Exoplanets down to the size of Earth have been found, but not in the habitable zone—that is, at a distance from the parent star at which water, if present, would be liquid. There are planets in the habitable zone of stars cooler than our Sun, but for reasons such as tidal locking and strong stellar activity, they are unlikely to harbour water–carbon life as we know it. The detection of a habitable Earth-mass planet orbiting a star similar to our Sun is extremely difficult, because such a signal is overwhelmed by stellar perturbations. Here we report the detection of an Earth-mass planet orbiting our neighbour star α Centauri B, a member of the closest stellar system to the Sun. The planet has an orbital period of 3.236 days and is about 0.04 astronomical units from the star (one astronomical unit is the Earth–Sun distance).
Since the discovery of the first exoplanet orbiting a solar-type star in 19951, the number of known planets has not stopped growing: at present, there are more than 750 confirmed planets2 with minimum mass estimates, and over 2,300 transiting planet candidates detected with the Kepler satellite that are awaiting confirmation. Two main detection techniques have led to this impressive number of discoveries: the radial-velocity technique, which measures the change in the velocity of the central star due to the gravitational pull of an orbiting planet; and the transit method, which measures the small reduction in flux when a planet passes in front of its host star. These two techniques are complementary; the former gives the minimum mass of a planet (minimum because the orbital inclination of the planet is unknown), whereas the latter gives a planet’s radius.
One of the major challenges in the search for exoplanets is the detection of an Earth twin, that is, an Earth-mass planet orbiting in the star’s habitable zone. In this regard, α Centauri B is one of the most interesting targets. At a distance of 1.3 parsecs, it is a member of the closest stellar system to the Sun, composed of itself, α Centauri A and Proxima Centauri. It also exhibits low stellar activity, similar to the solar activity level, usually associated with a small perturbing contribution of intrinsic stellar activity to the measured radial velocities. α Centauri B is cooler than the Sun (effective temperature3,4,5,6 5,214 ± 33 K, spectral type K1V), and has a smaller mass that our parent star7 (0.934 ± 0.006 solar masses). These two conditions ease the detection of a potentially habitable planet using radial velocities; the relative coolness implies a habitable zone closer to the star, and the smaller mass leads to a stronger radial-velocity variation for a similar-mass planet. In addition, theoretical studies show that the formation of an Earth twin is possible around α Centauri B8,9. Finally, the brightness of the star (visual magnitude 1.33) would allow for an efficient characterization of the atmosphere of potentially orbiting planets.
An Earth twin induces a typical radial-velocity variation of a few tenths of a metre per second on a star like α Centauri B. Such detections, technically possible with the most stable high-resolution spectrographs, are however challenging due to the presence of intrinsic stellar signals inducing a radial-velocity ‘jitter’ at the level of a few metres per second, even for quiet stars.
We report here the discovery of a planetary companion around α Centauri B, unveiled by a radial-velocity signal with a semi-amplitude K of 0.51 m s−1, a period P of 3.236 d, and a semi-major axis a of 0.04 astronomical units (au). This planet, with a minimum mass similar to that of Earth, is both the lightest orbiting a solar-type star and the closest to the Solar System found to date. Being much closer to its parent star than Earth is to the Sun, it is not an Earth twin. However, the small amplitude of the signal shows that the radial-velocity technique is capable of reaching the precision needed to detect habitable super-Earth planets around stars similar to our Sun, or even habitable Earths around cooler stars (that is, M-dwarfs). In addition, statistical studies of exoplanets suggest that low-mass planets are preferentially formed in multi-planetary systems10,11,12. There is therefore a high probability that other planets orbit α Centauri B, perhaps in its habitable zone.
High-precision radial velocities
High-precision measurements were obtained for α Centauri B between February 2008 and July 2011 using the HARPS spectrograph (Supplementary Information section 1, and Supplementary Data). HARPS is a high-resolution (R = 110,000) cross-dispersed echelle spectrograph installed on the 3.6-m telescope at La Silla Observatory (ESO, Chile). This instrument has demonstrated a long-term precision of 0.8 m s−1, thereby becoming the most powerful machine with which to hunt for exoplanets using the radial-velocity technique12,13,14. α Centauri B was observed with HARPS following an intensive observational strategy, optimized to sample high- and medium-frequency intrinsic stellar signals15, which makes it possible to model and consequently remove their perturbing contributions. The star was observed every possible night three times, with exposure times of ten minutes, and with measurements optimally separated by two hours14.
The raw radial velocities of α Centauri B (Fig. 1a) exhibit several contributing signals that we could identify. Their origin is associated with instrumental noise, stellar oscillation modes, granulation at the surface of the star, rotational activity, long-term activity induced by a magnetic cycle, the orbital motion of the binary composed of α Centauri A and B, light contamination from α Centauri A, and imprecise stellar coordinates.
In the following, we will consider each of these contributions separately, modelling and removing them one by one, from the largest to the smallest amplitude. The model parameters estimated for each contribution will then be used as initial conditions for a global fit that will remove all the identified radial-velocity signals. In the residuals, we will be able to search for small-amplitude planetary signals.
