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 5. Staying in the Game

IN EARLY 1994, a couple of astronomers got to talking. Brian Schmidt had just completed a doctoral thesis on supernovae at Harvard's Center for Astrophysics, and he was thinking about ideas for his next project as a postdoc. Nicholas Suntzeff had been an astronomer at the Cerro Tololo Inter-American Observatory in Chile since 1986, and he had been working on a supernova survey since 1989. As supernova specialists they had both been following the efforts of Berkeley's supernova project. Now, as they sat in the air-conditioned computer room at the observatory headquarters in the Chilean coastal town of La Serena, Schmidt mentioned that he'd been thinking about putting together a team to go up against LBL's.

Suntzeff didn't hesitate: "Can I be part of that?"

Now that, Schmidt thought, is the mark of a good problem in science. It's not when people say, "Oh, that's interesting." It's when they say, " Ooo, can I be part of that?"

Schmidt had to give Saul Perlmutter and the Berkeley team credit. They had seen that, thanks to advances in technology, supernovae might finally be used to do cosmology, and they were succeeding against enormous odds. They had been in the right place at the right time. But were they the right team?

Like many other astronomers, Schmidt had been skeptical that physicists—even physicists-turned-astrophysicists—would be able to consistently find distant supernovae. But even after the LBL team had found its first supernova, Schmidt and other astronomers remained skeptical that physicists-turned-astrophysicists—no matter how brilliant—could perform the kinds of follow-up observations and analyses that routinely strained even their own hard-won expertise. Seemingly everybody in the supernova game had been on the receiving end of a middle-of-the-night phone call from Saul, asking them to drop everything and perform a follow-up observation of a supernova candidate. Perlmutter had gotten a reputation in the community for being preternaturally persistent. But in Suntzeff's experience, every time he slewed his telescope to Perlmutter's target, the field was empty. "Must be too faint," Suntzeff would say diplomatically.

Schmidt and Suntzeff grabbed the nearest blue-and-gray sheet of IBM computer printout, flipped it over, and began scribbling. They continued the conversation in Suntzeff's office later that day and the next day as well, laying out their plan of attack.

Suntzeff, they decided, would be in charge of the observing. He would find the supernova candidates and do the follow-up measurements. Schmidt would be in charge of the analysis. He would take some existing software and create a new code that would clean the images, do the subtraction, and isolate the supernovae.

Suntzeff turned to Schmidt. "How long will it take you to write the new code?"

Schmidt could carry himself as a cocky young astronomer, but his wiseacre's side-of-the-mouth way of talking suggested not so much arrogance as irony. Suntzeff preferred to think of him as constitutionally optimistic. Yet even Schmidt had to hesitate. Then he reminded himself: Saul's doing it.

"Two months," he answered.

On his return to Harvard, Schmidt disappeared into his office for hours at a time, day after day, week after week, writing the code. But he also circulated through the halls, stopping colleagues and dropping into offices, letting a select group know that he and Nick Suntzeff were putting together a team to catch Saul. In each case he got the same response, expressed with the same level of eagerness: Can I be part of that?

Robert Kirshner wouldn't even have to ask. He had been a student of supernovae since 1970, longer than some of his students had been alive. At forty-four, he was now an elder statesman in astronomy, the chairman of the Astronomy Department at Harvard. He had long experience getting money out of the National Science Foundation, reserving time on the world's best telescopes, and helping to set policy for the Space Telescope Science Institute, the science and operations center for the Hubble Space Telescope. He was one of the world's foremost supernova experts, as well as the mentor to several generations of supernova experts—the graduate students he had recruited and the postdocs he had hired for his private duchy within the Harvard-Smithsonian Center for Astrophysics, half a mile up Garden Street from Harvard Square. When Nature received the Danish group's preliminary results in 1988, it was Kirshner whom the journal asked to privately review the paper and then publicly write an accompanying news analysis. When Berkeley's Center for Particle Astrophysics convened an External Advisory Board and needed a supernova guru, it was Kirshner who got the call. When Perlmutter et al. submitted a paper to the Astrophysical Journal Letters analyzing their 1992 supernova, it was Kirshner whom the editors asked to serve as referee.

To all his peer evaluations Kirshner brought a deep skepticism, born of his own decades of experience, about the ability of anyone to perform the near-surgical task of supernova analysis. Kirshner could be amusing; in casual conversation he often made exaggerated facial expressions, adopted accents, whinnied at his own jokes. His talks at conferences were reliably witty and well-attended. But when it came to supernovae, and to what you needed to know to do supernova astronomy, Kirshner could be exacting, even bruising.

But he had a point—several points, actually. If you wanted to do supernovae, you had to know spectroscopy—the analysis of an astronomical object's spectrum of light that identified its chemical composition as well as its motion toward or away from you. You had to know photometry—the tedious, difficult determination of an object's brightness. You had to account for dust, either within the supernova's host galaxy or somewhere along the line of sight between the supernova and the observer. Sometimes dust was there, sometimes not. If it was there, it would dim or redden the light from the supernova. And if you didn't know the extent to which dust was polluting the light, you wouldn't know how much to trust your data.

