The 4 Percent Universe

by Richard Panek

Turner was slightly more diplomatic during his own talk, but no less needling. "I am anxiously awaiting the results of the two deep searches for supernovae," he said, referring to the rival teams. "I think they're going to shed some important light on this. To draw any conclusion now would be to take away from their thunder later."

  The SCP submitted their data on their first seven supernovae to the Astrophysical Journal that August. If they made the standard assumption regarding lambda—"a A = 0 cosmology"—then omega was 0.88. But given the margin of error, you could reasonably interpret that result as omega equaling 1. If they made the less likely assumption that the universe was flat with a possible component of lambda, then omega, at 0.94, was even closer to 1, while lambda would be 0.06, a negligible amount and, given the margins of error, presumably 0.

  The universe was flat. Matter alone was enough to get omega to 1. And we didn't need lambda. Or at least that interpretation was, as the paper said, "consistent" with their results.

  Unfortunately for the SCP team, that interpretation wasn't consistent with their own next round of data.

  The Astrophysical Journal accepted the paper in February 1997 and published it in the July 10 issue. By then the SCP team was finishing their analysis of the two supernovae they'd examined with HST. Because HST photometry would be so superior to ground-based analyses, the team would place special emphasis on whatever it had to tell them.

  Peter Nugent had been hired as a postdoc by Perlmutter a year earlier, part of a campaign to bring astronomers onto the project. Nugent had written his thesis on Type Ia supernovae, and Perlmutter had assigned him to perform photometry. Nugent had a forceful style. He wouldn't have been out of place at the University of Chicago; his bearing and attitude were reminiscent of a David Schramm or a Rocky Kolb: a can-do, answer-any-question, know-the-restaurants-with-the-best-wine-lists spirit. On June 30 he finished the photometry on the two HST supernovae, giving him their magnitudes, the standard measure of luminosity for celestial objects. Spectroscopic analysis had already yielded the redshifts for the two supernovae. Now Nugent plotted the two values against each other, redshift on one axis, magnitude on the other.

  You would expect the points on this plot to fall pretty much on the usual 45-degree-angle straight line—the relationship among nearby galaxies that Hubble discovered in 1929. The straight line itself represents a universe that is expanding uniformly, experiencing no effects of gravity—in other words, a universe without mass, a universe with nothing in it. Eventually, at some great distance across space and back in time, the points will have to begin to deviate from the straight line to represent a universe that does have mass. But which kind of universe? The extent to which the most distant points deviate downward from the straight line will be minor, but it will tell you how much brighter the objects are than you would expect them to be at their particular redshifts—the brighter the supernovae, the higher the value of omega. And that value will tell you the weight, shape, and fate of the universe: open, closed, or flat; saddle, globe, or plane; Big Chill, Big Crunch, or Goldilocks.

  Nugent began plotting the two HST supernovae. First he looked along the redshift axis—the measurement that corresponded to their distances. Then he moved up the graph until he reached the magnitudes his photometry had given him. He assumed the two points would fall along the deviation—the particular downward curve—consistent with the conclusion that the SCP's latest paper had reached: a flat, all-matter, omega-equals-1 universe. But that's not where these two supernovae fell. They were landing on the other side of the straight 45-degree-angle Hubble relationship, on what would be an upward curve. The difference between what their luminosities should be at their redshifts and what their luminosities were, was approximately half a magnitude, meaning that the two supernovae were 1.6 times fainter than he expected.

  "There goes the universe," he wrote in the e-mail to his team. Not that he was ready to draw any conclusions about cosmology. After all, as he wrote, "it's only two data points." And he wasn't the team member responsible for determining the omega and lambda measurements. But the discrepancy between the magnitudes he expected and the magnitudes he measured was unequivocally jarring. "Hopefully this will be enough from me to get the paper out this week," he added. "I do think it has to go out now since the other group is most likely going to submit something soon (very soon) [about their own HST results]. It's good with the data—as-is. Lets get the damn thing out there!"

  But they didn't. The team quickly realized that they needed to decide not only whether to publish, but what.

