© 1999 Macmillan Magazines Ltd

In the 1990s, the quest to find planets out-side our own Solar System finally succeed-ed. The first extrasolar planets to be identi-

fied are in orbit around a pulsar1, a rapidlyrotating neutron star which is quite differentfrom our own familiar Sun. The first plan-etary companion to a main-sequence star — ahydrogen-burning star like the Sun — wasdiscovered in 1995, using Doppler measure-ments of the reflex motion (wobble) that theplanet’s orbit induces in the star 51 Pegasi2.Nearly 20 more main-sequence stars havesince been found to possess a single compan-ion of roughly Jupiter mass3. But now we havenews of something else again — the discovery,announced last week by Paul Butler and hiscolleagues4, of a system of three massive planets orbiting the star yAndromedae.

All of the planets known to orbit main-sequence stars other than the Sun have beenidentified using the Doppler technique,which measures changes in the star’s velocitycomponent along the line of sight. Theseobservations can be inferred as perturbationof the star by a companion planet. They yieldthe planet’s orbital period and orbital eccen-tricity, and the product of its mass and thesine of the inclination of its orbital plane tothe plane of the sky, Msin(i). Because sin(i)cannot exceed unity, the measured radial-velocity variations of the star provide a lowerlimit to the planet’s mass.

Determining the masses and orbits ofthree planets travelling about a star usingDoppler data is a highly complicated business,because the data have significant intrinsicscatter and the signatures of the planets inter-fere with one another. Butler et al.4 looked attwo sets of data, obtained respectively at theLick Observatory in California and by theAdvanced Fibre-Optic Echelle (AFOE) spec-trometer at the Whipple Telescope in Arizona.They analysed these data independently, andwhen combined, to derive three sets of plan-etary parameters. The similarity of the valuesof these parameters (see Fig. 1) lends confi-dence to the three-planet solution.

The star y Andromedae has a mass about30% greater than that of our Sun, is abouthalfway through its six-billion-year main-sequence lifetime and is about 44 light yearsfrom our Solar System. The innermost plan-et was first discovered in 1997 (ref. 5). It has anearly circular orbit with a period of 4.6 days,

and a minimum mass about 0.7 times that ofJupiter; these properties are similar to thoseof the only known planet of 51 Pegasi and toother star-grazing planets.

What about the two newcomers identi-fied by Butler et al.? The middle planet has aperiod of about 242 days, a mass about twicethat of Jupiter and an orbital eccentricity, e,of approximately 0.22 (comparable to theeccentricities of Mercury and Pluto, andlarger than the orbital eccentricities of allother planets known in our Solar System).The orbital period of the outer planet is~1,270 days; e ~0.36; and mass ~4.1 timesthat of Jupiter. These latter values are not wellconstrained, because most of the Dopplerdata were taken within a 6.5-year interval,less than two orbital periods of the planet.But the masses and eccentricities of the outer

two bodies are typical of half-a-dozenknown extrasolar giant planets with orbitalperiods of around a year.

Additional constraints on the y Androm-edae planets come from astrometric andphotometric data, as well as from numericalorbit integrations. The Hipparcos astromet-ric satellite observed the position of yAndromedae on the plane of the sky severaltimes during the early 1990s. No wobble isdetectable in these data, which implies thatthe outermost planet is less than ten andprobably no more than five Jupiter masses.Photometric measurements of the bright-ness of y Andromedae around the times thatthe inner planet would pass in front of (tran-sit) the star if its orbital plane were nearlyalong the line of sight (sin(i) close to 1) ruleout transits deeper than about 0.0002 to0.0003 magnitude (G. W. Henry, personalcommunication). This means that the incli-nation of the planetary orbit must be lessthan 837 or that we are dealing with anextremely dense, exotic object.

As far as numerical integrations are con-cerned, the preliminary results of calcula-tions by myself and a colleague, E. J. Rivera,imply that a system having the parametersderived from the AFOE data only, with sin(i)of approximately 1, would tear itself apartwithin a million years or so (Fig. 1). (Thechaotic nature of such a trajectory precludes

NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com 659

news and views

Three planets forUpsilon AndromedaeJack J. Lissauer

In 1997, a planet of roughly Jupiter’s mass was discovered in orbitaround the star y Andromedae. Two more have now been identified.

