Mirror Earth (15 page)

If there are two planets orbiting the star, it gets more complicated. There are two radial-velocity curves now, with different strengths and periods, but they're superimposed. Each measurement gives you the combined effects of both planets' gravity on the star. Fischer appreciated this, but even so, she said, “it was horrible. It wasn't coming out right at all.” Finally, she took the two curves that best fit the data, bad as they were, and subtracted them from the overall signal. When she did that, she recalled, “I saw the data doing this unbelievable extra sine curve. It sent chills down my spine. It looked like there
was a
third
planet in there, and no one had expected it. I remember holding the plot,” she said, “and walking across the campus to Geoff's office, thinking, ‘Look at this, there's a planet with a 4-day orbit, there's one with a 240-day orbit, and there's one with a 2.5-year orbit, and they are
big
.” It was hard to imagine how such a system could be stable, with three huge, tightly packed planets tugging not just on their star but on one another. “It has completely changed our vision of how planets form, how much space they really need, that sort of thing.”

At just about that time, Bob Noyes called up from Harvard to say he'd been looking at the same star, and thought he had enough data to confirm two planets. “I knew it wasn't two planets, I knew it had to be three,” said Fischer, and she was really distressed when Marcy proposed that the two teams combine their data and publish a joint paper. Once they saw her analysis, she realized, the Harvard team would see the third planet too, and get some of the credit. “I look back,” she said, “and it's silly to have been so disappointed. But I was.” Officially, the Harvard and Berkeley teams did get equal credit for discovering the first three-planet solar system beyond the Sun (a fourth may have now been found). Unofficially, without authority, and surely against their wishes if I should be foolish enough to ask permission, I hereby award full credit to the Berkeley group.

At the time, Marcy and Butler had been able to hone their technique to the point where they could measure a star's motion to a precision of three meters per second. That was good enough to find Jupiter- and even Neptune-mass planets around
other stars, but not good enough to find an Earth-mass planet, even in a very tight orbit. (Butler's group in Australia would remain part of the Berkeley team, and when Butler later moved to the Carnegie Institution of Washington, the collaboration would continue.) The astronomers weren't satisfied. They didn't know how much better they could do, but they would try. “I remember,” said Fischer, “Steve Vogt, Paul Butler, Geoff Marcy, and I would always sign our e-mails ‘OMPSD,' which stood for ‘one meter per second or death.' Steve, especially, loved that. But it was comical. Because of course you could never get to one meter per second.”

When Bill Borucki got the million dollars that allowed him keep honing the Kepler concept, his job wasn't just to convince NASA decision makers that his spacecraft would be stable, but also that it could reliably detect minute changes in light intensity. “They told us, ‘It's probably going to take you several years to build it, because nobody's ever done anything like it before,'” he told me. “So we went out and we bought up all of the invar we could find”—invar is an alloy of steel and nickel that expands and contracts very little when the temperature changes. The alloy, invented in 1896 by a Swiss metallurgist named Charles Édouard Guillaume, proved so useful for constructing high-precision scientific instruments that Guillaume won the Nobel Prize in 1920 as a result. Albert Einstein had to wait until 1921 for his.

Despite NASA's pessimistic projection, said Borucki, “we got the design together in a few months, and then we got all the machine shops in the Bay Area building the parts for us. Got this whole thing built in a year, and got it debugged six months later. Then we had to figure out ways to simulate transits
and measure them with high precision. We didn't care about accuracy.”

For anyone but a scientist or an engineer, this probably sounds nutty, but it turns out to make perfect sense. Kepler didn't need to measure how bright a given star is. The scientists didn't care if this star is exactly as bright as that other star, or brighter or dimmer. They could afford to be inaccurate about that kind of measurement. All the satellite had to do was measure, with extremely high precision, the
change
in brightness when a planet passed in front of the star. “It's very much like what Geoff Marcy does with radial velocities,” said Borucki. Marcy didn't care about a star's overall motion—it could be speeding toward the Earth or speeding away or sitting there like a bump on a log. All he cared about was the
change
in motion caused by an orbiting planet. “Geoff's accuracy … well, I would hardly call it poor, but he considers it poor,” said Borucki. “But his relative precision is one part in a hundred million.”

So how hard would it be to achieve the kind of precision Borucki needed to find an Earth—a precision of between ten and one hundred parts per million, or a millionth to a ten-millionth of a percent? “Think about those holes in a metal plate that represent stars,” he said. “Now I slide a piece of clear glass in front of one of them. How much will the decrease in light be? For regular glass, it's 4 percent. That's way, way above ten or a hundred parts per million. So we take the glass and add an antireflective coating.” That makes less light bounce off the glass and more pass through. “Well, now it's, you know, 1 percent or 0.5 percent. You're still hundreds of times too high.”

No good. So in order to test the sensitivity of their detectors,
they had to think of an entirely new way to create minuscule changes in light. David Koch, Borucki's longtime collaborator, came up with one. “We drill laser holes in a steel plate,” said Borucki, “and we take a very fine wire and run it across the hole. Then we run a little current through the wire; the wire heats up and expands and it blocks more light.” They built it. It behaved perversely. “Of course,” he said, “when we applied the current, we got
more
light coming through, not less. Now, how can that possibly be?” It turned out that if the wire wasn't absolutely straight to begin with, the expansion would make it bend. It was no longer covering the widest part of the round opening, so it was actually letting more light through. “So we had to go back and redrill the holes as squares.”

