© S. Dalal et al. 2021

1 Introduction

The first exoplanet orbiting a “solar-type star”, 51 Pegasi b (Mayor & Queloz 1995), was discovered using the radial velocity (RV) technique. This technique is currently the second-most successful planet detection method, having as of Jan. 21, 2021, detected 913 exoplanets ranging from Earth-mass to more massive than Jupiter¹. It remains much more efficient than the transit method at detecting long-period planets, and thanks to improved RV precision and increasingly long temporal baselines, RV surveys are ideal for discovering analogs of Jupiter and Saturn. The ELODIE and SOPHIE exoplanet survey at Observatoire de Haute Provence (France) is one of the longest operating RV surveys, with a total time baseline of over 25 yr which extends from before the discovery of 51 Pegasi b.

This paper presents discoveries of new substellar and low-mass stellar companions of solar-type stars – cool Jupiters (CJs), brown dwarfs (BDs), and M-dwarfs or low-mass stars (SCs) – from two ongoing exoplanet detection surveys with the SOPHIE spectrograph (Sect. 2). These substellar objects are classified according to their masses. CJs are defined as massive planets with masses above 0.3 M_J and orbital periods above 100 days (Wittenmyer et al. 2020). BDs are substellar objects in the 13–75 M_J mass range and occupy the domain between massive planets and stars. SCs are low-mass stars with masses above approximately 75 M_J (0.072 M_⊙). The mass limit that separates BDs from massive planets is conventionally the minimum mass required to fuse deuterium in the core of the substellar object, that is 13 M_J (Boss et al. 2003; Chabrier et al. 2014), and the limit between BDs and SCs comes from the minimum mass for hydrogen fusion in the core, that is 75 M_J (Chabrier et al. 2000). Both boundaries depend to a small extent on the metallicity of the object (Chabrier & Baraffe 1997; Spiegel et al. 2011).

Despite the discovery of thousands of giant planets and BDs, the statistics on their occurrence and properties are still very incomplete. Most statistical studies of massive planets are for hot and warm Jupiters (Howard et al. 2010; Fernandes et al. 2019), and CJs, which are suspected to be more abundant than hot Jupiters, are poorly characterized as a population (Wittenmyer et al. 2020). Like our Solar System’s Jupiter and Saturn, they are dynamically dominant in their system and influence the formation and evolution of any interior planets, including habitable worlds (Raymond et al. 2006; Morbidelli et al. 2012; Raymond & Izidoro 2017; Bryan et al. 2019). Detecting more CJs will help to provide detailed statistics on massive planets with a range of periods from a few days to a decade.

The two leading planet formation models are core accretion (Pollack et al. 1996; Mordasini et al. 2009; Guilera et al. 2010) and disk instability (Cai et al. 2010; Boss 2011). Previous studies (Santos et al. 2017; Narang et al. 2018; Schlaufman 2018) suggest that giant planets might divide into two distinct giant planet populations. The metallicity of the host stars of giant planets with a mass above 4 M_J is, on average, lower than that of the host stars of giant planets with mass under 4 M_J. This hints towards possibly two distinct planet formation scenarios for these two populations. Planet-formation models can probably form bodies up to 40 M_J (Ida & Lin 2004; Alibert et al. 2005; Mordasini et al. 2009), which suggests that low-mass BDs might form like massive planets, through the disk gravitational instability scenario. High mass BDs, by contrast, are likely to form like stellar binary systems, through molecular cloud fragmentation (Ma & Ge 2014).

Brown dwarfs are interesting as the transition between the formation mechanisms of giant planets and stars probably runs through their population. One interesting characteristic of the BD population is that few are detected at short orbital periods, a feature known as the BD desert (e.g., Halbwachs et al. 2000; Grether & Lineweaver 2006; Sahlmann et al. 2011b). New detections have shrunk this desert in recent years (Csizmadia & CoRot Team 2016; Wilson et al. 2016; Kiefer et al. 2019; Šubjak et al. 2020), but there is still a detection deficit for orbital periods under 100 days and masses between 30 and 60 M_J (Ma & Ge 2014; Ranc et al. 2015; Kiefer et al. 2019). One obvious path towards a better understanding of substellar mass objects is to detect additional objects in and around the desert to better characterize its shape, which in turn will constrain the giant planet and BD formation and evolution processes.

This paper combines the RV technique with astrometry to detect substellar objects and constrain their mass. The RV technique can, in most cases, only determine the minimum mass of the companion (m sini, where m is the true mass and i the orbital inclination of the planet). As mass is what distinguishes giant planets from BDs, and from SCs, the inclination ambiguity must be lifted to asserting the true nature of a substellar companion (e.g., Díaz et al. 2012; Curiel et al. 2020; Kiefer et al. 2021). As just one example, HD 5388 b was first announced as a likely gas giant (Santos et al. 2010) but turned out to be a BD companion when Sahlmann et al. (2011a) detected its astrometric signature in the HIPPARCOS measurements of HD 5388. Here as well, we combine astrometry and RV measurements to overcome the sin i ambiguity and constrain the true mass of the companions (McArthur et al. 2010; Tokovinin & Latham 2017).

Section 2 presents the two SOPHIE surveys which produced these new detections, while Sect. 3 explains how the observations were performed and the data reduced. Sections 4–6 respectively discuss the spectral analysis, the stellar activity, and the RV analysis of the SOPHIE observations. In Sect. 7 we analyze the HIPPARCOS and Gaia astrometric measurements. In Sect. 8 we review the new detections of CJs, BDs and SCs. Finally, Sect. 9 discusses and summarizes our results.

2 Description of the SOPHIE surveys

The SOPHIE volume-limited survey for giant planets and BDs observes a catalog of about 2300 FGK stars in the northern sky (δ > + 00: 00 : 00). These targets are within 60 pc of the Sun and have B−V between 0.35 and 1.0. Around 2000 of those stars have SOPHIE observations at this point (Hébrard et al. 2016; Kiefer et al. 2019). Two ongoing programs with the SOPHIE spectrograph contribute new substellar companions to this paper.

2.1 Giant planets survey

The goal of this ongoing volume-limited program is to increase the number of detections of giant planets orbiting nearby FGK stars, and to identify candidates for follow-up studies: multiplanetary systems for dynamics and transiting systems for structure characterization. This survey constrains the distributions of exoplanet parameters, which helps understand the diversity of planetary systems. The new CJs presented in this paper are the continuation of work by Boisse et al. (2010), Moutou et al. (2014), Díaz et al. (2016), and Hébrard et al. (2016).

2.2 Brown dwarfs survey

The giant planet survey stops observing any star with a companion which is clearly outside the planetary mass range (i.e., ≥13 M_J), and these stars are transferred to the BD survey which has looser RV precision requirements. The goal of the BD program is to obtain an unbiased inventory of companions within and about the BD mass regime for orbital periods up to 10 yrs. This includes stellar companions with mass (or m sin i) > 75 M_J (0.072 M_⊙), because detection of these stellar companions is inevitable while aiming for completeness for BD and because they probe the connections between massive BDs and low-mass stellar companions. The new BDs and SCs presented in this paper are the continuation of work by Díaz et al. (2012), Bouchy et al. (2016), Wilson et al. (2016), and Kiefer et al. (2019).