For the Berkeley supernova group, however, Kirshner reserved a special level of skepticism. As far as he was concerned, they were doing poor work that was giving his area of expertise a bad name.

From the start, Kirshner had his doubts about a bunch of particle physicists trying to do astronomy, adopting it as if it were a hobby rather than a science you needed to perfect over a lifetime. So far he'd seen nothing to ease those concerns. In the 1980s, Richard Muller had diverted time from the supernova survey at the Leuschner Observatory to pursue his Nemesis project. The discovery of a companion star to the Sun, if he made it, would be momentous, but it was so unlikely that the effort seemed almost a capricious use of precious telescope time. In 1989, Muller, Pennypacker, and Perlmutter got the attention of astronomers around the world by concluding that the famous supernova 1987 A—the first naked-eye supernova in four hundred years—had left behind a pulsar, a neutron star spinning hundreds of times per second. The "evidence" turned out to be an instrument error. And then came the embarrassment that Kirshner got to witness for himself, as a member of the External Advisory Board: a three-year attempt to find distant supernovae at the Anglo-Australian Telescope that had come up empty.

Hundreds of thousands of dollars: money enough to fund dozens of more modest and more practical astronomical projects: empty.

It wouldn't be entirely fair to say that particle physics operates according to the principle Get funding first, ask questions later. But it wouldn't quite be inaccurate, either. Projects in particle physics routinely involve dozens, hundreds, even thousands of participants, and require machines that manufacture ultra-energetic pyrotechnics that the universe hasn't seen since its megacompact first fraction of a second of existence. The Berkeley supernova search wasn't operating on that scale, but other projects at LBL were, and the lab itself had long been the world's foremost proponent of that work ethic. Particle physicists can somewhat afford to bulldoze ahead, confident that between their billion-pound hardware and their collective brainpower they'll find the answer to any question they might ask. And the first question that the LBL team had asked was: Can we find distant supernovae?

It was, Kirshner thought, the wrong question to ask first. The right one was whether distant supernovae were worth finding. Could they really serve as standard candles?

The recent history of astronomy held a couple of cautionary tales for the standard-candle-bearer. Having discovered evidence for the expansion of the universe, Edwin Hubble spent much of the last twenty years of his life working under the assumption that galaxies might be standard candles, even though they weren't entirely uniform. Maybe they were similar enough that he could use them to discern the universe's shape and fate. Walter Baade, one of his Mount Wilson colleagues as well as Fritz Zwicky's collaborator on the 1934 "super-nova" paper, argued that Hubble had it backward: "You must understand the galaxies before you can get the geometry right." Allan Sandage, Hubble's protégé and, upon his death in 1953, his successor at the Mount Wilson and Palomar Observatories, would later write, "Hubble clearly understood this, but rather than be stopped because this part of his subject was 30 years before its time, he pushed ahead with an abandonment known to pioneers in any milieu who try to reach Everest without proper equipment."

Then it was Sandage's turn. For a quarter of a century, he and the Swiss astronomer Gustav A. Tammann pursued an alternate candidate for standard candles. If galaxies themselves weren't uniform enough, then maybe clusters of galaxies were—or, more precisely, the brightest galaxy within each cluster. But this proposition, too, suffered from an insufficient understanding of galaxy mechanics. Some galaxies would grow dimmer with age, as their stars died out, while other galaxies would grow brighter with age, as they merged with smaller galaxies. Unable to reliably tell the difference, and wary of other factors they couldn't begin to guess, Sandage and Tammann turned back from the summit. "Essentially," Sandage announced to his colleagues in cosmology in 1984, at a conference on the expansion rate of the universe, "we have failed."

Kirshner never passed up an opportunity to point out that the same fundamental lack of understanding of underlying processes could easily sabotage the usefulness of supernovae for cosmic measurements. Already astronomers had determined that supernovae belong to two classes—and possibly more.

One class was the kind that Zwicky and Baade had prophesied—one that results in the birth of a neutron star. It was the kind Zwicky assumed he was finding in his 1930s survey of "star suicides." In 1940, however, Rudolph Minkowski at Mount Wilson took a spectroscope of a supernova that was different from the spectroscopic analyses of Zwicky's supernovae. Minkowski's supernova showed the presence of hydrogen. Zwicky's supernovae did not. They were clearly different types of supernovae.

Since then astronomers had come to think that one type of supernova—the type that Zwicky and Baade had predicted in 1934, that Zwicky thought he was observing in 1936 and 1937, and that Minkowski did observe in 1940—was the result of a chain of nuclear processes in a star several times the mass of the Sun, leading to a 40,000-miles-per-second implosion.