There was plenty more of this sort of thing, but by the time Natalie Batalha joined the group in 2000, Borucki and Koch had finished their experiments. They had convinced themselves that they could make precise (albeit inaccurate) measurements of the tiny changes in starlight that would signal the presence of a faraway, invisible world. They were writing up yet another proposal for submission to NASA's Discovery program. But it wasn't enough simply to prove that their space telescope would work properly; they also had to convince NASA that planetary transits were observable in principle. A planet would pass directly in front of its star only if the orbit were precisely edge-on. Since planetary orbits could come in any orientation at all, the Kepler team had to show that enough of them would line up correctly, purely by chance, to allow the spacecraft to find them in enough numbers to justify spending hundreds of millions of dollars on the mission.

Borucki had promised NASA that Kepler would measure the dimming of starlight with a precision nobody had ever achieved. To do that, however, the team needed a crucial piece of information. Say you measure the light curve of a transiting planet, and it dims the star's light by 1 percent. That tells you that the planet's diameter is 1 percent as big as the star's.
But how do you know how big the star is in the first place?
Without knowing that information, you can't learn anything.

Fortunately, Kepler can get the answer, by using a technique called astroseismology, which borrows its name from plain old seismology, the study of earthquakes. When an earthquake goes off, the shock propagates outward, but also downward. The downward shock waves travel toward the center of the Earth, then bounce when they run into a change in density—where the semi-molten rocky mantle meets the iron core, for example. Around the world, seismic stations pick up those bouncing shock waves, and seismologists use them to deduce the inner structure of the planet.

The Sun isn't solid like the Earth; it's a huge ball of gas. It's so dense, however, that it acts something like a huge blob of incandescent Jell-O. Turbulence in its upper layers makes the entire blob vibrate, in many overlapping frequencies at once. Decades ago, solar astronomers began to study the vibrations in the Sun by looking for radial-velocity differences from one part of the Sun to another. These differences are caused by the rising and falling of its surface under the influence of the vibrations. The pattern of vibrations depends very sensitively on the Sun's inner structure—on the temperature and pressure of its inner layers, mostly—but it also depends on the Sun's size.

With Kepler, astronomers could start doing the same sort of analysis for stars, and since knowing a star's size is crucial to calculating the size of a transiting planet, this was part of the Kepler Mission right from the start. As the project progressed through the 2000s, though, it began to run over budget. “NASA told us, ‘You've got to cut things,'” said Borucki. “One of the things that came up for cutting was astroseismology.” So the Kepler team cut a deal with a coalition of European stellar astrophysicists. The Europeans would get access to the Kepler observations before they were released to the public, and the Kepler team would get the astroseismology readings it needed in return.

The team also had to give a good answer to another obvious problem. We know the Sun has sunspots—dark blotches caused by the Sun's magnetic fields. We know that Sun-like stars have them too; they're called starspots, unsurprisingly enough. So how would Kepler distinguish between a dark starspot on the surface cutting down on the brightness of a star and the dark silhouette of a planet passing in front if it? This was Batalha's first assignment on the Kepler Mission. “I tackled the problem,” she said, “from a stellar populations perspective.” As a stellar astrophysicist, she knew that young stars tend to rotate quickly and have a lot of spots. That's problematic in two ways: First, the spots can masquerade as planets. In older stars like the Sun, that's not such a problem, since sunspots get fewer and a star rotates more slowly as it ages. The Sun rotates about once a month at its equator; a planetary transit should last only a few hours. It's easy to tell the difference. So Batalha's first job was to figure out what percentage of the Sun-like stars Kepler would look at might be too young
to be trusted. “It came out to be about 25 to 30 percent of the sample,” she said. “We basically showed that there would be enough stars left over that you'd still be able to detect substantial numbers of Earth-size planets.”

Batalha and the rest of the Kepler team also had to figure out how they would deal with an even more common source of confusion. It turns out that the majority of stars in the Milky Way are part of multiple-star systems—doubles, triples, even quadruples, orbiting one another as they move through the galaxy. Our own Sun, with no companion star at all, is a little bit of an oddball. Theorists have long believed that single stars like the Sun are the best places to look for planets; the complex, ever-changing gravity of binary or triple stars would, they argued, make planetary formation difficult, or even impossible. So Kepler would keep its electronic eyes trained on singles. But since double and triple stars are so common—doubles, especially—it wouldn't be at all surprising if they were lurking in the background or the foreground, on the same line of sight as a target star.

At least some of these multiple stars would be orbiting each other edge-on with respect to Earth, purely by chance, each star eclipsing the other in turn. (It's called a transit only if one object is much smaller than the other—or appears much smaller. When the Moon passes in front of the Sun, it's called an eclipse because from our perspective, the Moon is big enough to block the Sun's light entirely.) If one of these eclipsing binaries was in the field of view, you'd be seeing the light from three stars in total, which would dip down to two stars during an eclipse, then back up to three, over and over.

If the single, target star and the eclipsing binary are very close together in the sky, Kepler won't see them as separate stars. “Say a background star is one thousand times fainter than Kepler's target star,” Geoff Marcy explained, “but then it winks out by 30 percent. Then voilà! If you think you're looking at just one target star, the total dimming is a few parts in ten thousand.” That's the same dimming you'd expect from a small planet. One way around this would be to take a second look with a more powerful ground-based telescope equipped with an adaptive-optics system. This is a technology originally developed by the military for spy satellites looking down on Soviet missile sites and other strategic targets: It uses a flexible mirror that constantly changes its shape to cancel out the blurring caused by Earth's atmosphere. Adaptive optics was declassified in the 1990s, and it turns out to be just as effective when you're looking up at a star rather than down at a missile silo. The Hubble Space Telescope is much smaller and less powerful than many ground-based telescopes; its major advantage is that it doesn't have to deal with the blurry atmosphere. With adaptive-optics systems, which are now common, telescopes on the ground don't either. Up to a point, anyway; adaptive optics isn't perfect. But unless the target star and the eclipsing binary are practically right on top of each other, adaptive-optics telescopes can separate two objects that Kepler sees as one.