3 Spectroscopic data

3.1 Observations

We present new observations of 27 stars with the SOPHIE spectrograph, a cross-dispersed, environmentally stabilized echelle spectrograph at the 1.93 m telescope of Observatoire de Haute Provence (OHP). SOPHIE has been in operation since 2006 and covers the 3872–6943 Å wavelength range (Perruchot et al. 2008; Bouchy et al. 2009a). The spectrograph is fed through two optical fibers, one of which is always illuminated by starlight from the telescope. Our observations illuminate the second fiber with light from the background sky to estimate its contribution to the on-star spectrum (obj AB mode), and they were carried out in the high-resolution (R = 75 000) mode of the spectrograph. Wavelength calibrations and drift measurements were obtained approximately every two hours during the night, as well as at the beginning and end of each night. In 2011, the circular-section fiber was replaced with octagonal-section fiber in the fiber link to improve the stability of the spectrograph illumination (Perruchot et al. 2011; Bouchy et al. 2013), and the pre and post upgrade data have distinct characteristics. This work, therefore, distinguishes two SOPHIE datasets, labeled SOPHIE and SOPHIE+, depending on whether the spectra were taken before or after this SOPHIE upgrade.

3.2 Data reduction

The SOPHIE pipeline extracts the spectra and cross-correlates them with a numerical mask (Bouchy et al. 2009a). The cross-correlation functions (CCFs) are produced by considering masks corresponding to their stellar type and incorporating all of the spectral orders. The CCFs were then fitted with Gaussians to derive the radial velocities (RVs) (Baranne et al. 1996; Pepe et al. 2002). The exposure time was adjusted to reach a signal-to-noise ratio (S/N per pixel at 550 nm) of at least 50 for the giant planet survey and 30 for the BD survey (observations obtained while a BD target was still in the giant planet survey targeted S∕N= 50), under diverse weather conditions. Spectra that are significantly contaminated by the Moon were discarded, as were all spectra with less than half of the median S/N for a given target or a large uncertainty on the RV measurements. Spectra for the stars of the giant planet survey and the BD survey have an average S/N of 53.2 and 43.8, respectively. The RVs were also corrected from the CCD charge transfer inefficiency following Bouchy et al. (2009b). Parameters such as Full Width at Half Maximum (FWHM), contrast, stellar rotational velocity (v sin i_⋆, where i_⋆ is the inclination of the star’s rotational axis with respect to the line of sight), and the Bisector Inverse Span (BIS), were also derived from the CCF by the SOPHIE reduction software following the method of Boisse et al. (2010). A 5 m s⁻¹ systematic uncertainty is added in quadrature to the uncertainty of the RV measurements obtained before the June 2011 upgrade to account for the poor scrambling properties of the early exposures. The main characteristics of the SOPHIE and SOPHIE+ data sets are summarized in Table A.1.

4 Spectral analysis

For each star, we performed spectral analyses of an optimally weighted average of all the SOPHIE spectra unaffected by Moon pollution. We used the ARES+MOOG², following closely what was done in Santos et al. (2004) and Sousa et al. (2008) to derive the effective temperature T_eff, the surface gravity logg, and the metallicity [Fe/H]. Using those derived spectroscopic parameters as input, stellar masses were derived from the calibration of Torres et al. (2010) with the correction of Santos et al. (2013). Their uncertainties were computed from 10 000 random draws of the stellar parameters within their error bars and assuming Gaussian distributions. Table A.2 lists the resulting stellar parameters, as well as the and v sin i, which were obtained following the approach of Boisse et al. (2010).

5 Stellar activity analysis

Activity in the atmosphere of the star can alter the shape of stellar lines (Queloz et al. 2001), as can face-on binaries (e.g., Díaz et al. 2012; Wright et al. 2013). This gives rise to apparent variations in RV signatures which can mimic a planetary signal. We use various indicators, such as BIS, FWHM, and , to probe whether the observed RV signal stems from spectral-line profile changes related to stellar activity.

We evaluate the expected activity-related RV scatter σ_a (Table A.3) from our measurement of the mean index (Table A.2) using the Santos et al. (2000) σ_a vs. relation. The most active star in the giant planet survey sample is HD 204277, with −4.50 ± 0.11 and σ_a ≃ 21.8 m s⁻¹. The dispersion of its measured RVs is on the order of this σ_a value, and Sect. 8.1 presents a detailed discussion of the nature of its signal. We exclude it from the rest of our analysis and the next sections are dedicated to the 26 remaining stars. Their σ_a range between 5 and 8 m s⁻¹ for the giant planet survey targets and from 5 to 22 m s⁻¹ for the BD survey targets. The dispersion of the measured RVs of these 26 stars is significantly larger than their estimated σ_a (see Table A.1 for dispersion of the measured RVs).

To further investigate whether the observed RV signals can be caused by stellar activity, we looked for correlations between the measured RVs and two probes of the line shape, namely the FWHM and BIS. We calculated the Pearson correlation coefficients and the significance of the correlation (p-value) (Table A.3) and find that none of them is significant.

6 Radial velocity analysis

We use the Yorbit software (Ségransan et al. 2011; Bouchy et al. 2016) to fit the Keplerian RV signal induced by a companion. Yorbit uses a genetic algorithm to produce starting values for a Levenberg-Marquardt optimization, which in turn provides the priors for a Markov chain Monte Carlo (MCMC) estimation of the error bars following Díaz et al. (2014).

The first step of the RV data analysis is to identify significant periodic signals in the data. This is done by computing the Generalized Lomb-Scargle (GLS) periodogram algorithm of the RV measurements (Zechmeister & Kürster 2009). In the case of giant planet survey targets, we then estimate the false-alarm probabilities (FAPs) of the tentative signals through a bootstrap permutationof the data. The GLS periodogram is however known to fail for the signals induced by companions in highly eccentricity orbits (Zechmeister & Kürster 2009). For two of the stars (BD+631405 and HD 331093), we, therefore, used the PLANETPACK software to compute Keplerian periodograms (Baluev 2013). We ran its kpow command for a range of orbital periods with a frequency step of 0.01 and an upper limit on eccentricity. The latter is needed to control the computational cost, as the closer emax approaches unity, the longer is the computational time.

When a significant period is identified, the RVs are first fit using a single Keplerian orbital model initialized at that period. The following parameters are varied while fitting the single Keplerian model: P the orbital period, K the RV semi-amplitude, e the orbital eccentricity, ω the orbital argument of periastron, T_p the time of passage through pericenter, γ_S and γ_S+ the RV offsets for SOPHIE and SOPHIE+ datasets, respectively. To obtain robust confidence intervals for the free parameters, we use 1000 MCMC iterations.

