being done by an acquaintance of his. Stirling Colgate, toothpaste scion by birth and thermonuclear physicist by choice, had planted a 30-lnch telescope on a surplus Nike missile turret in the New Mexico desert. The plan was to program it to automatically look at a different galaxy every three to ten seconds. The telescope would then transmit the information through a microwave link to the memory of an IBM computer on the campus of the New Mexico Institute of Mining and Technology, eighteen miles away. There, software would search the images for supernovae.

Automated astronomy wasn't entirely new. The era of the lone observer standing on a mountaintop, eye to eyepiece, staring into the abyss, was coming to an end. Since the invention of the telescope in the first decade of the seventeenth century, astronomers and telescope operators had guided their instruments by hand. Now astronomers were writing computer programs that manipulated the motions of the telescope—and did so with far greater sensitivity than the human hand, and eventually without the need for a human presence on site at all. The University of Wisconsin had an automated telescope. So did Michigan State University, and MIT too.

Supernova searches weren't new, either. They had been a staple of astronomy since Zwicky's initial observing program in the 1930s. The 18-inch telescope on Mount Palomar that he helped design was still in use for that purpose, as was a nearby 48-inch, along with telescopes in Italy and Hungary.

What distinguished Colgate's project was the combination of the two ideas—automated telescope, supernova search. An astronomer looking for a supernova has to compare images of the same galaxy several weeks apart to see whether a dot of light has emerged. Traditional supernova searches required developing photographic plates by hand, then comparing them to previous exposures by eye, both time-consuming processes. In Colgate's system, a television-type sensor would replace photographic plates, and a computer would compare images almost instantly. In the end, his experiment didn't work. His hardware was fine; he had been able to make a telescope that could point where and when he wanted it to point. The problem was the software. The computer code required the equivalent of a hundred thousand FORTRAN statements, and Colgate was working more or less alone. Still, Alvarez saw, the potential was there. No: The inevitability was there—the era of the remote supernova survey.

Alvarez looked up from the magazine. "Stirling has been working on this project," he said, handing the article to a postdoc and former graduate advisee of his, Richard Muller. "And I think he's abandoning it. Talk to him. See what's going on."

By the time Perlmutter arrived at LBL in 1981 as a twenty-one-year-old graduate student, the stories of Muller's battles with the bureaucracy were legion. Year after year, LBL leadership would tell Muller that this was the last time he'd be getting funding, because if the Department of Energy was going to cut anything from the LBL budget, it was going to be the speculative astrophysics project, and then that money would be gone from the LBL coffers for good. But Muller had read Jim Peebles's book Physical Cosmology, and he understood that the astronomy of the very big and the physics of the very small were becoming closer, even indistinguishable. It was the era of the cosmic microwave background. Of quasars, those mysterious sources of extremely high energy from the depths of the distant universe. Of pulsars, which provided evidence not only that Zwicky and Baade had been right all those decades earlier about the existence of neutron stars, but that these stars were spinning at a rate of hundreds of times a second. You couldn't study any of these phenomena without thinking about high-energy physics. Astronomy might not be the kind of high-energy physics that the lab had ever pursued, but it was quickly becoming the kind that the lab had always done. From the point of view of a Luis Alvarez or a Richard Muller, they weren't drifting toward a new discipline. The discipline was drifting toward them.

If anything, Perlmutter thought of himself as a born physicist—someone who wanted to know how the world worked on the most fundamental level, to discover the laws that united all of nature. Science courses had always been the easy part of his education; he almost hadn't needed to think about them. He studied science, enjoyed it ... and then had plenty of time to indulge other interests. And what he thought about, when he wasn't thinking about the laws that united nature, were the "languages" that united humanity—literature, math, music. So maybe he was a born philosopher, too. His parents were both professors—his father taught chemical engineering, his mother social work—who had elected to raise their family in an ethnically mixed neighborhood in Philadelphia, and he grew up listening to his parents and their friends talking about social issues and the latest books and films. He began practicing violin in the second grade. He joined the chorus. In high school he challenged himself to learn how to think like a writer—to learn the nature of narrative. And when he arrived at Harvard College in 1977, he assumed he would double-major in physics and philosophy.

The "physics stuff," he soon realized, was starting to get hard. College physics bore little relation to high school physics. He concluded that if he pursued both physics and philosophy, he would have no time for other courses, let alone a social life. He would have to make a choice.

