The Life of Super-Earths (5 page)

 

FIGURE 4.2
.
Planetary systems are inclined at random angles. When we observe them from Earth, few of them will be aligned so that we can see a planet transit. The closer the orbit of the planet to its star, the higher the probability that we can see a transit.

With such a low probability, tens of thousands of stars must be monitored patiently to detect periodic dimming due to planet transits in a handful. In this only the gravitational lensing method rivals the transiting method. Perfecting the transiting method and making it work in practice took a lot of work, but the effort was worthwhile because of the method's added benefits: when we can measure both transits and Doppler wobble, we can deduce the planet's radius and mass, and hence mean density. But that was just the bonus! The transiting method turned out to be the best path to discover really small planets, whether super-Earths or Earth analogs.

By 1999, more than twenty-five extrasolar planets had been discovered by the Doppler shift method, most of them hot Jupiters. Because hot Jupiters orbit so close to their stars, the probability that they will transit increases.
Figure 4.2
shows how for the same planetary system an inner planet might transit, while another planet on a larger orbit would not. For such close-in planets there is a 5 to 10 percent chance that they will be seen in transit. In other words, for the more than twenty hot Jupiters discovered by the Doppler shift method, we should expect at least one of them to transit. The relatively high probability of finding a transiting planet makes the hunt for them a little more competitive. As planet-hunting teams discover new planets, they may keep their existence a secret until they check to see if the planet is transiting.

The lucky hot Jupiter turned out to be HD 209458b, an otherwise ordinary system of a planet orbiting a Sun-like star
every 3.5 days, about 150 light-years from Earth. The planet was discovered by the Doppler shift method in the summer of 1999 by a collaboration of the Geneva Observatory and planet hunters from the Harvard-Smithsonian Center for Astrophysics. By September 1999 they handed their data to a Harvard graduate student, David Charbonneau, who was spending time in Boulder, Colorado, with the small photometric telescope setup built by Tim Brown of the National Center for Atmospheric Research. David and Tim detected a transit, and so did the team of Geoff Marcy and Paul Butler, who had been racing to do the same. They had handed their own Doppler shift data to Gregory Henry of Tennessee State University, who did the photometric measurements with an automated telescope in Arizona.
⁷

This success was a watershed for two reasons: it confirmed beyond doubt the planetary nature of the extrasolar planets that had been found with the indirect Doppler shift method since 1995, and it boosted the effort to use transits as a method of discovery.

In the early 2000s the transiting method seemed to have a clear recipe for planet discovery: (1) do photometry of tens of thousands of stars simultaneously and (2) wait until you find a star that “blinks” in a regular fashion. If the star's light dims for a couple of hours by about 1 percent once every few days, then you have discovered a transiting hot Jupiter similar to HD 209458b. Two things seemed crucial: being able to measure a very large number of stars simultaneously and being able to do it with better than 1 percent precision. The former
meant using either a small telescope that can see a lot of sky or a regular large telescope but measuring only faint stars. The latter meant mostly improving the software and the details of the photometric measurements.

Many astronomers rushed to discover the first planet with the transiting method. The expectations were high and the predictions were very optimistic.
⁸
Measuring the light dimming due to the transit was thought to be sufficient to confirm the planet. Consequently even teams with very limited resources could compete with Michel Mayor's team in Geneva and Geoff Marcy's team in California in discovering new extrasolar planets. The recipe was easy; reality turned out to be much more difficult. Three years passed after the discovery of HD 209458b with no new transiting planets.

As it turns out, the problem is that a few stars do “blink” regularly, but for the wrong reasons. For example, sometimes two stars in a close orbit would eclipse each other and a third star nearby would dilute the effect of that deep eclipse and cause it to appear shallow—say a 1–2 percent deep, as if due to a much smaller planet-size body. All three stars would appear as a single dot of light even in our best telescopes. Or a very small star would orbit a star slightly larger than our Sun and the eclipses would also be about 2 percent deep and difficult to distinguish from a planet transit. The list of different scenarios continues. The realization gradually dawned that these false positives were quite pervasive. David Latham, the pioneer planet hunter at the Harvard-Smithsonian Center for Astrophysics, was helping a couple of teams confirm
possible transiting candidates with quick-look small telescope spectroscopy. Instead of planets, he kept uncovering false positives among the photometric transit candidates.

The problem came to a head in 2002. The OGLE team that we met earlier had equipped its telescope in Chile with a new large camera the year before. Before continuing with their primary experiment of detecting stellar gravitational lensing in our Galaxy, team members turned their telescope toward several patches of southern sky rich in stars and observed them nonstop every night for about four weeks, hoping to catch planetary transits. After a few months of dealing with the gigabytes of data that they gathered, the OGLE team found about sixty “blinking” stars. The “blinks” looked like planet transits, as the stars dimmed by just 1 to 3 percent; the trick was to make sure which, if any, were not false positives.

The photometric data obtained by the OGLE telescope alone was not sufficient to identify the kind of stars that showed the regular dimming. More information, such as stellar spectroscopy data or even distances from Earth, was not available because these stars were all faint and distant and previously unknown. The OGLE team had published its entire list of transiting candidates on the Internet (before publication in a journal) and invited the world community of astronomers and planet hunters to sort it out.
⁹
A true race began.