The other type—the type that Zwicky observed—begins life as a hydrogen-rich star like our own Sun. As it ages, the Sun will shed its outer hydrogen layer while its core contracts under gravitational pressure. In the end, only the core will remain—a shrunken skull called a white dwarf, with the mass of the Sun packed into the volume of Earth. If a white dwarf had a companion star (and most stars in our galaxy do), then at this point it might start to siphon gas off the other star. In the 1930s, the Indian mathematician Subrahmanyan Chandrasekhar calculated that when a star of this kind reaches a certain size—1.4 times the mass of the Sun, or the Chandrasekhar limit—it will begin to collapse of its own weight. The gravitational pressure will destabilize its chemical composition, leading to a thermonuclear explosion.

Through a telescope on Earth, the two types would look the same, even though one is an implosion and the other is an explosion. But a spectroscope would show the difference—hydrogen or no hydrogen, Type II or Type I. For astronomers, the uniformity of Type I supernovae offered the possibility that this type might be a standard candle. Since these supernovae all began as a single kind of star, a white dwarf, that had reached a uniform mass, the Chandrasekhar limit, maybe their explosions had the same luminosity.

In the 1980s, however, the clear distinction between Type I and Type II began to blur. Spectroscopic analysis of three supernovae—one each in 1983, 1984, and 1985—showed that they consisted of huge amounts of calcium and oxygen, consistent with the interiors of massive stars that end their lives as Type II supernovae, but no hydrogen, consistent with white dwarfs that end their lives as Type I supernovae. Some astronomers, including Kirshner, suggested that they were seeing a third type of supernova, essentially a hybrid of the other two. It was the product of a core collapse that had already lost its outer shell: a hydrogen-free implosion.

They added this specimen to the Type I column, calling it Type Ib. The old Type I, a thermonuclear explosion with no hydrogen, was now Type Ia.

In 1991, even that classification—Type Ia—began to blur. On April 13, five amateur observers in four locations around the world discovered a supernova designated 1991T.* On December 9, an amateur astronomer in Japan discovered a supernova designated 1991bg. Follow-up spectroscopic observations by professional astronomers—including Kirshner, on April 16, for 1991T—showed that they were both Type Ia supernovae. But their luminosities differed widely. Supernova 1991T was much brighter than the usual Type Ia at its particular distance, and 1991bg was much dimmer than the usual Type Ia at its particular distance. Astronomers could rule out the possibility that they were simply miscalculating distances: The dimmer supernova was ten times dimmer than a supernova observed in 1957 in the same galaxy.

Astronomers began to suspect that while each supernova in the universe might be a Type Ia, Type Ib, or Type II, the types themselves might be more like families. The supernovae within a family share traits, but they're not identical; they're more like siblings than clones. For astronomers hoping to adopt Type Ia supernovae as standard candles, Kirshner wrote, the problem "was serious and real." You couldn't ignore it.

And the Berkeley group didn't ignore it. In her 1992 doctoral thesis a team member summarized the collaboration's general attitude toward the problem: "There is still some contention" about "whether individual SNe Ia do not fit the model," but, she added, echoing the chorus that Kirshner had heard from the LBL group again and again, "it is clear that the overwhelming majority of SNe Ia are strikingly similar."

Clear? Not to Kirshner, and he was the expert—a "realist," as he liked to call himself, not a wishful thinker.

In his role as a member of the External Advisory Board of the Center for Particle Astrophysics since the late 1980s, Kirshner emphasized that the Berkeley search team hadn't yet found a supernova, needed to be careful about photometry, couldn't account for dust—and didn't know whether Type Ia supernovae were standard candles.

Then in 1992 the LBL group found their first supernova. In his referee's report for Astrophysical Journal Letters, Kirshner complained that they still needed to be careful about photometry, still couldn't account for dust—and still didn't know whether Type Ia supernovae were standard candles. All they had shown, he thought, was that one could find supernovae distant enough that one could, in principle, do cosmology with them. But the Danes had done that, too, and they'd done it four years earlier. What the LBL team hadn't shown, in Kirshner's reading of the paper, was that one could find supernovae distant enough that one could in fact do cosmology with them.

He sent the paper back for a simple reason: "They hadn't yet learned anything about cosmology"—basically, that you couldn't assume exploding white dwarfs were perfect standard candles. They weren't perfect standard candles. The best you could hope was that somebody, someday, would figure out whether Type Ia supernovae, however imperfect, might be just good enough.

In high school in Marin County in the late 1960s, Boris Nicholaevich Suntzeff Evdokimoff played on the same varsity soccer team as his good friend Robin Williams. At Stanford in the 1970s, he regularly competed on the tennis court with—and lost to—Sally Ride. What was really cool, though, was that as a Carnegie Fellow in the early 1980s he got to talk astronomy with Allan Sandage.

