Mirror Earth (21 page)

If
it's there, he said. The problem, as he well knew by the time we spoke, was that many exoplaneteers had become convinced that it wasn't. “The first principle,” Richard Feynman said, “is that you must not fool yourself—and you are the easiest person to fool.” Peter van de Kamp had fooled himself in the 1960s into thinking there was a planet orbiting Barnard's Star and never admitted he was wrong. Andrew Lyne fooled himself into thinking he'd found planets orbiting a pulsar, but salvaged his reputation by discovering the mistake himself. Now, even without Vogt's hasty remark about life, a consensus was rapidly developing that Vogt and Butler had probably fooled themselves. “There are e-mails flying around. People are frankly aghast,” one exoplaneteer told me. “When I read their paper,” said another, “red flags popped up. I was literally blown away. I couldn't believe it.”

The reason, this anonymous astronomer said, was that Debra Fischer had written a definitive paper a couple of years earlier outlining the standard technique for extracting the signals of multiple planets from a complex jumble of radial-velocity measurements. “The methodology was very clear,” he said, “and the pitfalls you have to watch out for were carefully outlined.” Butler and Vogt were two of Fischer's co-authors on that paper.
They had even gone on to debunk another planet claim using that same standard technique. “Now here we are,” said the exoplaneteer with the red flags, “and Butler and Vogt have ignored the Fischer paper. They're using techniques that have proven to be unreliable and actually dangerous to use.”

All of these doubts were confined to the planet-hunting community until about two weeks after the press conference. Then Francesco Pepe, a member of Mayor's team, spoke up at a conference in Turin, Italy. The Swiss group had gone back and reanalyzed their own observations of Gliese 581 with the HARPS spectrograph, folded in the observations Vogt and Butler had made with the High Resolution Echelle Spectrometer Vogt had built for the Keck Observatory, and found … nothing. “This does not prove there is no planet,” Pepe told me. “All it shows is that we cannot see the signal. It's really a matter of how you interpret the data, how much confidence you have. In this case, we think there has been a mathematical misinterpretation.” An American exoplaneteer was much less charitable. “This is not just a wrong result, this is a result from somebody who has gone off some deep end, and, what do you do? It's not to the advantage of the scientists who know the story to tell it. There is no value in it. It looks bad. Everybody loses, everybody from the scientists to the funding agencies to the individual scientists themselves loses if it looks like there was shoddy work being done.”

There was a deeper issue, however, than whether Vogt and Butler had used the right statistical techniques. In 2007, after more than fifteen years of collaboration and an extraordinary track record of discovering hundreds of planets together,
Marcy and Butler had gone their separate ways, with Steve Vogt joining Butler in an independent venture they called the Lick-Carnegie Exoplanet Survey. This sort of realignment of research teams is often perfectly amicable. In this case, however, according to people familiar with the situation, it wasn't. Nobody involved in what astronomers have described as a “horrible breakup, like the worst divorce” was willing to talk about it on the record, but it was evident from a number of conversations that the major problem was clashing egos. Butler and Vogt were always seen as junior members of the team, with Marcy getting most of the attention, especially from the press.

There was, in short, no love lost between Marcy on the one hand and Butler and Vogt on the other. For that reason, Marcy and his current collaborators had to be careful about piling on to their former partners. “The intensity of what's going on is very harsh and negative,” Debra Fischer told me. “These guys have a track record better than what the blogs are suggesting, but the penalty for being wrong in this business is extraordinary.” Still, she said, “if Geoff or I say anything about Gliese 581 g, people might just put it down to the ugly breakup and bad karma between the two groups.” When word got around that John Johnson, a new member of Marcy's collaboration, was writing a paper debunking Gliese 581 g, Fischer told me, “I had a Skype session with him telling him he shouldn't do it.” Any takedown of the Vogt-Butler discovery would better come from outside the radial-velocity community entirely. It should be done instead, she said, by astronomers who specialized in statistics, like Eric Ford, of the University of Florida,
or Phil Gregory, at the University of British Columbia. Mindful of the overlapping rivalries between the Swiss, Marcy, and Vogt-Butler teams, I realized I needed to go outside that community myself for an independent assessment. When I asked Steve Vogt whom he trusted for such an assessment, he, too, brought up Ford.