If he chose philosophy, he couldn't do physics. But if he chose physics, he would still be doing philosophy, because in order to do physics you had to ask the big questions. Before science was science (the study of nature through close observation), it was philosophy (the study of nature through deep thought). Even as science had accumulated all manner of empirical scaffolding over the past few centuries, the guiding impulse of the scientist had remained constant: What is our relationship to the natural world? When Perlmutter joined Richard Muller's physics group at LBL in 1982 and had to choose among the eclectic programs Muller was directing that were now acceptable to the institution—using planes to sample carbon in the atmosphere, measuring the gravitational deflection of starlight by Jupiter—he selected the nascent supernova survey because it seemed the kind of project that might lead to the biggest questions of all. Just as you were automatically doing philosophy if you were doing physics, you were automatically doing physics if you were doing astronomy. Instead of the nature of narrative, Perlmutter would be exploring the narrative of nature.

Alvarez had handed the idea of an automatic supernova survey to Muller on a whim, and now Muller handed that whim to his own postdoc, Carlton Pennypacker. By 1984, the Berkeley Automatic Supernova Search (BASS) team—Muller, Pennypacker, Perlmutter, and a few other graduate students—was operating on the 30-inch telescope at the university's Leuschner Observatory, in the hills a half-hour drive northeast of campus. Further technological advances that were particularly useful for supernova hunting were coming along all the time. When astronomers using photographic plates hunted for supernovae by eye, they used an optical device called a comparator. By rapidly switching back and forth between two images of a galaxy taken several weeks apart, the comparator would allow an astronomer to see whether any new pinpoint of light had appeared in the Interim. The comparator blinked the two images. New computer technology, however, allowed astronomers to take all the light from the earlier image and remove it from the later one. It subtracted the first image from the second. If the computer signaled that a telltale bit of light remained, then a real live human being analyzed the data. Sometimes the source of the bit of light was "local"—a fluctuation in the output, a cosmic ray from space striking the instrument, a subtraction error. Sometimes the source of the light was "astronomical"—asteroids, comets, variable stars. But once in a while the blip of light was a star erupting in a farewell explosion that stood out even against the background of all the tens or hundreds of billions of other stars in its host galaxy combined: that is, a supernova.

Or a Nemesis. In 1980, Luis Alvarez, along with his son Walter Alvarez, had hypothesized that the mass extinction of the dinosaurs 65 million years ago, at the cusp of the Cretaceous and Tertiary periods, had been caused by a comet or asteroid impact that had disrupted the global ecosystem. Then, in 1983, a pair of paleontologists announced that they had discovered evidence of a cycle of mass species extinctions every 26 million years. The following year, Muller and some colleagues published a paper speculating on the existence of a companion star to the Sun—Nemesis. Every 26 million years, they wrote, the highly elliptical orbit of Nemesis would bring it relatively near the Sun, and its gravitational influence would draw comets from the farthest reaches of the solar system into the orbital paths of the planets nearest the Sun, including Earth.

The idea wasn't as fanciful as it might seem; studies of Sun-like stars had shown that about 84 percent were in binary systems, meaning that the Sun, if solo, would be an anomaly. Muller assigned Perlmutter the task of looking for Nemesis (or the Death Star, as the media called it), and in 1986 Perlmutter completed his thesis, "An Astrometric Search for a Stellar Companion to the Sun." But the two projects shared a telescope and some other hardware, some of which Perlmutter helped design, as well as the search software, much of which he wrote, and when he was invited to stay at LBL as a postdoc, Perlmutter looked to supernovae for inspiration. A few months earlier, on May 17, 1986, BASS had bagged its first supernova, and that was good enough for him.

In 1981 the team had predicted a detection rate of one hundred supernovae per year, but scientific proposals were often overoptimistic. Besides, BASS supernovae were relatively nearby; they weren't going to be immediately useful for the big questions. Any changes in the universe's rate of expansion wouldn't be discernible unless astronomers could find standard candles in galaxies significantly farther than Hubble's sample, the deeper the better. How much those supernovae—and therefore their host galaxies—deviated from the straight-line Hubble diagram would tell astronomers the rate of deceleration. Thanks to BASS, Pennypacker and Perlmutter now knew they could do an automated supernova search; they had added two more supernovae in 1986 and another in 1987. But could they do an automated supernova search at cosmologically significant distances?

