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

Part III. The Face of the Deep

Chapter 9. The Tooth Fairy Twice

MIKE TURNER WAS following in David Schramm's footsteps. He was walking along the hallways and footpaths, among the blackboards and picnic benches, of the Aspen Center for Physics, a summer retreat for theorists under head-clearing blue skies. One look at the mountains, one deep breath, and you could see why a big-as-all-outdoors guy like Schramm had fallen in love with the place at first sight in 1976, enough to make Aspen his second home. Eventually he'd served as the chairman of the board for the Aspen Center, from 1992 until shortly before his death. But Schramm was gone now, and Turner had agreed to take his place opposite Jim Peebles in the "Nature of the Universe Debate" at the Smithsonian, so when Turner ran into Peebles at the Aspen Center, he had a question for him. For obvious logistical reasons, the organizers had bumped the event from April 1998 to October—and just as well. Turner and Peebles needed a new topic.

"Are you still willing to debate non-flat?" Turner asked.

Peebles shrugged.

"Jim, debates have a yes-or-no question. Correct me if I'm wrong, but you and Dave were supposed to debate whether or not the universe is flat. He got flat and you got non-flat." Turner asked his question again. Did Peebles really want to argue publicly that the universe wasn't flat?

"No."

The answer didn't surprise Turner. Both theorists knew that defending a non-flat universe in late 1998 would feel like defending the Steady State cosmology in late 1965. Within months of the January AAS press conference where Perlmutter unveiled the SCP's forty-two supernovae, and within weeks of the February UCLA meeting where Filippenko made his announcement, a consensus had emerged in the Big Bang community of astronomers, astrophysicists, cosmologists, and theorists: The universe wasn't what it used to be.

Since Hubble's discovery of evidence for the distance-velocity relation, astronomers had been following a syllogism: One, the universe is expanding; two, the universe is full of matter attracting other matter through gravity; therefore, the density of matter will affect the rate of expansion. So: How much was the expansion slowing? This question was what the two supernova teams had dutifully set out to answer, and they had succeeded: It wasn't.

The expansion wasn't slowing. The universe the two teams observed wasn't one where distant Type Ia supernovae were brighter than they should be at this particular redshift or that particular redshift, and therefore nearer. It was one where they were dimmer, and therefore farther. It wasn't a universe that was doing what an expanding universe full of matter acting under the influence of mutual gravitational attraction should be doing. It was doing the opposite.

The expansion of the universe was speeding up.

James Glanz broke the story in the February 27 issue of Science. Even though he had hinted at the possibility of a positive lambda in two previous articles, in October and January, such a result would be so difficult to accept that the community, rightly, was treating the possibility with what he considered "a preponderance of skepticism." An agreement between the two teams, however, might change that dynamic. Even before Filippenko's announcement at the UCLA meeting, Glanz had begun rounding up quotes from the High-z team. "To be honest," Bob Kirshner said, "I'm very excited about this result." Adam Riess said he was "stunned." Most quotable of all, from Brian Schmidt: "My own reaction is somewhere between amazement and horror." He elaborated: "Amazement, because I just did not expect this result, and horror in knowing that it will likely be disbelieved by a majority of astronomers—who, like myself, are extremely skeptical of the unexpected."

By the end of the day that Science published Glanz's article, February 27, Riess had appeared on CNN and PBS. (NewsHour interviewer: "Why did some scientists react with what one called amazement and horror to these conclusions?") Articles in magazines and newspapers appeared around the world in the following week, culminating in a 1,600-word feature in the New York Times ("'My own reaction is somewhere between amazement and horror,' said Dr. Schmidt, the team leader").

At Berkeley Lab, the SCP team also responded with amazement and horror. Amazement because they had succeeded in finding nothing less than the fate of the universe, and horror because acceleration itself was what everyone suddenly wanted to talk about—and the High-z team was getting the credit for it. As Perlmutter said in a lab press release, the two teams were in "remarkably violent agreement."

"Basically, they confirmed our results," Gerson Goldhaber told the New York Times. "But they won the first point in the publicity game."

"Hey, what's the strongest force in the universe?" Kirshner said in the same article. "It's not gravity, it's jealousy."*

In early March, the High-z team submitted "Observational Evidence from Supernovae for an Accelerating Universe and a Cosmological Constant" to the Astronomical JournaL. The first week of May—a few days before the paper was even officially accepted—Fermilab convened a conference on the two supernova teams' results. A straw poll of the sixty or so attendees showed that forty were willing to accept the evidence.