Suntzeff loved historical connections in astronomy. A great-uncle of his had gone to school in Russia with Otto Struve, the descendant of a line of prominent astronomers. Struve fled Russia and the Bolsheviks at the time of the revolution and wound up in Turkey, impoverished, until a relative put him in touch with the director of the Yerkes Observatory in Wisconsin, who offered him a job as a spectroscopist. Struve later became director of the observatory, as well as McDonald Observatory in Texas and Leuschner Observatory in Berkeley. Suntzeffs family also fled Russia, though they headed in the other direction, to China and, eventually, San Francisco. There Suntzeffs grandmother reunited with Otto Struve. Small world.

And now Nick Suntzeff would be doing his part to make astronomy a bit more intimate. He had applied for a Carnegie Fellowship for just that reason: to spend time with Sandage at the headquarters of the Carnegie Institution's Mount Wilson and Palomar Observatories. There, on an unassuming residential stretch of Santa Barbara Street in Pasadena, Edwin Hubble had figured out, in 1923, that the Milky Way was just one among a multitude of galaxies in the universe, and then, in 1929, that the universe was expanding. Allan Sandage arrived there in 1948, at the age of twenty-two, as a graduate student at Caltech. Over the next four years Sandage advanced from apprentice to assistant to Hubble's heir.

"There are only two numbers to measure in cosmology!" Sandage often said to Suntzeff, evoking the title of an influential article he'd written for Physics Today in 1970, "Cosmology: The Search for Two Numbers." The first number was the Hubble constant. The 45-degree straight line that Hubble plotted for the distances of galaxies and their redshifts—the farther the galaxy, the greater its velocity receding from us—implied a relationship you could quantify. If you knew how distant a galaxy was, then you should be able to know how much faster it would appear to be receding, and vice versa.

In the 1930s, Hubble himself estimated that galaxies were receding at a rate that was increasing 500 kilometers for every megaparsec (a unit of length in astronomy equal to 3.262 million light-years). That rate, unfortunately, corresponded to a universe that would be about two billion years old—which would make the universe younger than the three billion years that geologists had pegged as the age of the Earth. This disparity did nothing to help cosmology's reputation as a nascent science. But Hubble himself regarded his observations only as a "preliminary reconnaissance"; to do cosmology properly, he would have to keep seeking nebulae as far as the 100-inch telescope on Mount Wilson would allow, and then, eventually, as far as the 200-inch telescope on Mount Palomar, outside San Diego, would allow.

The 200-inch Hale Telescope saw first light in 1948, which happened to be the same year Sandage arrived at the Carnegie Observatories. But Sandage's timing was fortuitous in another way as well. The "monks and priests," as he called the first generation of Carnegie astronomers, were ready to retire. Up there, at the observatory on Mount Wilson, Sandage could dwell among his gods, astronomers who knew they'd "arrived," as Sandage would say, when they found their napkin not clipped to a clothespin but tucked inside a wooden ring inscribed with their name. And down here, on Santa Barbara Street, Sandage could inspect for himself "the plates of Moses"—the vast archives of photographic records, and an apt metaphor in more ways than one. Like Moses, Edwin Hubble had come down from the mountain bearing new laws of nature. But also like Moses, Hubble had to wander the desert for decades, only to die within sight of the Promised Land.

Hubble suffered a major heart attack in the summer of 1949, at the age of sixty, just six months after making his initial observations with the 200-inch telescope. To his assistant fell the responsibility for executing one of the most ambitious scientific programs in history. Sandage found that the distances to the nearest galaxies were greater than Hubble had calculated, a correction that in turn affected Sandage's interpretation of more distant galaxies, which in turn affected his interpretation of even more distant galaxies. Distance dominoes fell as far as the 200-inch Palomar telescope could see. After Hubble's death in 1953, Sandage and his collaborators derived a Hubble constant of 180—a value he continued to revise downward over the decades, until, by the time Suntzeff arrived at the Carnegie Observatories in the early 1980s, he had satisfied himself that the Hubble constant was around 50 to 55.

Despite its name, the Hubble constant wasn't a constant—a value unchanging over time. It told you only how fast the universe was expanding now—its current rate of expansion—and for this reason astronomers sometimes referred to it as the Hubble parameter. It told you nothing, however, about how much the expansion rate was changing over time. That value—Sandage's second number—astronomers called the deceleration parameter because it would tell you to what extent the universe was slowing down. From the Hubble parameter you could extrapolate backward into the past and, depending on the amount of matter in the universe, derive the universe's age. From the deceleration parameter you could extrapolate forward into the future and, depending on the amount of matter in the universe, derive the universe's fate. In that sense, there were only two numbers to measure in cosmology: the alpha and the omega of the universe.

Both measurements would require a standard candle, and at the time that Suntzeff received his Carnegie Fellowship in 1982, Sandage (along with Gustav Tammann) had settled on supernovae. Sometimes Suntzeff and Sandage would be in Chile at the same time, at the Carnegie's Las Campanas facility, Sandage working on one telescope, Suntzeff another, and Sandage would ask Suntzeff to check whether a speck on a photographic plate was really a supernova. A dozen times Suntzeff swung the telescope to perform follow-up observations, and eleven times he had to break the news to Sandage: no supernova. In the end, Sandage figured out that he literally lacked the "proper equipment." The photographic plates were flawed. When he couldn't get Kodak to meet his exacting specifications for supernova searches, he abandoned the project.

