Press Release
Free Access
Issue
A&A
Volume 613, May 2018
Article Number A25
Number of page(s) 9
Section Planets and planetary systems
DOI https://doi.org/10.1051/0004-6361/201731973
Published online 28 May 2018

© ESO 2018

1 Introduction

Clever observing strategies and techniques, together with technological progress, are moving comparative exoplanetology towards increasingly Earth-like planets. The coolest stars, in particular, offer clear observational advantages: compared to FGK stars, and everything else being equal, planets around M dwarfs have larger reflex motions, deeper transits (for well-aligned systems), and more favourable star-planet contrast ratios. This has long motivated our radial velocity (RV) search for planets around M dwarfs, which started with the discovery of the first planet around such a star (GJ876b; Delfosse et al. 1998)1. This now amounts to almost 40 detections, which include a few Earth-mass planets and a few super-Earths located in the habitable zones of their host (e.g. Astudillo-Defru et al. 2017b). M dwarfs have also been the focus of several other planet searches with spectacular discoveries, including Proxima Cen b (Anglada-Escudé et al. 2016), TRAPPIST-1 planets (Gillon et al. 2017), and LHS1140b (Dittmann et al. 2017).

Considering their number and their well-characterised selection function, these detections provide us with statistical insights into planet formation (Bonfils et al. 2013). At the same time, many of these individual detections, and all the more so when the planetary properties such as liquid water might exist on their surface, call for follow-up studies to characterise their atmosphere and constrain their structure, composition, and chemistry.

For the subset of planets that transit, transmission and occultation spectroscopy are the characterisation methods of choice. James Webb Space Telescope (JWST) transmission spectroscopy of a few dozen coadded transits of the TRAPPIST-1 planets b, c, and d, for instance, is expected to have sufficient sensitivity to detect O3 in putative Earth-like atmospheres for these planets (Barstow & Irwin 2016). This makes these planets strong candidates for a biomarker detection within the next few years, but one should remember that the TRAPPIST-1 star emits intense extreme ultraviolet (XUV) radiation and frequently produces powerful stellar flares, which together might have sterilised, if not completely stripped out, the atmospheres of at least its closer-in planets (Bourrier et al. 2017; Vida et al. 2017). With stellar activity factored in, the quiet M dwarf LHS1140 and its temperate super-Earth become an appealing alternative. Both TRAPPIST-1 and LHS1140 have been given top priority for JWST Guaranteed Time observing.

Planets that do not transit are generally more difficult to characterise, but can be found closer to our Sun. This translates into both increased brightness and wider angular separation and the closest non-transiting exo-Earths might thus be amenable to characterisation. The maximum angular separation between Proxima Cen and its b planet, for instance,is 37 milli-arcsec and can be resolved at visible wavelength by an 8 m class telescope. To match the daunting 10−7 planet-to-star contrast ratio, Lovis et al. (2017) have proposed to couple the SPHERE extreme adaptive optics system and the ESPRESSO high-resolution spectrograph, which combines the contrast enhancements that one can achieve with high-resolution spectroscopy and high-contrast imaging (Snellen et al. 2014, 2015). Under slightly optimistic assumptions, Lovis et al. (2017) have concluded that a few dozen observing nights at the VLT would detect O2, H2 O, and possibly CH4, which like TRAPPIST-1 represents a historic opportunity to detect biomarkers in the near future. Like TRAPPIST-1, however, Proxima Cen flares strongly and often, which likewise challenges the habitability of its planet (Davenport et al. 2016).

In that context, we report the detection of a planet orbiting a 2½ times more distant but much quieter M dwarf, Ross 128. The planet is only slightly more massive than our Earth, is temperate, and orbits a very nearby, slowly rotating, quiet M dwarf. We discuss the properties of the star in Sect. 2, present the data in Sect. 3, and use archive photometry in Sect. 4 to determine the stellar rotation period. In Sect. 5, we analyse the RVs and demonstrate the presence of a planet and an additional periodicity likely caused by stellar activity. The final model parameters are derived from a Markov chain Monte Carlo (MCMC) algorithm with Gaussian processes in Sect. 6. In Sect. 7, we conclude that, although K2 photometry excludes transit, the low stellar activity and moderate distance from Earth make Ross 128 b a good target for biomarker searches with forthcoming telescopes.

