369, 572 581 doi:1.198/rsta.21.245 Exoplanets: the quest for Earth twins BY MICHEL MAYOR*, STEPHANE UDRY, FRANCESCO PEPE AND CHRISTOPHE LOVIS Observatoire de Genève, Université de Genève, Geneva, Switzerland Today, more than 4 extra-solar planets have been discovered. They provide strong constraints on the structure and formation mechanisms of planetary systems. Despite this huge amount of data, we still have little information concerning the constraints for extra-terrestrial life, i.e. the frequency of Earth twins in the habitable zone and the distribution of their orbital eccentricities. On the other hand, these latter questions strongly excite general interest and trigger future searches for life in the Universe. The status of the extra-solar planets field in particular with respect to very-low-mass planets will be discussed and an outlook on the search for Earth twins will be given in this paper. Keywords: exoplanets; super-earths; Earth-type planets; habitable zone 1. Our Solar System and the diversity of extra-solar planetary systems The discovery of 51 Pegasi b in 1995 [1] immediately brought attention to a crucial physical ingredient of planetary formation: the orbital migration acting during the life of the accretion disk. Despite its early discovery by Goldreich & Tremaine [2], the central role of the orbital migration during the formation of planetary systems was not taken into account before the discovery of the class of hot Jupiters, Jovian planets with short orbital periods. With the growing sample of discovered exoplanets, another important global feature of planetary systems became evident: the diversity of their structure. The architecture of the Solar System seems to be rather marginal compared with the majority of known planetary systems. This point is best illustrated in figure 1, where we plot the orbital eccentricity of exoplanets versus their orbital period. This diagram mostly includes systems with Jovian planets. We can remark on the presence of hot Jupiters with quite short periods, as well as the large scatter of orbital eccentricities, at least for periods larger than a few days. On the same plot, we have also indicated the position of the four largest planets of the Solar System. Opposite to exoplanets, they all have almost circular orbits. Considering the gaseous giant planets, the architecture (P and e) of the large majority of detected planetary systems seems to be quite different compared with the Solar System. However, we have to be cautious with such a statement. The frequency of solar-type stars with detected giant gaseous planets (more *Author for correspondence (michel.mayor@unige.ch). One contribution of 17 to a Discussion Meeting Issue The detection of extra-terrestrial life and the consequences for science and society. 572 This journal is 211 The Royal Society
Exoplanets: the quest for Earth twins 573 1..8.6 eccentricity.4.2 1 1 1 1 1 4 period (days) Figure 1. Orbital eccentricity versus periods for exoplanets. Gaseous giant planets are illustrated by open pentagon symbols, stellar binaries by dots and the four largest planets of the Solar System by starred symbols. The position of the Earth is also especially marked. massive than about the mass of Saturn and with periods smaller than 5 years) is about 7 per cent. Some stars belonging to the complement (93%) exhibit small radial velocity drifts (linear or curved), possibly associated with gaseous giant planets with orbital periods larger than the span of current exoplanetdetection programmes using Doppler spectroscopy. It is important to note that a hypothetical planetary system containing a giant planet on a Jupiter orbit, but with a third of its mass, would have low probability to be detected and would most probably belong to the present 93 per cent of solar-type stars with apparently no giant planets. Simulations of planetary formation of gaseous giant planets carried out by Ida & Lin ([3] and references therein) and Mordasini et al. [4,5] can help understand the large diversity of planetary-system structures. However, we have to keep in mind that these models do not include the full complexity of planetary formation yet. For example, models still do not take into account the simultaneous formation of several gaseous giant planets. More recently, by using the HARPS spectrograph installed on the 3.6 m telescope at the European Southern Observatory (ESO) La Silla [6], we have discovered an extremely rich population of planets with masses in the range of 5 3 M [7 9]. Preliminary observed characteristics of this population of lowmass planets are given in 2. Super-Earths (1 1 M ) and Neptune-mass planets (1 3 M ) seem to be present at least around 3 per cent of solar-type stars
