Five Billion Years of Solitude (10 page)

On September 29, 2010, Vogt, Butler, and four other collaborators declared they had beaten those odds. Their announcement coincided, perhaps not by chance, with Marcy’s fifty-sixth birthday—a final spiteful gift from former friends. By combining their own old HIRES observations with publicly available HARPS data from the Swiss, Vogt and Butler claimed the RV detection of two planets around the red dwarf star Gliese 581, located some 20 light-years away in the constellation of Libra. One of the planets was between three and four Earth masses, in a 37-day orbit that placed it squarely in the center of the habitable zone. According to convention, its official name was Gliese 581g—the “g” denoted that the world was the sixth planet discovered around that particular star, though it could have equally well stood for “Goldilocks,” since the planet was potentially rocky, and in an orbit very clearly neither too hot nor too cold for life as we know it. Vogt preferred another appellation. He called it “Zarmina’s World,” after his wife, and said in a press conference that he believed the chances for life on the planet were “one hundred percent.” Butler opted for the more conservative statement that “the planet is the right distance from the star to have water and the right mass to hold an atmosphere.”

The announcement left the Swiss team collectively scratching their heads. They had previously found four small planets orbiting Gliese 581, including two borderline-habitable worlds hugging each end of the star’s habitable zone. How had they missed two more planets? When they expanded and reanalyzed their own HARPS observations of the star, they readily confirmed once again the four planets their team had previously detected, but there was no signal of Zarmina’s World or of the other potential planet, Gliese 581f. They argued that Vogt’s less-precise HIRES data had introduced phantom worlds into their higher-quality HARPS observations of Gliese 581. Multiple independent examinations of the public HARPS and HIRES data came to divergent conclusions, with some finding evidence for the new planets and others dismissing them, depending on a shifting host of assumptions. If a certain variety of statistical analysis was used to extract
the RV signals, only four emerged, rather than six—the new planets were false alarms! In their dynamical simulations, Vogt and Butler changed the orbits of the planets to nearly circular rather than moderately elongated, and found the results more stable with six worlds than with four—the new planets were real! But, adding up the various ways that the orbits of the four confirmed planets could interact, it appeared they could partially obscure and dilute RV signals of the two disputed worlds.

With uncertainty clouding the case of Gliese 581g, the prize of finding the first undisputed terrestrial world in a habitable zone remained up for grabs. In July of 2012, Vogt, Butler, and another colleague released a rebuttal to their critics, a two-steps forward, one-step back reanalysis of the HARPS data that scarcely mentioned Gliese 581f and modified Zarmina’s World into a smaller 2.2-Earth-mass planet in a shorter 32-day orbit, still within the habitable zone. The potential world, they wrote, had approximately a 4 percent chance of being illusory, though since its signal was so weak much more data would be required for an airtight confirmation. To the Swiss, a 4 percent chance was still too high to bequeath confirmation upon Zarmina’s World. Even a 1 percent chance would be considered borderline for such a planet. Extraordinary claims, they argued, required extraordinary evidence—evidence, they tacitly suggested, that would only come from a spectrometer at least as precise as HARPS.

In the summer of 2011, one of Butler’s occasional collaborators, a thirty-two-year-old Spanish astronomer named Guillem Anglada-Escudé, began developing his own alternative data-analysis software to derive RV signals from HARPS spectral measurements, which the Swiss made public after a two-year proprietary period in accordance with ESO policies. Unlike the analytic methods of the Swiss, which threw away a significant amount of a star’s raw spectrum, Anglada’s software harvested a greater portion of a star’s spectral data, extracting more signal from the noise to further boost RV precision for certain varieties of stars—particularly red dwarfs. Soon he was running his code on samples of the HARPS data, hoping to find borderline planetary
signals that the less-precise Swiss analyses might have missed. His postdoc at Carnegie was almost finished, and he was looking for another job in the field—he thought a few planets under his belt could only help his chances. The first dozen data sets he examined came up empty of any new signals.