2 Star

Ross 128 entered the literature as the 128th entry in the Ross (1926) catalogue of high-proper motion stars, and has since acquired denominations including Proxima Virginis, FY Virginis, GJ 447, HIP 57548, and LHS 315. The spectral type of this object is M4 and, owing to its proximity, it is one of the brightest representatives of this subclass (V mag = 11.15, Jmag = 6.51, Hmag = 5.95, Kmag = 5.65). With a distance of just 3.4 parsec (π = 295.80 ± 0.54 mas; Gaia Collaboration 2016), Ross 128 is the closest star in the Virgo constellation (α = 11h47m44.4s, δ = +00o4816.4; Epoch = 2000). Including brown dwarfs, it is the 13th closest (sub-)stellar system to the Sun. Ross 128 is moving towards us and will actually become our closest neighbour in just 71 000 yr from now (Dca = 1.9 pc; García-Sánchez et al. 2001).

Mann et al. (2015) have derived its effective temperature Teff = 3192 ± 60 K, mass M = 0.168 ± 0.017, radius 1 R = 0.1967 ± 0.0077, and metallicity [Fe ∕H] = −0.02 ± 0.08. Accordingly, its luminosity is L = 0.00362 ± 0.00039 L. In Astudillo-Defru et al. (2017a), we measured a low Ca II emission level log(R HK)=5.573±0.082$\log (R'_{\textrm{HK}})=-5.573\pm0.082$. The calibration between log(R HK)$\log (R'_{\textrm{HK}})$ and the stellar rotation period Prot in the same paper converts this low calcium-line emission to an estimated rotation period of approximately 100 days, which is indicative ofan age of the order of a few Gyr (Newton et al. 2016).

Table 1

Observed and inferred stellar parameters for Ross 128.

3 Data

From July 26, 2005 (BJD = 2 453 578.46) to April 26, 2016 (BJD = 2 457 504.7), we collected 158 observations with the HARPS spectrograph (Mayor et al. 2003; Pepe et al. 2004). Exposure times were fixed to 900 s. We discarded the 159th measurement that appears in the ESO archives, which is a just a 5 s exposure (March 23, 2015; BJD = 2 456 740.68). We used the high-resolution mode (R = 115 000), with the scientific fibre illuminated by the target and calibration fibre either unused or illuminated by the sky. The data reduction followed the same steps as in all our recent papers. Spectral extraction and calibration relied on the on-line pipeline (Lovis & Pepe 2007), which also gives an initial guess for the RV. An offline processing then refines the RV measurements and their uncertainties (e.g. Astudillo-Defru et al. 2015, 2017b). The line spread function changed significantly when the May 2015 upgrade of HARPS replaced its fibre link with octogonal fibres. In this work, we treat the pre- and post-upgrade data as independent time series, which appear in the figures in red and blue, respectively. Table 5 gives the RV time series in the barycentric reference frame. Before proceeding to the next section, however, we removed the small but significant secular acceleration (dRV/dt = 0.14 m s−1 yr−1), which we computed using the distance and proper motion of Ross 128 (μα = 0.60526∕yr, μδ = −1.21926∕yr; van Leeuwen 2007) and Eq. (2) of Zechmeister et al. (2009).

To complement our HARPS observations, we used archive photometry from both ASAS and K2. The All Sky Automated Survey (ASAS; Pojmanski 1997) observed Ross 128 for over nine years. We retrieved its V -band photometry extracted through the smallest ASAS aperture, ASAS MAG 0. The K2 mission (Howell et al. 2014) observed Ross 128 for 82 days in its Campaign 1. We retrieved the K2 light curves detrended with the EVEREST (Luger et al. 2016) and POLAR (Barros et al. 2016) pipelines from the Mikulski Archive for Space Telescopes (MAST)2.

4 Stellar rotation

Since inhomogeneities such as spots, plages, or inhibition of the convection at the surface of a rotating star can induce apparent Doppler shifts, prior knowledge of the stellar rotation helps eliminate false positive planets. The low log(R HK)$\log (R'_{\textrm{HK}})$ of Ross 128 already indicates that its rotation period is long, ~100 days. Here, we used ASAS and K2 photometry to refine its value.