As soon as the first list of OGLE candidates appeared on the Internet, my younger colleague Kris Stanek walked into my office and challenged me to find planets orbiting the stars. He felt that the Harvard-Smithsonian Center for Astrophysics
might have the resources and—most importantly—the expertise to pull this off. I agreed with him about the expertise; after all, the first planet showing transits, HD 209458b, had been our local success just two years earlier. But finding planets on the OGLE list needed a new approach. In fact, the entire transiting method of finding planets needed to be sorted out. The simple recipe of the 1990s—using photometry to look for blinking stars—had not produced any results.

The first hint of what needed to be done came from the OGLE list itself. A few of the transit candidates on it showed changes in their light between consecutive blinks. I had seen this many times before, but it had nothing to do with planets. When two stars, known as binary stars, orbit each other very closely—in orbits similar in size and period to those of the hot-Jupiter planets—the stars literally pull each other into pear-shaped forms. Their asymmetric forms mean that the shape of the surface we distant observers can see differs over the course of the orbit, and, as a result, the amount of light we see varies too. In addition, stars also illuminate each other, and that adds to the light variation. Now, if the two orbiting stars happen to be aligned just right, we also see the stars eclipse each other. There was only one problem—stars are large and their eclipses are very deep (so deep that the first such binary star was discovered by the unaided eye of John Goodricke in 1782).
¹⁰
The OGLE eclipses were ten times smaller; how could that be?

At this point my background in stellar physics came in handy. These OGLE candidates were certainly eclipsing stars,
but with one of two possible differences from the typical binary pair. Either a third star was also in the picture or a very small star was in an orbit with a very big star. In the first case, the third star washes out the depth of the eclipse, making it appear shallow; in the second case the eclipse is shallow to begin with. One or two of the stars in the OGLE list even showed the telltale sign of a mismatched pair of stars, as there were barely visible eclipses due to the smaller star. The problem was figuring out whether the rest of the OGLE list was composed of similar eclipsing binaries as well.

The resolution to our primary problem—distinguishing false positives from real transiting planets—lay in fifty years of understanding of how stars work, known as the theory of stellar evolution. Stars of different masses have highly predictable temperatures and luminosities at any given age, and binary stars are of the same age by definition. Convinced that this basic stellar knowledge could be used to solve the problem, I began to lay out the steps needed to confirm that a blinking star was in fact being dimmed by a transiting planet.

I am a theorist—about stars—and although I love using telescopes and their instruments, I wasn't able to manage the OGLE challenge by myself. The problem needed a team. Guillermo Torres (a.k.a. Willie), an expert on binary stars and spectroscopic observation, was ready for the challenge, and we agreed that we should talk to one of our graduate students who knew how to do spectroscopy of faint stars, or their explosions, as it happens. That student was Saurabh Jha, who then was spying on very distant supernovae to understand
dark energy. Saurabh was already excited about extrasolar planets; he had collaborated with our senior graduate student, David Charbonneau, in observing the transits of HD 209458b. In the meantime, Dave had already been building his own telescope for seeking out transiting planets, while working at Caltech.

Willie, Saurabh, and I got to work right away. Our new method involved multiple steps. After using photometry to identify the potential transits, we obtained a single low-resolution or medium-resolution spectrum of the star, in order to identify whether it was a massive star. If it was, we excluded it. If it was not, we obtained more spectra, to look for a large Doppler-shift wobble. A large wobble would indicate that the “transits” are due to a star, not a planet. Next, we used more sensitive instruments to look for a wobble due to a planet. That step required the largest telescopes on Earth, with the best spectroscopy possible. If we detected the small wobble, the next step was to analyze the spectra for distortions in the absorption spectral lines. If such distortions were not present, and no second set of spectral lines was visible either, we put all the accumulated observations together and compared them to the predictions of the range of stellar models with different possible configurations of foreground and background intervening stars. At the end of the day, if all this cohered consistently, the planet was confirmed, and its size and mass were precisely determined.
¹¹
Willie and I also had to prepare the stellar and binary star models that we would need to analyze the systems and determine if they were stars or planets.

But at Harvard we had no access to a large telescope with a precise spectrograph for Doppler shift measurements of the very faint OGLE stars. Only the largest—the Keck telescope—would do, and the lion's share of observing time belonged to the partner institutions, the University of California and Caltech, that had built it. Fortunately, I had an ongoing collaboration on both Keck telescopes with colleagues at Caltech, the wizards of optical astronomy. At Caltech, Maciej Konacki—a young researcher working with my colleague Shri Kulkarni, and also a Pole (like the rest of the OGLE team)—was more than excited and ready to bring in the Keck at the last step of our transiting method. The Keck instrument—the old reliable HIRES spectrograph designed by Steve Vogt and used by Geoff Marcy to discover many extrasolar planets with the Doppler shift method—would be made available to us, and Maciej made sure we completed the crucial last step for planet discovery.

We had a busy summer, first doing the observations in Chile and Hawaii, where the Keck observatory sits atop Mauna Kea on the Big Island, and then analyzing the data as fast as possible in between. We were in a hurry not only because we knew that there was a race and the competition was fierce, but also because the core of our idea was to complete our steps in the right order, eliminating false positives along the way, and to make the best use of the precious time on Keck at the end. The results from the first few steps, done in Chile, were stunning: most of the OGLE transiting candidates were not transits and not planets, eliminating more than 90 percent of the “planets” on the OGLE list.