A SUBSTELLAR COMPANION IN A 1.3 yr NEARLY CIRCULAR ORBIT OF HD 16760^*,^†

During the past decade, precise Doppler surveys have observed 3000 of the closest and brightest main-sequence stars and have detected more than 250 extrasolar planets so far (e.g., Butler et al. 2006; Udry & Santos 2007). The planets show a surprising diversity in their masses, orbital periods, and eccentricities, and the statistical distribution and correlation of these parameters now begin to serve as the test cases for theories of planet formation and evolution (e.g., Ida & Lin 2004a, 2004b, 2005).

Stellar metallicity is one of the key parameters that controls planet formation. It is well known that frequency of giant planets becomes higher as stellar metallicity increases (e.g., Santos et al. 2003; Fischer & Valenti 2005, and references therein). Such a trend favors a core-accretion model as the main mechanism of giant planet formation, because high metallicity likely increases the surface density at the midplane of the protoplanetary disk, making it easier to build metal cores massive enough to accrete gas envelope (e.g., Ida & Lin 2004b; Alibert et al. 2005). It has also been noted that there is a possible flat tail in the low-metallicity regime of the frequency distribution (Santos et al. 2004; Sozzetti et al. 2009). This might suggest the existence of a distinct formation mechanism, disk instability, which is not dependent on metallicity (Boss 2002). Correlations between metallicity and orbital parameters are inconclusive. No significant trends have been found in period–metallicity and eccentricity–metallicity distribution, although hints of some weak correlations have been pointed out (e.g., Santos et al. 2003; Fischer & Valenti 2005).

It has also been pointed out that in addition to the presence of gas giant planets, the mass of the detected planet or the total mass of all planets in a system are correlated with high metallicity (e.g., Santos et al. 2003; Fischer & Valenti 2005), as predicted by core accretion. Massive companions with ≳10 Jupiter-mass (M_JUP) could be formed by core-accretion depending on the assumed truncation condition for gas accretion (Ida & Lin 2004b; Alibert et al. 2005; Mordasini et al. 2007). Or they could be formed by gravitational disk instability as has been suggested for the formation of brown-dwarf companions. In fact, an upper limit of planet mass has not been well established theoretically or observationally. In core accretion, the limit is regulated by a balance between gap opening (e.g., Ida & Lin 2004a; Crida et al. 2006), the inflow rate due to viscous diffusion and dissipation timescales of protoplanetary disks (e.g., D'Angelo et al. 2003; Dobbs-Dixon et al. 2007; Tanigawa & Ikoma 2007). Planet mass might also depend on subsequent evolution of planets by mechanisms such as giant impacts (Baraffe et al. 2008). In disk instability, planet mass depends primarily on the mass ratio between the disk and central star. Observational properties of planets help to constrain the planet formation models. Currently, no significant difference in stellar metallicity has been found for planets above and below ∼10 M_JUP (e.g., Santos et al. 2003; Fischer & Valenti 2005). Indeed, very high-mass planets (i.e., greater than 10 M_JUP) are rare; a larger sample would improve statistics for extracting information regarding the upper limit of planet mass and correlations between planet mass and metallicity.

The N2K consortium began precise Doppler surveys at Keck, Magellan, and Subaru in 2004, targeting a new set of 2000 metal-rich solar-type stars (Fischer et al. 2005). The main purpose of the survey was to search for short-period planets with a high-cadence observational strategy in order to find prospective transiting planets. The sample was biased toward high-metallicity stars in order to increase the detection rate of the planets. From the collective N2K surveys, we have discovered seven short-period (P ⩽ 5 days) planets so far including: HD 88133 (Fischer et al. 2005), the transiting planet HD 149026 (Sato et al. 2005), HD 149143 and HD 109749 (Fischer et al. 2006), HD 86081, HD 224693, and HD 33283 (Johnson et al. 2006).

