The Discovery of Extrasolar Planets

By Geoffrey Marcy and R. Paul Butler

In the vastness of the universe, are we humans alone? The answer depends on whether there are other planets that are endowed with the warm climate, diverse chemicals, and stable oceans that provided the conditions for biological evolution to proceed here on Earth.

During the 4th century BC, two great philosophers, Aristotle and Epicurus, opposed each other about the existence of worlds besides Earth. Epicurus asserted that the universe must be infinite and hence contain an infinity of worlds. Aristotle argued that Earth was placed at the center of the universe, making it unique in the universe. For over 2000 years, the question remained: Does the universe contain other worlds, like Earth? Scientists have learned that our sun is simply 1 star among 100 billion in our Milky Way Galaxy. Is the Milky Way a heavily populated metropolis of intelligent creatures, or is it a virtual desert, with precious few Earth-like oases?

Astronomers are still searching for the answers to these questions. We and other astronomers recently took an important step toward addressing some of these questions when we reported finding that planets do exist outside our own solar system. Since October 1995, we and other astronomers have announced the detection of eight planets orbiting sunlike stars. In April 1997, a group of astronomers reported that they had detected another new planet. Astronomers at the Smithsonian Institution's Astrophysical Observatory in Cambridge, Massachusetts; the National Center for Atmospheric Research in Boulder, Colorado; and Pennsylvania State University in State College, Pennsylvania, found evidence for a planet orbiting around the star Rho Coronae Borealis in the Northern Crown constellation.

Already the properties of these extrasolar planets have defied expectations, upsetting existing theories about how planets form. Investigation of planets outside our own solar system has just begun, and scientists are working on new telescopes that should tell us more about the planets detected to date and enable a more comprehensive search for other extrasolar planets.

Scientists began to gain some understanding of the structure of our own solar system in the 1500s, when Polish astronomer Nicolaus Copernicus argued that the sun was at the center of the universe and that Earth revolved around it. Many people, like the unfortunate Italian philosopher Giordano Bruno, quickly deduced that other stars might also harbor planets. He was burned at the stake in February 1600 for his many beliefs that conflicted with the teachings of the Roman Catholic Church at that time.

During the ensuing centuries, scientists discovered that nine planets orbit our sun and developed a theory that described how the planets formed from a flat disk of dust and gas surrounding the sun while it too was forming. Thus, the planets ended up in orbits that all lie in the same plane, all orbiting around the sun in the same direction. The theory naturally suggests that planets will form around other stars equally easily. But until 1995, no planets had been discovered orbiting other stars like the sun.

It took a long time to find extrasolar planets because detecting them from Earth is extraordinarily difficult. Unlike stars, which like our sun glow brightly from the nuclear reactions occurring within, planets shine primarily by light that is reflected off them from their host star. Planets only weakly emit infrared light (light with a wavelength just longer than visible light) from their cool surfaces. Indeed, the planets in our solar system are about one billion times fainter than the sun in visible light and one million times fainter in infrared light.

How Planets Were Finally Detected Around Other Stars

Astronomers finally gained the means to find planets around other stars with a clever new technique that involves searching for a telltale wobble in the motion of a star. When planets orbit a star, they exert a gravitational force of attraction on it. The force on the star causes it to be pulled around in a small circle or oval in space. The circle or oval is in fact simply a miniature replica of the planet's orbital path. Two embracing dancers similarly pull each other around in circles due to the attractive forces they exert on each other. This wobble of a star gives away the presence of an orbiting planet, even though the planet cannot be seen directly.

However, this stellar wobble is very difficult to detect from far away. An observer watching our sun from 30 light-years away would see the sun travel around in the sky in a tiny circle with a radius of only one-seventh of one-millionth of one degree. (A light-year is the distance that light travels in one year, equal to 9.5 trillion km or 5.9 trillion mi.) For comparison, the measure of a great circle drawn across the sky from horizon to horizon is 180 degrees. The tiny circle made by the sun would appear as large in the sky as a quarter seen from 10,000 km (6000 mi) away.