574 M. Mayor et al. 5 4 planetary fraction (%) 3 2 1 1 1 1 M (Earth mass) 1 Figure 2. Histogram of known exoplanets with periods less than 1 days, and scaled by a factor 8 for each exoplanet with masses less than 3 M to compensate for the difference in size of surveys on both sides of that limit. Low-mass planets (super-earths and Neptune-type planets) turn out to be much more frequent than giant planets. Note that the part of the histogram below 1 M is still subject to strong detection biases. [8,9]. Owing to the existing detection bias of the Doppler technique, we have mostly detected low-mass planets with orbital periods smaller than about 1 days. (Neptunes with larger periods have been discovered as well.) The degree of planetary multiplicity among systems with short-period, lowmass planets is extremely high. A first estimation points towards a multiplicity rate of the order of 8 per cent for systems hosting low-mass planets on tight orbital periods. Once again, we are very far from the structure of our Solar System and planets able to sustain living organisms. If we only consider the domain of periods shorter than 1 days (a domain not too much affected by detection biases), we can have a first estimation of the mass distribution from super-earths to gaseous giant planets. To establish such a histogram, we have to take into account the fact that giant planets have been discovered by many teams (a global sample of maybe 3 stars), but lowmass objects with masses less than 5 Earth mass are mostly issued from the HARPS high-precision sample. Figure 2 illustrates the relative mass distribution for planets from a few M to several Jupiter masses. At least for the domain of short periods (less than 1 days), we observe the overwhelming predominance of low-mass planetary companions compared with gaseous giant planets. Quite obviously, we would expect to encounter a similar mass distribution of planets orbiting solar-type stars in the range of periods from 1 days to about 1 year. With this aspect in mind, we will discuss, in 3, the prospect to detect rocky planets in the habitable zone (HZ) around their stars, some of them being potential sites to develop life. For smaller mass M dwarfs, we already have some hints that low-mass planets seem quite frequent at distances from one to a few
Exoplanets: the quest for Earth twins 575 astronomical units. Based on the limited number of low-mass planets detected by microlensing, first statistical estimates indicate a rather high frequency rate of around 3 per cent, although these estimations are still affected by large uncertainties [1]. 2. Super-Earths and hot Neptunes around solar-type hosts The start of operations, in 23, of the HARPS spectrograph at La Silla [6] marked the beginning of a new detection era. HARPS is a high-resolution (R = 11 ) vacuum echelle spectrograph, optimized to search for exoplanets, fed by an optical fibre from the ESO 3.6 m telescope. The spectrograph has an extreme thermal control to achieve a stability of a few milli-kelvins on one night and a few hundredths of Kelvin over the full year. The achieved radial velocity precision is well below 1 m s 1 for non-active stars [11]. As part of the guaranteed observing time (GTO) allocated to our consortium for the construction of the HARPS instrument, we have dedicated about 25 observing nights over 5 years to explore the domain of low-mass planets. This intensive programme has allowed us to discover a rich population of super-earths and Neptunes orbiting solar-type stars [7]. A large programme of a comparable size (28 nights over the next 4 years) is now continuing with HARPS. The goal is to fully characterize the very large number of planetary systems with low-mass planets. This observational effort is needed owing to the small velocity amplitudes and the complex multiplicity of these systems with up to five low-mass planets orbiting some of these solar-type stars. The observing strategy for our programme is designed to detect and study very-low-mass planets. It also requires integrations long enough to decrease the influence of stellar acoustic modes which we would suffer as additional noise to a level below.2 m s 1. With the set of already discovered systems, we can provide preliminary characteristics of the population of low-mass planets and their star hosts. (a) The statistical occurrence of low-mass planets The HARPS sub-programme devoted to the detection of low-mass planets included originally a sample of about 4 slowly rotating nearby FGK stars. After a selection of the less-active stars in the sample (log(r HK) < 4.8), the number of stars diminished down to 28 stars. For the moment, we still do not have enough measurements for 117 stars, but we are already in a position to have preliminary statistics based on the 163 stars with a number of precise radial velocities large enough, and over a time span of several years. The current status suggests that between 39 (conservative) and 58 per cent (optimistic) of solar-type stars of the HARPS high-precision survey host planets with masses below 5 M. Most of these systems with low-mass planets (maybe as many as 8%) and on rather tight orbits are multi-planetary. (b) The correlation with the metallicity of host stars The correlation between the frequency of gaseous giant planets and the metallicity of host stars is striking and extremely well established by different teams [12 14]. If we examine, however, the metallicity of systems having planets