The metallicity-biased sample is expected to contain many long-period planets as well as short-period ones. Therefore, we have continued to observe stars with significant radial velocity variations and we have also detected 13 intermediate-period (18–1405 days) planets: HD 5319 and HD 75898 (Robinson et al. 2007), HD 11506, HD 125612, HD 170469, HD 231701 and the transiting planet HD 17156 (Fischer et al. 2007), the double planet system HIP 14810 (Wright et al. 2007), HD 205732 and HD 154672 (Lopez-Morales et al. 2008), HD 179079 and HD 73534 (Valenti et al. 2009). These planets help to investigate correlations between properties of planets and stellar metallicity in more detail, especially in the high-metallicity regime.

Here, we report the discovery of a substellar companion to the G5 dwarf star HD 16760 in a 1.3 yr nearly circular orbit from the N2K sample. The companion has a minimum mass of 13.13 M_JUP, which is just above the deuterium-burning threshold that is often used to distinguish brown dwarfs from planets.

HD 16760 (HIP 12638) is listed in the Hipparcos catalog (ESA 1997) as a G5V star with a visual magnitude V = 8.7 and a color index B − V = 0.715. The revised Hipparcos parallax of π = 22.00 ± 2.35 mas (van Leeuwen 2007) corresponds to a distance of 45.5 pc and yields the absolute visual magnitude of M_V = 5.41. The star is probably a visual binary system having a Hipparcos double star catalog entry. The secondary is separated by 14.6 arcsec which corresponds to a projected separation of about 660 AU. A high-resolution spectroscopic analysis with Spectroscopy Made Easy (SME; Valenti & Piskunov 1996) described in Valenti & Fischer (2005) derives an effective temperature T_eff = 5629 ± 44 K, a surface gravity log g = 4.47 ± 0.06, rotational velocity vsin i = 0.5 ± 0.5 km s⁻¹, and metallicity [Fe/H] = 0.067 ± 0.05 dex for the star. The bolometric luminosity is L_star = 0.72 ± 0.43 L_☉ calculated using the M_V and a bolometric correction of −0.108 based on Flower (1996). The radius of 0.81 ± 0.27 $R_☉$ for the star is derived from the Stefan–Boltzmann relation using L_star and T_eff.

To estimate a stellar mass, we interpolated the metallicity, effective temperature, and luminosity onto the stellar interior model grids computed by Girardi et al. (2002). We used the three-dimensional interpolation method described by Johnson et al. (2007), and adopt their estimated 7% uncertainty based on a comparison among different stellar model grids. Using the SME-derived stellar parameters, we estimate 0.78 ± 0.05 M_☉ for HD 16760.

As an indicator of chromospheric activity, we measured S_HK, the core emission in the Ca ii HK lines relative to the continuum, to be 0.176 for the star. The ratio of flux from S_HK to the bolometric stellar flux, log R'_HK, was derived to be log R'_HK = −4.93, which indicates that the star is chromospherically inactive. The expected stellar "jitter," which is intrinsic variability in radial velocity as an additional source of astrophysical noise, was estimated to be 2 m s⁻¹ for the star based on the activity and the spectral type from empirical relation by Wright (2005). We applied the jitter to radial velocities from both Subaru and Keck when fitting a Keplerian model in Section 3. The stellar parameters are summarized in Table 1.

Ten radial velocity observations of HD 16760 were obtained with the High Dispersion Spectrograph (HDS) on the 8.2 m Subaru Telescope (Noguchi et al. 2002) from 2004 December to 2008 February. We used an iodine (I₂) absorption cell (Kambe et al. 2002) to provide a fiducial wavelength reference for precise radial velocity measurements. We adopted the setup of StdI2b for all the data, which simultaneously covers a wavelength region of 3500–6100 Å by a mosaic of two CCDs. The slit width was set to 08 for the first four data (2004–2005) and 06 for the last six ones (2006–2008), giving a reciprocal resolution (λ/Δλ) of 45,000 and 60,000, respectively. Typical signal-to-noise ratio (S/N) was 140–200 pix⁻¹ with exposure time of 110–300 s depending on the observing conditions. Radial velocity analysis for an I₂-superposed stellar spectrum was carried out with a code developed by Sato et al. (2002), which is based on a technique by Butler et al. (1996) and Valenti et al. (1995), giving a Doppler precision of about 4–5 m s⁻¹.