A new technique for detecting the wobble of stars has proven to be extraordinarily successful for detecting planets. The key is the Doppler effect—the change in the appearance of light waves and other types of waves from an object that is moving away from or toward a viewer. When a star wobbles toward Earth, the star's light appears from Earth to be shifted toward the blue, or shorter wavelength, of the visible light spectrum than it would have if the star had not moved toward Earth. When the star wobbles away from Earth, the opposite effect occurs. The wavelengths are stretched. Light from the star appears to be shifted toward the red, or longer wavelength, end of the spectrum in a phenomenon known as red shift.

Astronomers can determine the velocity of a star from the Doppler shift because the Doppler shift is proportional to the speed with which the star approaches or recedes from a viewer on Earth. However, the Doppler shift caused by the wobbling of stars with companion planets is a tiny effect; the length of the stellar light waves changes by only about 1 part in 10 million due to a large, Jupiter-like planet. For example, the sun's wobble speed is only about 12.5 m/sec (about 41 ft/sec). To detect planets around other stars, the errors in the measured speed of stars must be smaller than about 10 m/sec, or about 36 km/h (about 33 ft/sec, or about 22 mph).

At San Francisco State University, we have perfected a technique that measures the speeds of stars, using the Doppler effect, with a precision of plus or minus 3 m/sec (10 ft/sec)—bicycling speed! The technique involves putting iodine vapor in a glass bottle (an "absorption cell") near the focus of a telescope. As starlight passes through the iodine, the iodine absorbs certain signature colors, or wavelengths, of light. The spectrum of colors from the star are suddenly missing particular wavelengths of light at very specific wavelengths.

The entire spectrum of light from the star, spread out according to its wavelength, will be shifted to the blue or the red depending on the star's motion relative to Earth, slowly sliding toward bluer or redder colors against the constant reference of the dark wavelengths at which the iodine has absorbed light. These dark wavelengths along the spectrum of colors serve in the role of "tick marks" on a ruler, providing an unchanging scale against which to observe the Doppler shift of the stellar spectrum.

This Doppler technique is so sensitive that shifts in wavelengths as small as 1 part in 100 million can be discerned using this technique. These shifts are recorded by sending the starlight into specialized spectrometers—devices that measure the wavelengths of light. The spectrometers consist of giant prisms, mirrors, and diffraction gratings (surfaces scored with fine slits) and cost several millions of dollars. A computer analysis of a star's spectrum reveals the minute shifts in color of a star's spectrum relative to the stationary iodine markers.

Astronomers believe they have discovered eight planets orbiting other sunlike stars using this Doppler technique. Swiss astronomers Michel Mayor and Didier Queloz of the Geneva Observatory in Switzerland reported finding the first planet with this technique. They noticed that the Doppler shift of the star 51 Pegasi in the constellation Pegasus varied in a periodic way, first shifted toward the blue and then toward the red. The timing of the Doppler shift suggested that the star was wobbling due to a closely orbiting planet that completes a full revolution around its star in only 4.2 days. This means that the planet is orbiting its star at a speed of about 134 km/sec (83 mi/sec) or 482,000 km/h (299,000 mph), more than four times faster than Earth's velocity around the sun.

A survey by us of 120 sunlike stars has revealed six planets, one of which (around the star 16 Cygni B) was independently discovered by astronomers William Cochran and Artie Hatzes of the University of Texas McDonald Observatory on Mount Locke in western Texas. A seventh object, which orbits the star known by its catalog number HD114762, was first discovered in 1989 by astronomer David Latham of the Harvard-Smithsonian Center for Astrophysics in Cambridge, Massachusetts, and his collaborators. This companion has a large mass (more than 10 times that of Jupiter), not unlike the planet around 70 Virginis, which has a mass more than 6.5 times that of Jupiter. Both HD114762 and 70 Virginis are so large that most astronomers are not sure whether to consider them large planets or low-mass brown dwarfs. A brown dwarf is an object intermediate in mass between a planet and a star.

