illion :l as- erial !r, in nical time final simi- fther and mto Probingthe GiantPlanets rtion f 106 have very eccentric orbits while our and own giant planets orbit the Sun in se- date circles. Could it be that "otdi- nary" circular orbits are in fact quite extraordinary by Galactic standards? Is our solar system an unlikely out- come of the general mechanism of TristanGuillot planet formation? Or perhaps, might extrasolar planets be formed by a dif- ferent mechanism? llefore 1995, most astronomers expected giant extra- These are but a few ofthe questions that confront us. lJsolar planets to orbit their stars in quasicircular or- There's no shortage of suggested answers, but none thus bits at a distance of more than a few astionomical units. far has given satisfaction. A large part ofthe problem is (1AU is the mean b distance between Earth and the Sun.) that the so-calledradiovelocimetry method used until now qr tt our Solar System, the orbits of the four giant planets- to discover extrasolar planets provides only a lower limit Jupiter, Saturn, IJranus, and Neptune-have semimajor on the planetary mass M. Relying on the tiny Doppler mod- axes ranging from 5.2 to 30 AU and eccentricities no ulation as the star is tugged to and fro by its orbiting la larger than 5.6Vo. planet, the method determines the orbit's diameter and ec- Since then, more than 100 extrasolar planets have centricity. But the inclination angle i between the normal been discovered, all ofthem giants with at least 107othe to the orbital plane and the line of sight is unknown, and mass of Jupiter (0.1 M,, about twice the mass of Uranus radiovelocimetry only determines the product M sini. r-a or Neptune). Much smaller Earthlike ("terrestrial") extra- Of course, we do have interesting additional meas- ral solar planets would not be massive enough to be detected urements. For example, it is becomingclear that h6 stars with by current methods. planets tend to be enriched in heavy elements-meaning, Ein The newly found planets are strange-not at all what loq in astronomers'jargon, anything other than hydrogen and we expected.A signifrcant fraction ofthe extrasolar giants rcm helium. That tells us something about planetary forma- tDr, detected thus far orbit extremely close to their star, that tion, namely that planets probibly grew more rapidly in is, at less than 0.1AU. Some semimajor axes are as small environments rich in dust. (Most heavy elements form n,K as 0.04AU, implying a period of revolution around the star chemical speciesthat condenseat low temperatures.) But ofonly about three Earth days. The archetype ofthese so- "hot we do not know two crucial properties ofthe extrasolar gi- called Pegasi planets (also called Jupiters") is the first ants: their exact masses and their sizes. Except for one M7 one to have been discovered: 51 Pegasi b, a roughly serendipitous casea-the planet HD209458b. flx, Jupiter-mass giant orbiting at only 0.05 AU from a sun- like star in the constellation Pegasus.l A planet in transit ,219 The other fraction of extrasolar planets discovered HD209458 is a faint star in Pegasus,barely visible to the since 1995 may be even stranger. As shown in frgure 1, naked eye. Like 51 Pegasi, its planet has a mass close to many of them exhibit very eccentric orbits,2 quite unlike that of Jupiter, and a tiny semimajor axis, a mere 0.045AU. 54if our own giant planets. It may be, however, that Solar Sys- The particularity of the HD209458 system is that, as seen tem analogs are common rigin but hard to discover.After all, from Earth, the planet passesin front ofthe star every 3.5 Jupiter and Saturn U. of are so far from the Sun that they take days. Therefore, we can measure the time it takes for the 12 and 29 years, respectively, to complete an orbit. A few planet to transit the stellar surface and the consequent :M. planets in frgure t have large orbits with low eccentricity. dimming of the star (about 2Vo).Thus we can also calcu- Discovering more objects like these will require patience late the planet's size, the inclination ofits orbit, and there- and improved observational techniques. fore its mass.5 Ithasbeen very dillicult to 'the). construct a coherent sce- The radius measured for the giant planet HD20g4b8b nario that would explain the formation of giant planets as is about SSVolargerthan that ofJupiter, although its mass \riz. we seethem both within and beyond our solar system. The is 30Voless than Jupiter's. This apparent contradiction is presenceofclose-in planets is generally thought to be due understood by realizing that gaseousgiant planets ;, P. tend to to early migration of planets forming in a disk of gas and cool slowly by infrared emission and thus contract. and dust surrounding the young