  In the jargon of science, the two HST supernovae were "fighting" the earlier, all-matter, omega-equals-1 result. That summer, the team threw out two of the first seven supernovae—one that further analysis determined to be a core-collapse supernova rather than a Type Ia, and another one that was an obvious outlier. They also eliminated the 1996 HST supernova because they felt that, while the individual measurements were probably accurate, they didn't have enough observations—enough points on the light curve—to subject the supernova to peer review. But the similarity in results between the 1996 and 1997 HST supernovae did reinforce the team's confidence in the 1997. By September they had settled on six supernovae in total as well as a conclusion—albeit one inconsistent with a paper they had published only two months earlier.

  One member of the collaboration wrote to Nugent that he "must realize that we will look very bad if we change our limits every time we add *one* SN to the total sample without a discussion [in the paper]. How can anybody trust what we say if they know we are going to say something else in a few months time without any explanation?" After all, the two papers weren't dependent on two separate samples or a significantly larger set of data. The new paper had two fewer supernovae than the previous paper. The only addition to the data was a single supernova from HST.

  The point of the paper, Nugent argued back, would be to demonstrate what HST could do for a distant supernova search— "NOT" to declare that the universe has certain values of omega and lambda, "says God." The number of supernovae didn't matter. "I've never given a rat's ass about one data point (or even a number under 10 for that matter) in my life when the error bars are so large." The method was what was worth reporting.

  Still, writing a paper that reverses a result even implicitly was going to require some finesse. Not until September 1997 did the team have a draft they would submit to Nature, and by then they had larded the prose with enough qualifiers to choke even a Kirshner: "we use the words 'preliminary', 'initial' and 'if...' all over this paper," Nugent reassured a colleague in a September 27 e-mail. And when the paper got to the omega and lambda part, it delivered a double qualifier: "these new measurements suggest that we may live in a low mass-density universe" (emphases added).

  The team submitted its paper on the HST supernova to Nature the first week of October. Sure enough, just as Nugent had fretted at the end of June, the High-z team followed with its own HST supernova paper, with Garnavich as the lead author, posting it on the Internet on October 13. The High-z paper reported that its sample, too, "suggests that matter alone is insufficient to produce a flat Universe." Clearly the two groups were converging on the result that had motivated their supernova searches: the fate of the universe.

  If there was no cosmological constant, then omega was low and the universe was open—destined to keep expanding for all time. Even if there was a cosmological constant, then omega was still low and the universe was flat—slowing to a virtual halt, but not collapsing. Either way, the expansion of the universe would continue forever. That fall the American Astronomical Society invited both teams to participate in a press conference at the AAS meeting in January 1998. The press department at the AAS usually organized four or five press conferences during the course of the semiannual five-day meetings, and a discussion of the fate of the universe seemed like the kind of topic that would draw a crowd. Sure, the two teams told the AAS, we'd be glad to send representatives to a press conference.

&nb
sp; But a subtler, and certainly more esoteric, question remained: Was there a cosmological constant?

  Gerson Goldhaber, anyway, thought there was. On September 24 he showed the group the histograms compiling all the supernovae, one for a no-tambda universe, and one for an omega-plus-tambda-equals-1 flat universe. For a measurement as delicate as the one the team was trying to make, binning supernovae into broad categories wasn't going to be as persuasive as plotting individual points. But a trend was clearly developing. The more supernovae the team analyzed, the lower the value of omega seemed to be heading. Two weeks later, the minutes from another team meeting reflected the trend: "Perhaps the most disturbing thing is that the first 7"—the bunch on which the team had based their previous paper—"were consistent within themselves but the next 31 Sne give what seems to be a consistent answer that is lower."

  In the 1930s Fritz Zwicky had discovered a set of supernovae that he assumed were examples of the implosion process that he and Walter Baade had predicted; in retrospect, those supernovae all turned out to be examples of an explosion process that hadn't yet been discovered. Now the SCP team was realizing that they, too, had defied the odds. Even after eliminating the obvious outlier and the Type II from the original set of seven, those five initial supernovae still appeared to be on the bright side. As a result, the addition of the dozens of fainter supernovae was driving the value of omega down. In the histogram analysis of the data, a sharp peak was developing around an omega of 0.2.