576 M. Mayor et al. with masses less than about 2 M, we obtain a different result. Despite the still limited number of such systems, it seems that we do not have any correlation between the presence of these low-mass planets and the host-star metallicity, a result already mentioned by Udry et al. ([15]; see also [16]). It is interesting to compare these preliminary observational results with existing planetaryformation models and simulations. For instance, it has to be noticed that for the domain of Neptune-mass planets, the simulations also predict a lack of correlation between the planet frequency and the host-star metallicity [5]. (c) The mass distribution If we consider all the low-mass planets presently detected in the HARPS survey, we can already establish a more detailed histogram of planetary masses smaller than 4 M. It confirms the prediction of planet-formation models, suggesting a steep rise of the distribution towards the lowest mass planets, a feature that already appears in the distribution of the already known sample (figure 2; see also [17]). It must be noted, however, that our survey for low-mass planets is still in progress, and that, owing to the present detection biases, the obtained distribution is preliminary and obviously incomplete for masses smaller than 1 M and low-mass planets on long-period orbits. (d) The period distribution For Jovian planets, we observe a peak of their period distribution in the range of 3 5 days, the peak of the so-called hot Jupiters. For super-earth and Neptunetype planets, we already have sufficient detections to obtain a preliminary view of their orbital-period distribution. In contrast to gaseous giant planets, the distribution is regularly rising from very short periods, without a peak at 3 days, up to the point where detection limitations start to play a significant role (approx. 2 days). This observation has important implications for our understanding of the migration of light bodies (type-i migration) in protoplanetary discs. Another striking characteristic of these multi-planetary systems with several very-low-mass planets is the lack of resonances. Their orbital periods are not proportional to the simple ratio of integers. A possible scenario to explain the formation of them has been explored by several teams. The type-i migration of large planetesimals on tight orbits produces collisions and coalescences, inducing the formation of a multi-planetary system with a few low-mass planets. The structure of these multi-planetary systems with several low-mass planets on close-in orbits suggests interesting new constraints to the formation of planetary systems (e.g. [18]). 3. Searching for Earth-type planets in the habitable zone Is it possible to detect terrestrial planets in the HZ of neighbouring stars? For planets orbiting M dwarfs, this is already feasible. The detection of a low-mass planet of 7 M inside the HZ of the M4V-star GJ 581 [19,2] is a good example of this reality. Figure 3 illustrates the HZ in the mass versus the semi-major axis plot for different types of stars (courtesy of F. Selsis). The respective locations of the four planets of the GJ 581 system are indicated as well. The outermost planet,
Exoplanets: the quest for Earth twins 577 1. sun Mstar/Msun (a)(b) (c) (d) GJ 581.1.1 1. orbital distance (AU) 1 Figure 3. Habitable zone (HZ) around dwarf stars with the position of the planets around GJ 581 (F. Selsis 29, personal communication). Light grey region, HZ; thick grey region, possible extension of the HZ owing to uncertainties related to clouds, opacities, three-dimensional circulation. GJ 581 d (7 M, period of 66 days), appears to lie in the HZ. If it was formed by the migration-coalescence mechanism of planetesimals, this planet is probably solid. However, we cannot expect to have enough rocky material in the initial disk of the M4 star to form such a massive super-earth purely from rocks. GJ 581 d is therefore probably largely constituted of icy material, and given its present position, could be an example of an ocean planet [21]. Moreover, a transiting planet with a mass as small as 6 M and an orbital period of only 1.5 days in orbit around a low-mass M4 star (GJ 1214) has recently been detected by Charbonneau et al. [22]. The mean density of the planet suggests a structure with