HIGH-CONTRAST IMAGING OF INTERMEDIATE-MASS GIANTS WITH LONG-TERM RADIAL VELOCITY TRENDS

We obtained RV data for the targets, except for ι Dra, with the 1.88 m telescope and the High Dispersion Echelle Spectrograph (HIDES; Izumiura 1999) at OAO between 2001 and 2014. We used an iodine absorption cell (I₂ cell; Kambe et al. 2002) for precise RV measurements, which provides a fiducial wavelength reference in a wavelength range of 5000–5800 $\mathring{\rm A}$. We used the HIDES-slit mode setting with a slit width of the spectrograph of 200 μm ($0\buildrel{\prime\prime}\over{.} 76$), which corresponds to a spectral resolution ($R=\lambda /{\rm{\Delta }}\lambda$) of 67000 by about 3.3 pixel sampling. Reduction of the echelle data (i.e., bias subtraction, flat-fielding, scattered-light subtraction, and spectrum extraction) was performed using the IRAF software package.³²

For precise RV analysis, we modeled I₂-superposed stellar spectra (star+I₂) by the method detailed in Sato et al. (2002) and Sato et al. (2012), which is based on the method by Butler et al. (1996) and Valenti et al. (1995). In the method, a star+I₂ spectrum is modeled as a product of a high resolution I₂ and a stellar template spectrum convolved with a modeled instrumental profile (IP) of the spectrograph. The stellar template spectrum is obtained by deconvolving a pure stellar spectrum with an IP estimated from a B-star or flat spectrum taken through an I₂ cell. We achieved a long-term RV precision of about 4 m s⁻¹ over the entire span of the observations. The measurement error was derived from an ensemble of the velocities from each of the ∼300 spectral segments (each ∼3 $\mathring{\rm A}$ long) in every exposure. We show the derived RVs for 18 Del, γ Hya, and HD 109272 in Figures 1 and 2, and have listed them in Tables 3–5 together with the estimated uncertainties. The RVs for 18 Del were updated and extended from those presented in Sato et al. (2008). The RVs for HD 5608 and HD 14067 presented in Sato et al. (2012) and Wang et al. (2014), respectively, were used for the analysis in this paper.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

After the first announcement of the discovery of a planet around 18 Del by Sato et al. (2008), we collected 31 more epochs of RV data for the star in five years and updated its orbital parameters including a possible linear velocity trend (∼4 m s⁻¹ yr⁻¹) suggested in Sato et al. (2008). The updated orbital parameters and the uncertainties were derived using the Bayesian Markov chain Monte Carlo (MCMC) method (e.g., Ford 2005; Gregory 2005; Ford & Gregory 2007), following the analysis in Sato et al. (2013a). An extra Gaussian noise factor representing stellar jitter for the data and a linear velocity trend were incorporated as free parameters. We generated 10 independent chains having 10⁷ points with an acceptance rate of about 25%, the first 10% of which were discarded, and confirmed that each parameter sufficiently converged based on the Gelman–Rubbin statistic (Gelman & Rubin 1992). We derived the median value of the merged posterior probability distribution function (PDF) for each parameter and set the 1σ uncertainty as the range between 15.87% and 84.13% of the PDF. We plot the derived Keplerian orbit together with the RV points and their measurement errors including the jitter in Figure 1, and list the orbital parameters and the uncertainties in Table 6. We confirmed the linear velocity trend for the star to be $\dot{\gamma }=-2.8\pm 0.7$ m s⁻¹ yr⁻¹ with 4σ confidence.

Table 6.  Updated Orbital Parameters for 18 Del

Parameter	18 Del b
P (days)	982.2 ± 3.4
K1 (m s−1)	121.7 ± 2.2
e	${0.016}_{-0.011}^{+0.017}$
ω (degree)	−210${}_{-73}^{+84}$
Tp (JD−2450000)	−310${}_{-200}^{+230}$
${a}_{1}\mathrm{sin}i$ (10−3 au)	10.98 ± 0.20
${f}_{1}(m)$ (10${}^{-7}\;{M}_{\odot }$)	${1.83}_{-0.097}^{+0.10}$
${m}_{2}\mathrm{sin}i$ (${M}_{{\rm{Jup}}}$)	10.2
a (au)	2.5
jitter (m s−1)	${12.9}_{-1.2}^{+1.3}$
$\dot{\gamma }$ (m s−1 yr−1)	−2.8 ± 0.7
${N}_{{\rm{obs}}}$	82
rms (m s−1)	13.4

Download table as:  ASCIITypeset image

For γ Hya and HD 109272, we fit the linear velocity trends using the method of least squares. The trends of our targets are summarized in Table 7.