Part of the rush to consensus in the community was sociological. Corroboration for any scientific result is always necessary, and if only one team had reached a surprising result, the response of the community would have been intense skepticism. That two teams had independently arrived at the same conclusion was notable. So, too, that the two teams had used mostly independent sets of data (very few of the same supernovae), had relied on several independent methods of analysis (including the corrections for dust), and had arrived at a conclusion that was the opposite of what they expected. But two teams that met all those criteria and had been infamously competitive? "Their highest aspiration," Turner said, "was to get a different answer from the other group."

And part of the rush to consensus was aesthetic. Just as inflation in 1980 solved the flatness and horizon problems, a positive lambda in 1998 made the universe understandable again. As Turner said, "It made eve ry thing fit!"

Those measurements of the Hubble constant on the "wrong" side of 60 that displeased Allan Sandage because they suggested a universe younger than its oldest stars? Problem solved. Those large-scale structures of supercluster filaments that seemed too mature for such a young universe? Problem solved. The universe was "too" young only if you assumed that the expansion rate had been decelerating throughout the history of the universe, or at least holding steady. A car that had been accelerating from 50 to 60 miles per hour and was only now reaching 65 would have needed more time to cover the same stretch of road than a car that had already been cruising at 65 miles per hour or slowing down from 70. If the expansion were decelerating, hitting the brakes, it would have been going faster in the "recent" past, and therefore taking less time to reach the present, than if it had just been constant. But an expansion that was accelerating today, hitting the gas, going faster and faster, would have been going less fast in the recent past, taking more time to reach the present. Thanks to acceleration, the age of the universe seemed to be, roughly, in the range of fifteen billion years, safely in the older-than-its-firstborn, old-enough-to-have-mature-filaments range.

But what made the supernova results perhaps most aesthetically pleasing wasn't just the presence of a positive value for lambda but the value itself.

If you were Bob Dicke or Jim Peebles in the late 1970s and you wanted the observation of a uniform cosmic microwave background to make sense, you wanted a theoretical explanation for homogeneity and isotropy. And then you got one: inflation. If you were Mike Turner or Rocky Kolb in the 1980s and you wanted the theory of inflation to work, you wanted an observation that revealed a flat universe. And then you got one—or half of one, anyway. COBE indicated that inflation was correct, meaning that the universe had to have an omega of 1. But numerous other observations indicated that the amount of mass in the universe was less than critical density, meaning that omega had to be less than 1—significantly so.

But now lambda explained away that contradiction. The amount of matter in the universe wasn't enough to halt the expansion, but the amount of matter and energy in the universe was. According to Einstein, matter and energy are equivalent, so while the mass, whether in the form of dark matter or regular matter, might well fall short of the critical density, the energy causing the acceleration—lambda—could make up the difference. A mass density of 40 percent or so plus an energy density of 60 percent or so added up to 100 percent of the critical density, or an omega of 1.

The universe did have a low matter density.

The universe was flat.

"Admit it," Jim Peebles once teased Brian Schmidt, "you didn't know what you had. You'd never heard of inflation."

"Inflation!" Schmidt answered. He informed Peebles that at Harvard he'd shared an office with the theorist Sean Carroll while Carroll was writing "The Cosmological Constant," one of the influential pre-1998 papers that explained how lambda could save inflation. "Alan Guth used to drop by once a week!" he added.

A positive lambda solved so many problems that when Turner approached Peebles in Aspen in 1998, he already knew what he wanted to debate: "Is Cosmology Solved?"

It was a debate Turner had already been framing. That March, for a dark-matter workshop in Gainesville, Florida, he had titled his talk "Cosmology Solved? Maybe." The following month, for a conference in Kyoto, Japan, he had dropped the qualifier from the title and went with a more straightforward "Cosmology Solved?" The published versions of both papers included the same sentence in the abstract: "These are exciting times in cosmology!" For the Smithsonian debate he took the exclamation point out of the body of the talk and promoted it to the title: "Cosmology Solved? Quite Possibly!"