We only retained the last seven years of the more than nine years of ASAS photometry, since Ross 128 was sampled infrequently prior to BJD = 2 452 500. We subtracted the median value of each observing season, clipped out all 4σ outliers, and computed the generalised Lomb-Scargle periodogram (GLS; Zechmeister & Kürster 2009). As seen in Fig. 1 (top panel), the GLS has obvious power excess for periods around 121 days with power pmax = 0.08. We evaluated the power threshold for a given false alarm probability (FAP) on virtual data sets generated by bootstrap with replacement. The 1% FAP threshold is p1% = 0.07 and the 121 days periodic signal is therefore significant. The phased photometry (middle panel of Fig. 1) shows a ~ 1% semi-amplitude.

The K2 photometry has orders of magnitude better precision than the ground-based measurements and provides quasi-continuous observations during 80 days, but does not cover a full stellar rotation. The ~0.4% trend of this photometry over 80 days is compatible with ~1% variations on a 121 day period (bottom panel of Fig. 1).

thumbnail Fig. 1

Photometry of Ross 128. Top: periodogram of the ASAS V -band photometry. Middle: ASAS photometry phase-folded to P = 121.2 days. Grey filled circles are median values in 0.1-phase bins. Bottom: K2 photometry extracted with the Everest pipeline (Luger et al. 2016) as a function of time. We only show corrected photometry FCOR with the highest quality flag.

5 Evidence for an orbiting planet and additional stellar activity

The raw pre- and post-upgrade RV time series (Fig. 2, top panel) have root mean square (rms) dispersions of 2.1 and 3.0 m s−1, respectively, i.e. well in excess of the ~1.2 m s−1 expected from the photon noise on the individual measurements, and a constant model has a Bayesian information criterion BIC = 618. The GLS periodogram shows a prominent power excess around period of 9.9 days and several other significant peaks (Fig. 2, middle panel). The maximum power pmax = 0.28 is well in excess of the 1% FAP threshold p1% = 0.17 and the detection of a periodic signal is thus highly significant. For a visual sanity check, we phase the RVs to a 9.9 day period (Fig. 1, bottom panel) and see that the signal is well sampled at every phase.

The 9.9 day period is comfortably away from the 121 day stellar rotation period (Sect. 4) and its first few harmonics, which by itself already lends considerable confidence to its interpretation as a planet detection. A Levenberg-Marqardt adjustment of a Keplerian model has rms residuals of 1.9 and 2.6 m s−1 for the two time series and an overall BIC = 429. The planet’s orbital period is Pb =9.9 days, RV semi-amplitude is K1 = 1.7 m s−1, and eccentricity is compatible with zero.

The residuals from the 9.9 day Keplerian model thus remain well in excess of the dispersion expected from pure photon noise, and we searched for periodicities in those residuals using both GLS and Keplerian-GLS (KGLS; Zechmeister & Kürster 2009) periodograms. Whereas the GLS measures the power of a sine fit at each period, the KGLS does so for Keplerian signals, therefore exploring periodicities for a wider range of functional shapes. The GLS (Fig. 3, top panel) has its maximum power, pmax = 0.23, at a P = 51.8 day period, andmultiple other peaks above the 1% FAP threshold. The ~52 days peak remains significant in the KGLS (Fig. 3, middle panel), but the most powerful peak (pmax = 0.28) is now at P = 123 days. This period is very close to the ~121 day stellar rotation period inferred from the ASAS photometry (Sect. 4), and is certainly compatible with it after accounting for the effect of differential rotation. The residuals phase-folded with a P = 123 day period (bottom panel of Fig. 3) suggest a coherent signal at that period with an approximate symmetry around phase 0.5. This approximate symmetry predicts excess power in the 123/2 day second harmonic of the rotation period, as is indeed observed, and the ~52 day peak of the GLS additionally matches a 1 year alias of 123/2 days. The power excess in the residuals of the Keplerian fit is therefore entirely consistent with two-spotted stellar activity modulated by stellar rotation with a ~120 day period. Since the spot configuration is likely to have evolved over the ~11 years of the HARPS measurements, the next section models the effect of this configuration using Gaussian process regression rather than a deterministic physical model.

thumbnail Fig. 2

HARPS radial velocities: as a function of time (top), periodogram (middle) and phase-folded to P = 9.9 days (bottom). The red and blue points represent pre- and post-upgrade measurements, respectively.

thumbnail Fig. 3

Radial velocity residuals after subtraction of the best-fit Keplerian and drift. Top: GLS periodogram is shown. Middle: KGLS periodogram is shown. Bottom: residuals phase-folded to a P = 123 days period are shown. The vertical red dashed line indicates the 121 day rotation period inferred from the ASAS photometry. The red and blue points represent pre- and post-upgrade measurements, respectively.

thumbnail Fig. 4

Transit search in the K2 photometry.