Muller himself thought such a project might be premature. But he was also a scientist who for years had entertained Luis Alvarez's imaginative flights and was willing to risk his own scientific reputation on a search for a Death Star. He gave his consent, and Pennypacker applied for funding for a camera he wanted to mount on a telescope in Australia. Or, rather, Pennypacker commissioned the camera, then applied for the funding. But the heavens opened over Berkeley and the NSF showered millions of dollars on the Center for Particle Astrophysics, and even though the names on the supernova proposal were Richard A. Muller and Carlton Pennypacker, the search for the fate of the universe was, from the start, Pennypacker's and Perlmutter's.

Supernovae remained attractive as potential standard candles for a couple of reasons. They're bright enough to be visible from the farthest recesses of space, meaning that astronomers can use them to probe deep into the history of the universe. And they operate within human time frames, their luminosity rising and falling over the course of weeks, meaning that, unlike most astronomical phenomena (such as the formation of a solar system or the coalescing of galaxies into a cluster), supernovae offer a soap opera that astronomers can actually watch.

But supernovae were also problematic for at least three reasons. As the LBL group put it, "they are rare, they are rapid, and they are random." In our own Milky Way galaxy, supernovae pop off at an average rate of maybe once a century. So astronomers pursuing supernovae have to devise a way to look at a great number of galaxies, whether individually in quick succession or in great gulps of the sky all at once—or, ideally, both: great gulps in quick succession. Supernovae also require a fast response. Once astronomers identify a supernova, they have to move quickly to do the necessary follow-up studies—not always possible when time on telescopes is assigned months in advance. And they're random. You never know where or when one's going to go off, so even if you could reserve follow-up time on telescopes months in advance, you wouldn't know whether you would have a supernova worth studying on that date.

Perlmutter and Pennypacker were already testing new subtraction software on the 60-inch telescope at Mount Palomar when they first heard that they weren't alone. The idea of chasing supernovae for clues about cosmology was half a century old, and now the widgets were out there to make that chase a reality, so they weren't entirely surprised that another group had been using distant supernovae to try to determine how much the expansion of the universe was changing over time.

Trying, and failing. From 1986 to 1988, three Danish astronomers, with the help of two British astronomers, took turns making a monthly trek to a 1.5-meter telescope at the European Southern Observatory in La Silla, Chile. They calculated that if they looked not at one galaxy at a time but at clusters of galaxies, they could beat the once-a-century-per-galaxy odds of finding the right kind of supernova. They selected clusters with well-established distances. And they timed their searches carefully, choosing the nights just before and after a new moon so that they were able not only to capitalize on dark skies but to compare images about twenty days apart, a period that, through happy coincidence, corresponds to the natural life (or, more aptly, death) cycle of the kind of supernova they wanted.

They found one. Possibly a second, on what would be their final night of observing, though they didn't bother to follow up that detection. They were ahead of their time, and, after two years, their time was up. Their telescope turned out to be too small for their purposes, their rate of discovery too low. Unless they could garner access to a larger telescope and a more powerful detector, they would need dozens of years to collect a suitable sample. Even at the rate of one good supernova a year, they would still need ten years, minimum, to complete their program.

For the members of the Berkeley supernova team, the news of this failure raised a potentially fatal question: How could they reassure the review committees at the Center for Particle Astrophysics that their group could succeed where the Danish collaboration had failed?

First, the team stressed that the Danes had succeeded in finding a distant supernova—so distant that it broke the redshift record for a supernova, 0.31, meaning that it exploded about a quarter of the way back to the beginning of the universe (or 3.5 billion years ago). Second, the Berkeley team would be using better instrumentation. The Anglo-Australian Telescope in Siding Spring, Australia, had a 3.9-meter mirror, more than double the diameter—or four times the aperture, the light-collecting area—of the one the Danish team had used in Chile. And LBL was commissioning a much larger camera.

For insurance, Pennypacker invited Gerson Goldhaber onto the project. In 1933, when Goldhaber was nine, his family fled Germany. He lived in Cairo and then Jerusalem before emigrating to the United States to attend graduate school at the University of Washington. Goldhaber had worked at LBL since 1953, playing key roles on the Bevatron and then, for the past twenty years, collaborating with the Stanford Linear Accelerator Center. He had made important discoveries. He had guided teams that won the Nobel in Physics. As Pennypacker reasoned, "They would never shut down anything he did."

Problems began even before they could start observing. The contractor constructing the camera delivered a mirror that didn't fall within "tolerances," as opticians call the allowable imperfections. The second cut was spoiled when cleaning fluid spilled on the mirror. Finally, the third cut of the mirror worked.