an important ice and water content. In this context, and by analogy, GJ 581 d should probably be considered as a mini-neptune. The important point to recall is, nevertheless, that these low-luminosity stars have an HZ close enough to the central star to allow the detection of super-earths in the HZ with already existing instruments. For Earth-type planets orbiting solar-type stars, the situation is obviously more challenging. Figure 4 illustrates the exoplanets discovered with radial velocities over the past 2 years. Not only should we note the high rate of discoveries after 1995, but also point out the progress in the precision achieved by Doppler spectroscopy. The lower limit of the masses of detected planets decreases from about 15 M in 1995 to less than 2 M in 29 (GJ 581 e; [5]). At the time of writing, the HARPS spectrograph has demonstrated a radial velocity precision and a stability better than.5 m s 1 over several years. About an order of magnitude in precision has still to be gained to detect Earth twins in the HZ of solar-type stars. The most important limitations come from the intrinsic variability of stellar atmospheres, such as acoustic modes, granulation noise and effects owing to magnetic activity in the stellar atmosphere (spots, plages and any anisotropies of the convective zone). Adequate observing strategies aiming at averaging the stellar noise of various time scales will play a very important
578 M. Mayor et al. 1 Jupiter mass (M ) 1 1 Saturn Neptune 1 Earth 199 1995 2 25 21 epoch Figure 4. Mass of exoplanets detected by Doppler spectroscopy as a function of the year of discovery. White circles are used for detections done from all observatories/instruments except HARPS, which are plotted in grey. We can notice the huge increase in planetary discoveries following the 51 Pegasi announcement. We can also observe the amazing lower limit as a function of time for the mass of discovered exoplanets. The recently discovered GJ 581 e (1.94 M ) seems to confirm that Earthmass planets are in view [2]. The HARPS contribution to the detection of planets with masses less than Neptune s one is clearly visible. role for the detection of Earth twins. Interesting progress has been made in this direction as well. Temporal binning of actual HARPS measurements obtained on 1 consecutive nights has demonstrated that it is possible to reduce the radial velocity residuals around long-period orbital solutions down to.35 m s 1 or less [23]. Recent simulations demonstrate that the velocity jitter induced by stellar noise can be averaged down to values as small as.1 m s 1 if an optimal observing strategy is adopted [24,25], and allow the detection of Earth-type planets in the HZ of K dwarfs (figure 5). On the side of instrumental precision [11], we have identified two limitations: on the one hand, slit-illumination variations might produce non-calibratable radial velocity errors at the level of.2 m s 1. New developments in fibre optics and light scrambling will be able to remove this limitation in the near future. On the other hand, the measurements are also affected by wavelength-calibration errors introduced by the finite quality of the spectral reference, which in turn can affect the long-term repeatability of the spectrograph. First experiments carried out on HARPS using a laser comb instead of Th-Ar lamps are extremely promising [26], and we expect this calibration problem to be fully solved within the next couple of years. Instruments providing cm s 1 precision seem feasible nowadays. Other programmes aim at searching for terrestrial planets around solartype stars (selected to have very low chromospheric activity) in the solar neighbourhood. Space missions like the National Aeronautics and Space Administration (NASA) Kepler satellite are looking for rocky planets by means of the method of photometric transits [27]. Taking advantage of the known orbital period and phase of such candidates, the determination of the mass for planets
Downloaded from http://rsta.royalsocietypublishing.org/ on May 7, 218 579 Exoplanets: the quest for Earth twins (a) 3 m per night each 3 nights, binning 1 day, M = 2.5 M, P = 2, sini = 1, log(r HK) = 4.9 1.5 1. RV (m s 1).5.5 1. 1.5 2. (b) 2 4 6 8 time (days) 1 12 14 3 m per night each 3 nights, binning 1 days, M = 2.5 M, P = 2, sini = 1, log(r HK) = 4.9.6 RV (m s 1).4.2 2..4.6 (c) 2 4 6 8 time (days) 1 12 14 3 m per night each 3 nights, binning 1 days, M = 2.5 M, P = 2, sini = 1, log(r HK) = 4.9.6 RV (m s 1).4.2 2..4.6.2.4 f.6.8 1. Figure 5. Simulation of the radial velocity (RV) measurements of a K dwarf hosting a 2.5 Earthmass planet in its HZ (p = 2 days), over more than 13 days. (a) Average of three observations per night. (b) Same but with binned radial velocities in 1 days bin. (c) Phased equivalent radial velocities with the orbital solution.