6 Modelling

Our model of the HARPS RVs consists of a single Keplerian function representing the effect of the planetary companion, and we explore the effect of including an additional distant body modelled as a linear velocity drift. Since the RV time series contains additional signals with frequencies close to the rotational rate of the star, its harmonics, and aliases, we modelled the error term as a multivariate Gaussian distribution with a covariance matrix produced by an appropriate kernel function. This includes the effect of correlation between the data points into the model.

For the kernel function, we chose a quasi-periodic kernel, kQP(ti,tj)=A2exp((titj)22τ22ϵsin2(π(titj)P)),\[ k_{\mathrm{QP}}(t_i, t_j) = A^2 \exp\left(-\frac{(t_i - t_j)^2}{2\tau^2} - \frac{2}{\epsilon}\sin^2{\left(\frac{\pi (t_i - t_j)}{\mathcal{P}}\right)}\right), \] which is known to represent adequately the covariance produced by active regions rotating in and out of view (e.g. Haywood et al. 2014; Rajpaul et al. 2015). This kernel function has four hyperparameters, corresponding to the amplitude of the covariance term (A), rotational period of the star ( P$\mathcal{P}$), covariance decay time (τ), and shape parameter (ϵ). To test the robustness of our results with respect to the choice of kernel function, we also explored models employing the simpler squared-exponential kernel, kSE(ti,tj)=A2exp(12(titj)2τ2),\[ k_{\mathrm{SE}}(t_i, t_j) = A^2 \exp\left(-\frac{1}{2}\frac{(t_i - t_j)^2}{\tau^2}\right), \] with only two hyperparameters, A and τ, and no periodic term. In addition, an extra white noise component was added to the model by adding the following term to each of the previous kernels: kWN(ti,tj)=δij[ σi2+SiσJ2+Si+(σJ+)2 ],\[ k_{\mathrm{WN}}(t_i, t_j) = \delta_{ij} \left[\sigma^2_i + S_i \sigma^2_J + S^+_i \left(\sigma^+_J\right)^2\right], \] where δij is the Kronecker delta function, σi is the internal incertitude of the data point taken at time ti; σJ and σJ+$\sigma^+_J$ are the width of the additional noise component for the pre-, and post-upgrade data, respectively; and Si is an indicator variable, whose value is one if observation i is taken before the HARPS fibre upgrade and zero otherwise, and vice versa for Si+$S^+_i$.

In summary, four models were tested and were constructed by combining the two variants for the data model, i.e. a single Keplerian (k1) or a Keplerian plus a linear velocity drift (k1d1), and the two options for the noise term model, i.e. the squared-exponential kernel (sek) and quasi-periodic kernel (qpk). For the Bayesian inference of the model parameters, we set the priors listed in Table 2. The pre- and post-upgrade velocities were treated independently with a different extra white noise amplitude for each, and an offset between these velocities.

The model parameters were sampled using the MCMC algorithm described in Goodman & Weare (2010) and implemented by Foreman-Mackey et al. (2013). The initial positions of 300 walkers were randomly drawn from the prior distribution.The algorithm was run for 40 000 iterations, and the walkers were evolved to different posterior maxima. The separate maxima were identified by clustering the samples in parameter space and the marginal likelihood ofeach mode was estimated using the importance sampling estimator described by Perrakis et al. (2014). In all cases, a single mode exhibited overwhelming evidence with respect to all other secondary maxima. The walkers in the secondary maxima were then replaced by new walkers initiated in the main maximum and the algorithm was run until no further evolution of the samplers was seen. This step took between 15 000 and 40 000 iterations, depending on the model. Then 100 000 additional steps were run, on which the final inference was performed.