Figure 1 The reported orbital parameters for the planets around y Andromedae4 are given on the leftside of the plot. The orbit of each planet is represented by three points — the periastron andapoastron (closest and farthest distances from the star, bottom and top points) and the semimajoraxis (roughly, the average orbital distance). Triangles represent the outer planet, crosses the middleplanet and circles the inner planet. Error bars are shown for the outer two planets; observationaluncertainty is far less for the innermost planet. Data from the Lick Observatory are given in greenand from the AFOE (Whipple Telescope) in red; combined data are in black. The rest of the plotshows the temporal evolution of the periastra and apoastra of the three planets using the nominalparameters on the left as initial conditions. The systems using the combined and Lick data setscontinue to be stable for the entire ten-million-year integration interval. (Calculations and figure byE. J. Rivera of SUNY, Stony Brook, and J.J.L. using the SwIFT symplectic MVS package8.)

0 105 2x105 3x105 4x105 5x105

t (yr)

4

3

2

1

0

Rad

ial d

ista

nce

from

sta

r (A

U)

© 1999 Macmillan Magazines Ltd

a precise calculation.) The star, and thus pre-sumably its planets, are about three billionyears old, and it would be extremely unlikelythat the system just happens to be so near to the end of its expected lifetime. What thecalculation implies, then, is that the nominalAFOE parameters are unlikely to be correct.Nonetheless, the characteristics of the system could well be within the error barsquoted for the AFOE data.

By contrast, systems using the Lick orcombined Lick/AFOE data sets appear to bestable for at least ten million years and possi-bly much longer. Additional integrations,not shown in Fig. 1, imply that systems withmore massive planets (smaller sin(i)) aregenerally less stable, especially if the orbitsare substantially inclined to one another, and that changing the orbital period of theouter planet within the range allowed by thedata can have a big effect on the stability ofthe system.

Neither the star y Andromedae nor any ofits three planets appear in themselves to beparticularly unusual, but the system as awhole stands out in the currently knownmenagerie of extrasolar planets. Its existenceshows that several massive planets can orbitfar closer to a star than do the gas giants(Jupiter and Saturn) and smaller ice giants(Uranus and Neptune) of our Solar System.So if giant planets all form far enough out in aprotoplanetary disk for ice to condense (theconventional, but not exclusive, view of the-orists6), then the migration or mutual scat-tering (or both) that brings them in towardstheir star is not always violent enough to

destroy themselves or their brethren. The y Andromedae system is just one

example of a wide variety of planetary con-figurations that can be expected to exist inour Galaxy. All of the extrasolar planets thusfar discovered orbiting main-sequence starsare more massive than Saturn, and mosteither orbit very close to their stars or travelon much more eccentric paths than do any ofthe major planets in our Solar System. TheSun’s planets are all either low in mass ortravel on distant orbits from their star, andthey are therefore more difficult to discoverusing the Doppler technique. So it could bethat most planetary systems will turn out tobe like ours.

There is much, much more to come. Withthe projected advances in detection technol-ogy, including schemes to find Earth-likeplanets7, we are at the beginning of a GoldenAge of extrasolar planetary studies.Jack J. Lissauer is in the Space Science Division,NASA Ames Research Center, Moffett Field,California 94035, USA.e-mail: JLISSAUER@ringside.arc.nasa.gov 1. Wolszczan, A. & Frail, D. A. Nature 355, 145–147 (1992).

2. Mayor, M. & Queloz, D. Nature 378, 355–359 (1995).

3. Marcy, G. W. & Butler, R. P. Annu. Rev. Astron. Astrophys. 36,

57–97 (1998); http://www.physics.sfsu.edu/~gmarcy/

planetsearch/planetsearch.html

4. Butler, R. P. et al. Astrophys. J. (submitted); http://www.physics.

sfsu.edu/~gmarcy/planetsearch/upsand/upsand.html;

http://cfa-www.harvard.edu/afoe/

5. Butler, R. P., Marcy, G. W., Williams, E., Hauser, H. & Shirts, P.

Astrophys. J. 474, L115–L118 (1997).

6. Wuchterl, G., Guillot, T. & Lissauer, J. J. Protostars and Planets

IV (eds Mannings, V., Russel, S. & Boss, A. P.) (Univ. Arizona

Press, Tucson, in the press).