After the first three observing runs at Subaru, we identified significant radial velocity variations of the star and then began follow-up observations with the 10 m Keck telescope. Seventeen Keck radial velocity data were obtained with the HIRES spectrograph (Vogt et al. 1994) between 2006 January and 2009 January. We used the B5 decker (086 width with a resolution of about R = 65, 000). Exposure times ranged from 150 to 240 s depending on the observing conditions, resulting in a consistent S/N of about 150. An I₂ cell was inserted in the light path to provide the wavelength solution and the instrumental point-spread function (PSF; Marcy & Butler 1992; Butler et al. 1996) in our model of the observed spectrum. The typical Doppler precision for the Keck observations is 1.5 ms⁻¹.

The Doppler measurements are listed in Table 2 along with the time of observation, the estimated uncertainties, and the origin of the observation (Subaru or Keck). The radial velocities were modeled with a Keplerian orbit using a Levenberg–Marquardt fitting algorithm to obtain a minimum chi-squared solution by varying the free parameters (orbital period, time of periastron passage, eccentricity, velocity amplitude, and omega—the orientation of the orbit reference to the line of nodes). The combined radial velocities from Subaru and Keck are plotted in Figure 1 and the best Keplerian fit is overplotted as a solid line. The expected stellar jitter of 2 m s⁻¹ was added in quadrature to the velocity uncertainties when fitting the Keplerian orbit and are included in the velocities plotted in Figure 1.

Zoom In Zoom Out Reset image size

The best-fit orbital parameters are listed in Table 3. The uncertainty for each orbital parameter was determined using a bootstrap Monte Carlo approach, subtracting the theoretical fit, scrambling the residuals, adding the theoretical fit back to the residuals, and then refitting. The radial velocities are best fit by a Keplerian model with a period P = 466.47 ± 0.35 days, a velocity semiamplitude K₁ = 407.71 ± 0.84 m s⁻¹, and an eccentricity e = 0.084 ± 0.003. An offset of 37 m s⁻¹ was applied to the Subaru data in order to minimize reduce chi-squared $(\sqrt{\chi _{\nu }^2})$ when fitting a Keplerian model to the combined Subaru and Keck velocities. The rms scatter of the residuals to the Keplerian fit is 4.3 m s⁻¹, and the reduce chi-squared is $\sqrt{\chi _{\nu }^2}=1.3$. We found no significant additional periodicity in the residuals. Adopting a stellar mass of 0.78 ± 0.05 M_☉, we obtained for the companion a minimum mass M_psin i = 13.13 ± 0.56 M_JUP, which is close to the planet/brown-dwarf transition, and a semimajor axis a = 1.084 ± 0.023 AU.

Table 3. Orbital Solution for HD 16760b

Parameter	Value
P (days)	466.47 (0.35)
Tp (JD)	2453337.0 (2.4)
Eccentricity	0.084 (0.003)
ω (deg)	242.9 (1.9)
K1 (m s−1)	407.71 (0.84)
a (AU)	1.084 (0.023)
a1sin i (10−3 AU)	17.420 (0.042)
f1(m) (10−6 M☉)	3.241 (0.021)
Mpsin i (MJUP)	13.13 (0.56)
Nobs (Subaru)	10
Nobs (Keck)	17
RMS (m s−1)	4.3
Reduced $\sqrt{\chi _{\nu }^2}$	1.3

Download table as:  ASCIITypeset image

If we assume that the orbit is randomly oriented, there is a 1.4% chance that the true mass exceeds 80 M_JUP (i < 95), the boundary between the brown dwarf and stellar mass regime. In this case, the projected semimajor axis is less than 2.1 mas based on the a₁sin i and the distance to the star, which is below the measurement error of Hipparcos. We measured a small projected rotational velocity for HD 16760 (vsin i = 0.5 km s⁻¹). This may suggest that a near pole-on viewing angle and thus high mass of the companion is possible if the orbital plane is coplanar to the stellar equator. High-precision space-born astrometric observations such as using Hubble Space Telescope, which can achieve ∼0.3 mas precision (e.g., Bean et al. 2007), are highly encouraged to set a stringent constraint on the mass of the companion.