Results are reported in Table 3 for each tested model. The inferred results on most planet parameters are independent of the choice of model (see e.g. the marginal posterior of the velocity amplitude in Fig. 5). The most notable exception is the orbital period, which exhibits a bimodal distribution, with modes centred on 9.86 and 9.88 days, in which the difference relative weight of the modes depends on whether the model includes a linear drift or not (Fig. 5). The other parameters that change slightly with the inclusion of a linear drift are the velocity zero-point, the mean longitude at epoch, the amplitude of the covariance, A, and the offset between pre- and post-upgrade velocities. The evolution timescale hyperparameter τ marginal distribution varies significantly between models with different kernel functions.

The relative merits of each model was studied by estimating the marginal likelihood of each model using the importance sampling estimator of Perrakis et al. (2014). This is a biased estimator, so we explored the evolution of the estimation for each model as the size of the sample increased (Fig. 6). After around 5000 samples, the estimator seems to have converged. All models are approximately equally good at explaining the data with a slight preference for the squared-exponential kernel. The final inference on the model parameters was carried out by combining the samples from the four models weighted by their posterior probability; this probability was, in turn, computed assuming all four tested models form an exhaustive set, i.e. their probabilities add up to the value one. The results are listed in Table 4 and the MAP velocity model is presented in Fig. 7.

Table 2

Prior distribution for the model parameters.

Table 3

Orbital elements inferred from each tested model.

thumbnail Fig. 5

Marginal posterior distribution of the orbital period (left) and RV semi-amplitude (right) for the four tested models and the weighted average of all four. The period marginal posterior distribution has a more marked bimodality for models including a non-zero acceleration term.

thumbnail Fig. 6

Evolution of the marginal likelihood estimation for the four tested models with sample size. The known bias of the importance sampling estimator used (Perrakis et al. 2014) is evident, but seems to become negligible for sample sizes larger than around 2000. The inset shows an enlargement of the region with sample sizes between 1000 and 10 000, where a slight preference for the squared exponential models is seen. A small random noise was added in the x-direction to facilitate viewing.

Table 4

Orbital elements inferred from the model mixture.

thumbnail Fig. 7

Radial velocity induced by the planet on its host star, phase-folded to the maximum-a posteriori (MAP) orbital period. The effect of the secular acceleration and the Gaussian process prediction were subtracted from the data (using the MAP parameter values). The orange and blue points correspond to the pre- and post-upgrade velocities, respectively. The inferred MAP additional white noise term (σJ and σJ+$\sigma_J^+$) were added in quadrature to the velocity uncertainties. The grey shaded region extends between the 5th and 95th percentile ofthe model at each orbital phase, computed over 10 000 samples of the merged posterior distribution.

7 Discussion

The msini = 1.35 m Ross 128 b planet orbits Ross 128 with a 9.86 day period, and at 0.049 AU is ~ 20 times closer to its star than the Earth is to the Sun. Since the star is ~ 280 times less luminous than the Sun, Ross 128 b receives just 1.48 times more energy than our Earth. For assumed albedos of 0.100, 0.367, or 0.750, its equilibrium temperature would thus be 294, 269, or 213 K. Using theoretically motivated albedos, the Kopparapu et al. (2017) criteria place the planet firmly outside the habitable zone, while Kopparapu et al. (2013), Yang et al. (2014), and Kopparapu et al. (2016) find it outside, inside and just at the inner edge of the habitable zone. The precise location of the inner edge is therefore still uncertain, as it depends on subtle cloud-albedo feedbacks and on fine details in complex GCM models. The habitable zone most likely will not be firmly constrained until liquid water is detected (or inferred) at the surface of many planets. Meanwhile, it is probably preferable to refer to Ross 128 b as a temperate planet rather than as a habitable zone planet.

A planet just 3.4 parsec away either having liquid water or just shy of having some makes an extremely appealing characterisation target. From the occurrence rate of temperate planets measured by Kepler, Dressing & Charbonneau (2015) estimated that the closest habitable zone planet that transits its star is approximately 11 parsec away. Yet, a stroke of luck could certainly align a closer temperate planet to undergo transits from our position in space, and all RV detections are therefore worth following up with photometry. As for Ross 128 b, existing K2 photometry readily answers whether it transits or not. We phase-folded the de-trended, low-frequency filtered, POLAR K2 photometry (Barros et al. 2016) to the ephemeris computed in the previous section (Fig. 4). Ross 128 b unfortunately does not transit, with central transits of any planet bigger than 0.19 R excluded at least at the 99% confidence level. Non-grazing transits of a more realistic 0.5- or a 1.0-R planet are excluded with very high confidence.