7. http://www.kepler.arc.nasa.gov/; http://ast.star.rl.ac.uk:80/darwin/

8. Levison, H. F. & Duncan, M. J. Icarus 108, 18–29 (1994).

ions of the salt and the molecular pore inthe a-haemolysin channel. When a bias isapplied across the membrane, the currentflow is tiny (at the picoampere level), but itcan nonetheless be measured with modernroom-temperature semiconductor elec-tronics.

The innovative step here1 is to sensitizethe nanopore to specific organic chemicalspecies by using b-cyclodextrin as anadapter. b-cyclodextrin, a doughnut-likemolecule made of seven sugar units, diffusesinto the a-haemolysin channel, partiallyobstructing its water-filled pore. Theauthors show that the cyclodextrin moleculeis exquisitely sensitive to different guest mol-ecules. Thus, when bound to cyclodextrin,members of the adamantane family ofpetroleum derivatives can be distinguished,as can members of the group of tricyclicpharmaceuticals that include imipramineand promethazine. These guest moleculeseach bind to b-cyclodextrin for millisecondsand make their presence known by alteringthe electrical conductivity of the a-haemolysin pore in which the b-cyclodex-trin resides.

In neurobiology and biophysics, it is rou-tine to measure picoampere ionic currentspassing through single protein channels inbiological membranes or planar lipid bilay-ers3. Such measurements are usually made todetermine the properties and behaviour ofthe channel, but they can also reveal detailsabout the concentration and other proper-ties of molecules that are able to passthrough or into the channel. Very smallalterations in the distribution of charges lin-ing the pore can give rise to relatively largechanges in the ionic flux, so a membranechannel can serve as an amplifying trans-ducer. Such a transducer allows one to detectextremely small modulations (*0.02 kT) inthe energy barrier to ionic transport4, anddoes so at bandwidths that permit high-speed (*50 msec) measurements. Thus,membrane channels have been used to effecthigh-speed molecular counting andsizing5,6, and show promise for high-speedsequencing of polynucleotides7. Channelshave also been modified to incorporatedesign elements that convert them into sen-sitive, analyte-driven switches8, or genetical-ly engineered to create selective, divalentmetal sensors9.

Bayley and his colleagues have longpuzzled over how to engineer or redesignthe a-haemolysin channel as a biosensorthat can discriminate between differentorganic molecules9. They reasoned that thesensitivity and extraordinarily rich informa-tional content of single-channel recordingsreduce the requirement for the channel tocontain a highly selective binding site. Byfashioning a single-channel detector, ratherthan one that integrates signals from numer-ous sensor molecules, several independent

news and views

660 NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com

Biochemical sensors

Adapting to nanoscale eventsDaniel Branton and Jene Golovchenko

Ahallmark of twentieth-centuryscience has been the continual devel-opment of experimental strategies

to observe individual atomic-scale ‘events’.These strategies ultimately rely on signifi-cantly amplifying the consequences of aselective microscopic interaction, for exam-ple the chemical development of a silverhalide grain in a photographic emulsion, thecondensation of a droplet around a singleion in a cloud chamber, or the charge ampli-fication in electron multiplier devices.

New strategies for detecting and charac-terizing single-molecule events are nowemerging in the biochemical sciences. Thelatest example comes from Gu et al. on page686 of this issue1. This group, working inHagan Bayley’s laboratory, show howmeasurement of ionic transport through asingle, atomic-scale pore in an insulatingmembrane can be used to detect organicmolecules of relative molecular mass as lowas 100. Coupled with highly sensitive semi-

conductor electronics, the membrane–poresystem can amplify these currents ontimescales commensurate with the interac-tion times between the molecule and thepore. The system thus reveals the presence,nature and interaction of single moleculeswith the pore. Remarkably, a single proteinchannel can be adapted for simultaneousanalysis of a mixture of organic molecules.

Gu et al.1 assemble the nanoscale chemi-cal and mechanical building blocks of theirdetector from the toolbox of biochemistry;the key components are shown in Figs 1 and2 of the paper on page 687. Seven a-haemolysin molecules (a bacterial toxin)self-assemble to form a channel with a 15-Å-diameter aqueous pore through a 50-Å-thick lipid bilayer membrane2. The mem-brane separates two chambers filled with aconducting salt solution. Because the lipidbilayer is a nearly perfect insulator, the d.c.electrical conductance across the membraneis determined by the interaction between