In mostcases, a single companion (planet, BD, or stellar) on a Keplerian orbit is a good description of the RV measurements. For HD 124330 and BD+550362, however, linear and quadratic drifts are considered in addition to one Keplerian. Six targets, HD 8291, HD 25603, HD 76332, HD 187057, HD 211961, and HD 352975, have too few RV measurements with either SOPHIE or SOPHIE+ to constrain both γ_S and γ_S+. Fitting for both parameters, therefore, produced unrealistically high values for γ_S+- γ_S : they are on the order of 100 m s⁻¹ when the RV offsets between SOPHIE and SOPHIE+ measurements are known to be bounded by 50 m s⁻¹ (Bouchy et al. 2013; Kiefer et al. 2019). We therefore fix γ_S+- γ_S to 0 when deriving Keplerian solutions for those six stars.

Section 8 discusses the results of the RV analysis of the newly detected objects in detail, and these objects are divided into CJs, BDs and SCs based on their minimum mass. Three tables summarize the Keplerian orbital elements as well as the m sin i and a derived parameters for each category. Table 1 reports the orbital parameters of the six CJs, while Fig. 1 plots their Keplerian fits as a function of time along with their RV measurements and residuals. Table 2 and Fig. 2 provide the same information for the four BD candidates, and Figs. B.3 and B.4 display phase-folded plots for respectively the CJs and BDs. Finally, Table A.4 reports the orbital parameters of the sixteen SCs and Fig. B.5 shows their Keplerian fits and RV measurements as a function of the orbital phase.

7 Astrometry analysis

Astrometry can complement the RV orbital information and measure the inclination of the systems. We, therefore, used the HIPPARCOS and Gaia DR1 data to perform the astrometric analysis presented below. The true mass of the companion (either CJ, BD, or SC) is expressed by M_c.

7.1 HIPPARCOS astrometry

7.1.1 Selecting the orbit candidates

As the input sample for SOPHIE giant planet survey was selected from the HIPPARCOS catalog, all 26 stars listed in Table A.5 were observed by the HIPPARCOS satellite (Perryman et al. 1997). After a preliminary examination of all 26 systems, we selected 12 systems with indications of significant orbital motion³. This subset contains all sources with nonstandard HIPPARCOS solution types (‘1’ for stochastic solutions, ‘5’ for standard solutions, and ‘7’ or ‘9’ for accelerated solutions). We list upper mass limits in Table A.5 for those of the 26 systems where one could be derived.

7.1.2 Analysis of the HIPPARCOS astrometry

We analyze the Intermediate Astrometric Data (IAD) of the most recent HIPPARCOS reduction (Van Leeuwen 2007) for signatures of orbital motion, following Sahlmann et al. (2011b) where a detailed description of the method can be found. Fixing the other orbital elements to their values from the RV orbit (Tables 1, 2, and A.4) we adjust seven free parameters to the IAD of each star: the inclination of the orbit I_p, the longitude of its ascending node Ω, the parallax ϖ, and offsets to both coordinates (Δα^⋆, Δδ) and both proper motion components (, Δμ_δ). We search a two-dimensional grid in the two nonlinear parameters, I_p and Ω, for its global χ²-minimum. The false-alarm probability of the derived astrometric orbit was then evaluated through a permutation test employing 1000 pseudo-orbits. The uncertainties in the solution parameters were derived by Monte Carlo simulations which also propagate the uncertainties in the RV parameters. This method has a good track record in reliably detecting orbital signatures in the HIPPARCOS IAD (Sahlmann et al. 2011b,a; Díaz et al. 2012; Sahlmann & Fekel 2013).

Table A.5 lists the target names and the basic parameters of the HIPPARCOS observations relevant for the astrometric analysis. The solutiontype S_n indicates the astrometric model adopted by the (Van Leeuwen 2007) reduction. The code is ‘5’ for the standard five-parameter solution, whereas it is ‘7’ and ‘9’ when the model included proper-motion derivatives of first and second order, respectively. N_orb represents the number of orbital periods covered by the HIPPARCOS observation time span, N_Hip is the number of astrometric measurements, and σ_Λ is their median precision. Outliers in the IAD can very substantially alter the outcome of the astrometric analysis and therefore need to be clipped.

Even when the astrometric data detected no orbital signal (i.e., the derived significance is low), we can set an upper limit on the companion mass by determining the minimum astrometric signal a_min that would be detectable for the individual target. When the data cover at least one complete orbit, Sahlmann et al. (2011b,a) showed that an astrometric signal-to-noise of S∕N ≳ 6−7 is required to obtain a detection at the 3σ level, where and a is the semi-major axis of the detected orbit. We conservatively use a S/N-limit of 8 to derive an upper limit on the astrometric signal of (1)

where the factor 1 − e² accounts for the most unfavorable orientation of I_p = 90° and Ω = 90°, in which the astrometric signal is given by the semi-minor axis of the orbit. The last column in Table A.5 lists the corresponding companion mass upper limit for systems with N_orb ≳ 1.

We detect the astrometric orbit of 7 sources with a significance of at least 2σ, as determined by the permutation test. Those are listed in Table 3, where we also include BD+031552 even though its permutation significance is low. However, the null probability is small, the decrease in residual amplitude is large, the derived orbit looks visually good. Given that this source also has an accelerated HIP solution type, we indicate this solution as viable despite the failure of the permutation test. The 8 sources with orbit determinations then include all 7 sources with nonstandard HIPPARCOS solution types.

Table A.6 lists the updated parallaxes, proper motion, inclination, and ascending node of the orbits of those 8 stars. The 8 orbits detected from HIPPARCOS astrometry include no CJ, and we astrometrically detect the orbit of one of the BD candidates, HD 205521 (Fig. 3). Its low inclination however shifts the companion mass from the BD to the SC domain, as discussed in Sect. 8.4. The remaining astrometric orbits are for SCs are plotted in Fig. B.6.

Fig. 1 Keplerian orbit of the RV variations for the six CJs. SOPHIE and SOPHIE+ RV measurements are indicated in blue squaresand red circles, respectively.

7.2 Gaia astrometry

Recent studies have shown that information from the first Gaia Data Release (DR1) can constrain the mass of RV-detected nontransiting companions (Kiefer et al. 2019; Mugrauer & Michel 2020). The GASTON code uses the astrometric excess noise, as published in DR1, to determine the amplitude of the astrometric motion of the host star of a known RV-detectedcompanion (Kiefer 2019; Kiefer et al. 2019, 2021). This measures, or at least constrains from below, the inclination of the orbit, and thus resolves inclination ambiguity on the true mass. Starting from the RV-derived parameters of the companion, the parallax of the star and an estimation of its mass as priors, GASTON uses a MCMC algorithm to explore the space of astrometric orbits compatible with the measured astrometric excess noise, hereafter ε_DR1, and hence to constrain the possible inclinations. It accounts for the systematics in the Gaia DR1 catalog and distinguishes between stars belonging to the TGAS (Tycho-Gaia Astrometric Solution) and the secondary subset of the DR1 catalog (Lindegren et al. 2016; Kiefer et al. 2021). For the targets in the TGAS dataset, the inputs to the 5-parameter fit include a 24-yr older Tycho-2 or HIPPARCOS-2 astrometric point in addition to 14 months of Gaia measurements, leading to an improved proper motion accuracy. Targets in the secondary dataset only have Gaia observations, and the 14 months time baseline available for DR1 was insufficient to reliably decouple parallax from proper motion. DR1 only published positions (and magnitudes)for those stars and we used their Gaia DR2 (Gaia Collaboration 2018) absolute parallax as a replacement for the missing DR1 parallax, with a conservatively allocated 10% relative uncertainty. Lastly, GASTON accounts for the effect of the companion’s light on the position of the photocenter, by modeling the luminosity of both components from their mass (Appendix A in Kiefer et al. 2021). Here we additionally impose that the secondary contributes no more than 10% of the total luminosity since the systems we study are single-lined spectroscopic binaries.