central star.3 But that's not that planets 952 heavily irradiated by their stars contract the only possibility, and it doesn't explain why Jupiter and more slowly. Jupiter itself is estimated to be contracting its sisters didn't share such a fate. at a rate ofabout 3 mm per year. nof Similarly, it is not clear why extrasolar planets often The observations of HD209458b, a hundred times TX closerto its star than is Jupiter, confrrmed the general the- oretical picture and showed that the planet, discoveredin tuf,, 200O,is made mostly of hydrogen and helium.6 The study I of this transiting Jovian giant was the first confirmation @ 2004 American Instituteof Physics,S-003i-9229-0404-040-2 April 2004 PhysicsToday 63 and coworkers,s in 200L, detected the presence of sodium in the atmosphere of HD209458b. The sodium, it turned out, was less abun- dant than expected. But be- cause we know so little about the atmospheres of these ex- otic planets, it's not yet clear what that means. The sodium shortfall could, for example, be due to colder atmgspheric temperatures than expected. Or it might result from at- mospheric dynamics or non- equilibrium effects. In any case, the observation was a milestone in the study of ex- trasolar planets, .because it showed that we can detect constituents in the atmos- pheres of planets millions of times further from us than Jupiter. Another crucial recent discovery is that HD209458b is slowly losing mass. Last year,Alfred Vidal-Madj ar and coworkers found that when one observes the transiting planet at the LIV wavelength of the Lyman-a absorption line of hydrogen, it appears three times larger than at other wavelengths.eThis suggests the presence of an ex- tended, tenuous, envelope ofhydrogen escaping from the planet as a result ofheating ofthe upper atmosphere and bombardment by ionized stellar wind. The magnitude of this mass loss appears to be consistent with estimates based on Jupiter, but scaled up by a factor 10 000 because the star is 100 times closer to the planet than the Sun is to Jupiter.GThe Lyman-a measurements give us the first possibility of quantiffing such processes more precisely planet that our models of formation were not completely and examining how gaseousplanets manage to suryive so off-track. Had HD209458b been found to have a smaller close to their stars. diameter than Jupiter, it would have meant that the new Impressive as they are, the HD209458b measure- planet consists mostly of heavy material, which would ments tell us only about the upper atmospheres of giant patently contradict our formation models. gas planets. How can we better constrain their interior However, the hope that we would be able to learn more compositions?To do so,we have to learn how surface meas- precisely the composition of HD209458b has dwindled rap- urements relate to interior compositions.And to that end, idly. The planet's radius turned out to be slightly larger we must look more carefully to the gas giants closestto us. than expected. So we're probably missing some enerry source that prevents the planet from cooling faster.TIt ap- Our own giant planets pears that tides may be dissipating heat into the planet's We can, of course, measure very accurately the masses, interior and thus slowing its contraction. The energ'y sizes, and hence the mean densities'of Jupiter, Saturn, source might be the orbital enerry of an unseen eccentric IJranus, and Neptune. The densities turn out to be quite giant companion planet, or it might be winds generated in low-ranging from 0.7 glcml for Saturn to 1.6 g/cm3 for the planet's atmosphere by the strong stellar irradiation. Neptune. For comparison, the four inner terrestrial plan- Alternatively, the atmosphere may be hotter than most ets have mean densities near 5 glcml. One might have at- models predict (seefigure 2). We know little about meteor- tributed the difference to the conjecture that the giant ology and tides on gaseousplanets. Tens ofEarth masses planets have the same composition as Earth but very much of heavy elqments could be mixed in with the hydrogen and hotter. Alternatively, one could supposethat they are very helium without our being able to tell. Better understand- much colder but made of light material. Of course, we've ing the giant planets will require more examples of tran- known for a hundred years that only the second explana- siting extrasolar planets, with different massesand orbital tion is valid. distances. That's an important goal of spacemissions such Indeed, a look at the atmospheres ofthe giant planets as COROT, Kepler and, we hope, Eddingtoa (seebox 1). shows that they are made mostly of hydrogen and helium, In the meantime, the power of transit observations with only traces of heavier elements.