Table 7.  RV Trends of our Targets

Name	$\dot{\gamma }$ (m s−1 yr−1)
γHya	4.1 ± 0.2	This work
ι Dra	−13.65 ± 0.75	Kane et al. (2010)
18 Del	−2.8 ± 0.7	This work
HD 5608	−5.51 ± 0.45	Sato et al. (2012)
HD 14067	−22.4 ± 2.2	Wang et al. (2014)
HD 109272	2.4 ± 0.2	This work

Download table as:  ASCIITypeset image

Direct-imaging observations were conducted from 2011–2014 as part of the SEEDS survey (Tamura 2009) using the High Contrast Instrument for the Subaru Next Generation Adaptive Optics (HiCIAO; Suzuki et al. 2010) on the 8.2 m Subaru Telescope. We used the adaptive optics system AO188 (Hayano et al. 2008) together with HiCIAO, and the target stars themselves were used as the natural guide stars in our observations. In addition, we used an atmospheric dispersion corrector, which helped mitigate the drift of the stellar point-spread functions (PSFs) on the detector (Egner et al. 2010). Furthermore, the angular differential imaging (ADI) method (Marois et al. 2006) was applied to our observations to improve the high-contrast performance. All six targets were first observed in 2011–2012 in the H-band (∼1.6 μm) and four targets that have companion candidates were followed up in 2014 by employing J- (∼1.2 μm), H-, and Ks- (∼2.1 μm) band filters.

In order to maximize the sensitivity of our observations, it is necessary to use an occulting mask or to saturate the central star's PSF. Such masked or saturated images (i.e., science images) can be taken over relatively long integration times and are used for the companion survey. However, we require unsaturated and unmasked PSFs of the central stars (i.e., calibration images) for our measurements of the contrast limit, central star centroid. Therefore, we made unmasked observations with neutral density (ND) filters to avoid PSF saturation for each target, selecting ND filters of an appropriate transmission based on the central star's brightness: HiCIAO has ND filters with transmittances of 0.854%, 0.063%, and 0.016%. We obtained the unsaturated and unmasked PSFs before and after observing the science images for each target. The observation log is summarized in Table 8.

Our data reduction procedure is as follows. First, we removed stripe patterns (see Suzuki et al. 2010) appearing on each observed frame and subsequently corrected hot and bad pixels. Then, hot-pixel masks were generated from dark frames obtained in each observing run. To create a hot-pixel mask, we used the L.A. cosmic algorithm (van Dokkum 2001) for the data taken before 2012 September, while our originally developed routine was applied to the data taken after 2014 April. Flat-fielding was performed following these procedures.

Next, we corrected the distortion of the images, since the distortion correction of our observed images is crucial to achieve reliable astrometry. To measure the distortion map, we usually obtained images of the globular cluster M5 or M15 in each observing run. The distortion map was made by comparing the stellar positions on the M5/M15 images taken by HiCIAO with those on the images taken by the Hubble Space Telescope/Advanced Camera for Survey whose distortion is well-corrected (Brandt et al. 2013). Using the measured distortion map, we applied the geometric transformations to the post-flat-fielded frames. This procedure fixes the plate scale of images to be 9.5 mas.

We estimated the centroids of the primary star in the distortion-corrected images using the unsaturated data whose distortions have been also corrected, and shifted the positions of the central stars to the centers of the arrays. Then, we assumed that the stellar position on the detector did not drift. We also calculated the stellar frame-to-frame centroids by fitting the Moffat function to the masked PSFs and confirmed that the PSF drifts are less than 1 pixel (=9.5 mas) during the observations.

Next, we carried out ADI reductions after subtracting the central star's radial profile. We used the locally optimized combination of images (LOCI) algorithm (Lafrenière et al. 2007b), which allows further improvement in our capability to detect faint companions. We evaluated the self-subtraction effect caused by the LOCI algorithm by embedding artificial PSFs. To determine a realistic detection limit, the final image was convolved with a circular aperture with a diameter equal to the PSF FWHM (Lafrenière et al. 2007a). Finally, we checked the achieved 5σ contrast ratio by calculating the standard deviation within 2 pixel wide rings, from the center to the outer region every 4.5 pixels.