Peebles would have to handle "quite possibly not," which was fine by him. It wasn't in his nature to argue passionately for a specific side of an unresolved issue, if only because having convictions about unresolved issues was unscientific. If he felt passionately about anything, it was that, in the absence of facts, you shouldn't feel too passionately. Fourteen years after he himself had used inflation's prediction of a flat universe as the basis for a lambda argument, Peebles still thought the community's embrace of inflation was premature—"distasteful," even. When he thought about physics, he divided its practitioners into classicists and romantics. The classicists were inventive but followed the rules; the romantics were respectful of the rules but followed their intuition. The romantics waved their hands and came up with a homogeneous, isotropic universe, and then, if they were lucky, an observation came along that could test their assumptions and predictions. A classicist looked to that observation—the one that suggested an expanding universe—then made a prediction of a temperature for the background radiation that observations would test. Then it was the romantics' turn again, waving their hands and invoking inflation and dark matter and, now, "missing energy"—the explanation that some classicist was going to have to invent for whatever physical presence in the universe corresponded to a positive lambda and caused the expansion to accelerate. Jim Peebles liked to think of himself as a classicist.

Mike Turner liked to think of Jim Peebles as "half enthusiast, half curmudgeon."

The debate took place on a wet Sunday afternoon in October at the National Museum of Natural History on the National Mall in Washington. The setting was Baird Auditorium, the same hall where the astronomers Heber Curtis and Harlow Shapley had "debated" in 1920 whether the Milky Way was the universe in its entirety or whether other "island universes" existed outside of it. Back then, Vesto Slipher's spectroscopy showing redshifts of nebulae was less than a decade old. Einstein's cosmology was only three years old and still applied to a static universe, thanks to his insertion of lambda. Hubble's discoveries that the nebulae were separate island universes and that, when their distances were graphed against their velocities, they seemed to be receding, lay in the decade to come—and with them, the apparent obsolescence of lambda. But now, some seventy years later, lambda was back. On their way into the auditorium, audience members received buttons bearing "A." If they were bewildered by the symbol, they weren't for long.

Like the earlier debate, the 1998 version wasn't going to solve anything; its purpose was to educate and entertain. And in terms of showmanship, as Peebles would have known in advance from having attended numerous talks by Turner, the debate was over before it began. All Turner had to do to win over the audience was to display one of his usual colorful viewgraphs, complete with Keith Haring-like dancing silhouettes:

COSMOLOGY is EXCITING!... for at least the next 20 years
STRONG FOUNDATION: Hot Big-bang Model
BOLD IDEAS DEEPLY ROOTED IN FUND. PHYSICS: Inflation + CDM
FLOOD OF DATA

(And all Turner had to do to make Peebles wince was say the words "precision cosmology.")

When the audience left four hours later, the drizzle might have felt like exclamation points dancing over their heads. But the question "Cosmology Solved?" was, by Turner's own admission, "ridiculous." As he acknowledged at the end of the debate, he was being "purposefully provocative." Debates might need a yes-or-no question, but Turner couldn't answer "Yes" and Peebles couldn't answer "No" without seeming foolish.

In a way, their roles on stage, while suiting their personalities, were almost perversely reversed. Turner argued, "I believe we will ultimately refer to 1998 as a turning point in cosmology as important as 1964"—the year that Wilson and Penzias inadvertently detected the cosmic microwave background at a temperature that Peebles himself had predicted. He cited the progress in establishing the most fundamental numbers in cosmology—the two that Sandage had always cited, plus the third that inflation had introduced. Astronomers were converging on a Hubble constant in the mid-60s. They were agreeing to a matter density of 0.4, give or take. And despite that seeming shortfall, they had discovered observational evidence that bumped the ratio between the overall density and the critical density—between the matter/energy density and the density necessary to keep the universe from collapsing—up to 1.

Yet Turner himself acknowledged that there was a problem that a positive value of lambda didn't solve. Cosmology had a new syllogism: One, the expanding universe was full of matter attracting other matter through gravity; two, the expansion was speeding up; therefore, something other than matter, dark or otherwise, had to be overwhelming the influence of gravity. So: What was it?

Cosmology solved? Hardly!

To astronomers, lambda was just a fudge factor, a symbol in an equation. It might equal zero. It might not. But if you had confidence in the usefulness of Type Ia supernovae for cosmology, and if you satisfied yourself that you'd checked your results, then you accepted its value. Brian Schmidt had been aware of the implications of a positive lambda for the theory of inflation, but Adam Riess, for instance, had not. In the days after Riess's computer code told him that the universe had negative mass unless he balanced it with a positive lambda, he'd had to educate himself—happily—about all the problems that a cosmological constant would settle.