Out of 26 targets studied in the present paper, 25 targets figure in the DR1 catalog, with HD 211961 as the sole exception. Table A.7 summarizes for those 25 stars the DR1 inputs to GASTON. Fortargets in the TGAS (respectively secondary) dataset of the DR1, the reported value of ε_DR1 is considered significant beyond noise, if above 0.85 mas (resp. 1.2 mas) (Kiefer et al. 2021). Targets with ε_DR1 above these thresholds can lead to, but does not guarantee, a true mass measurement from an allowed orbit inclination range that excludes 90°. Targets with ε_DR1 below the thresholds cannot be distinguished from noise, so can only provide an upper-limit (lower-limit) on the mass (inclination).

Tables 4, A.8, and A.9 summarize the results of this analysis. The results for duplicate sources should be considered with care, as the Gaia observations for a single source can be mistakenly divided between two Gaia “sources” with different IDs (Gaia Collaboration 2016; Lindegren et al. 2016). The astrometric solution in such cases is thus based on an incomplete astrometric data series.

Similar to what happens with the HIPPARCOS astrometry, the 8 orbits detected with the Gaia astrometry include 7 SCs, a single BD candidate, HD 205521, which Gaia DR1 likewise demonstrates is actually an SC (Sect. 8.4), and no CJs. The remaining detected orbits are detected for the SCs. As discussed at length in Sect. 8, the RV-orbit of the HD 76332 system is incompatible with its observed ε_DR1 of 2.19 mas: with the orbital elements of the RV orbit of for HD 76332 as prior, all GASTON Monte Carlo draws produce an astrometric excess noise significantly below 2.19 mas.

Finally, a word of caution. GASTON does not take into account both companions of multiple systems at once, since this would require a totally unpractical number of MCMC steps – and thus a ridiculous time – to converge. We thus examine multiple companions with GASTON, one-by-one, which means that the derived mass of each is overestimated when both contribute significantly to the astrometric signal. In the HD 114762 system (Kiefer et al. 2019) the wide orbit companion HD 114762 B with a_B ~ 130 au has little effect on the astrometry of the host star compared to the closer companion with P ~ 84 days, but a larger mass for the companion would lead to larger perturbations on the star’s orbit.

Fig. 2 Orbital solutions of the RV variations of four BDs with 25.05 MJ ≤ m sin i ≤ 60.27 MJ are depictedhere. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles, respectively.

7.3 Comparison HIPPARCOS and Gaia astrometry results

We detect astrometric orbits for eight stars in each of the HIPPARCOS and Gaia astrometric analyses, with five being common to both. The results of the two astrometric analyses agree within 3σ for the true mass of these five companions, with the exceptions of HD 26596. As discussed in Sect. 8.4, the Gaia analysis finds a significantly lower mass for HD 26596 B.

8 Results

In this section, we discuss each SOPHIE RV detection in the light of the astrometric data from HIPPARCOS and Gaia, and we classify them as CJs, BDs, or SCs based on their minimum mass (or true mass for some companions). After this RV and astrometric analysis, we have six CJs, three BD candidates, and 17 low-mass stellar companions. We also present a detailed analysis of the RV signal from HD 204277, previously presented as a tentative planetary candidate and which we conclude is due to magnetic activity.

8.1 HD 204277 : activity rather than a planet

HD 204277 is a V = 6.7 magnitude F8V star located at a distance of 33 parsecs from the Sun and has a stellar mass of 1.14 ± 0.08 M_⊙. It is an active star which has = −4.50 ± 0.11 and therefore has σ_a ≃ 21.8 m s⁻¹. Butler et al. (2017), using the HIRES spectrograph, listed this system as an SRC⁴ with a period of 30 days. The Generalized Lomb-Scargle periodogram of the SOPHIE RV measurements for HD 204277 instead shows a strong signal around a 250 days period. However, the periodograms of both BIS and FWHM peak at the same period as the RV periodogram, and so does the periodogram of (Fig. 4). This suggests starspots, magnetic cycles, and other stellar activities as the likely origin of the RV signal, and there is thusa low probability that it is due to a planet. We conclude that the observed RV variations are likely due to the stellar activity, and will discuss the nature of this signal in a dedicated forthcoming paper.

8.2 Cool Jupiters

8.2.1 BD+450564

BD+450564 is a K1 type star with mass 0.81 ± 0.07 M_⊙ and of −4.98 ± 0.11. It was only observed after the SOPHIE upgrade and the final dataset has 14 SOPHIE+ RV measurements. Its GLS periodogram (Fig. B.1) has a clear peak around 300 days, with a FAP below 0.1%. The Keplerian fit of the 14 RVs has a period of 307.8 ± 1.5 days and a semi-amplitude of 47.7 ± 2.8 m s⁻¹, and it corresponds to a minimum mass of m sini = 1.36 ± 0.12 M_J. The orbit of BD+450564 b is quasi-circular, with e = 0.12 ± 0.06. Due to the low (and possibly null) eccentricity, we fit for the time of possible transit instead of the time of periastron. The 3.3 m s⁻¹ dispersion of the RV residuals is consistent with the typical uncertainty of the RVs.

No significant astrometric orbit is detected in HIPPARCOS data. The Gaia astrometric excess noise ε = 0.25 leads to a 3σ upper limit on the true mass of the giant planet companion, that is M_c < 31.4 M_J.

8.2.2 BD+550362

BD+550362 is a K3 type star with mass of 0.91 ± 0.10 M_⊙ and = −5.11 ± 0.12. Its 22 RV measurements were all obtained after the SOPHIE upgrade. The GLS periodogram (Fig. B.1) shows two peaks around 260 and 2000 days with FAPs below 0.1 and 1%, respectively. We tested three models for this target: k1 (one-planet Keplerian), k1d2 (one-planet Keplerian and a quadratic drift), and k2 (two-planet Keplerian). An F-test comparison of the k1 and k1d2 models reveals that the k1d2 model gives a better fit than the k1 model with confidence of 96% (F-value = 24.3). The F-test comparison of the k1d2 and k2 models finds no significant improvement for the k2 model (F-value = 3.8, Probability = 84.5%). The k2 model fits the shorter period planet well, but its orbital parameters for the longer period planet are very uncertain because our data do not sample its period well. We, therefore, adopt the k1d2 model for our final fit.