58 M. Mayor et al. as small as 1 or 2 M seems possible. Evidently, the apparent magnitude of the majority of the transiting planets to be discovered by Kepler will be a serious problem for the precise radial velocity follow-up. Moreover, in most cases, the distances of these systems will be too large and their magnitudes too faint to be interesting for missions searching signatures of life in their atmospheres. To overcome these difficulties, the ESA PLATO mission aims at detecting transiting planets (down to Earth size) for the large majority of bright stars on the sky. Photometric space missions searching for Earth-type planets around bright stars are moreover especially important to explore the diversity of internal structures of low-mass planets. A new perspective is delivered by the combination of future very-low-mass transiting planet candidates and their independent follow-up by the technique of radial velocity. In order to be able to find and characterize Earth-mass planets in the HZ of a solar-type star, we need even better instruments. With this respect, we have to mention the ESPRESSO project, a second-generation HARPS-type spectrograph presently under study and to be installed at the incoherent focus of the Very Large Telescope 8.2 m telescopes at Cerro Paranal (ESO Chile). The instrument is designed to achieve a precision and long-term stability better than.1 m s 1. Despite the intrinsic source of stellar noise, we are confident that Doppler spectroscopy will be able to detect rocky planets in the HZ of solartype stars well before the launch of DARWIN- or terrestrial-planet-finder-type space satellites and thus provide us with the required targets for these challenging missions. The detection of terrestrial planets orbiting the closest solar-type stars has the potential to prepare the list of targets to be scrutinized by future space missions. Both the Doppler spectroscopy and the transit method have the potential to detect Earth twins in the HZ. The detection of terrestrial planets orbiting the closest solar-type stars will offer the possibility to prepare a list of targets to be scrutinized by future space missions. References 1 Mayor, M. & Queloz, D. 1995 A Jupiter-mass companion to a solar-type star. Nature 378, 355 359. (doi:1.138/378355a) 2 Goldreich, P. & Tremaine, S. 198 Disk-satellite interactions. Astrophys. J. 241, 425 441. (doi:1.186/158356) 3 Ida, S. & Lin, D. 28 Toward a deterministic model of planetary formation. V. Accumulation near the ice line and super-earths. Astrophys. J. 685, 584. (doi:1.186/5941) 4 Mordasini, C., Alibert, Y. & Benz, W. 29 Extrasolar planet population synthesis. I. Method, formation tracks, and mass-distance distribution. Astron. Astrophys. 51, 1139. (doi:1.151/ 4-6361/28131) 5 Mordasini, C., Alibert, Y., Benz, W. & Naef, D. 29 Extrasolar planet population synthesis. II. Statistical comparison with observations. Astron. Astrophys. 51, 1161. (doi:1.151/ 4-6361/281697) 6 Mayor, M. et al. 23 Setting new standards with HARPS. Messenger 114, 2 24. 7 Mayor, M. & Udry, S. 28 The quest for very low-mass planets. Phys. Scr. 13, 141. 8 Lovis, C., Mayor, M., Bouchy, F., Pepe, F., Queloz, D., Udry, S., Benz, W. & Mordasini, C. 29 Towards the characterization of the hot Neptune/super-Earth population around nearby bright stars. In Proc. of the IAU Symp., 19 23 May 28, vol. 253, pp. 52 55. Cambridge, UK: Cambridge University Press.