Perturbing signals for planet searches
α Centauri B is a quiet star among the targets monitored in searches for low-mass planets. However, the very high precision of HARPS allows us to discern in the measurements different perturbing signals at the metre-per-second level. Compared to the radial-velocity signal induced by terrestrial planets, a few to a few tens of centimetres per second, these perturbing signals are non-negligible and must be modelled and mitigated before searching for small-mass planets.
Guiding noise and other possible instrumental noise are not considered in the data error bars. Their global effect is estimated to be 0.7 m s−1, given the typical dispersion obtained for the most stable stars of the HARPS high-precision programme14.
Stellar oscillation modes
α Centauri B exhibits high-frequency oscillation modes16,17, with typical periods of less than five minutes. An exposure time of ten minutes thus averages out efficiently, to a level of a few centimetres per second, the signal due to oscillation modes.
α Centauri B is a solar-type star and has therefore an outer convection zone responsible for a granulation pattern on its surface. Depending on temperature, granulation cells have positive or negative radial velocities, resulting in a non-zero global radial-velocity signal when their individual contributions are integrated over the disk of the star, weighted by the luminosity of the cells. The granulation effect introduces radial-velocity variations on timescales ranging from 15 min to several hours18,19. For α Centauri B, models of granulation20 suggest an r.m.s. radial velocity of 0.6 m s−1.
Rotational activity signal
Owing to stellar rotation and the Doppler effect, one side of the star has a positive radial velocity compared to the average, while the other side has a negative one. However, if a spot (darker or brighter than the mean stellar surface) is present on one side of the star, the velocity balance will be broken and a residual radial velocity will be measured. With stellar rotation, a spot will move from one side of the stellar disk to the other, introducing periodic signals at the stellar rotational period and the corresponding harmonics21. The lifetime of spots on the stellar surface is typically a few rotational periods22, so after several rotations, the configuration of spots will be different, thus changing the phase and amplitude of the signal.
The radial velocities of α Centauri B show a clear signal at 38.7 d (Figs 1d and 2b), which corresponds to the rotational period of the star23. An efficient way to model rotational activity effects is to select radial-velocity measurements over time intervals of a few rotational periods, and fit sine waves at the rotational period and the corresponding harmonics21 (Supplementary Information section 2). The best fit for the rotational activity signal for each observational season can be seen in Fig. 3.
Long-term activity signal
During a solar-like magnetic cycle, the number of spots on the stellar surface (dark spots, plage faculae) varies from zero to several hundreds. Inside these spots, a strong magnetic field is present, which freezes the convection24,25,26,27,28. For the Sun, as for other stars similar to α Centauri B in spectral type29, convection induces a blueshift of the stellar spectra30,31,32. Therefore, no convection means no convective blueshift inside these regions, and so the spectrum of the integrated stellar surface will appear redshifted. Because a redshift means a measured positive radial velocity, a positive correlation between the magnetic cycle variation and the long-term radial-velocity variation is then expected.
α Centauri B shows signs of weak but detectable chromospheric activity, evidenced by the re-emission in the centre of the Ca ii H and K lines (the log(R′HK) activity index). α Centauri B exhibits a magnetic cycle with a minimum amplitude of AR′HK ≈ 0.11 dex (Fig. 2a). To correct the radial-velocity effect due to the magnetic cycle (see Fig. 1b and c), we assume a linear correlation between the log(R′HK) activity index and the activity-related radial-velocity variation33 (that is, both variations have the same shape; Supplementary Information section 3).
The orbital period of the binary composed of α Centauri A and B is PAB = 79.91 yr (ref. 7). The HARPS observations of α Centauri B cover an interval of only four years. The orbit of the system over such an interval can then adequately be approximated by a second order polynomial (see Fig. 1a).
Owing to the close separation on the sky between α Centauri A and B, the spectra of B can be contaminated by light coming from A when the observing conditions are poor. The resulting effect on the radial-velocity measurements was estimated and problematic observations discarded (Supplementary Information section 4).
Imprecise stellar coordinates
When estimating stellar radial velocities with regard to the barycentre of the Solar System, we need to remove the component of the velocity of the Earth in the direction of the star. Imprecise coordinates will then result in an imprecise correction and therefore in a residual signal in the radial velocities. This effect was first pointed out when searching for planets around pulsars, when the times of arrival were varying periodically in time owing to imprecise pulsar coordinates34. Owing to the circular orbit of the Earth around the Sun, this signal will be a sinusoid with a one-year period. α Centauri A and B are gravitationally bound, resulting in a binary orbital motion, which has to be corrected to obtain precise coordinates for α Centauri B (Supplementary Information section 5).