The wobble motion of a star with a planetary companion can provide a great deal of information about the star's companion planet, including an estimate of its mass and the size and frequency of its orbit. The orbital period of a planet (the time it takes the planet to complete one full revolution around its star) is equal to the time it takes the star to finish one wobble cycle.

The size of the stellar wobble is also proportional to the size of the planet's orbit. To find the distance between the star and its companion object, we apply German astronomer Johannes Kepler's third law of planetary motion, which states that the cube of the average distance between two orbiting bodies equals the square of the orbital period. Knowing that the orbital period is the same as the period of the star's wobble, one can calculate the average distance between a star and its companion.

For example, Jupiter, the largest planet in our solar system, has one-thousandth of the mass of the sun. Therefore, every 11.8 years (Jupiter's period), the sun wobbles in a circle that is one-thousandth the size of Jupiter's orbit. The mass of Jupiter is about 318 times the mass of Earth. The other eight planets also cause the sun to wobble, but by smaller amounts.

The Doppler shift can also reveal the shape of the planet's orbit. If the Doppler shifts change smoothly back and forth in time like a perfect sine wave (the graph of a fundamental trigonometric function, which cycles smoothly back and forth between 1 and -1), then the orbit is a circle. If the shift is not so regular over time, the distortion in the graph of Doppler shift versus time can be analyzed to determine how elongated the oval shape of the orbit is, a property known as the eccentricity of the orbit.

The amount of Doppler shift at its peak determines the minimum mass of the planet, which can be calculated using English physicist Sir Isaac Newton's laws of motion. The size of the stellar wobble is proportional to the ratio of the mass of the planet to the mass of the star: the greater the Doppler shift and thus the wobble of the star, the more massive the planet.

However, there is an uncertainty about the planet's mass. Orbital planes that astronomers can view edge-on will give the true mass for the planet. But orbital planes that are tilted or inclined produce a Doppler shift that is reduced due to the diminished amount of approach and recession as viewed from Earth. This effect yields an inferred mass that is smaller than the true mass. The unknown orbital inclination means that astronomers can only determine a lowest possible value for the planet's mass. The true mass can be larger, but even with this uncertainty astronomers know that the actual mass of a planet is almost certainly within a factor of 2 of the inferred mass.

The Diversity of Planets

Are all planetary systems alike? The eight planets found around other stars have shocked astronomers because their orbits are so different from those they expected. Astronomers expected that the orbits and masses of other planets would be similar to those in our home solar system. After all, the only examples of planet formation came from the nine planets here. So, all the theories to explain planet formation were designed to reproduce the orbits and masses of our planets.

Astronomers have a very clear theory of planet formation that they designed to explain some of the remarkable characteristics of our solar system. These characteristics include the fact that all the planets orbit the sun more or less in the same flat plane, which is known as the ecliptic. Second, all the planets revolve around the sun in the same direction.

The conventional theory of planet formation holds that the planets formed out of a spinning disk of gas and dust that has flattened out in the equatorial plane of the sun, much as a pizza dough flattens out in the air as it is tossed and spun. Material in the disk would all be rotating in the same plane, in the same direction, in a circular orbit around the sun, as the planets do today. According to the theory, planets would not form in the disk too close to the star, because the material would be too hot to coalesce, nor would they form too far from the star, because the material would be too cold and thin.

In retrospect, it seems silly that astronomers had adopted narrow expectations about planets in the rest of the universe. The planet orbiting the star 47 Ursae Majoris in the Big Dipper constellation is the only one that is similar to what was expected. It has a minimum mass of 2.4 Jupiter masses, and its orbit is circular with a radius of 2.1 astronomical units (AU). One AU is the distance between Earth and the sun, or 150 million km (93 million mi). This planet is only somewhat more massive than our Jupiter, orbits farther than Mars does from the sun but not as far as Jupiter and the other outer planets, and has a circular orbit like the planets in our solar system. If placed in our solar system, this new planet would fit in nicely as Jupiter's big brother.