Exoplanets: the quest for Earth twins 581 9 Mayor, M. et al. 29 The HARPS search for southern extra-solar planets. XIII. A planetary system with 3 super-earths (4.2, 6.9, and 9.2 M ). Astron. Astrophys. 493, 639 644. (doi:1.151/4-6361:281451) 1 Gould, A. et al. 21 Frequency of solar-like systems and of ice and gas giants beyond the snow line from high-magnification microlensing events in 25 28. Astrophys. J. 72, 173. (doi:1.188/4-637x/72/2/173) 11 Pepe, F. & Lovis, C. 28 From HARPS to CODEX: exploring the limits of Doppler measurements. Phys. Scr. 13, 147. (doi:1.188/31-8949/28/t13/147) 12 Santos, N. C., Israelian, G. & Mayor, M. 21 The metal-rich nature of stars with planets. Astron. Astrophys. 373, 119. (doi:1.151/4-6361:21648) 13 Santos, N. C., Israelian, G. & Mayor, M. 24 Spectroscopic [Fe/H] for 98 extra-solar planethost stars. Exploring the probability of planet formation. Astron. Astrophys. 415, 1153 1156. (doi:1.151/4-6361:234469) 14 Fischer, D. & Valenti, J. 25 The planet-metallicity correlation. Astrophys. J. 622, 112. (doi:1.186/428383) 15 Udry, S. et al. 26 The HARPS search for southern extra-solar planets. V. A 14 Earth-masses planet orbiting HD 438. Astron. Astrophys. 447, 361 367. (doi:1.151/4-6361:25484) 16 Sousa, S. G., Santos, N. C., Mayor, M., Udry, S., Casagrande, L., Israelian, G., Pepe, F., Queloz, D. & Monteiro, J. P. F. G. 28 Spectroscopic parameters for 451 stars in the HARPS GTO planet search program. Stellar [Fe/H] and the frequency of exo-neptunes. Astron. Astrophys. 487, 373 381. (doi:1.151/4-6361:289698) 17 Udry, S. & Santos, N. C. 27 Statistical properties of exoplanets. Astron. Astrophys. 45, 397 439. (doi:1.1146/annurev.astro.45.5186.11529) 18 Ogihara, M. & Ida, S. 29 N-body simulations of planetary accretion around M dwarf stars. Astrophys. J. 699, 824. (doi:1.188/4-637x/699/1/824) 19 Udry, S. et al. 27 The HARPS search for southern extra-solar planets. XI. Super-Earths (5 and 8 M ) in a 3-planet system, Astron. Astrophys. 469, 43 46. (doi:1.151/4-6361:277612) 2 Mayor, M. et al. 29 The HARPS search for southern extra-solar planets. XVIII. An Earthmass planet in the GJ 581 planetary system. Astron. Astrophys. 57, 487 494. (doi:1.151/ 4-6361/2912172) 21 Léger, A. et al. 24 A new family of planets? ocean planets. Icarus 169, 499 54. (doi:1.116/j.icarus.24.1.1) 22 Charbonneau, D. et al. 29 A super-earth transiting a nearby low-mass star. Nature 462, 891 894. (doi:1.138/nature8679) 23 Lovis, C. et al. 26 An extrasolar planetary system with three Neptune-mass planets. Nature 441, 35 39. (doi:1.138/nature4828) 24 Dumusque, X., Udry, S., Lovis, C., Santos, N. C. & Monteiro, M. In press. Planetary detection limits taking into account stellar noise. I. Observational strategies to reduce stellar oscillation and granulation effects. Astron. Astrophys. (http://arxiv.org/abs/11.2616) 25 Dumusque, X., Santos, N. C., Udry, S., Lovis, C. & Bonfils, X. In press. Planetary detection limits taking into account stellar noise. II. Effect of stellar spot groups on radial velocities. Astron. Astrophys. 26 Wilken, T., Lovis, C., Manescau, A., Steinmetz, T., Pasquini, L., Lo Curto, G., Hänsch, T. W., Holzwarth, R. & Udem, Th. 21 High-precision calibration of spectrographs. Mon. Not. R. Astron. Sos. 45, L16 L2. (doi:1.1111/j.1745-3933.21.85.x) 27 Borucki, W. J. et al. 21 Kepler planet-detection mission: introduction and first results. Science 327, 977 98. (doi:1.1126/science.118542)