3.1.1. γ Hya

We discovered a companion candidate with an H-band contrast of ΔH = 7.24 located 16 from γ Hya, as shown in Figure 3. Two years after the first observation, a follow-up observation enabled us to confirm that the companion candidate, γ Hya B, shares a common proper motion with the central star (Figure 5). Our astrometric and photometric results are shown in Table 9. Considering γ Hya's age (0.37 Gyr, Takeda et al. 2008) and consistency of the mass derived from J-, H-, and Ks-band photometry, we adopted the 400 Myr NextGen model to convert the measured photometry into mass. The mass of γ Hya B is ${0.61}_{-0.14}^{+0.12}\;{M}_{\odot }$ by averaging four independent mass estimates.

Zoom In Zoom Out Reset image size

Table 9.  Astrometric and Photometric Results for γ Hya B

Name	Date (UT)	Filter	Sep. ('')	P.A. (degree)	Δmag	Mass (${M}_{\odot }$)
γ Hya B	2012 May 13	H	1.623 ± 0.011	194.4 ± 0.2	7.24 ± 0.08	0.63 ± 0.06
	2014 Apr 25	J	1.611 ± 0.004	195.2 ± 0.2	7.97 ± 0.24	0.53 ± 0.14
	2014 Apr 25	H	1.611 ± 0.004	195.2 ± 0.2	7.39 ± 0.20	0.61 ± 0.12
	2014 Apr 25	Ks	1.626 ± 0.006	195.3 ± 0.1	7.07 ± 0.30	0.65 ± 0.21

Download table as:  ASCIITypeset image

3.1.2. HD 5608

We found two companion candidates around HD 5608. The first candidate has ΔH = 9.40 with a separation of 06 and the second has ΔH = 13.1 with a separation of 74, as shown in Figure 4. The time intervals of our three observations are long enough to allow for common proper motion tests (Figure 5). We conclude that the close companion candidate HD 5608 B is co-moving and the other companion candidate at 74 is a background star. Our astrometric and photometric results for HD 5608 B are shown in Table 10. Considering that HD 5608 B has an age of 2.5 Gyr (Takeda et al. 2008), we used the interpolation between the 2 and 3 Gyr Dusty models to estimate the mass of HD 5608 B. The mass derived from Dusty is $0.106\pm 0.002\;{M}_{\odot }$ from the weighted mean of three observational results, and that derived from the NextGen model is $0.13\pm 0.01\;{M}_{\odot }$. These indicate that HD 5608 B is a low-mass star, and the Dusty model is a model to reproduce the luminosity of the low-mass stars. Thus, we adopted the interpolation between the 2 and 3 Gyr Dusty models, and took the mass of HD 5608 B to be $0.10\pm 0.01\;{M}_{\odot }$.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 10.  Astrometric and Photometric Results for HD 5608 B

Name	Date (UT)	Filter	Sep. ('')	P.A. (degree)	Δmag	Mass (${M}_{\odot }$)
HD 5608 B	2011 Dec 31	H	0.627 ± 0.009	58.9 ± 0.4	9.40 ± 0.11	0.11 ± 0.02
	2012 Sep 12	H	0.627 ± 0.022	59.9 ± 1.0	9.70 ± 0.10	0.10 ± 0.02
	2014 Oct 7	H	0.588 ± 0.012	55.7 ± 0.6	9.55 ± 0.20	0.11 ± 0.02

Download table as:  ASCIITypeset image

3.1.3. HD 109272

We discovered one companion candidate (ΔH = 7.18) at a separation of 12 from HD 109272 (Figure 3). Follow-up observations allowed a common proper motion test for the candidate (Figure 5), which suggested that the companion candidate HD 109272 B is gravitationally bound to the central star. Table 11 shows the astrometric and photometric results for HD 109272 B. Its mass was calculated using the 1 Gyr Dusty model based on an age of HD 109272 B of 1.4 Gyr (Takeda et al. 2008). By averaging two mass estimates derived from two observations, we find that HD 109272 B has a mass of $0.28\pm 0.06\;{M}_{\odot }$.