But by then Suntzeff himself had become intrigued by supernovae. On cloudy nights at the observatory he would retire to the library and catch up on the supernova literature, or seek advice from Uncle Allan, as everybody called Sandage. The time was coming for Sandage to pass down to the next generation the program that Hubble had passed down to him. He had lost some sight in his right eye, the one he had pressed to an eyepiece for four decades, and his sense of balance was off, a hazard on an observing platform dozens of feet off a concrete floor. Soon he would have to pack up his eyepiece, pocket his napkin ring, and come down from the mountain.

Besides, Sandage could see that his way of doing astronomy was coming to an end. For the first two centuries after the invention of the telescope, astronomers had to rely on nothing other than the light that hit their eyes at any one moment, and then that light was gone. Astronomers could draw what they had seen. They could capture it in words. They could record measurements to designate the location of an object or describe its motions. But what they saw—the light itself, the visual representation of the object in a moment in time—was gone.

The invention of photography in the mid-1800s radically changed that relationship between observers and their observations. For astronomy, photographs had an obvious advantage over the eye. A photograph preserved what an astronomer saw. It preserved the light itself, and therefore the image of the object at one particular moment. Astronomers could refer back not only to what they had drawn or captured in words or recorded as math, but to what they had actually seen. And then so could any other astronomer, now or in the future.

But photography didn't just allow astronomers to collect light. It allowed them to collect light over time. Light didn't just land on the photographic plate; it landed and stayed there, and then more light landed and stayed there, and then more light. The sources of light were so faint your eyes couldn't see them, even with the help of a telescope, but the photographic plate could, because it was acting not like a moment-to-moment sensor but like a sponge. It could soak up light all night long. The longer the exposure, the greater the amount of light on the plate; the greater the amount of light, the deeper the view.

But now the charge-coupled device, or CCD, promised to do for the photographic plate what the photographic plate had done for the eyeball. A CCD consists of a small wafer of silicon that collects light digitally; one photon creates one electrical charge. A photographic plate is sensitive to 1 or 2 percent of the available photons; a CCD can approach 100 percent. For any aspect of astronomy, the advantages were obvious. Digital technology meant that you could process the images with computers, and more light meant that you could see farther and collect data faster. But for supernova searches, as Sandage explained to Suntzeff, the CCD came with a bonus.

The usefulness of a supernova for cosmology depends in large part on its light curve, a graph that shows the rise and fall of the luminosity of a supernova over time. Every supernova light curve rises abruptly over a matter of days as the supernova climbs toward maximum luminosity, then falls gradually as the supernova fades. But because each type of supernova releases its own distinctive cocktail of elements (hydrogen or no hydrogen, for example) and emerges out of a specific process (explosion or implosion), its light rises and falls in a signature pattern. To trace that pattern, you want to know when the curve peaks—when the brightness reaches maximum—so you need to be fortunate enough to discover the supernova on its way up. To chart the curve, you then need to make multiple follow-up observations—the more observations, the more data points you can plot on the graph; the more points, the more reliable the curve. But those observations are reliable only insofar as you can be sure how bright the light from the supernova is, and the accuracy of that measurement depends on how well you can distinguish the supernova light from the light of the host galaxy. A technology that allows you to make more observations and then quantifies those observations pixel by pixel can go a long way toward reducing the margin of error. The speed and precision of CCD technology, Sandage said, were going to make a light curve into a graceful, unambiguous arc—to the eye of a photometrist like Suntzeff, a work of art.

Suntzeff was already familiar with CCD technology. When he completed his Carnegie Fellowship in 1986, he became a staff astronomer at the Cerro Tololo Inter-American Observatory (a division of the U.S. National Optical Astronomy Observatory) in Chile; he was recruited by Mark Phillips, a good friend from graduate school—they'd both been at the University of California, Santa Cruz, back in the 1970s. Suntzeffs first mission was to install a CCD on a telescope, and he teamed up with Phillips to test the equipment on supernova 1986G. Suntzeff would do the observing and photometry, and Phillips would do the comparisons with light curves from other supernovae.

Suntzeff expected the result to be historic. As far as he and Phillips knew, theirs was the first "modern" light curve, meaning the first one obtained with a CCD. Historic though it was, the result was disappointing. The light curve for 1986G seemed to be significantly different from other Type Ia light curves. The supernova appeared to be fainter than it should be at its redshift, and the light curve looked as if it rose and fell more steeply than other Type Ia curves.