For particle physicists, however, a positive lambda didn't solve a problem. It created one.

From a particle physics perspective, lambda wasn't just a number. It was a property of space. And space, in particle physics, wasn't empty. It was a quantum circus, a phantasmagoria of virtual particles popping into and out of existence. Not only did those particles exist, as experiments had shown, but they possessed energy. And energy, in general relativity, interacts with gravity. The result of quantum particles possessing energy that interacts with gravity was what physicists called the Casimir effect, after the Dutch physicist Hendrik Casimir. Put two parallel plate conductors closer and closer together, Casimir proposed in 1948, and you could measure the increase in the vacuum energy. Numerous experiments since then had found agreement with his predictions. As the mathematician Stephen Fulling noted, "No worker in the field of overlap of quantum theory and general relativity can fail to point this fact out in tones of awe and reverence."

So positive energy itself wasn't a surprise. And theorists even had two forms of vacuum energy in mind—or two names for them, anyway. One form of vacuum energy would be constant over space and time, and they would call it the cosmological constant. Another would vary over space and time, and they would call it quintessence (the fifth element in ancient Greek physics). In order to discourage astronomers from assuming that the terms "lambda" and "cosmological constant"—which they'd been using nearly interchangeably—were identical, Turner started testing other terms. "Funny energy" he auditioned at the Fermilab conference in May 1998, but that didn't stick. His next try—"dark energy," with its deliberate echo of "dark matter"—did.

The problem with the supernova result of a positive energy density in the universe, however, was that quantum mechanics predicted a value larger than the 0.6 or 0.7 that astronomers measured. A lot larger. Ten-to-the-power-of-120 larger. That's of 1,000,000,000,000, 000,000,000,000,000,000,000,000,000,000,000,000,000,000,000, 000,000,000,000,000,000,000,000,000,000,000,000,000,000,000, 000,000,000,000,000,000 times larger. As the joke went, even for cosmology that's a big discrepancy. The stretching of space under the influence of such a ridiculously large energy density would be so extreme that, as Turner said, "you wouldn't be able to see the end of your nose." Not that the universe would be here for your nose to have an end on: A density that high would have cooled the cosmic microwave background below 3 K in the first 1/100,000,000,000,000,000, 000,000,000,000,000,000,000,000th of a second after the Big Bang. So given the choice between an energy density with a value of 10120 and one with a value of 0.7, most particle physicists would have been perfectly content to assume that somehow someone someday would manipulate the math, or figure out how particles were annihilating one another in just the right proportion, in order to make the result be what everyone had always been perfectly content to assume it was: A = 0.

Skeptics liked to quote a saying: "You get to invoke the tooth fairy only once"—meaning dark matter—"but now we have to invoke the tooth fairy twice"—meaning dark energy. An epithet became inescapable, or at least a commonplace on the conference circuit: epicycles. Were astronomers and their inflation-theorist enablers simply saving the appearances, like the ancients and their desperate measures to make the math correspond to the motions in the heavens?

One could cite modern precedents. Scientists in the nineteenth century figured that the phenomenon of waves of light propagating across space like waves across water didn't make sense unless they inferred the presence of a cosmic pond. The Scottish physicist William Thomson, eventually Lord Kelvin, spent the entirety of his career trying to find equations to describe this "ether." In 1896, on the occasion of the golden jubilee of his service to the University of Glasgow, he wrote to a friend, "I have not had a moment's peace or happiness in respect to electromagnetic theory since Nov. 28, 1846." He died in 1907, two years after Einstein established the theory of special relativity by eliminating the need for absolute space, thereby making the ether "superfluous."

Would future generations look on the whole of modern cosmology as a similar lesson in the limitations of inferences from indirect evidence? The motions of galaxies didn't make sense unless we inferred the existence of dark matter. The luminosities of supernovae didn't make sense unless we inferred the existence of dark energy. Inference can be a powerful tool: An apple falls to the ground, and we infer gravity. But it can also be an incomplete tool: Gravity is...?

Dark matter is...?

Dark energy is...?

Astronomers might not have been able to identify dark energy, but some theorists knew what it was: an inference too far. Just because a positive lambda would solve many problems didn't mean it existed.