Transit spectroscopy being excluded, we turn to the potential of measuring phase curves. Snellen et al. (2017) estimated that five days of JWST observations could detect the putative atmosphere of Proxima Cen b (see also Kreidberg & Loeb 2016). Ross 128 b is notas favorable however, since its host star is 1.4 times larger and, at near- or mid-infrared wavelengths, 3−4 times fainter than Proxima Centauri. Similar JWST observations for Ross 128 b are thus likely to be prohibitively expensive.

The best odds of characterising Ross 128 b are most likely through combining the contrast improvements achieved with high-angular resolution and with high-spectral dispersion (Sparks & Ford 2002). Snellen et al. (2015) investigated the potential of this strategy for rocky planets around our nearest neighbours and found that a putative temperate exo-Earth orbiting Proxima Cen could be detected in just 10 h on the European ELT (E-ELT). A year later, Anglada-Escudé et al. (2016) detected an actual planet with very similar properties using RV measurements, and the technique was immediately contemplated to characterise that planet. Lovis et al. (2017) proposed to upgrade the SPHERE adaptive optics system of the VLT and inject light from the location of the planet into the ESPRESSO high-resolution spectrograph to detect the planet in few tens of nights, and possibly detect its atmospheric O2 with 60 nights of observations. Ross 128 b again is not quite as favorable as Prox Cen b, since it cannot be resolved by a 10 m-class telescope. Its 15 mas angular separation, however, will be resolved by the 39 m E-ELT at optical wavelengths (> 3λD in the O2 bands) and its expected contrast is similar to that of Prox Cen b, owing to their similar radii and semi-major axes. The two host stars have similar optical apparent magnitudes, leading to similar planetary apparent magnitudes. A realistic investment of E-ELT resources can therefore most likely detect Ross 128 b with high-angular resolution plus high-dispersion spectroscopy, although not as easily as Prox Cen b.

On the flip side, Ross 128 is one of the quietest stars to host a temperate exo-Earth. Newton et al. (2017) measured an Hα equivalent width EW Hα = −0.068  Hα which makes Ross 128 one the most quiescent M dwarfs. They classified stars as active when EW Hα < −1  and, for comparison, measured EW Hα = −4.709  for Proxima Cen b. Stellar activity is probably the highest concern regarding the emergence of life, and even the survival of an atmosphere, on planets orbiting M dwarfs. Restricting the target list to quiet stars would disqualify Proxima Cen b and leave Ross 128 b as the best temperate planet known to date. This will certainly make this new temperate exo-Earth a top target for characterisation with the ELTs.

Acknowledgements

We wish to thank the referee for his remarks that led to an improved manuscript. X.B., J.M.A., and A.W. acknowledge funding from the European Research Council under the ERC Grant Agreement No. 337591-ExTrA. N.C.S. acknowledges the support from Fundação para a Ciência e a Tecnologia (FCT) through national funds and by FEDER through COMPETE2020 by grants UID/FIS/04434/2013&POCI-01-0145-FEDER-007672 and PTDC/FIS-AST/1526/2014&POCI-01-0145-FEDER-016886. N.C.S. was also supported by FCT through Investigador FCT contract reference IF/00169/ 2012/CP0150/CT0002.