Table 11.  Astrometric and Photometric Results for HD 109272 B

Name	Date (UT)	Filter	Sep. ('')	P.A. (degree)	Δmag	Mass (${M}_{\odot }$)
HD 109272 B	2012 Apr 11	H	1.187 ± 0.005	53.0 ± 0.2	7.18 ± 0.14	0.30 ± 0.04
	2014 Apr 23	J	1.168 ± 0.004	52.0 ± 0.1	7.36 ± 0.33	0.28 ± 0.06
	2014 Apr 23	H	1.166 ± 0.004	52.2 ± 0.1	7.22 ± 0.28	0.30 ± 0.09
	2014 Apr 23	Ks	1.170 ± 0.006	52.0 ± 0.1	7.40 ± 0.16	0.24 ± 0.04

Download table as:  ASCIITypeset image

3.2.1. 18 Del

We found a faint companion candidate of ΔH = 16.9 with a signal-to-noise ratio (S/N) of ∼ 5 at 75 from 18 Del A (Figure 4). In the 2012 July observation, we were not able to detect the candidate because the exposure time was not sufficient to detect the candidate. We detected it again on 2014 June 10 and we carried out a common proper motion test. The result indicates that the companion candidate traces the track expected for a background star (Figure 5). The achieved contrast ratio is shown in Figure 6. The detectable mass limits derived from the COND 0.8 Gyr model (Baraffe et al. 2003) for an age of 18 Del of 0.79 Gyr (Takeda et al. 2008) is displayed in the right panel of Figure 6. We exclude the a $\sim 0.13\;{M}_{\odot }$ object at 05, a $\sim 0.05\;{M}_{\odot }$ object at 10, and a $\sim 0.03\;{M}_{\odot }\approx 31\;{M}_{\mathrm{Jup}}$ object beyond 20 from the central star.

Zoom In Zoom Out Reset image size

3.3.1. ι Dra

We were not able to detect any objects on 2012 May 14 in the H-band for ι Dra beyond 06. Figure 7 plots the PARSEC isochrone (Bressan et al. 2012) model within the range of the uncertainty of ι Dra's mass derived by Baines et al. (2011). The position of ι Dra in the figure agrees with an age of 2 Gyr. Hence we use the 2 Gyr COND model (Baraffe et al. 1998) to evaluate the detectable mass limits. Although the age estimation for ι Dra may not be accurate, we note that there is not a large difference in the results even if the difference in the adopted age is ±1 Gyr. The excluded object mass range is shown in Figure 6. We exclude the a $\sim 0.09\;{M}_{\odot }$ object at 10 and a $\sim 0.05\;{M}_{\odot }\approx 52\;{M}_{\mathrm{Jup}}$ object over 2''.

Zoom In Zoom Out Reset image size

3.3.2. HD 14067

Combining the RV trend for the primary star with the projected separation of the detected companions, we can calculate the minimum dynamical mass that would be required to produce the RV trend with the following equation (Torres 1999; Liu et al. 2002):

$$\begin{eqnarray}{M}_{\mathrm{dyn}} & = & 5.34\times {10}^{-6}\;{M}_{\odot }{\left(\displaystyle \frac{d}{\mathrm{pc}}\displaystyle \frac{\rho }{\mathrm{arcsec}}\right)}^{2}\\ & & \times \left|\displaystyle \frac{\dot{v}}{{\rm{m}}\;{{\rm{s}}}^{-1}\;{\mathrm{yr}}^{-1}}\right|F(i,e,\omega ,\phi ),\end{eqnarray} \tag{ 2 }$$

where d is the distance to the target, ρ is the observed angular separation of the companion (see also Knutson et al. 2014), and F is a function that depends on the orbital parameters (inclination i, eccentricity e, longitude of periastron ω, and orbital phase ϕ) of the companion. Torres (1999) determined that the minimum value of $F(i,e,\omega ,\phi )$ is $3\sqrt{3}/2$. We use this equation to calculate the minimum mass limited by the RV trend. If the mass estimated from photometry exceeds the dynamical minimum mass derived from the RV trend, then we can conclude that a detected companion is responsible for the observed RV trend, and if not, the companion is not responsible for the observed RV trend. Additionally, we calculate a physical (unprojected) separation of the detected companion from the central star, consulting Howard et al. (2010) who derived a true separation of a stellar companion around HD 126614 by combining the companion's estimated mass with the central star's RV trend. There are two solutions for each of the three companions. As a simplification, we do not consider projection effects on the orbit, but assume that the projected separation is equal to the semimajor axis when comparing imaging and RV limits. As shown in e.g., Brandeker et al. (2006), the statistical mean conversion factor between the semimajor axis and the projected separation is close to 1 for eccentricity distributions representative of wide binaries, which supports this approximation.