"You observational astronomers," a theorist told Alex Filippenko in 1998, "are wasting a lot of valuable Keck and Hubble time, because your result must be wrong. We have no theory that could be compatible with a tiny non-zero vacuum energy"—tiny in the sense that lambda would be equal to 0.6 or 0.7 of critical density, rather than 10120—"and there's no theory that could possibly be compatible with this."

"Look," Filippenko said, "this is an observational result. I only know what end of the telescope to look through. You're a lot smarter than I am. But with additional observations, we will either confirm this, or we will find that we were wrong—hopefully for some subtle reason, and not '2 plus 2 equals 5' in some computer program."

In other words: Just because a positive lambda created a problem didn't mean it didn't exist.

In the end, sociology—the fact that two intense rivals had independently reached the same surprising result—wasn't going to be enough to convert the skeptics or, for that matter, to convince appropriately cautious astronomers that they weren't fooling themselves. Neither would aesthetics—whether the result solved problems or created problems. Not even the honor of being Science's "Breakthrough of the Year" for 1998. Filippenko's point was that only science, only further observations, could test a positive value for lambda.

And so astronomers did what scientists do in such circumstances: They set out to prove that the effect didn't exist. What problems might they have overlooked that could cause distant supernovae to appear dimmer than they should? Two possibilities immediately presented themselves.

One was an exotic kind of dust. Astronomers knew that regular dust within galaxies makes the light redder, and they knew how to correct for that dust—thanks in large part to Riess. His paper on the MLCS—multicolor light-curve shapes—correction method for dust won the 1999 Trumpler Award, an honor that recognizes a recent PhD thesis of unusual importance to astronomy. But now astronomers were mentioning the possibility of gray dust, and positing its presence between galaxies.

"Nobody has ever seen gray dust between galaxies," Riess thought. "But," he reminded himself, "nobody has ever seen a cosmological constant either."

Or what if the unusually faint appearance of supernovae at great distances was the result of supernovae being different back then, when the universe was younger and less complicated? What if the nature of Type Ia supernovae had changed over the life of the universe, and the recipe for a relatively nearby supernova was different from the recipe for a distant supernova? Maybe more distant supernovae had a simpler cocktail of elements, making them intrinsically fainter and giving the illusion that they were more distant.

There was one way to find out. If the interpretation of the supernova evidence was correct, then we were living at a time when dark energy was dominant over matter; the anti-gravitational force of dark energy was winning a tug of war with the gravitational force of matter. In that case, the expansion of the universe would be accelerating, and, as the two teams found, distant supernovae would appear dimmer than we would expect.

In earlier eras, however, the universe would have been smaller and therefore denser. The earlier the era, the smaller and denser the universe; the denser the universe, the greater the cumulative gravitational influence of matter. If astronomers could see far enough across the universe—far enough back in time—they would reach an era when dark matter was dominant. At that point, the gravitational influence of dark matter would have been winning the tug of war with the anti-gravitational force of dark energy. The expansion would have been decelerating, and supernovae from that era would therefore appear brighter than we would expect.

Not so the supernovae that we would see through gray dust, or that had a simpler cocktail of elements in the early universe. Those supernovae would just keep appearing fainter and fainter, the farther and farther we looked.

To distinguish between the two scenarios—dark energy versus either gray dust or changing cocktail—you would need to observe a supernova distant enough that it had exploded during that far earlier, far more distant era. You would need a supernova that had exploded before the expansion of the universe "turned over"—before the universe had made the transition from deceleration to acceleration, back when matter, not energy, was winning the tug of war. You would expect that supernova to be brighter than it "should" be. Plot it on the Hubble diagram—way out there, far beyond the nearby supernovae from Calán/Tololo, beyond the high-redshift supernovae that the two teams had discovered—and the slight upward deviation from the 45-degree straight line that High-z and the SCP had graphed would "turn over," too, just like the universe. It would dip down.

And if it didn't, you'd have to rethink dark energy.

Ground-based telescopes, however, couldn't see that far across the universe. The Hubble Space Telescope could, and it could even discover supernovae at that distance. From December 23 to 27, 1997, Ron Gilliland and Mark Phillips had used HST to try to prove that you could do just that—detect supernovae from the earliest epochs of the universe. For their search, they chose a familiar, even famous, speck of sky: the Hubble Deep Field. Two years earlier, in 1995, HST had made the most distant image of the universe. For ten days the telescope had drilled a hole in the sky the size of a grain of sand at arm's length, just soaking up photons, seeing deeper and deeper across space and therefore farther and farther back in time. In the end the Hubble Deep Field contained about three thousand galaxies, some faint blue and among the first in the universe. Gilliland and Phillips wanted to make a repeat visit and do what supernova hunters had been doing since the 1930s—compare the earlier images with a current image and see what had changed. Did any of the galaxies in 1997 contain a speck of light—a supernova—that hadn't been there two years earlier?