But the remaining planets around other stars defy expectations. Three planets have orbits that are oval-shaped, notably those around 16 Cygni B, 70 Virginis, and HD114762, which have eccentricities of 0.6, 0.4, and 0.3 respectively. An eccentricity of zero is a perfect circle, while an eccentricity of 1.0 is a long, flattened oval. The largest eccentricities in our own solar system are for Mercury and Pluto, both about 0.2, while all other planets have nearly circular orbits (eccentricities less than 0.1). So why do these newly discovered planets have such large eccentricities?

The eccentric orbits sent astronomers back to the blackboards to revise their theories. Within two months, astronomers had concocted new embellishments and tacked them on to the standard planet-formation theory. Astronomers Pawel Artymovicz of the University of Stockholm in Sweden and Pat Cassen of the National Aeronautics and Space Administration (NASA) Ames Research Center in Moffett Field, California, have recalculated the gravitational forces at play during the birth of planets from the disks of gas and dust that are observed swirling around young sunlike stars.

The calculations of Artymovicz and Cassen suggest one mechanism by which planetary orbits might be pushed away from pure circular motion. Their work indicates that the gravitational forces exerted by protoplanets (planets in the process of forming) on the disks of gas and dust surrounding the young star create spiral waves of alternating greater and lesser density in the gas. These density waves, which are shaped not unlike the "arms" seen in spiral galaxies, in turn exert subtle forces back on the formative planet, driving the planet away from a pure circular motion.

Artymovicz and Cassen's work suggests that during a million years, departures from circular motion can be quite significant, leading to eccentric orbits. Further calculations are needed to determine if eccentricities as high as 0.6 can actually result from these spiral waves.

A second theory that accounts for the large orbital eccentricities seen with some of the recently detected planets seems equally likely. Suppose, for a moment, that Saturn had grown to a larger mass than it actually did. After all, what caused Saturn to stop growing at one-third of a Jupiter mass? No one really knows. Conceivably, all four giant planets in our solar system—Jupiter, Saturn, Uranus, and Neptune—might have grown to larger masses, if our original protoplanetary disk had contained more mass or had existed longer. In that case, our solar system would contain four super-Jupiters. They would exert gravitational forces on each other, resulting in persistent perturbations (disturbances) of their orbits, causing the orbits to eventually cross. The planets would at some point approach each other closely as their orbits crossed, exerting strong gravitational forces on each other, and causing them to drastically change their orbital paths.

After this slinging effect, some of the super-Jupiters could be thrown inward to smaller orbits and some thrown outward—perhaps even ejected from the planetary system. This scattering is similar to the "break" of the balls in a pool game—they go every which direction chaotically. The scattered Jupiters would now orbit with extreme eccentricities, just as we have discovered for the three eccentric planets. Interestingly, this billiard model for eccentric planets predicts that eventually we should be able to detect the other Jupiters that caused the scattering and the eccentric orbits. These large planets may orbit farther out than the one planet we have detected. A variation on this theme is the idea that a companion star, rather than other planets, might be responsible for the gravitational scattering.

The Mysterious 51 Pegasi-Type Planets

The most bizarre of the new planets are the four so-called "51 Peg" planets. They are characterized by orbital periods shorter than 15 days. The four members of this class are 51 Peg itself, Tau Bootis, 55 Cancri, and Upsilon Andromedae, having orbital periods of just 4.2, 3.3, 14.7, and 4.6 days, respectively.

All of these orbits are small, with radii less than 0.11 AU (less than one-tenth the distance between Earth and the sun). These are much smaller orbits than that of the closest planet to the sun in our solar system, Mercury's orbit having a radius of 0.38 AU and an orbital period of 88 days.