Eric Ford was a graduate student at Princeton in the early 2000s, a young, cheerful guy you probably wouldn't have pegged as someone who would become prominent. Ed Turner, a senior astrophysicist at Princeton, knew better. “It was quite obvious that he was exceptional even then,” Turner said. Ford's primary research at Princeton was on something called the Space Interferometry Mission, or SIM. It was a complex space telescope NASA was cooking up to try to measure the side-to-side wobbles a planet would induce in its star if a solar system presented itself to us face-on. It was these side-to-side, or astrometric, wobbles that Peter van de Kamp thought he saw in Barnard's Star, and which NASA had talked about searching for before Bill Borucki and Geoff Marcy began their own planet-detection projects. “Probably two thirds of my thesis was on thinking about how you'd design a planet search for SIM,” Ford recalled. Ford's thesis adviser was Scott Tremaine, the head of Princeton's astrophysics department at the time, and later John Bahcall's successor at the Institute for Advanced Study across town. Unfortunately for Ford, Tremaine went on sabbatical for a year in the middle of the research. “It was a sort of scary experience,” Ford said. “Scott would sort of give you some ideas. You'd start to work on them, and suddenly he's off … I think he was hiking the
Adirondacks or something … some mountain range, I forget. And you sort of get to a point where like, okay, I'm really not making that much more progress. Instead of beating my head against a wall, maybe I'll start something new. So, I kind of made up something.”

Six months later, Tremaine came back. “I go, ‘Well, hi, Scott, I've been working on this other project. I hope you like it.' It made me a little bit nervous,” said Ford, “but fortunately, he did.” This other project was an attempt to come up with new statistical methods to analyze radial-velocity signals from multiple-exoplanet solar systems—exactly the problem Debra Fischer had solved for the Upsilon Andromedae system, and exactly the problem, except far more complicated, with Gliese 581. “So, it actually worked out amazingly well. Sometimes, either a little bit of luck or sort of looking for something where you say, ‘What can I do that's different, where is the unique opportunity for me in this field?' Amazingly, it sometimes pans out.”

By the time Vogt and Butler made their announcement, Ford was still young, energetic, and cheerful, but now he was also verging on eminence. When I asked him about the controversy over Gliese 581 g, he asked me in turn whether this was for a news story or for something more substantial. If it was for a story, he didn't want to comment. “One single quote is going to be misleading no matter what I say,” he told me. I reassured him that it wasn't for a news story, and he went ahead.

“Okay, we're human,” he began, carefully and gently. “We make mistakes. Sometimes we're under pressure to get results so we have a better chance of getting grants.” In principle, he
thought the paper's referees—the scientists a journal consults to see if a result is worth publishing—should have been skeptical about the Vogt-Butler paper all along. “But you can get a busy referee,” said Ford, “or someone who doesn't have the relevant knowledge or experience to evaluate it properly.” If the claim is boring, nobody might ever notice. “But if you make an exciting claim, you can be pretty sure someone will check on it.”

Ford paused. Then he said, choosing his words very carefully, “I think the current data do not support the claim for this planet to a level of significance that meets the standard. I suspect that the analysis was not done as carefully as it could have been. I was skeptical as soon as I read it. It's a little bit unfortunate, because I'm obviously aware of the history [of the Marcy-Butler-Vogt breakup]. It may be that they're good scientists and do good work, but, in this case there was one part of the analysis where maybe they weren't as strong as when they were part of a larger team.” But whatever the merits of Vogt and Butler's claims, he said, “I'm confident with time, science will figure out what's going on. It's a fascinating system whether or not this planet exists. The process will play itself out and come to a consensus.”

Six months later, Phil Gregory, the second statistical expert Fischer had invoked, took a new set of observations from the European HARPS spectrograph, added them to all known observations of Gliese 581 from both the Europeans and the Americans, and analyzed them all. “I don't find anything,” he said in an interview with
Wired
magazine. “My analysis does not want to lock on to anything around 36 days [the period
claimed by Vogt and Butler for Gliese 581 g]. I find there's just no feature there.”