Two did. Those specks got the designations SN 1997ff and SN 1997fg. Without follow-up observations, Gilliland and Phillips couldn't do the photometry that would allow them to construct light curves. But they'd made their point. You could use HST to discover supernovae at distances inaccessible from telescopes on earth.

What you couldn't do, however, was what astronomers needed to do in order to test dark energy: make multiple reference images, then return to the same field in weeks to come in the hope of discovering the most distant supernova yet, for which you would have already reserved time for follow-up observations in the weeks and months ahead. You couldn't guarantee that you wouldn't be wasting HST time.

Still, those two supernovae—SN 1997ff and SN 1997fg—bugged Adam Riess. He couldn't stop thinking about them. By 2001, while remaining a member of the High-z collaboration, he was a staff scientist at the Space Telescope Science Institute, so HST results and possibilities were always on his mind. But SN 1997ff and SN 1997fg were a particularly poignant reminder of a lost opportunity. They were at a distance sufficient to test the deceleration-before-acceleration period of the dark-energy cosmological model—when the expansion was still slowing down under the dominant influence of dark matter, rather than speeding up under the dominant influence of dark energy. If only Gilliland and Phillips had been able to do follow-up work on SN 1997ff or SN 1997fg, Riess thought, astronomy would already have been able to put dark energy to a particularly compelling test.

In early 2001, Riess realized he could reframe the question.

What if one of those supernovae had been followed up? Not deliberately, by Gilliland and Phillips, but serendipitously, by HST during some other observation?

So he called up the HST search page on his office computer. He typed in the coordinates. Right ascension: 12h36m44s.11. Declination: + 62012'44".8. He requested dates that would correspond to the period during which SN 1997ff and SN 1997fg would have brightened and dimmed: December 27, 1997, to April 1, 1998.

Riess understood that the possibility of his finding what he wanted was extremely remote. HST looked at a lot of space; what were the chances that it had been staring at a particular dot of deep space during a particular period of time?

"Nobody's that lucky," he told himself.

Adam Riess was that lucky.

Earlier in 1997, Space Shuttle astronauts had added a couple of instruments to HST, including the Near Infrared Camera and Multi-Object Spectrometer, or NICMOS. By seeing in the infrared, NICMOS was particularly sensitive to distant objects whose light was so redshifted that, by the time it reached our patch of the universe, it had left the visible part of the electromagnetic spectrum. The NICMOS team had decided to test their instrument on a particularly distant patch of space with features they could easily identify: the Hubble Deep Field.

The observation program didn't begin until January 19, but the camera took some test images in the interim. And there it was: 1997ff. It was in the HST archives of the NICMOS test run, on December 26, January 2, January 6. Once NICMOS started taking data for real, 1997ff appeared in frame after frame, right at the edge. Sometimes it fell off the edge. But usually it was there. Riess spent the early part of 2001 examining SN 1977ff, establishing from the redshift that it had exploded about 10.2 billion years ago—far earlier than the period when the expansion of the universe would have gone from slowing down to speeding up. If indeed the universe had gone from deceleration to acceleration. If indeed dark energy existed.

Every spring the Space Telescope Science Institute hosted a symposium. One previous symposium topic had been the Hubble Deep Field; another had been stellar evolution. The topic in 2001 happened to be "The Dark Universe: Matter, Energy and Gravity." It was a chance for more than a hundred astronomers from around the world to reflect on their seemingly oxymoronic mission—what the symposium organizer and astrophysicist Mario Livio called "astronomy of the invisible."

STScI occupied a low, modern—and somewhat modest, considering its NASA provenance—building on a winding road in a far corner of the Johns Hopkins University campus in Baltimore.* It looked as if it were ducking its head so it could fit under the trees. At the rear of the building, outside the door to the auditorium on the first floor, was a wall of glass overlooking a creek. No Fermilab-style stampedes-around-the-accelerator-track-followed-by-a-barbecue-for-hundreds here; wine and cheese or the occasional overcaffeination was as wild as an STScI meeting would get.