Removing the various signals
The approaches used to remove or mitigate the effects of the various signals potentially masking the existence of a low-mass planet have been described in the preceding paragraphs. For contamination coming from α Centauri A, we removed observations with a too-high level of contamination. For instrumental noise and granulation that cannot be easily modelled, the estimated radial-velocity contribution from each source is quadratically added as white noise to the raw error bars. For the other effects, parametric models have been proposed. A global fit to the data, including the binary signal, the long-term activity signal and the rotational activity effect, involves 23 free parameters (Supplementary Information section 6).
A planet with a minimum mass that of the Earth
The generalized Lomb-Scargle periodogram35 of the radial-velocity residuals shows two peaks at 3.236 and 0.762 d, with a false alarm probability (FAP) lower than a conservative 1% limit (Fig. 4a). These two periods are aliases of one another. A careful analysis of the structure in frequency of the periodogram suggests that the peak at 3.236 d is the true signal (Supplementary Information section 7).
The global model makes use of parameters associated with different timescales. One could thus worry whether the signal at 3.236 d could have been introduced during the process of eliminating the stellar signals. We performed Monte Carlo simulations to determine if this could be the case, and concluded that the signal is real and not an artefact of the fitting process (Supplementary Information section 8).
The peak at 3.236 d in the radial-velocity residuals is significant with a FAP of 0.02%. Using a Markov chain Monte Carlo algorithm coupled to a genetic algorithm to characterize the Keplerian solution, we obtained a signal with a well-constrained period and amplitude. The eccentricity is poorly constrained but is compatible with zero within a 1.4σ uncertainty (Supplementary Information section 9). To fit this planetary signal simultaneously with the other contributions to the radial velocities, we added a sinusoidal signal representing the circular planet orbit to the global fit (Supplementary Information section 6). The observed dispersion of the residuals around the final solution is 1.20 m s−1 and the reduced χ2 value is 1.51 (with 26 parameters for 459 radial-velocity points). The semi-amplitude of the planetary signal is K = 0.51 ± 0.04 m s−1, which corresponds to a planet with a minimum mass of 1.13 ± 0.09 Earth masses considering a stellar mass of 0.934 solar masses and with an orbital period of P = 3.2357 ± 0.0008 d. The orbital and planet parameters are given in Table 1. In the residuals of the global fit, a signal with an FAP of 0.3% is present; however, it could have multiple origins (Supplementary Information section 6).
In Fig. 5, we show the radial-velocity measurements corrected for stellar and binary effects, folded in phase with the 3.236-d period, superimposed on the derived solution for the planetary signal. In Fig. 4b we show that the 3.236-d signal conserves its phase for each observational year, which is expected for a planetary signal.
An important piece of information about the inner composition of an exoplanet is obtained when the planet is transiting its parent star, allowing its radius to be measured. Combined with the real mass estimate, the radius leads to the average density of the planet. In the present case, given a stellar radius36 of 0.863 times the solar radius and assuming the radius of the planet is that of the Earth, the planet transit probability is estimated at 10%, with a transit depth of 10−4. The detection of a planet transit, only possible from space, would allow us to confirm the expected rocky nature of the detected planet around α Centauri B.
The r.m.s. radial velocity induced by the stellar rotational activity amounts to 1.5 m s−1 on average. The detection of the tiny planetary signal, with a semi-amplitude K = 0.51 m s−1, thus demonstrates that stellar activity is not necessarily a definitive limitation to the detection of small-mass planets. Using an optimized observational strategy and the present knowledge about activity-induced radial-velocity effects, it is possible to model precisely and mitigate activity signals, and therefore improve considerably the planet detection limits.
With a separation from its parent star of only 0.04 au, the planet is orbiting very close to α Centauri B compared to the location of the habitable zone. However, the observed radial-velocity semi-amplitude is equivalent to that induced by a planet of minimum mass four times that of Earth in the habitable zone of the star (P = 200 d; ref. 37). The HARPS spectrograph therefore has the precision required to detect a new category of planets, namely habitable super-Earths. This sensitivity was expected from simulations of intrinsic stellar signals15, and actual observations of planetary systems14.
The optimized observational strategy used to monitor α Centauri B is capable of reaching the precision needed to search for habitable super-Earths around solar-type stars using the radial-velocity technique. However, it requires an important investment in observation time, and thus only few targets can be observed over several years. Recent statistical analyses and theoretical models of planetary formation suggest that low-mass rocky planets and especially Earth twins should be common12,38,39,40. We are therefore confident that we are on the right path to the discovery of Earth analogues.
The data presented here were obtained with the ESO 3.6-m telescope at La Silla Paranal Observatory, Chile. We thank the Swiss National Science Foundation (FNRS) for continuous support. We thank R. Mardeling for English revision. N.S. and X.D. acknowledge support by the European Research Council/European Community under FP7 through Starting Grant agreement number 239953, as well as from Fundacao para a Cîencia e a Tecnologia (FCT) through programme Cîencia 2007 funded by FCT/MCTES (Portugal) and POPH/FSE (EC), and in the form of grants PTDC/CTE-AST/098528/2008 and PTDC/CTE-AST/098604/2008.
This text file contains all data used in our analysis, in tab-separated format.
About this article
Acta Astronautica (2019)