Yet these planets are comparable in mass or even more massive than the largest planet in our solar system. The masses of these planets range from 0.45 Jupiter (for 51 Peg) to 3.7 Jupiters (for Tau Bootis). The Doppler wobbles for these planets are smooth sine waves, showing that the planets move in circular orbits. The 51 Peg planets defy conventional planet-formation theory, which predicts that giant planets like Jupiter, Saturn, Uranus, or Neptune would be formed in the cooler outskirts of a protoplanetary disk, 5 AU from the star.

A revised planet-formation theory comes from University of California at Santa Cruz astronomers Douglas Lin and Peter Bodenheimer and astronomer Derek Richardson at the University of Washington in Seattle. Their theory involves an extension of the standard model. A young protoplanet that is just precipitating out of a massive protoplanetary disk around its star will carve out a groove in the disk, separating the disk into an outer and inner section. The inner disk and newly forming planet will lose energy and angular momentum to the outer disk, and the protoplanet will slowly spiral inward. According to Lin and Bodenheimer, the planet would be dragged right into the star, if something did not stop it at death's door.

The planet's savior may be the rotation of the young star. Young stars spin rapidly, making a complete revolution in about five to ten days. As a planet approaches the star, tides would be raised on the star, just as the moon raises tides on Earth. But the young star would be spinning more quickly than the planet would be orbiting around the star and the bulge raised by the planet would begin to move in front of the motion of the planet, rather than staying on the star's surface directly in line with the planet. As this happened, the gravitational pull from the bulging tide on the star would pull the planet forward, adding angular momentum to the planet, which moves it into a larger orbit and halts its death spiral into the star.

The protoplanet can remain in a stable orbit, delicately balanced between the tidal drag from the disk and the tidal pulls forward from the spinning star. Lin actually predicted, prior to the discovery of the 51 Peg planets, that our own Jupiter should have spiraled into the sun long ago. If so, why did our Jupiter survive? Perhaps our solar system contained previous "Jupiters" that indeed spiraled into the sun. Then, finally, after many "Jupiters" went "down the drain," the last Jupiter formed as the disk had largely dissipated. This last Jupiter is the one that survives today, when tidal forces from a protoplanetary disk no longer pull it toward the sun.

One might wonder why our solar system has no 51 Peg-like planet orbiting in close to the sun. Perhaps our Jupiter formed toward the end of the lifetime of our protoplanetary disk, after the gas in the disk had thinned. Alternatively, our protoplanetary disk may never have had enough gas and dust to exert sufficient tides in the first place. Indeed, protoplanetary disks are observed to have a wide range of masses, from a few Jupiter-masses to hundreds of Jupiter-masses, as measured by the emission of radio waves from the disk material. So the extraordinary diversity of the new planets may correspond to different disk masses or disk lifetimes, and perhaps to different environments, such as the presence or absence of nearby radiation-emitting adult stars.

A challenge to the very existence of the 51 Peg planets has been offered by astronomer David Gray of the University of Western Ontario in London, Canada. He has suggested that those four stars are themselves oscillating—in much the same way as a water balloon bulges and vibrates when nudged—and have no planets at all. In his view, the oscillations of the star cause the periodic Doppler shifts, rather than a planet.

The reality of these oscillations is heavily debated. The strongest argument against the oscillations stems from the single period and frequency seen in the Doppler variations from the star. Most oscillating systems, such as tuning forks, display a set of harmonics, or several different oscillations occurring at different frequencies, rather than just one frequency. But the 51 Peg stars show only one period each, quite unlike harmonic oscillations. Further, basic physics predicts that the strongest and most readily detected oscillation would occur at a much higher frequency than that of the observed wobble of any of these stars. Also, the 51 Peg stars show no brightness variations, which indicates that their sizes are not changing. But future observations of the 51 Peg stars will determine if they reveal any sign of oscillations.