The k1d2 Keplerian model fit of the RVs gives a 25.1 ± 1.7 m s⁻¹ semi-amplitude, and a minimum mass of m sini = 0.72 ±0.08 M_J. The orbit of BD+550362 b has a 265.6 ± 1.0 days period and a significant eccentricity of e = 0.27 ± 0.06. The 3.8 m s⁻¹, dispersion of its residuals is consistent with the noise of the RV measurements. The RV drift is well fitted by a quadratic trend and we obtained quadratic (d2) and linear (d1) drift coefficients of 1.4365 ± 0.0005 m s⁻¹yr⁻² and 1.76 ± 0.04 m s⁻¹yr⁻¹. If one forces a circular orbit for this additional companion, the k2 fit converges to a CJ with an orbital period of at least 3600 days and a mass of at least 2.1 M_J. Other orbital solutions, with longer periods, higher eccentricity, and higher masses, are however equally consistent with the measurements.

Additional RV data will therefore be needed to reveal the nature of this outer companion. The 3σ upper limit on the mass of BD+550362 b from the GASTON astrometric analysis of its host star, M_c < 72.45 M_J, is not sufficiently tight to nail down its planetary nature.

8.2.3 BD+631405

BD+631405 is a K0 type star with a mass of 0.82 ± 0.08 M_⊙ and has low magnetic activity ( = −4.93 ± 0.13). Its 16 RVs were all obtained after the SOPHIE spectrograph upgrade. Figure B.2 shows both a GLS periodogram and a Keplerian periodogram of those velocities. We produced the Keplerian periodogram using the kpow command of the PLANETPACK package for orbital periods between 3 and 1500 days with a frequency step of 0.01 days⁻¹ and an emax = 0.91 upper limit on the eccentricity. The GLS periodogram has no obvious peak, but its Keplerian counterpart shows a clear one around 1200 days.

A one-planet Keplerian fit of the RVs finds a highly eccentric orbit (e = 0.88 ± 0.02) of period 1198.5 ± 60.8 days. The semi-amplitude, 186.0 ± 14.9 m s⁻¹ of the Keplerian fit corresponds to a planet with minimum mass, m sini = 3.96 ± 0.31 M_J. The poor sampling of the fast RV changes at periastron and the incomplete coverage of the orbital period are the main reasons for the large uncertainty on the orbital period and T_p (the time of possible transit is reported in Table 1 rather than the time of periastron). The 4.6 m s⁻¹ dispersion of the residuals is consistent with the noise of the RV measurements. From the ε = 0.39 in the Gaia DR1 astrometry, we set a 3σ upper limit on the true mass of the giant planet of M_c < 40.23 M_J.

Fig. 3 Top panel: astrometric orbit of HD 2055213σ. North is up and East is left. The solid red line shows the model orbit and the open circles mark the individual 1-dimensional HIPPARCOS measurements. Bottom panel: O–C residuals for the normal points of the orbital solution (filled blue circles) and of the five-parameter model without companion (open squares).

Fig. 4 Lomb-Scargle periodogram of the observed radial velocities, BIS, FWHM and for HD 204277. The black dashed line represents the 0.1% False Alarm Probability.

8.2.4 HD 124330

HD 124330 is a G4IV type star which has a mass of 1.15 ± 0.08 M_⊙ and = −5.27 ± 0.12. Its 58 RV measurements were all secured after the SOPHIE upgrade. The GLS periodogram shows a clear peak around 270 days (Fig. B.1). The k1d1 (one-planet Keplerian and a linear drift) model fit of the radial velocities gives a significant eccentricity, e = 0.34 ± 0.05 and an orbital period of 270.7 ± 1.2 days. This Keplerian has a semi-amplitude of 22.8 ± 1.2 m s⁻¹ and corresponds to a minimum planet mass of m sini = 0.75 ± 0.06 M_J. The 5.4 m s⁻¹ dispersion of the residuals is slightly larger than the noise of the RV measurements.

The 2.95 ± 0.07 m s⁻¹yr⁻¹ linear drift of the RVs indicates that this system contains an additional outer companion. For a circular orbit and the 3900 days minimum period defined by the extent of the measurements, its mass is at least 0.85 M_J, but longer orbital periods and large eccentricities are obviously also compatible with the RV measurements. Neither HIPPARCOS nor Gaia detects any significant astrometric orbit and we set a 3σ upper limit on the true mass of the giant planet of M_c < 53.76 M_J.

8.2.5 HD 155193

HD 155193 is a magnetically quiet F8IV type star with = −5.14 ± 0.25 and a 1.22 ± 0.08 M_⊙ mass. Its 73 RV measurements were all acquired after the SOPHIE upgrade. Their GLS periodogram (Fig. B.1) peaks around 350 days. The one-planet Keplerian fit to the radial velocities has a 352.6 ± 2.6 days orbital period, a 19.5 ± 1.4 m s⁻¹ semi-amplitude, a modestly significant eccentricity of e = 0.21 ± 0.08, and it corresponds to a m sini = 0.75 ± 0.06 M_J minimum mass. Although the Keplerian fit appears robust, the nearly one year orbital period induces a large phase gap in our coverage of the planetary orbit, which limits the precision on orbital parameters such as ω and T_p. Table 1 reports the time of possible transit instead of the time of periastron. The 7.05 m s⁻¹ dispersion of the residuals is slightly above the estimated accuracy of the measurements. The ε = 0.55 Gaia astrometric excess noise translates into a M_c < 63.21 M_J 3σ GASTON upper limit for the true mass of the companion.

8.2.6 HD 331093

HD 331093 is a magnetically quiet K0 type star with = −5.10 ± 0.13 and a 1.03 ± 0.08 M_⊙ mass. It wasobserved before and after the SOPHIE upgrade for a total of 20 RV measurements. Figure B.2 shows both their GLS periodogram and a Keplerian periodogram, which we produced using the kpow command of the PLANETPACK package for orbital periods between 3 and 1400 days, with a 0.01 days⁻¹ frequency step, and an emax = 0.7 upper limit on the eccentricity. Similar to the also eccentric BD+631405 system, the GLS periodogram is featureless but the Keplerian periodogram shows a clear peak, here around 600 days.

The Keplerian fit of the radial velocities gives a 621.6 ± 16.1 days orbital period, a high eccentricity of e = 0.59 ± 0.03, a 43.6 ± 2.2 m s⁻¹ semi-amplitude, and it corresponds to a m sini = 1.5 ± 0.1 M_J minimum mass. The dispersion of the residuals, 3.3 and 3.2 m s⁻¹ for respectively SOPHIE+ and SOPHIE, is consistent with the typical uncertainty on the RV measurements. The Gaia astrometric excess noise only provides a very loose 3σ upper limit on the true mass of the companion, M_c < 270.5 M_J, and therefore contains no additional insight on the nature of HD 331093 b.

8.3 Brown dwarfs

The BDs presented in this section are “BD candidates”, as the outcome of their astrometric analysis is compatible with a substellar mass but does not demonstrate one.