  On December 14, 1997, Goldhaber presented his findings at a seminar at the Institute for Theoretical Physics at UC Santa Barbara. Kirshner was in residence at the institute that fall, on sabbatical from Harvard, and as usual Goldhaber found him to be "antagonistic." Kirshner interrupted the presentation: An omega of 0.2; so what else is new? But Goldhaber thought he was making an argument that omega could be 0.2 only if accompanied by lambda. At least the director of the institute, David Gross, seemed to understand, though when he asked Goldhaber why he believed the results, all Goldhaber could offer was that he had a long history of interpreting histograms. "I'm convinced," he said.

  Perlmutter, too, was presenting preliminary results in public that fall, carrying his transparencies of low-omega scatter plots from colloquium to colloquium—the first on October 23 at the Physics Department at UC San Diego, the second on December 1 at the Physics Department at UC Berkeley, and a third on December 11 at the Physics Department at UC Santa Cruz. As in the Nature paper, he was careful to qualify his comments, but he also made sure to let his audiences know that the data contained the possibility of "some rather striking consequences for physics," as he said at the Berkeley colloquium. "In particular, if you consider the flat-universe case—the case of the inflationary universe that's favored—a mass of this sort, a mass density of this sort, means that the cosmological constant has to be contributing a cosmological constant's energy density of about 0.7." In case the non-cosmologists in the audience were missing the point, the astrophysicist Joel Primack stood up at the end of Perlmutter's talk at Santa Cruz to say the results were "earthshaking." Then he added the crucial caveat: "If true."

  For the High-z team, Adam Riess was now "it." Riess knew that his team was at a disadvantage concerning the quantity of supernovae, if only because Peter Nugent kept reminding him. The two of them were in a group that got together on weekends in a city park to play a variation on football called, for obvious reasons, mudball. Sometimes the trash talk took the form of my-distant-supernova-search-is-better-than-yours. One day Riess decided he was tired of hearing how many supernovae the SCP was raking in and how far behind the High-z search was. If you couldn't beat the SCP on quantity, he figured, you could beat them on quality.

  For his master's thesis Riess had tackled the problem of dimness. His light-curve shape method proposed a mathematical solution to deriving luminosity from the rise and fall of light-curve shapes. For his PhD thesis Riess had approached the problem of dust. If you're trying to determine the distance of a supernova by measuring its redshift, then you need to know to what extent dust is contributing to the reddening of the light (just as dust in the atmosphere reddens a sunset). In Riess's multicolor light-curve shape method, or MLCS, the observations of light in several color filters would provide a cumulative measure of the effect of dust, allowing you to derive a more accurate determination of distance.

  As his team's resident expert on correcting for intergalactic dust between the supernova and the observer, he might be able to clean up the supernovae in such a way that they provided a tighter margin of error than the SCP's. He wouldn't even need a greater number of distant supernovae, though they were always welcome. Even nearby supernovae would do the trick. If he could anchor the lower end of the Hubble diagram with sufficiently reliable data, then the higher-redshift supernovae—while fewer in number than the SCP's—would be more reliable as well. And he knew where he could get nearby supernovae: observations he had already made, as part of his thesis research, at the 1.2-meter telescope on Mount Hopkins, in Arizona. Twenty-two supernovae in all. None of them yet published.

  The addition of those supernovae, however, created a new problem. Never mind a universe with no matter. His calculations were producing a universe with negative matter.

  "I'm only a postdoc," Riess told himself. "I'm sure I've screwed up in ten different ways." Computers, he thought, don't know physics. They know only what we program them to know. Clearly he had programmed his computer with impossible physics. So Riess checked his math, and he checked the computer code he'd written, and he couldn't find any mistakes. Of course, Einstein's equations allowed for another option—a universe with a positive lambda. Plugging his data into that universe brought the amount of matter up, into the positive range. But that option, he knew, wasn't palatable to most astronomers—for instance, the team leader, Brian Schmidt, who liked to say that astronomers who talk about the cosmological constant are astronomers without many friends.

  Riess sent his results to Schmidt.