Late one August evening, after interviewing for a postdoc position at the University of Göttingen in Germany, Anglada returned to his hotel room and looked at another batch of HARPS data, 143 RV measurements taken between 2004 and 2008 for the red dwarf star GJ 667C, part of a triple star system some 22 light-years away from Earth. He fed his reduction of the data into Laughlin’s Systemic software, and waited as the program looked for patterns. It first found the signature of a planet in a 7-day orbit that the Swiss had announced in 2009, but Anglada could see what looked like residual structure in the pointillistic clusters of measurements. He ran the data through Systemic again, and the software found a strong 91-day trend in the data—another possible planet, but also perhaps a cyclic stellar pulsation related to the star’s estimated rotational period of 105 days. Anglada ran Systemic once more, nulling out the 7- and 91-day signals, then with trembling hands lit a cigarette and stared in disbelief at his laptop’s glowing screen. Another sine wave snaked through the measurements, looking for all the world like the signal of a 4.5-Earth-mass planet, likely terrestrial, in a 28-day orbit firmly within GJ 667C’s habitable zone. If the planet proved to be real, it would be assigned the name GJ 667Cc.

“It was very strange to find an unpublished, unclaimed, potentially habitable planet in a three-year-old public dataset,” Anglada recalled to me. “So I looked again at the measurements using [the Swiss] method—the 28-day signal was there, but with a false-alarm probability that looked to be substantially greater than one percent.” Too high, that is, to cross the HARPS team’s traditionally ultra-stringent thresholds for announcing a discovery. Anglada’s analysis of the 28-day signal, by contrast, yielded a false-alarm probability of three hundredths of 1 percent. He shared his findings with Butler, who excitedly agreed to gather more
data on the star. Butler obtained twenty new RV measurements for GJ 667C, and Vogt supplied twenty older measurements from Keck’s HIRES archives, both of which strengthened the 28-day signal. The team had soon modeled the putative system to examine its dynamical stability, and began drawing up a paper to announce the discovery. In the meantime, Anglada had decided he should bolster his case by seeking new data from HARPS, at the time still the best source in the world. Against the warnings of Butler and Vogt, who did not trust the HARPS team and urged him to publish with the data in hand, on September 28, 2011, Anglada submitted a proposal to ESO for twenty nights of HARPS observing time. The proposal did not explicitly announce Anglada’s potential discovery but included GJ 667C on its short target list, as well as a figure discussing the star’s 7-day, 91-day, and 28-day signals.

After submitting, Anglada closely watched the incoming traffic to his personal website, reasoning that he could gauge when his proposal was examined by looking for members of the ESO review committee who would visit to examine his credentials. In mid-November, his site received a pulse of traffic associated with computers in Munich, where the ESO review committee was based, as well as from computers in Geneva, Porto, Paris, and Santiago—all cities hosting HARPS team members. Each visitor browsed for minutes, then departed and did not return. On November 21, the HARPS team uploaded a 77-page preprint to a public online repository. The preprint was a draft version of a paper submitted to a prominent peer-reviewed journal, summarizing six years of HARPS observations, from 2003 to 2009. The astronomer Xavier Bonfils, a senior HARPS team member, was the first author. In a table on page three and a paragraph on page eight, the team noted its detection of a super-Earth in a 28-day orbit around GJ 667C, and referred interested readers to a more detailed forthcoming paper that was in the midst of preparation. Later, Anglada would learn his HARPS proposal had been rejected.

Vogt was the first to see the Swiss preprint claiming discovery of GJ 667Cc. He immediately sent a terse e-mail to Anglada and Butler:
“We’ve been scooped.” Anglada was crestfallen; he read the preprint, then left his Carnegie office for a long walk. That night he couldn’t sleep.

“I was very upset,” he recalled. “So I reread the preprint again and started cataloging strange things. It didn’t include any detailed analysis of the system’s dynamics, and it barely discussed the 91-day signal. It said that the evidence for GJ 667Cc would be presented in a paper that was in preparation that would be submitted in future, yet it also said this paper had already announced GJ 667Cc—it was a self-contradictory statement that didn’t seem to be a proper way to formalize the discovery.” Looking at the table listing the planet’s orbital parameters, Anglada noticed something odd. The table listed GJ 667Cc’s orbital period as 28 days, but the size of its listed orbit erroneously corresponded to that of an orbit of 91 days, as if GJ 667Cc’s entry had at one time concerned the 91-day rather than the 28-day signal.