Part of the problem with being a scientific pioneer is that you have a compromised historical sample. The only light-curve comparisons that Phillips and Suntzeff could make had to come from photographic plates. They didn't know whether their odd CCD light curve said more about Type Ia supernovae or about CCD technology. Still, the two astronomers were confident enough in their corrections to the data that they concluded, in a paper they published the following year, that Type Ia supernovae probably varied too much in luminosity to serve as standard candles.

But as frustrating as the result was, Phillips and Suntzeff also sensed an opportunity. Their job would be to convince the community that Type Ia supernovae weren't standard candles—or to convince themselves that they had been wrong, and that Type Ia were standard candles after all. Either way, the two astronomers were in the supernova game now.

Their timing couldn't have been better. On February 23 of the following year, 1987, a supernova went off right overhead. SN 1987 A appeared in the Large Magellanic Cloud, one of the few galaxies visible to the unaided eye—and only from the Southern Hemisphere. It was the first unaided-eye supernova since 1604, and among astronomers it prompted a worldwide viewing party. It wasn't a Type Ia, the explosive kind of supernova that Phillips and Suntzeff had studied. It was a Type II, the implosive kind. Still, on the basis of their access to a CCD in the Southern Hemisphere and their co-authorship on the SN 1986G paper, they found themselves assuming the role of what they facetiously called "the local supernova experts."

In July 1989, they attended a two-week supernova workshop at their alma mater, UC Santa Cruz. The topic for the first week was 1987 A, but since the workshop was sure to attract just about every supernova expert in the world—all fifty of them—the organizers added a second week on supernova topics other than 1987 A. By this point, virtually everyone at the conference had been working on 1987 A nonstop for more than two years. They had results: observations, interpretations, theories. But they also had core-collapse-supernova fatigue. What about explosive supernovae? What about Type Ia? The first week of the meeting would be for work; the second week, fun.

Sometimes at conferences the most productive work happens in the hallway between sessions, or over a beer in the evening. For Suntzeff, it happened during a conversation with an old friend. "There are only two numbers to measure in cosmology!" Uncle Allan boomed at him, and while Suntzeff didn't think much of the comment at the time, he recalled it later, back home in La Serena, when a junior staff member at the observatory mentioned an idea for a project.

Mario Hamuy had arrived at Cerro Tololo as a research assistant on February 27, 1987—three days after 1987 A blossomed in the Large Magellanic Cloud. The original, pre-1987 A plan for Hamuy was that, as the new hire, he would go to the mountain and spend a few days acclimating himself to the instruments. Instead, the director of the observatory sent him to the mountain to observe 1987A and only 1987A. By the time he returned to La Serena a month later, Hamuy was, if not yet a bona fide supernova expert, at least a supernova enthusiast above and beyond his characteristic enthusiasm.

Now he explained to Suntzeff and Phillips that he had attended a talk at Santa Cruz by Bruno Leibundgut, a Swiss astronomer with a fresh PhD. Suntzeff and Phillips knew Leibundgut from observing runs in Chile in the early and mid-1980s, when he was a graduate student working with Gustav Tammann. They had attended his talk, too, and Leibundgut told them afterward that he'd been a late addition to the schedule. Bob Kirshner had recently hired him as a postdoc at Harvard, starting that fall, and at some point during the meeting Kirshner had turned in his seat and casually asked Leibundgut when he was giving a talk. Leibundgut answered that he wasn't; Kirshner told him that he was, now: He could have Kirshner's slot. And so Leibundgut wound up telling the world's supernova experts about his doctoral thesis: a template for Type Ia supernovae suggesting that they might be standard candles after all.

For Phillips and Suntzeff, the talk was part of a long-term, ongoing conversation in the supernova community. They had made their own contribution through their work on 1986G. For Hamuy, however, the talk provided a vision for the future. Listening to Leibundgut, he recalled that his graduate school advisor at the University of Chile, José Maza, had coordinated a supernova survey in the late 1970s and early 1980s. Maybe the time had come to revive the idea of a supernova survey from the Southern Hemisphere, this time using the superior CCD technology. On his return from Santa Cruz, he had approached Maza with the idea, and Maza agreed to help. Now Hamuy wanted to know what Phillips and Suntzeff thought.

Phillips told him he thought it might be a good idea, but, he cautioned, a supernova survey from the Southern Hemisphere had to be something more than a supernova survey from the Southern Hemisphere. At which point Suntzeff thought, "There are only two numbers to measure in cosmology."

Maybe they were wrong in thinking that Type Ia supernovae were not standard candles. Maybe Leibundgut, who after all had been studying other Type Ia while they were busy with the Type II 1987 A, was right. And if he was right, then maybe they could use nearby Type Ia supernovae to measure the Hubble parameter—the current rate of the universe's expansion. And if that program worked, they could go to farther supernovae to measure the deceleration parameter—the rate at which the expansion was slowing down.