By now, even the most ardent dark-energy skeptics had learned to accommodate the findings from a series of balloon experiments that had been launched from the outskirts of Antarctica and the Atacama Desert in Chile. The balloons had floated to an altitude of 100,000 feet and scraped the underbelly of outer space, at which point the on-board detectors had surveyed the cosmic microwave background. The goal was to refine COBE's measurements of the differences in temperature between points on the sky. If the differences in temperature were greatest between points separated by less than 1 degree, then the universe was open; by more than 1 degree, then the universe was closed; by 1 degree, then the universe was flat. So far, the verdict was all flat.

But saying that the universe sure looked flat wasn't quite the same as saying that the expansion of the universe was accelerating. You couldn't rely on an argument by subtraction—an omega of 1 minus a mass density of 0.3 equals a lambda of 0.7. That math showed only the same seeming paradox that had existed pre-1998: an apparently flat universe via COBE, an apparently open universe via other observations. The balloon experiments made COBE's flat universe much more compelling, to the point that a flat universe was quickly becoming cosmological orthodoxy. But acceleration? Especially if you were a particle physicist, that result still didn't make sense—still left you pining for alternatives.

Among the observers in attendance was Vera Rubin, opening the conference with a historical overview of dark matter—or, actually, a historical overview of the idea of dark matter, since, as she pointed out, until you know what dark matter is, you can't really know its history. She recalled predicting in 1980 the discovery of dark matter within ten years, and she said she was amused to see the British astronomer Martin Rees recently making the same prediction. She said she knew what Fritz Zwicky would have said about the current state of cosmology: "Epicycles!"

Among the theorists in attendance was Michael Turner, exhorting the congregation to indulge in "irrational exuberance" and embrace the era of "precision cosmology." To a fellow theorist complaining about the 10 problem, Turner responded with exasperation: "Can't we be exuberant for a while?"

Saul Perlmutter was there too, talking up the possibilities of a space telescope dedicated to supernovae, and a couple of dozen other presenters were there to promote their own prospective research projects and report on their latest observations and postulate extravagant possibilities about the identity of dark energy. But mostly everybody was there to try to answer the question that Mario Livio had written on a transparency for his talk summarizing the symposium: "Accelerating Universe—Do We Believe It?"

Which is why it was Adam Riess who stole the show.

He choreographed his presentation, on the third day of the four-day conference, as a striptease. He had to do something to spice it up, since everybody in the auditorium knew what he would be showing. Two days earlier, on the first day of the symposium, Riess had attended a NASA press conference in Washington to announce his discovery. And one day earlier, that announcement had made the front page of the New York Times as well as other newspapers around the world. Still, now was his chance to let his fellow cosmologists examine this new evidence for themselves.

He would be using an overhead transparency. He kept most of it covered at first, while he explained what he would be showing. It was a Hubble diagram—redshift against brightness—of the supernovae from both the SCP and High-z teams. The points in this case represented not individual supernovae but averages of supernovae at similar redshifts.

Riess revealed the first three dots on the transparency: here, and here, and here, the averages of the nearby supernovae from the Calán/ Tololo survey.

Then, moving to the right, the next three dots: here, and here, and here, the averages of the distant supernovae from the SCP and High-z searches.

The dots were beginning to describe the now-familiar gentle departure from the straight line, the upward turn toward the dimmer. In six dots Riess had taken his audience from a few hundred million light-years across the universe, to a billion, then two billion, three, four. Now, he said, he had the point that represented SN 1997ff. He had determined its redshift to be about 1.7, the farthest supernova to date by a long shot, a distance of about eleven billion light-years.

They knew what they were going to see, but the hundred or so astronomers in the auditorium couldn't help themselves. They shifted in their seats. Leaned forward. Held back. Crossed arms.

There: SN 1997ff.

A gasp.

The gentle upward curve was gone. In its place was a sharp downward pivot. The supernova was twice as bright as you would naively expect it to be at that distance. The universe had turned over, all right.

While Riess went on to explain that the result ruled out the hypothetical effects of exotic gray dust or a change in the nature of supernovae at a confidence level greater than 99.99 percent, the evidence continued to loom on the screen behind him. His audience couldn't take their eyes off it. For the astronomers of the invisible, it was something to see.