“It could all have been coincidence,” Anglada told me. “But I couldn’t help feeling suspicious. If they had seen this signal in their data back in 2008, why did they wait three years, only to announce the planet in such a curious way the week after my HARPS proposal was reviewed? Why did the orbital size and period not match? I started to feel angry, and decided I should go on and push ahead with what I had found.” Within a week, Anglada had completed his paper and submitted it to the
Astrophysical Journal Letters
, which published it in February 2012—beating the HARPS team to peer-reviewed publication. UC Santa Cruz issued a press release crediting Anglada, Butler, Vogt, and their collaborators with GJ 667Cc’s discovery.

Bonfils and the rest of the HARPS team were aghast. They were the planet’s true discoverers, they argued, as established by their November preprint. The controversy remained unaddressed until June, when Anglada and Bonfils agreed to a private meeting at a coffee shop outside a conference in Barcelona. Bonfils told Anglada that the HARPS team had already known of GJ 667Cc back in 2009 when they had announced their discovery of the system’s other world in a 7-day
orbit. They had submitted their 77-page survey paper for peer-reviewed publication in April of 2009, Bonfils said, but feedback from one of the reviewers had delayed the preprint’s public release until November 2011. Anglada replied that the timing of the preprint was irrelevant, because it did not contain enough detailed information to support the HARPS team’s discovery claims. Anyone could report wobbling stars, but to prove the wobbles were planets they had to show their analytical work. If published analysis was the test, Bonfils countered, then it was one Anglada’s paper had still failed, because it contained the same mistake Butler and Vogt had made with the Gliese 581 system. Pooling HARPS data with that from lesser spectrometers such as HIRES, Bonfils maintained, would degrade the RV data and only increase the likelihood of false alarms; in contrast, the HARPS team’s preprint was a valid discovery paper. By the time they finished their coffees, neither man had given any ground, and the tension between them had only grown.

I reached Bonfils by telephone a month after his meeting with Anglada. He sounded pained.

“They are trying to take credit for a discovery they did not make. It is as basic as that. It’s not by hazard that we found this planet—it was on purpose. GJ 667C is one of our survey’s most-sampled stars. That’s why [Anglada] looked at it. HARPS was built by our team, and the scientific program and observations were done by our team. Most of the data reduction was already done and provided in our public data. I think it would be a pity if the guys who made the instruments and designed and performed the program of observations did not receive credit for their work. I’m a supporter of public data, but I had long feared someone would try to publish our data before us, and it has now happened. Right now this community rests on good behavior and gentlemen’s agreements.”

Nothing, nothing at all, Bonfils insisted, had precipitated the preprint’s release in November 2011 after years of delay. It had been simple
serendipity. “It was slow,” he acknowledged. “I’m not proud of how long it took.” He offered his own view of what had motivated the fight over public HARPS data. “Before, it was Marcy, Butler, Vogt, Fischer. Now they have split, the group has almost evaporated. I don’t know Vogt or Butler personally, I have met Anglada once only. But I think there is, how to say, tension? You see it in their papers, in the language and accusations. A hunger, an aggression. I think it has become more difficult for them to get the observing time they need.”

For the time being, Anglada told me, he had no plans to stop examining the public HARPS data for new discoveries. “People seem to think GJ 667Cc was a one-off thing, or that I got lucky looking where I did, but that’s not true,” he said. “This is really just the beginning of a bigger story. When you improve precision as I have, more things appear. The population of exoplanets is growing exponentially as we become sensitive to lower masses. I’ve looked at hundreds of systems now in their database. A lot of objects are showing up.”

Despite working closely with Vogt and Butler, Laughlin had managed to stay above the battle for the first potentially habitable exoplanets. He had not been directly involved with the detections, announcements, and criticisms of Gliese 581g or GJ 667Cc, and he preferred to keep it that way. He held a longer view on the controversy surrounding them. In his mind, the strife between teams and the explosive expansion of exoplanetology were just growing pains, symptoms of a field struggling with its own imminent maturity.