Hamuy devised the logistics. The survey would be a collaboration between two observatories, Cerro Calán, the university's observatory, in Santiago, in the middle of the city, and Cerro Tololo, his current employer—hence the Calán/Tololo survey. Ideally, a supernova search would combine the widest-field camera with the latest CCD technology, but that option wasn't available to the collaboration. Instead they had to choose between a telescope that couldn't accommodate a CCD camera but had a wide-field view and a telescope that could accommodate a CCD but had a narrow-field view. They chose the wide-field, no-CCD view, the 24-inch Curtis Schmidt Telescope on Cerro Tololo. When hunting prey as rare and elusive as supernovae, the more galaxies you can grab at a time, the better your chances of finding even one, and in identifying supernova candidates, quantity still trumped quality. The photographic plates were big—eight inches by eight inches—and covered a patch of sky equal to one hundred full moons. For the follow-up observations, they would use the narrow-field CCD, a telescope with only a one-moon view, but that window was wide enough for performing photometry and spectroscopy on a supernova for which you already knew the specific coordinates.

The workday for the Calán/Tololo collaboration would begin at sundown at the Curtis Schmidt Telescope, where the Cerro Tololo team would take the images and develop the photographic plates. At sunrise, they would put the plates on a truck, which would take the plates to a passenger bus, which would arrive, via the coastal highway, seven or eight hours later in Santiago. There the bus would be met by research assistants from Cerro Calán, who would bring the plates back to the observatory and then blink the previous night's images with reference images from a few weeks earlier. By the time the sun was setting and the dome was opening in Cerro Tololo, Hamuy, Phillips, and Suntzeff would have a list of supernova candidates they would be chasing that night with the CCD.

The survey, however, wouldn't just discover supernovae. It would also improve the field's quality of observation and analysis by following up on the supernova discoveries of other astronomers, both professional and amateur. The team looked at the two odd 1991 supernovae—the surprisingly bright 1991T and the surprisingly dim 1991bg; Phillips became the lead author on an Astronomical Journal paper analyzing 1991T. Those two supernovae only reinforced his and Suntzeff's earlier suspicion that supernovae weren't standard candles. You could see the disparity at a glance—the light curves were just that different. The light curve belonging to the surprisingly bright 1991T rose and fell more gradually than the typical Type Ia light curve. The light curve belonging to the surprisingly dim 1991bg rose and fell more abruptly than the typical Type Ia light curve.

The bright one declined more gradually. The dim one declined more abruptly.

Bright ... gradually. Dim ... abruptly.

The correlation jolted Phillips. Would it hold if he examined the light curves from a range of supernovae? If so, then maybe Type Ia didn't have to be identical in order to be useful for cosmology. Maybe how gradually or abruptly a light curve rose and fell could serve as a reliable indicator of its brightness relative to other Type Ia supernovae. And if you knew the relative luminosities among supernovae, then, through the inverse-square law, you would also be able to figure out relative distances. You would be able to use supernovae to do cosmology.

Phillips recalled that while working on 1986G—the first CCD supernova light curve—he had consulted papers by Yuri Pskovskii from 1977 and 1984 positing a relation between the rise and fall of light curves and their absolute luminosities. But Phillips knew that the uneven quality of photographic plates made Pskovskii's hypothesis untrustworthy. Now, Phillips figured, he could use non-photographic studies to resolve the question.

Throughout 1992 he collected light curves, including some of his own, that he felt satisfied the most stringent observational criteria, and then he subjected them to months-long analysis. One morning late that year, he felt his preparations were over. The time had come to take the light curves, nine in all, and plot the data.

One of the advantages of living in a relatively small city like La Serena, Chile, Phillips thought, was that you could walk home and have lunch with your wife. He usually didn't discuss his work with her. She wasn't particularly interested in astronomy, and he didn't particularly feel like talking about it at home. But that afternoon he made an exception.

"I think," he told his wife, "I've discovered something important."

One of the first astronomers to write Phillips with congratulations was none other than Bob Kirshner—a blessing from afar, a benediction from above. The Danish and Berkeley teams had both asked whether one could discover supernovae at distances sufficient to do cosmology. Their answers, in 1988 and 1992, were: Yes. Now the Calán/Tololo team had taken what Kirshner considered the scientifically responsible first step and answered the question of whether Type Ia were standard candles: No. But they might be the next best thing: candles you could standardize. You could correlate the decline rate of the light curve with the supernova's absolute magnitude.

The next question, then, was: Could one detect distant Ia supernovae on a regular and reliable basis?

In March 1994, out of nowhere, the Berkeley team answered that question decisively: Yes. In a stunning announcement, they said that between December 1993 and February 1994 they had discovered six distant supernovae in as many nights.

When Schmidt sat down with Suntzeff in La Serena in late March and discussed the possibility of competing against Saul, the repercussions of the announcement were still reverberating in the supernova community, like one of the aftershocks perpetually roiling the Chilean countryside. Schmidt and Suntzeff themselves were reeling from a sickening realization. Since 1989, Calán/Tololo had collected fifty nearby supernovae, of which twenty-nine were Type Ia. The members of the collaboration would soon be publishing their value for the Hubble parameter, at which point their data would become public information, freely available. Eventually Berkeley was going to need nearby Type Ia to anchor the lower end of their Hubble diagram.