References

  1. Anglada-Escudé, G., Amado, P. J., Barnes, J., et al. 2016, Nature, 536, 437 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  2. Astudillo-Defru, N., Bonfils, X., Delfosse, X., et al. 2015, A&A, 575, A119 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  3. Astudillo-Defru, N., Delfosse, X., Bonfils, X., et al. 2017a, A&A, 600, A13 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  4. Astudillo-Defru, N., Forveille, T., Bonfils, X., et al. 2017b, A&A, 602, A88 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  5. Barros, S. C. C., Demangeon, O., & Deleuil, M. 2016, A&A, 594, A100 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  6. Barstow, J. K., & Irwin, P. G. J. 2016, MNRAS, 461, L92 [NASA ADS] [CrossRef] [Google Scholar]
  7. Bonfils, X., Delfosse, X., Udry, S., et al. 2013, A&A, 549, A109 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  8. Bourrier, V., Ehrenreich, D., Wheatley, P. J., et al. 2017, A&A, 599, L3 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  9. Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, The IRSA 2MASS All-Sky Point Source Catalog endrefcommentnewpage [Google Scholar]
  10. Davenport, J. R. A., Kipping, D. M., Sasselov, D., Matthews, J. M., & Cameron, C. 2016, ApJ, 829, L31 [NASA ADS] [CrossRef] [Google Scholar]
  11. Delfosse, X., Forveille, T., Mayor, M., et al. 1998, A&A, 338, L67 [Google Scholar]
  12. Dittmann, J. A., Irwin, J. M., Charbonneau, D., et al. 2017, Nature, 544, 333 [NASA ADS] [CrossRef] [Google Scholar]
  13. Dressing, C. D., & Charbonneau, D. 2015, ApJ, 807, 45 [NASA ADS] [CrossRef] [Google Scholar]
  14. Foreman-Mackey, D., Conley, A., Meierjurgen Farr, W., et al. 2013, Astrophys. Source Code Library [record ascl:1303.002] [Google Scholar]
  15. Gaia Collaboration (Brown, A. G. A., et al.) 2016, A&A, 595, A2 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  16. García-Sánchez, J., Weissman, P. R., Preston, R. A., et al. 2001, A&A, 379, 634 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  17. Gillon, M., Triaud, A. H. M. J., Demory, B.-O., et al. 2017, Nature, 542, 456 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  18. Goodman, J., & Weare, J. 2010, Comm. Appl. Math. Comput. Sci., 5, 65 [CrossRef] [MathSciNet] [Google Scholar]
  19. Haywood, R. D., Collier Cameron, A., Queloz, D., et al. 2014, MNRAS, 443, 2517 [NASA ADS] [CrossRef] [Google Scholar]
  20. Henry, T. J., Walkowicz, L. M., Barto, T. C., & Golimowski, D. A. 2002, AJ, 123, 2002 [NASA ADS] [CrossRef] [Google Scholar]
  21. Howell, S. B., Sobeck, C., Haas, M., & Still, M. 2014, PASP, 126, 938 [Google Scholar]
  22. Kopparapu, R. k., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131 [NASA ADS] [CrossRef] [Google Scholar]
  23. Kopparapu, R. k., Wolf, E. T., Haqq-Misra, J., et al. 2016, ApJ, 819, 84 [Google Scholar]
  24. Kopparapu, R. k., Wolf, E. T., Arney, G., et al. 2017, ApJ, 845, 5 [NASA ADS] [CrossRef] [Google Scholar]
  25. Kreidberg, L., & Loeb, A. 2016, ApJ, 832, L12 [NASA ADS] [CrossRef] [Google Scholar]
  26. Landolt, A. U. 1992, AJ, 104, 340 [NASA ADS] [CrossRef] [Google Scholar]
  27. Lovis, C., & Pepe, F. 2007, A&A, 468, 1115 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  28. Lovis, C., Snellen, I., Mouillet, D., et al. 2017, A&A, 599, A16 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  29. Luger, R., Agol, E., Kruse, E., et al. 2016, AJ, 152, 100 [Google Scholar]
  30. Mann, A. W., Feiden, G. A., Gaidos, E., Boyajian, T., & von Braun, K. 2015, ApJ, 804, 64 [NASA ADS] [CrossRef] [Google Scholar]
  31. Marcy, G. W., Butler, R. P., Vogt, S. S., Fischer, D., & Lissauer, J. J. 1998, ApJ, 505, L147 [NASA ADS] [CrossRef] [Google Scholar]
  32. Mayor, M., Pepe, F., Queloz, D., et al. 2003, The Messenger, 114, 20 [NASA ADS] [Google Scholar]
  33. Newton, E. R., Irwin, J., Charbonneau, D., et al. 2016, ApJ, 821, 93 [NASA ADS] [CrossRef] [Google Scholar]
  34. Newton, E. R., Irwin, J., Charbonneau, D., et al. 2017, ApJ, 834, 85 [NASA ADS] [CrossRef] [Google Scholar]
  35. Pepe, F., Mayor, M., Queloz, D., et al. 2004, A&A, 423, 385 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  36. Perrakis, K., Ntzoufras, I., & Tsionas, E. G. 2014, Comput. Stat. Data Anal., 77, 54 [CrossRef] [Google Scholar]
  37. Pojmanski, G. 1997, Acta Astron., 47, 467 [NASA ADS] [Google Scholar]
  38. Rajpaul, V., Aigrain, S., Osborne, M. A., Reece, S., & Roberts, S. 2015, MNRAS, 452, 2269 [NASA ADS] [CrossRef] [Google Scholar]
  39. Ross, F. E. 1926, AJ, 36, 124 [NASA ADS] [CrossRef] [Google Scholar]
  40. Snellen, I. A. G., Brandl, B. R., de Kok, R. J., et al. 2014, Nature, 509, 63 [NASA ADS] [CrossRef] [Google Scholar]
  41. Snellen, I., De Kok, R., Birkby, J. L., et al. 2015, A&A, 576, A59 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  42. Snellen, I. A. G., Désert, J.-M., Waters, L. B. F. M., et al. 2017, AJ, 154, 77 [NASA ADS] [CrossRef] [Google Scholar]
  43. Sparks, W. B., & Ford, H. C. 2002, ApJ, 578, 543 [NASA ADS] [CrossRef] [Google Scholar]
  44. van Leeuwen, F. 2007, A&A, 474, 653 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  45. Vida, K., ovári, K. Z., Pál, A., Oláh, K., & Kriskovics, L. 2017, ApJ, 841, 124 [NASA ADS] [CrossRef] [Google Scholar]
  46. Yang, J., Boué, G., Fabrycky, D. C., & Abbot, D. S. 2014, ApJ, 787, L2 [NASA ADS] [CrossRef] [Google Scholar]
  47. Zechmeister, M., & Kürster, M. 2009, A&A, 496, 577 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  48. Zechmeister, M., Kürster, M., & Endl, M. 2009, A&A, 505, 859 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]