In the cases where no companion could be found, the upper and lower limits for the companion as function of semimajor axis were calculated (see Figure 8). The lower-mass limit from the RV trend is calculated from Equation (2), and the upper mass limit is set by the detectable mass limit of the direct-imaging observation. The object that could cause the RV trend is then constrained between these two limits.

Zoom In Zoom Out Reset image size

Furthermore, the lack of curvature in a linear trend can be used to exclude the existence of inner companions. We assume that the time span of the observations must correspond to at most half an orbital period, or significant curvature would necessarily be seen. This sets a lower limit on the period, and thus an inner limit on the semimajor axis of the companion.

γ Hya. With an angular separation of 1623 (2012 May 13), the dynamical minimum mass of γ Hya B is 0.25 ${M}_{\odot }$. The mass estimated from photometry, ${0.61}_{-0.14}^{+0.12}\;{M}_{\odot }$, exceeds the dynamical minimum mass. Therefore, we conclude that γ Hya B is responsible for the observed RV trend. The physical separation is 67.5 ± 0.6 au or 159 ± 7 au.

HD 5608. The dynamical minimum mass of HD 5608 B is 0.095 ${M}_{\odot }$ (2011 December 31). Our photometric estimated mass is 0.10 $\pm 0.01\;{M}_{\odot }$, which is consistent with the dynamical minimum mass derived from the RV trend. We confirm that the companion is the RVTG. The calculated physical separation is 40 ± 1 au or 47 ± 3 au.

HD 109272. In the HD 109272 system, the dynamical minimum mass limit calculated with the angular separation on 2012 April 11 is 0.12 ${M}_{\odot }$. The estimated mass from the photometry of HD 109272 B is $0.28\pm 0.06\;{M}_{\odot }$. Therefore, we conclude that HD 109272 B is the RVTG for the observed RV trend in HD 109272 A. The true physical separation of HD 109272 B is 59.3 ± 0.9 au or 140 ± 6 au.

18 Del. Mugrauer et al. (2014) reported that 18 Del A has a distant companion 18 Del B outside the field of view of HiCIAO. The projected separation of 18 Del B is 2199 au and its mass is 0.19 ${M}_{\odot }$. The dynamical minimum mass at 2199 au is 181 ${M}_{\odot }$. Therefore, 18 Del B cannot be the source of the observed RV trend. The upper and lower limits for the RVTG are shown in Figure 8. In addition, the long-term linear RV trend would exclude the existence of inner objects. The semimajor axis range of the RVTG is $a\;\sim$ 10–50 au. A stellar companion at wide separation is ruled out, although a low-mass stellar companion at the inner region is possible. The minimum mass at $a\;=$ 10 au is ${m}_{p}\;\approx$ 4 M_Jup, so the RVTG is either a high-mass planet, a brown dwarf, or a low-mass stellar companion.

ι Dra. The absence of the detection of any companions around ι Dra is consistent with the result observed by Kane et al. (2014) at 692 and 880 nm. An analysis combining the RV trend and the HiCIAO result is shown in Figure 8. Considering the linear RV trend observed over a decade (Kane et al. 2010), the possible innermost object is ${m}_{p}\;\approx$ 16 M_Jup at $a\;=$ 9 au. On Figure 8, the intersection of the observation detectable line with the dynamical minimum line derived from the RV trend is $a\;\approx$ 31 au. Hence, the semimajor axis range of the RVTG is $a\;\sim$ 9–31 au. The mass range implies that the RVTG is a brown dwarf or a stellar companion at small separation.

HD 14067—From the observation result, we can determine limits for identifying the RVTG for HD 14067 (Figure 8). The linear RV trend over five years excludes objects in the inner region. The possible objects' orbital period is 10 years at least, which means that the innermost possible RVTG is at $a\;=$ 10 au in the HD 14067 system. The minimum dynamical mass at $a\;=$ 10 au in the system is ${m}_{p}\;\approx$ 32 M_Jup. In Figure 8, the outermost possible RVTG is at $a\;\approx$ 49 au, where the detectable limit line crosses the dynamical minimum line. The dynamical minimum mass at $a\;\approx$ 49 au is ${M}_{p}\;\approx$ 0.74 ${M}_{\odot }$. We can rule out a wide-orbit stellar companion, although a stellar companion at $a\;\sim$ 10–49 au is still possible. The RVTG for HD 14067 is a brown dwarf or a stellar companion at $a\;\sim$ 10–49 au.