Saul was going to be using their data to beat them at their own game.

Extending the Type Ia search to higher redshifts, upon the completion of the nearby survey, had always been a possibility. Now it had become urgent. But they would have to move fast if they were going to collect enough distant supernovae of their own.

Brian Schmidt was the beneficiary of an oil-money education: a high school in Alaska that hired PhDs to teach teenagers. Schmidt hadn't been the best physics student in his class; he reckoned that two other students were superior. But he was the one who wound up putting physics to use, and he attributed the difference to "passion." If he said he could write the code for a high-redshift supernova search in two months, then he would write the code in two months. And he did, sort of. He wrote some new code and patched in some existing code from Phillips and Suntzeffs CCD observations of 1986G, and on the basis of the resulting test "data," he and his team received three runs of two nights each on the 4-meter telescope at Cerro Tololo early the following year, in February and March of 1995. By that time, Schmidt was in the process of moving to Australia with his wife and their three-month-old—he would be living in Canberra and working at the Mount Stromlo and Siding Spring Observatories—and he didn't have travel money. But he figured he could manage the observations with the help of on-site collaborators and the Internet.

That's when he had his first epiphany: Test data ain't the real thing.

In February 1995 the team began taking the reference images—the initial images that they would subtract from later images. Sometimes the software wouldn't run. When it did run, Schmidt found himself trying to debug software and download images on a 100-byte-per-second link. Often he would have to guess what the problem was and do minuscule modifications of code, which his collaborators in Chile would implement, and then inevitably they'd all be talking on the phone, Cambridge and La Serena and Canberra, and Schmidt would wind up staying up all night while his wife, caring for the baby, hovered nearby. Finally Schmidt instructed his collaborators in Chile to send a set of data to him on tape so that he could examine it himself.

It never arrived. Somewhere between Santiago and Siding Spring, it vanished.

That's when he had his second epiphany: From now on, I will always go to Chile.

"From now on," though, presupposed that they would find a distant Type Ia. The first night of follow-up observations, February 24/25, was clear, but the "seeing"—the term astronomers use for atmospheric conditions—was poor. The second night, March 6/7, had excellent seeing and produced six candidates, but on closer inspection none turned out to be a supernova. The third night, March 24/25, was, except for one brief period, overcast.

In the Observing Proposal that the team had submitted the previous September, Schmidt and Suntzeff had written, "Based on the statistics of discovery from the Calán/Tololo SN survey, we can expect to find about 3 SNe Ia per month." Surely, they told themselves now, they should find at least one. Yet they were heading into their final night, and they'd gotten nothing. Were they doing something wrong? Did the software harbor a glitch? Or had they just been unlucky?

On the final night, March 29/30, a 16-pixel-by-16-pixel smudge of light squeezed through the Internet pipeline between Chile and Australia. Schmidt squinted at it, then squinted at it some more, but he couldn't be sure. He picked up the phone and called Phillips and Suntzeff and told them to take a look and tell him what they thought. They took a look and agreed that it sort of seemed like a supernova, maybe. But they couldn't know for sure until they got the results of the crucial follow-up spectroscopy, from the 3.6-meter New Technology Telescope in La Silla, Chile.

Bruno Leibundgut—whose talk in 1989 had inspired the earlier survey—was in charge on Sunday, April 2. Late that night he called Mark Phillips with a bad-news/good-news report.

First the bad news. "It's very faint," he said.

But the good news was that at least the smudge was still there, indicating that it was indeed a supernova.

So: They'd been merely unlucky.

The team continued to sort through the data from the March runs, and by Wednesday of that week they had found that the candidate supernova from the final night had also appeared on the last field from the third night of observing, just before the clouds closed them down. "This is very encouraging," Phillips e-mailed the team.*

Thursday morning, Chile time, Schmidt sent out a progress report to all the members of the collaboration. He suggested they start thinking about submitting the observation to the International Astronomical Union circular, a standard procedure. He also suggested that they start thinking of what to call the collaboration—"a catchy name (or at least an accurate one if we cannot be catchy)." They were moving forward as a team, though he still wished, he added, that they had a redshift for the galaxy.

That evening, they did. Mario Hamuy had examined the spectra that Leibundgut had obtained four nights earlier and reported the result to Phillips, who relayed it to the rest of the collaboration: The host galaxy, and with it the supernova itself, showed a redshift of 0.48, placing it at a distance of 4.9 billion light-years—and setting a new supernova record.

They couldn't yet tell if it was a Type Ia, meaning that they didn't yet know if it would be useful in determining the rate of deceleration. Leibundgut would have to keep pounding at the data before they could say anything with confidence.

But as Leibundgut wrote to Schmidt that day, "We are rolling again. What a change a supernova can make."

They had beaten Saul at beating them at their own game.