  "Adam is sloppy," Schmidt reminded himself. Brilliant, but prone to mathematical errors. Schmidt agreed to double-check the results. As a rule, mathematicians check each other's work not by looking back over the same calculations but by performing the calculations independently, so as not to be lulled into making the same mistakes. Schmidt and Riess soon developed a routine. Riess would e-mail a problem, and a day later Schmidt would respond. I started with this image, and my analysis said the supernova was this bright—how about you? Or We observed in this filter, and I found that the redshift was equivalent to this number—how about you? They signed their e-mails Pons and Fleischmann, after the two physicists who, in 1989, had "discovered" cold fusion, and who, after a long period of infamy, had fallen into obscurity. If you're Stephen Hawking and you make a major mistake, you're still Stephen Hawking. If you're a postdoc under thirty and you make a major mistake, you're history. Sometimes when Riess couldn't wait for an answer, the phone would ring in the Schmidt household. Schmidt's wife, sleepless from caring for a six-month-old, would say, "If that's Adam, tell him to—"

  Schmidt: "Hello, Adam."

  Riess: "Oh." A pause. "Is it early there?"

  Schmidt: "It's four in the morning."

  Riess: "Oh." A pause. "So, what do you know?"

  What Schmidt knew, night after night, was the same thing: So far, so good.

  Riess remembered now that one day when he was a graduate student at Harvard, Kirshner had brought Mike Turner and Alan Guth by his office and encouraged Riess to show them what he was working on. Riess had just taken the team's first Type Ia supernova, 1995K, and plotted it on the Hubble diagram. The supernova fell on the "bright" side of the 45-degree straight line, but its location didn't matter; it was only one point. What mattered was that the team actually had a point to plot. Still, Turner couldn't help mentioning that it was in the "wrong" part of the diagram.

  "How embarrassing," Riess had thought. "There's never been so much brainpower in P-306"—his office at the
time—"and we're probably showing them that we're not even doing the experiment right."

  But now, a couple of years later, he thought maybe the location of that point had mattered more than anyone knew. Maybe the answer to the fate of the universe had been right in front of them from the very first supernova.

  Riess was getting married in January 1998—the weekend at the end of the AAS meeting, in fact. When his future wife flew home from Berkeley to her family in Connecticut a few weeks early to take care of the final preparations, Riess sequestered himself in his office in Campbell Hall, on the Berkeley campus, and began writing a paper that would report the results—if they held up.

  The campus was empty for holiday break. The heating was off, and Riess had to bundle up; even in California, December can get chilly. But every day, walking past the locked office doors and under the unlit hallway lights, he went to work. On December 22, he wrote an outline and started a draft. Garnavich's HST paper, the first from the group, had been a short letter. Riess figured the next paper would have to be the War and Peace version, as scientists like to say; if you're claiming something surprising, you have to show all the work. In the coming days he also contacted Nick Suntzeff, down in Chile, and asked him to double-check some photometry, though he didn't say why so as not to prejudice the result. At one point he beckoned a colleague into his office.

  Alex Filippenko, who also had been taking advantage of the semester break to catch up on work, greeted Riess with his usual wide and deep smile, rectangular and cavernous. Nobody could be that happy all the time, and Filippenko wasn't. He had once been a member of the SCP, and as an astronomer on a team with a particle physics mentality, he had experienced the clash of cultures probably more acutely than anyone else on either team. He disliked the hierarchical structure that awarded Perlmutter lead authorship on the important papers; Filippenko would go to astronomy conferences and hear about "this supernova survey" that Saul had organized, and he'd have to inform his peers that he was actually part of that collaboration. He watched as his friends in the supernova game—Kirshner, Riess, Schmidt, Suntzeff—coalesced into a collaboration of their own. He complained to them that he had been warning his SCP colleagues about the possible non-standardness of Type Ia, about dust, about the difficulty of photometry and spectroscopy—all the concerns that Kirshner had been raising for years as a member of the External Advisory Board. He said he felt that the Berkeley Lab physicists regarded these concerns as if they were "irritations" and "annoyances" rather than supernova astronomy's swords of Damocles. He felt marginalized and ignored on the SCP collaboration, and he suspected that they kept him around only as the "token astronomer" who could get them time on telescopes.