Our high-contrast observations would exclude planetary RVTGs for five out of the six targets. The exception, 18 Del, could either be a planet or a low-mass star. To distinguish the nature of the RVTG of 18 Del, further RV monitoring and higher contrast imaging for the inner region using extreme AO (e.g., SCExAO; Martinache & Guyon 2009) would be important.

Two systems, ι Dra and HD 14067, for which we cannot identify RVTGs in the high-contrast imaging, are important. Vigan et al. (2012) reported that the frequency of brown dwarfs about intermediate-mass stars is as low as ${2.8}_{-0.9}^{+6.0}$% at a range of separation of 5–320 au. On the other hand, Duchéne & Kraus (2013) reported that the orbital period distribution for intermediate-mass multiple systems has two peaks at $P\;\approx$ 10 days and $a\;\approx$ 350 au. In either case, the two systems offer unique samples of brown dwarfs or low-mass stars around intermediate-mass stars. Further high-contrast imaging observations with deeper and inner sensitivity would be important not only to clarify the frequency of brown dwarfs and low-mass stars around intermediate-mass stars, but also to study the planet migration of the inner eccentric planets in those systems.

Several studies have revealed that the formation mechanism of eccentric planets cannot be explained by core accretion theory and Type I/II migration. The Kozai mechanism, which is a perturbation mechanism from a distant stellar companion (e.g., Wu & Murray 2003), planet–planet scattering (e.g., Nagasawa et al. 2008), and secular chaotic excursions (e.g., Wu & Lithwick 2010) are promising approaches to describe eccentric planets.

Four targets already have known inner planets (Table 2). We consider the Kozai mechanism to explain the eccentric planets, namely a perturbation due to an outer stellar companion periodically oscillates the eccentricity and inclination of an inner planet. The oscillation timescale of the Kozai mechanism is calculated by

$$\begin{eqnarray}&&{P}_{\mathrm{Kozai}}\sim \displaystyle \frac{{M}_{A}}{{m}_{B}}\displaystyle \frac{{P}_{B}^{2}}{{P}_{b,0}}{(1-{e}_{B}^{2})}^{3/2}\end{eqnarray} \tag{ 3 }$$

where M_A is the primary star's mass, m_B is the stellar companion's mass, P_B is the period of the companion, ${P}_{b,0}$ is the initial period of the planet, and e_B is the eccentricity of the companion (Holman et al. 1997).

We calculated the timescale of the Kozai mechanism for the significantly high eccentricity planets ι Dra b, HD 5608 b, and HD 14067 b. For ι Dra b, we assume that its companion has a mass of 0.18 ${M}_{\odot }$ and in a circular orbit at 31 au, which is the maximum mass and separation. We then assume that the initial period of the planet is equal to that of a circular orbit of the observed semimajor axis. The timescale is ${P}_{\mathrm{Kozai}}\;\sim$ 107 kyr, which is much less than the Gyr order of ι Dra's age. The timescale for HD 5608 b, assuming the circular orbit of HD 5608 B, is ${P}_{\mathrm{Kozai}}\;\sim$ 450 kyr or 277 kyr. The timescale for HD 14067 b is ${P}_{\mathrm{Kozai}}\;\sim$ 30 kyr with the 0.74 ${M}_{\odot }$ object at 49 au. These timescales are also sufficiently shorter than the system's age. It follows from Equation (3) that if the planets have migrated inward from an initially wider separation, then the initial Kozai timescales would be even shorter. Therefore, we conclude that the Kozai mechanism could be a plausible explanation for the eccentricity of the planets ι Dra b, HD 5608 b, and HD 14067 b. We note that an alternative mechanism for producing high eccentricities is planet–planet scattering (e.g., Nagasawa et al. 2008). This possibility can be tested by continuing RV and direct-imaging observations, in order to search for additional planets in the systems, as well as providing improved constraints for the orbits of the imaged companions.