8.3.1 BD-004475

BD-004475 is a G0 star with a 0.81 ± 0.10 M_⊙ estimated mass. It was observed both before and after the SOPHIE upgrade, for a total of 13 radial velocities. The Keplerian fit finds a 723 days eccentric orbit (e = 0.39) and corresponds to a m sini = 25-M_J BD candidate companion with a 1.48 AU semi-major axis. The dispersion of the residuals of the Keplerian fit is 6.19 m s⁻¹. We detect no significant astrometric motion in the HIPPARCOS data, and the 0.72 mas astrometric excess noise measured by Gaia translates into a 125 M_J (~0.12 M_⊙) upper limit on the true mass of the companion. BD-004475 b can therefore still be either a BD or an M-dwarf star.

8.3.2 HD 184601

The 15 RVs of HD 184601 were all obtained after the SOPHIE upgrade. It is a G0 type star with a 0.95 ± 0.07 M_⊙ mass. HD184601 b has an orbital period of 849 days and an eccentricity of 0.49, and its minimum mass is 60.27 ± 2.15 M_J. The dispersion of the residual of the keplerian fit to the RVs is 7.69 m s⁻¹, and compatible with their measurement noise. Neither HIPPARCOS nor Gaia detect any significant astrometric motion, and the upper limit on the true mass is very loose, 276 M_J (~0.2 M_⊙).

8.3.3 HD 5433

HD 5433 is a G5 star with a mass of 0.98 ± 0.07 M_⊙. Its 20 RV measurements identify a companion with an orbital period of 576 days, the highest eccentricity (e = 0.81) among the BD candidates presented in this paper, and a minimum mass of 49 ± 3.4 M_J. The astrometric analysis of the Gaia DR1 data only sets a loose upper limit on its true mass of 236 M_J (~0.23 M_⊙), which leavesthe true nature of HD 5433 b undetermined.

8.4 Low-mass stars

Table A.4 and Fig. B.5 present the Keplerian orbits of the 17 stellar companions with masses (or m sini) above 75 M_J (or equivalently 0.072 M_⊙). All of these except for HD 8291 are detected for the first time. Their orbital periods range from 30 to 4198 days and their eccentricities from 0.03 to 0.65, and we discuss a few of the more interesting ones below.

HD 205521

It is a G5 type star and has a mass of 1.10 ± 0.082 M_⊙. The Keplerian orbit has an orbital period of 2032.32 days, an eccentricity of 0.17, and a semi-amplitude of 406.91 ± 5.82 m s⁻¹. The 2.57 m s⁻¹ dispersion of its residuals is compatible with the measurement uncertainties of the RVs. The minimum mass of the companion is m sini = 26.62 ± 1.64 M_J and firmly into BD candidate territory, but our analyses of the HIPPARCOS and Gaia astrometric data both find that the orbit is close to face-on and that the companion is actually a star. The astrometric orbit is firmly detected in the HIPPARCOS time series, with a ~24 mas semi-major axis of the photocenter orbit around the center of mass. The 2.6 mas Gaia DR1 astrometric excess noise points towards a smaller value of ~10 mas (1σ confidence interval) but is compatible with the HIPPARCOS estimate at the 3σ level. The orbit of the companion is thus within 3–12° of face-on, and its true mass is within the 0.13–0.7 M_⊙ range. While initially classified as a BD candidate from its m sin i, it is a low-mass star.

BD+031552

It is a K5 type star with an uncertain stellar mass. Its substellar companion has a 879 days orbital period, an eccentricity of 0.47, and a minimum mass of 133.2 ± 49.0 M_J. The analysis of the HIPPARCOS time series of BD+031552 detects its astrometric orbit with low significance but sets a 3-σ upper limit of 0.22 M_⊙ on the mass of the companion. The upper limit from the Gaia DR1 astrometric excess noise is looser and therefore not informative.

HD 162735

It has a stellar companion with an orbital period of 4197 days, an eccentricity of 0.65, and a minimum mass of m sini = 227.5 ± 13.1 M_J. Neither HIPPARCOS nor Gaia detect any significant astrometric motion, and the 3σ upper limit on the true mass of the companion from the Gaia DR1 excess noise is a loose 0.73 M_⊙. HD 162375 B has the highest eccentricity and the longest orbital period of the 17 low-mass stellar companions presented in this work.

8.4.1 Updated parameters for HD 8291b

HD 8291

It is a well-known wide binary system with a low-mass (0.073 M_⊙) companion at a 2222 AU projected distance (Baron et al. 2015, and references therein). dos Santos et al. (2017) first reported the detection of a closer-in m sini = 124.6 ± 2.1 M_J companion on an eccentric orbit (e = 0.680 ± 0.009) of 1852.3 ± 3.2 days orbital period using HARPS and SOPHIE data (Table 3 in dos Santos et al. (2017)). Adding 15 new SOPHIE and SOPHIE+ RV data to the previous data, we refine this detection to an orbital period of 1862.53 ± 2.9 days, an eccentricity of 0.632 ± 0.001, and m sini = 99.48 ± 5.86 M_J. The poorly covered periastron in dos Santos et al. (2017) led to an overestimated eccentricity and therefore a generally less accurate Keplerian orbit.

This 5-yr period M-dwarf companion induces a motion of the star with semi-major axis >4.6 mas, which is indeed detected in Gaia DR1 as a 1.98 mas astrometric excess noise. To 1σ confidence GASTON finds i_c = 49 ± 12° and a companion mass of 142 M_J (~0.14 M_⊙), but the 3σ confidence region includes edge-on orbits.

8.4.2 Incompatible orbits

The companion of HD 76332 has an orbital period of 2489 days, an eccentricity of 0.14, and m sini of 216.01 ± 12.46 M_J. The semi-major axis of the reflex orbit is at least 13 mas, and thus expected to be well detected by both HIPPARCOS and Gaia. They indeed both detect significant nonlinear motion even though their measurements do not span a full orbit, but cannot agree on a common true mass for this companion. The HIPPARCOS time series detect the reflex orbital motion with high significance, and find a a = 25 ± 5 mas semi-major axis and a 3σ range of 0.24–0.84 M_⊙ for the companion mass. Masses above ~0.6 M_⊙ would give rise to an unobserved second peak in the CCFs of this system and can therefore be excluded. The Gaia DR1 astrometric excess noise of 2.2 mas is well below the minimum expected semi-major axis. This is qualitatively expected from Gaia DR1 only covering ~1/6 of the orbit (HD 76332 is not part of the TGAS dataset), but none of the GASTON simulations, whatever the mass of their companion, could produce a Gaia DR1 astrometric excess noise above ε = 1.55 mas. We note that HD 76332 has a duplicate source and that duplication may be the sign that something went wrong during the reduction for this specific source. For non-TGAS sources, the GASTON simulations fit the proper motion through the orbital motion observed with Gaia and we suspect that the best explanation for this discrepancy is an inaccurate proper motion fit in the DR1. If not fitting for the proper motion, GASTON leads indeed to I_c within 40–70°. We adopt the HIPPARCOS result as the best estimation of HD 76332 B’s true mass.