We have reported the discovery of a substellar companion around the G5 dwarf star HD 16760 in a 1.3 yr orbit from precise Doppler measurements at Subaru and Keck. Although the N2K survey originally targeted short-period planets with a few successive observations, based on the Fischer and Valenti metallicity correlation, we expect that 15% of the target stars harbor gas giant planets. The metallicity-biased sample is expected to contain many long-period planets as well as additional short-period ones.

HD 16760 b has a minimum mass of about 13 M_JUP, which is close to the border between planet and brown dwarf regimes. By definition, this is a brown dwarf because the mass is above limit for deuterium burning (13 M_JUP) and in this context it is considered to be part of a low-mass tail of the brown-dwarf distribution. However, such a companion is often called a "super-planet" because it could be a high-mass tail of the planet distribution too. Core-accretion models of planet formation can predict supermassive planets larger than 10 M_JUP or even up to 20–25 M_JUP under metal-rich environment, depending on the assumed truncation condition for gas accretion (Ida & Lin 2004b; Alibert et al. 2005; Mordasini et al. 2007). In fact, the upper limit of planet mass has not been well established theoretically, and it should be set based on the observational properties of planets.

Currently, seven companions with minimum mass of 13–25 M_JUP have been detected within semimajor axis of 5 AU around solar-type stars (FGK dwarfs with 0.7 ⩽ M/M_☉ < 1.6) by precise Doppler measurements including a transiting one CoRoT-Exo-3 b for which accurate mass of 21.7 M_JUP was obtained (Deleuil et al. 2008). The number of such companions is still small and their statistical properties have not yet been established. In Figure 2, the planet eccentricity is plotted against planet mass. As pointed out by Marcy et al. (2005), the more massive planets (≳5 M_JUP) have systematically higher eccentricities than lower mass planets. Most companions with >8 M_JUP (other than HD 16760 b and HD 168443 c⁸) have eccentricities larger than about 0.3, except for planets in short-period orbits where tidal circularization could be effective.

Zoom In Zoom Out Reset image size

Since massive companions have the largest inertial resistance to perturbations that drive them out of their initial orbits, they should retain their initial orbits. Therefore, massive planets with high eccentricity are difficult to explain within the framework of standard core-accretion model. This suggests a distinct evolutionary scenario that could result in eccentric orbits for massive companions, such as disk instability (e.g., Boss 1998), giant impacts between massive planets following core accretion (Baraffe et al. 2008), or orbital evolution by Kozai mechanism for binary stars (e.g., Wu & Murray 2003). The fact that HD 16760 is probably in a binary system but the companion has a small eccentricity is interesting from this view point of orbital evolution with respect to the Kozai mechanism.

The metallicity of the host stars may be one of the keys to distinguish between planet formation models; disk instability is not dependent on metallicity (Boss 2002), while accretion should be more efficient in high-metallicity disks. We do not observe a correlation between metallicity and eccentricity for the most massive companions at this stage. However, there are still only a small number of these objects.

In summary, HD 16760 b is unique among the known substellar companions because it resides in a nearly circular orbit with an intermediate orbital period and a mass that is greater than 13 times the mass of Jupiter. Detection of additional objects in this mass regime could provide deeper insight for theoretical models.

We thank Akito Tajitsu and Tae-Soo Pyo for their expertise and support of the Subaru HDS observations. We gratefully acknowledge the dedication and support of the Keck Observatory staff, in particular Grant Hill for support with HIRES. B.S. is supported by MEXT's program "Promotion of Environmental Improvement for Independence of Young Researchers" under the Special Coordination Funds for Promoting Science and Technology. DF gratefully acknowledges support from NASA grant NNG05G164G. This research has made use of the Simbad database, operated at CDS, Strasbourg, France. The authors extend thanks to those of native Hawaiian ancestry on whose sacred mountain of Mauna Kea we are privileged to be guests. Without their generous hospitality, the Subaru and Keck observations presented herein would not have been possible.