At last report, Vogt was still telling me, as he had from the beginning, that he'd give up on the planet if the evidence went against him. But nothing he'd heard yet, including Gregory's analysis, was enough to convince him. Most astronomers now believe that Gliese 581 g was nothing but a mistake.

Vogt and Butler didn't seem to care. They were standing by their planet.

By June 2010, even though Kepler had been in orbit for over a year, the team had announced only a handful of planets—five in its first presentation, nearly a year after launch, and just two more in the months afterward. But that didn't mean the satellite wasn't seeing anything. Plenty of the stars in Kepler's field of view were flickering regularly, and even when the false positives were thrown out, the pipeline had spit out 706 planet candidates in the first three months of observation alone. The Kepler team had released 306 of these in June 2010—not fully confirmed planets, but candidates they were pretty sure of. With NASA's permission, they held back the remaining 400 for further analysis.

They needed permission because, as a publicly funded mission, Kepler results had to be made public as soon as possible. The Kepler team got first crack, but eventually they had to be made freely available to other astronomers who could comb the data for discoveries of their own. “Obviously,” said Geoff Marcy of the sequestered four hundred objects, “we have good candidates, and you can bet that the implications of some of
them are profound. But if you say something publicly, it had better be unassailable. This is an unprecedented technical challenge, and the people at NASA Ames are working their butts off eighteen hours a day to make sure we get it right.”

It's not as though the 306 that did get released were boring. The paper Borucki and the Kepler team published was titled “Characteristics of Kepler Planetary Candidates Based on the First Data Set: The Majority Are Found to Be Neptune-Size and Smaller.” For Geoff Marcy, this was pretty mind-blowing. “Up to 1995,” he wrote in an e-mail, “I thought we might not live to detect even one planet, of any size. Incredibly, the past 15 years brought us over 400 planets, painstakingly found one by one, most being giants like Jupiter and Saturn, with a few Neptune-size ones. Now these early Kepler returns hint strongly of a vast reservoir of ever smaller planets, approaching a few times the size of Earth, if not smaller—over 300 of these planets, each with their coordinates in space. It's a planetary treasure map. You can point the Keck and Hubble at them, or even your home telescope. There is only one appropriate assessment of this technical accomplishment and the haul of prospective small planets: WOW!”

The deal Borucki and Batalha and the rest had worked out with NASA headquarters was that they'd release the four hundred sequestered planet candidates, and maybe more, at a press conference around February 1, 2011—no backing out this time. This didn't mean, however, that there wouldn't be any more news from Kepler until then. Within just two months of the June data release, a paper appeared in
Science
announcing two Kepler planets that had been confirmed in an entirely
novel way. It was so novel, in fact, that nobody on the Kepler team had even considered it when putting the mission together.

The most reliable way to move an object from the “planet candidate” column into the “planet” column was to detect a radial-velocity wobble, with the same period as the transit Kepler had seen. The two techniques were entirely independent, and no conceivable false positive could show up in both kinds of measurement. The added benefit, as Brown, Charbonneau, Marcy, and Henry had demonstrated with the first transit detection of HD 209458 b, was that by combining the two you could calculate the star's density. But most of the stars in the Kepler field were too faint to make radial-velocity measurements possible in any sort of reasonable time, even with a huge telescope like the Keck, in Hawaii.

The detection of HD 209458 b, however, had gotten a handful of exoplaneteers thinking—not about a new planet detection technique at first, but instead about a mystery concerning this first transiting planet from back in 1999. When both Marcy's and Brown's teams calculated the density of HD 209458 b, they found it was lower than they would have expected. The planet, a gas giant like Jupiter, appeared to be puffed up in size. “If I recall correctly,” said Matt Holman, an astrophysicist at the Harvard-Smithsonian Center for Astrophysics and the lead author of the
Science
paper, one day in his Cambridge office, “there was a paper at the time, exploring how to make the planet bigger than the model would predict.” One idea was that the planet had an eccentric, elongated orbit. That would create tidal forces that would squeeze and stretch the planet,
heating up the interior and making the whole thing expand. Since the tides would also tend to damp out eccentricity and circularize the orbit, something had to be interfering with that process—the gravity of a second, still-unobserved planet in the system, for example.