Pennypacker, however, had ordered a camera without a filter, figuring that the more light he got, the better—and not understanding that if you want to compare the brightness of an object on different days, you need to observe in different filters in order to "equalize" the light level. After the crack technical crew at LBL designed an after-the-fact filter wheel, Pennypacker handed it to a graduate student, sent her to Australia, and provided her with the number to give to customs. When she got to customs, she went up to the two clerks and said, "I have this number." They looked at each other, then back at her. "Number for what?" one of them said. (A few days later the director of the observatory in Australia managed to convince the authorities to unconfiscate the wheel.)

Even when the team did perform supernova searches, the subsequent logistics were daunting. The on-site computers didn't have sufficient bandwidth for the Berkeley project's purposes, so team members had to take the computer data to the Sydney airport, fill out volumes of paperwork, and put the cargo on the next plane to San Francisco. There, someone else had to fill out more paperwork before claiming the package and driving it across the bay to Berkeley. Total travel time: forty-eight hours. Then the physicists at Berkeley needed another two days to search the images for supernova candidates, and then another day to study finding charts—maps that show all the known objects in a section of the sky—to see if they really were supernovae. Five days is a long time to wait if you want to schedule follow-up observations of an object that is fast disappearing. Before long, however, they managed to figure out a way to get their data to Berkeley without air travel: A team member at LBL would call up NASA, specify when the supernova search would need to be transmitting data from Australia, and ask if someone at NASA would please turn on the Internet then.

Not that the quality of the data or a delay in transmission mattered. In the course of two and a half years, Pennypacker and Perlmutter secured a dozen nights on the Anglo-Australian Telescope; of those, nine and a half were cloudy or had poor atmospheric conditions, or were needed for testing. And although they did manage to identify six candidates, the final count of actual supernovae was worse than the Danes': zero.

The Berkeley team had come up with an ambitious three-year plan, and they'd executed it. They had stretched the existing technology as far as they could under the circumstances, tweaking each widget until it had nothing left to give, and their efforts simply weren't enough.

Every few months the supernova search had to justify its existence as part of the Center for Particle Astrophysics to an internal Program Advisory Committee. Every few months it also had to justify its existence to an External Advisory Board. That justification now took the form of the team's ability to secure time on the 2.5-meter Isaac Newton Telescope, in the Canary Islands, off the northwest coast of Africa. The telescope was slightly smaller than the one in Siding Spring, but the camera would be bigger, and the weather would be better. Muller himself made the pitch to Bernard Sadoulet, the director of the center: "Look, two or three years from now, we will be delivering supernovas. We will be making real measurements. We will have results. I guarantee that. By the time the initial funding for the Center for Particle Astrophysics runs out, we will have something real to show. And you must understand that. You must know that."

The supernova search got its reprieve. If you were a veteran of the project, you could consider yourself on probation yet one more time. Newcomers, however, wondered what they'd gotten themselves into. Group meetings were held in an office where there weren't enough chairs. You might find yourself sitting at a computer, quietly typing away, when a higher-up would tell you that the project had exceeded its computer allotment and to either shut down the computer now or someone would be along shortly to pull the plug. One graduate student read the recruitment brochure, liked the idea of "weighing the universe," and committed to the project—only to learn that the search had yet to produce one supernova. A postdoc arrived for his first day on the job to find a note on his desk from Perlmutter, saying that he'd gone to the Canary Islands and asking the postdoc to use a finding chart to choose fields for Perlmutter to target. The postdoc stared at the note. He had trained as a particle physicist; he didn't know what a finding chart was.

And then Pennypacker—his words—"blew up the budget." He had developed a habit of spending money he hoped would materialize in the next round of budgets. This time, however, he misread a ledger, spent money he didn't have, and then spent it again.

By his own admission, Pennypacker wasn't leadership material, at least not of the kind required to run an unorthodox and high-risk project. People loved collaborating with him; he was enthusiastic and affable and smart and visionary. The ability to translate those virtues into the words that a review committee or an administrator wanted to hear, however, eluded him. So did a fundamental understanding of administrative details. If the supernova project were to continue, he was made to understand, it would have to do so under different leadership.