HD 26596

It has astellar companion in an eccentric orbit (e = 0.4) at 890 days and a minimum mass of 121 ± 7 M_J. To 3σ confidence, GASTON derives an upper limit of <0.19 M_⊙ for the true mass and the HIPPARCOS analysis derives a range of 0.23–0.43 M_⊙. This source again has a duplicate in the DR1 database, and we suspect that its Gaia astrometric excess noise is underestimated. We therefore again prefer the HIPPARCOS confidence range for the mass of this companion to its likely underestimated GASTON upper limit.

9 Discussion and conclusions

This paper reports the discovery of 6 CJs, 3 BD candidates, and 16 SCs with the SOPHIE spectrograph at OHP. We also present updated orbital parameters for the low-mass star HD 8291 B. We analyzed the HIPPARCOS and Gaia astrometry to constrain the inclinations of the orbits and hence, the true masses of the companions. Figure 5 shows the period-mass distribution of these 26 companions.

The newly detected giant planets have periods that range from 266 to 1198 days, and minimum masses between 0.72 and 3.96 M_J. All 6 can therefore be classified as CJs. They span a wide range in eccentricity, from the possibly circular orbit of BD+450564 b (e = 0.12 ± 0.06) to the highly eccentric one of BD+631405 b (e = 0.88 ± 0.02. BD+631405 b is the fifth most eccentric known exoplanet, after HD 20782b (Jones et al. 2006; O’Toole et al. 2009), HD 80606 b (Naef et al. 2001; Moutou et al. 2009), HD 7449 A b (Dumusque et al. 2011), and HD 4113 A b (Tamuz et al. 2008), and is today the most massive of these 5 highly eccentric giant planets. Figure 6 shows the period-eccentricity distribution of all known giant planets (0.3 M_J ≤ m or m sini ≤ 13 M_J). Its red circles are taken from The Extrasolar Planets Encyclopaedia⁵ (Schneider et al. 2011) while the cyan circles represent the 6 new discoveries.

The radial velocities of BD+550362 and HD 124330 show additional drifts, hinting at an additional longer-period companion in both planetary systems. Additional observations of both targets will help characterize those companions. Most of the giant planet hosts in this paper are metal-rich, and their average metallicity is 0.07. We compare the metallicities of two samples of giant planet hosts (including the new detections), divided between planetary masses above and below 4 M_J. Figure 7 shows the cumulative metallicity distribution for host stars of these two samples, which a Kolmogorov-Smirnov (K-S) test⁶ shows has a very low p-value of 9.53 × 10⁻⁸ to be drawnfrom a single population. This reinforces the previous observations that the average metallicity of the host star decreases as the mass of the planet increases (Santos et al. 2017; Narang et al. 2018; Swastik et al. 2021). A possible interpretation is that the population of giant planets is bimodal, with giant planets (M_p ≤ 4 M_J) forming via core accretion and more massive giant planets (M_p > 4 M_J) forming via the disk gravitational instability.

The three new BD candidates have minimum masses between 25 and 60 M_J and orbital periods between 576 and 850 days. Figure 8 shows the period-eccentricity distribution of all known BDs. All three newly detected BD candidates have significant eccentricities. The lightest of the three, BD-004475, with a minimum mass of 25 M_J, has an eccentricity below 0.4, and the two BDs with masses 49 and 60 M_J have eccentricities of 0.5 and 0.8. These new detections are consistent with the observation by Ma & Ge (2014) that a significant number of the BDs with (minimum) masses below 42.5 M_J and with orbital periods between 300 and 3000 days have eccentricities below 0.4. The heaviest CJ and BD companions in our small sample havehigher eccentricities.

We also present RV and astrometric orbit solutions for 16 low-mass stellar companions heavier than 75 M_J or 0.072 M_⊙. These low-mass stars have orbital periods between 30 days and 11.5 years and have eccentricities between 0.04 and 0.65. These objects werefollowed with SOPHIE as potential BDs and turned out to be SCs. Information from the Gaia and HIPPARCOS astrometric space missions allowed us to constrain the orbital inclination and hence, the true mass of some of these objects.

The SOPHIE giant planet and BD surveys play an important role in identifying promising targets for future direct imaging detections. Furthermore, the SOPHIE sample is volume-limited, which will help to obtain unbiased statistics for these planetary systems, which is crucial for understanding the formation and evolution of these objects.

Fig. 5 Period - mass distribution: the CJs and BDs candidates are shown as cyan circles and green diamonds, respectively. The SCs with m sin i measurement are represented by yellow stars, while the SCs with true masses from HIPPARCOS data are represented by yellow squares. The indigo square represents HD 8291 B with its true mass inferred from the Gaia data. The red (at 13 MJ) and blue (at 75 MJ) dashed linesseparates BDs from CJs and SCs, respectively.

Fig. 6 Period - eccentricity distribution of giant planets: the new CJs presented in this paper are shown as cyan solid circles, while the light red circles represents all giant planets with 0.3 MJ ≤ m (or m sini) ≤ 13 MJ in the Extrasolar Planets Encyclopaedia,. The size of the circles is proportional to their mass (or minimum mass), with this proportion increased by a factor of 5 for the cyan circles to emphasize the newly detected CJs.

Fig. 7 Cumulative metallicity distribution of the host stars of giant planets having masses above (red line) and below (green line) 4 MJ (see text formore details).

Fig. 8 Period - eccentricity distribution of known BD companions. The new BD candidates from this paper are shown as solid diamonds. The circles represents the BDs from the combination of Table 1 in Ma & Ge (2014) and Table A.1 in Wilson et al. (2016). The size of all markers is proportional to the minimum mass (or mass represented by a black cross on top of the circles) of the corresponding object. The green color marks lower-mass BDs with m (or m sini) < 42.5 MJ, and the indigo color indicates high mass BDs with m (or m sini) > 42.5 MJ.

Acknowledgements

We thank the staff of Haute-Provence Observatory for their support at the 1.93 m telescope and on SOPHIE. This work is based on observations made with SOPHIE in the context of the program “Recherche de Planètes Extra - solaire” (PI: I. Boisse, F. Bouchy) and the Programme National de Physique Stellaire (PNPS), in particular the two programs “Recherche et caractérisation de planètes géantes avec SOPHIE” (PI : G. Hébrard) and “Exploring the Brown Dwarf Desert around FGK stars in the Solar neighborhood” (PIs: F. Kiefer, IDs: 2017B_PNPS008, 2018A_PNPS008, 2018B_PNPS013 ; S. Dalal, IDs: 2019A_PNPS008, 2020A_PNPS004, 2020B_PNPS DALAL). F.K. acknowledges support by fellowship grants from the Centre National d’Etude Spatiale (CNES) and from the Paris-Science-Lettre (PSL) University. This work was also supported by the PNPS of CNRS/INSU co-funded by CEA and CNES. X.B., X.D, T.F. and L.M. acknowledge funding from the French National Research Agency (ANR) in the framework of the Investissements d’Avenir program (ANR-15-IDEX-02), through the funding of the “Origin of Life” project of the Univ. Grenoble-Alpes. E.M. acknowledges funding from the French National Research Agency (ANR) under contract number ANR-18-CE31-0019 (SPlaSH). M.H. acknowledges support from ANID - Millennium Science Initiative - ICN12_009. N.A.-D. acknowledges the support of FONDECYT project 3180063. This work was supported by FCT - Fundação para a Ciência e a Tecnologia through national funds and by FEDER through COMPETE2020 - Programa Operacional Competitividade e Internacionalização by these grants: UID/FIS/04434/2019; UIDB/04434/2020; UIDP/04434/2020; PTDC/FIS-AST/32113/2017 and POCI-01-0145-FEDER-032113; PTDC/FISAST/28953/2017 and POCI-01-0145-FEDER-028953. O.D.S.D. is supported in the form of work contract (DL 57/2016/CP1364/CT0004) funded by FCT.