The Uniqueness of Our Solar System

Since the announcement of the detection of the planet around 51 Peg in October 1995, astronomers have discovered eight extrasolar planets. It is tempting to compare those planets to the nine planets in our solar system. Unfortunately, eight is too few to draw firm conclusions and our ability to detect all types of planets is quite limited. For example, planets having Earth-like sizes are not currently detectable. Also, the extrasolar planets found so far have orbital periods less than three years. This is not necessarily representative of planetary systems in general, but is due to the fact that astronomers have been searching for other planets with their improved techniques for only about ten years. With more time and improved Doppler precision, more planets may be found that have longer orbital periods and orbit farther from their stars.

Nonetheless, the surprises found among the planets have revealed that our own planetary system could have been very different than it actually is. The three planets in eccentric orbits are particularly eye-opening, and even frightening. Suppose that gravitational scattering of planets is a common occurrence in planetary systems. Our own solar system provides evidence that during its first billion years, planetesimals (fragmentary bodies of rock and ice) ran rampant. The cratering of the moon and the extreme tilt of the rotation axis of Uranus show that collisions were common, and some must have involved two large planet-sized objects. The neat record-groove orbits in our current, middle-aged solar system are the stable pattern that emerged from the collision-happy orbits of our solar system's youth. Perhaps we are lucky that our Jupiter ended up in a nearly circular orbit. If Jupiter had been thrust into an oval orbit, it would have collided with Earth, sending it careening out of the solar system. The onset of biology and the process of evolution that led to human beings has been made possible by the placement of both Jupiter and Earth in mutually stable, circular orbits.

The Future of Planet Hunting

After 2000 years, astronomers finally have developed the telescopes and instruments to test the claim by Epicurus that many planets exist in the universe. It is a reasonable guess from our detections so far that most stars have some planets orbiting them. Many additional questions remain that should be answered before the next 2000 years.

What fraction of stars have planets? Do they all have planets? We wonder if seemingly bizarre solar systems with planets in eccentric orbits or close-in orbits are in fact freaks. Could they be the norm? Perhaps our system is the freak. Do terrestrial planets occur commonly? Are there other Earths that have oceans and continents, providing a stable environment for life? Among the 100 billion stars in our Milky Way Galaxy, we would guess that somewhere an Earth-like planet with liquid water must exist.

By the year 2010 astronomers will have completed the first census of planets orbiting nearby stars. As of April 1997, we at San Francisco State University continue to monitor 120 stars from Lick Observatory. In July 1996, with support from NASA, we began a second Doppler survey of 400 stars using the 10-m (33-ft) Keck telescope at Mauna Kea Observatory in Hawaii. Mayor and Queloz have recently tripled the size of their northern hemisphere Doppler survey to about 400 stars, and in late 1997 they will begin a southern hemisphere survey of 500 more stars. Within the next five years Doppler surveys of several hundred additional stars will begin at the 9-m (30-ft) Hobby-Eberly Telescope located at the McDonald Observatory, directed by Cochran.

Also within five years the power of the two Keck telescopes on Mauna Kea will be joined together to form a device known as an optical interferometer that will provide sufficient precision to detect extrasolar planets the size of Earth. NASA is planning at least three space-borne telescopes to detect planets in infrared light. One proposed NASA space-borne interferometer, a second-generation telescope known as Planet Finder, should be able to obtain actual pictures of pale blue dots orbiting like clockwork around stars.

This future space-borne interferometer planned by NASA will be particularly valuable. It is undoubtedly the greatest telescope ever conceived. If funding is made available, Planet Finder will be able to take pictures of other Earths, starting in about 2010. More importantly, it will be able to analyze the planet's light with a spectrometer to determine the chemical constituents of the planet and its atmosphere. Astronomers may be able to determine whether biological activity is occurring. For example, the oxygen on Earth results primarily from the photosynthesis activity of plants. The NASA telescope will be designed to detect oxygen as a signature of life elsewhere. Perhaps alien beings on other planets are similarly planning their own space-borne telescopes to find out if they are alone.

About the authors: Geoffrey Marcy and R. Paul Butler, astronomers affiliated with San Francisco State University and the University of California at Berkeley, announced the discovery of six of the eight planets near sunlike stars that have been detected to date. They continue to search for additional planets in other solar systems.