1

Also detected by Marcy et al. (1998).

All Tables

Table 1

Observed and inferred stellar parameters for Ross 128.

Table 2

Prior distribution for the model parameters.

Table 3

Orbital elements inferred from each tested model.

Table 4

Orbital elements inferred from the model mixture.

All Figures

thumbnail Fig. 1

Photometry of Ross 128. Top: periodogram of the ASAS V -band photometry. Middle: ASAS photometry phase-folded to P = 121.2 days. Grey filled circles are median values in 0.1-phase bins. Bottom: K2 photometry extracted with the Everest pipeline (Luger et al. 2016) as a function of time. We only show corrected photometry FCOR with the highest quality flag.

In the text
thumbnail Fig. 2

HARPS radial velocities: as a function of time (top), periodogram (middle) and phase-folded to P = 9.9 days (bottom). The red and blue points represent pre- and post-upgrade measurements, respectively.

In the text
thumbnail Fig. 3

Radial velocity residuals after subtraction of the best-fit Keplerian and drift. Top: GLS periodogram is shown. Middle: KGLS periodogram is shown. Bottom: residuals phase-folded to a P = 123 days period are shown. The vertical red dashed line indicates the 121 day rotation period inferred from the ASAS photometry. The red and blue points represent pre- and post-upgrade measurements, respectively.

In the text
thumbnail Fig. 4

Transit search in the K2 photometry.

In the text
thumbnail Fig. 5

Marginal posterior distribution of the orbital period (left) and RV semi-amplitude (right) for the four tested models and the weighted average of all four. The period marginal posterior distribution has a more marked bimodality for models including a non-zero acceleration term.

In the text
thumbnail Fig. 6

Evolution of the marginal likelihood estimation for the four tested models with sample size. The known bias of the importance sampling estimator used (Perrakis et al. 2014) is evident, but seems to become negligible for sample sizes larger than around 2000. The inset shows an enlargement of the region with sample sizes between 1000 and 10 000, where a slight preference for the squared exponential models is seen. A small random noise was added in the x-direction to facilitate viewing.

In the text
thumbnail Fig. 7

Radial velocity induced by the planet on its host star, phase-folded to the maximum-a posteriori (MAP) orbital period. The effect of the secular acceleration and the Gaussian process prediction were subtracted from the data (using the MAP parameter values). The orange and blue points correspond to the pre- and post-upgrade velocities, respectively. The inferred MAP additional white noise term (σJ and σJ+$\sigma_J^&#x002B;$) were added in quadrature to the velocity uncertainties. The grey shaded region extends between the 5th and 95th percentile ofthe model at each orbital phase, computed over 10 000 samples of the merged posterior distribution.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.