Appendix A Tables

Appendix B Figures

Fig. B.1 Generalized Lomb-Scargle (GLS) periodogram of the RVs for the CJs BD+450564 b (top left), BD+550362 b (top right), HD 124330 b (bottom left), and HD 155193 b (bottom right). The three dashed black lines represent the 10, 1, and 0.1% false alarm probability in ascending order.

Fig. B.2 Generalized Lomb-Scargle (GLS) periodogram (left) and Keplerian periodogram (right) of the RVs for the eccentric CJs, BD+631405 b and HD 331093 b. The two dashed black lines in GLS periodogram represent the 10 and 1% false alarm probability in ascending order.

Fig. B.3 Phase folded RV curve for the six CJs. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles respectively.

Fig. B.4 Phase folded RV curve for the four BDs with m sin i in the range of 13–75 MJ. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles respectively.

Fig. B.5 Keplerian fit in the radial velocities (RV vs time with O-C residuals and RV vs. phase) for the targets with 16 SCs having m sin i ≥ 75 MJ (0.072 M⊙) obtained with SOPHIE spectrograph. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles respectively, except for HD 8291 where the blue squares correspond to the HARPS RV data. For targets, HD 8291, HD 25603, HD 76332, HD 187057, HD 211961, and HD 352975, SOPHIE and SOPHIE+ RV data are considered as one dataset and labeled “sophie” in top right corner of RV vs. time plot.

Fig. B.6 Top panels: astrometric orbits of 3σ detections projected on the sky. North is up and east is left. The solid red line shows the model orbit and open circles mark the individual HIPPARCOS measurements. Bottom panels: O–C residuals for the normal points of the orbital solution (filled blue circles) and of the five-parameter model without companion (open squares).

References

All Tables

All Figures

	Fig. 1 Keplerian orbit of the RV variations for the six CJs. SOPHIE and SOPHIE+ RV measurements are indicated in blue squaresand red circles, respectively.
In the text

	Fig. 2 Orbital solutions of the RV variations of four BDs with 25.05 MJ ≤ m sin i ≤ 60.27 MJ are depictedhere. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles, respectively.
In the text

	Fig. 3 Top panel: astrometric orbit of HD 2055213σ. North is up and East is left. The solid red line shows the model orbit and the open circles mark the individual 1-dimensional HIPPARCOS measurements. Bottom panel: O–C residuals for the normal points of the orbital solution (filled blue circles) and of the five-parameter model without companion (open squares).
In the text

	Fig. 4 Lomb-Scargle periodogram of the observed radial velocities, BIS, FWHM and for HD 204277. The black dashed line represents the 0.1% False Alarm Probability.
In the text

Fig. 5 Period - mass distribution: the CJs and BDs candidates are shown as cyan circles and green diamonds, respectively. The SCs with m sin i measurement are represented by yellow stars, while the SCs with true masses from HIPPARCOS data are represented by yellow squares. The indigo square represents HD 8291 B with its true mass inferred from the Gaia data. The red (at 13 MJ) and blue (at 75 MJ) dashed linesseparates BDs from CJs and SCs, respectively.

In the text

Fig. 6 Period - eccentricity distribution of giant planets: the new CJs presented in this paper are shown as cyan solid circles, while the light red circles represents all giant planets with 0.3 MJ ≤ m (or m sini) ≤ 13 MJ in the Extrasolar Planets Encyclopaedia,. The size of the circles is proportional to their mass (or minimum mass), with this proportion increased by a factor of 5 for the cyan circles to emphasize the newly detected CJs.

In the text

	Fig. 7 Cumulative metallicity distribution of the host stars of giant planets having masses above (red line) and below (green line) 4 MJ (see text formore details).
In the text

Fig. 8 Period - eccentricity distribution of known BD companions. The new BD candidates from this paper are shown as solid diamonds. The circles represents the BDs from the combination of Table 1 in Ma & Ge (2014) and Table A.1 in Wilson et al. (2016). The size of all markers is proportional to the minimum mass (or mass represented by a black cross on top of the circles) of the corresponding object. The green color marks lower-mass BDs with m (or m sini) < 42.5 MJ, and the indigo color indicates high mass BDs with m (or m sini) > 42.5 MJ.

In the text

	Fig. B.1 Generalized Lomb-Scargle (GLS) periodogram of the RVs for the CJs BD+450564 b (top left), BD+550362 b (top right), HD 124330 b (bottom left), and HD 155193 b (bottom right). The three dashed black lines represent the 10, 1, and 0.1% false alarm probability in ascending order.
In the text

	Fig. B.2 Generalized Lomb-Scargle (GLS) periodogram (left) and Keplerian periodogram (right) of the RVs for the eccentric CJs, BD+631405 b and HD 331093 b. The two dashed black lines in GLS periodogram represent the 10 and 1% false alarm probability in ascending order.
In the text

	Fig. B.3 Phase folded RV curve for the six CJs. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles respectively.
In the text

	Fig. B.4 Phase folded RV curve for the four BDs with m sin i in the range of 13–75 MJ. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles respectively.
In the text

Fig. B.5 Keplerian fit in the radial velocities (RV vs time with O-C residuals and RV vs. phase) for the targets with 16 SCs having m sin i ≥ 75 MJ (0.072 M⊙) obtained with SOPHIE spectrograph. SOPHIE and SOPHIE+ RV measurements are indicated in blue squares and red circles respectively, except for HD 8291 where the blue squares correspond to the HARPS RV data. For targets, HD 8291, HD 25603, HD 76332, HD 187057, HD 211961, and HD 352975, SOPHIE and SOPHIE+ RV data are considered as one dataset and labeled “sophie” in top right corner of RV vs. time plot.

In the text

	Fig. B.6 Top panels: astrometric orbits of 3σ detections projected on the sky. North is up and east is left. The solid red line shows the model orbit and open circles mark the individual HIPPARCOS measurements. Bottom panels: O–C residuals for the normal points of the orbital solution (filled blue circles) and of the five-parameter model without companion (open squares).
In the text