Robert Cahn, the new director of the Physics Division at LBL, first approached the senior researcher on the project, Gerson Goldhaber. But for Goldhaber the chance to work on supernovae had represented a freedom from the kind of responsibilities he'd held for four decades at behemoth particle accelerators. Muller had moved on. The next choice was Saul Perlmutter. Cahn consulted with Muller: Was the kid ready? Muller thought maybe so. Twice, Muller said, he'd had the experience of going to Perlmutter with what he thought was a conceptual breakthrough, and Saul had said, "Ah! That's a very interesting idea," then pulled out a notebook and flipped to the page where he'd already seen the idea through and found that it didn't work.

In March 1992, Perlmutter went to the Isaac Newton Telescope to take the first set of images. In late April and early May, Pennypacker went to the INT to make follow-up observations of the same fields while Perlmutter stayed in Berkeley and waited for data to arrive via the BITNET. As the sun was setting over the Atlantic—midafternoon in Berkeley—Perlmutter and a couple of team members would settle into the seats before the high-quality image display in the deliberately overcooled computer center in the basement at LBL, bundled in sweaters and jackets, and start to sort through the software. By ten or eleven in the evening, Perlmutter was alone and the images would begin to emerge on his screen. Each image held hundreds of galaxies; by the end of the night he would collect dozens of images. He printed out each one, just in case. Sometimes the computer told him that a blip of light had appeared that hadn't been there the previous month, and he would bend close to the screen and try to figure out what was wrong. The view of the wide-field camera distorted the geometry, so he didn't trust blips near the edge of the frame. Sometimes a blip would be too near the center of a galaxy, meaning that its light would be impossible to distinguish from the background light during follow-up observations. Sometimes the blip turned out to be an asteroid. One night he found a blip that he couldn't discount, and he had to wonder what obvious error he was overlooking, but he couldn't think of one, so he asked himself what subtle error he was overlooking, but he couldn't think of one, and then he wondered what he was doing wrong, when suddenly he realized, "Wait—this is what we're supposed to be looking for": the potential supernova you can't throw away.

For corroboration he had to wait until his collaborators showed up in the morning, and even then it wasn't a moment for celebration, partly because they couldn't be absolutely sure, and partly because the work had just begun. The data was worthless without follow-up observations that would tell them whether it was indeed a supernova, and if so, at what redshift.

Some of those observations they could do on their own, in the days that followed, while they still had time on the INT. Others required them to find out which astronomers were observing on the big telescopes around the world, figure out whether anyone in the LBL operation might be friends with them, then phone them in the middle of an observing run they had been planning for six months or a year, to plead with them to drop everything and point their telescope somewhere else. In this regard, Perlmutter was singularly talented. Nobody worked the middle-of-the-night phone calls to astronomers on other continents like he did. He was persistent, and he was persuasive, and he was impervious to rejection or insult. He literally wouldn't take no for an answer. Sometimes the plea elicited a laugh, sometimes an outburst of anger. But sometimes the plea elicited data—just enough to tell them that the blip was still there, and fading. They had a supernova.

Still no cause for celebration. Again, the data was meaningless for cosmology unless they knew how distant the supernova was—its redshift. For that, they would need a spectroscopic analysis. Twelve times, at four observatories around the world, astronomers agreed to make the follow-up observations. Eleven times the weather didn't cooperate. The twelfth, the instrument malfunctioned.

As spring stretched into summer, Pennypacker began to think of his team as characters in The Treasure of the Sierra Madre: fortune hunters wandering the desert in search of gold. And they find it—the prospectors discover their vein; the astronomers detect their supernova. And then the gold dust slips through their fingers and blows away in the wind. Walter Huston or Humphrey Bogart or Tim Holt says Thanks anyway to a pal in a faraway observatory, then slowly hangs up the phone.

One night in late August, Pennypacker and Perlmutter called Richard Ellis, a friend to the team as well as a British veteran of the Danish observations in the late 1980s. Ellis snapped at them. Didn't they know that observing conditions in the Canary Islands had been bad lately, and that he and the other observers were already inundated with requests for make-up observations from astronomers who had actually had time on the telescope—unlike the Berkeley team?

Then he went and made the observation. On August 29, 1992, Ellis took out his finding chart and, working on the 4.2-meter William Herschel Telescope, he and a postdoc took two half-hour spectra. When they were done, Ellis got Pennypacker on the phone in Berkeley.

The old record redshift, the one set by the Danish team, had been 0.31, corresponding to roughly 3.5 billion years ago. The new record redshift was 0.458, or 4.7 billion years ago.

Pennypacker let out a whoop. Six years after he and Perlmutter had discussed collaborating on a search for cosmologically significant supernovae, they hadn't found the weight or the shape or the fate of the universe.

But they were in the game.