THE PAN-PACIFIC PLANET SEARCH. VI. GIANT PLANETS ORBITING HD 86950 AND HD 222076

We find that the velocities for HD 29399 can be fit with a planet having $P=765\pm 15$ days and $K=52\pm 34$ m s⁻¹; the large uncertainty can be attributed to phase gaps which allow for a family of high-eccentricity solutions. As in our previous work, we checked the All-Sky Automated Survey (ASAS) V-band photometric data (Pojmanski & Maciejewski 2004) for variability due to intrinsic stellar processes. We analyzed a total of 984 epochs spanning 8.36 years, with a mean value of 6.07 ± 0.32 mag. This scatter is an order of magnitude larger than that found for previously confirmed planet hosts from the PPPS (Wittenmyer et al. 2015, 2016c). The generalized Lomb–Scargle periodogram of this photometry (Zechmeister & Kürster 2009) is shown in Figure 1, with the period of the candidate planet marked as a vertical dashed line. There is an extremely significant periodicity squarely at the ∼765 day period of our radial velocities. Given this evidence and the large photometric variability, we must conclude that the $K\sim 50$ m s⁻¹ signal in the radial velocities is intrinsic to the star and not due to an orbiting planet. While stellar rotation is frequently the culprit in radial velocity false positives (e.g., Robertson et al. 2015; Johnson et al. 2016; Rajpaul et al. 2016), it is extremely unlikely in the case of HD 29399. The projected rotational velocity determined by De Medeiros et al. (2014) is less than 1.2 km s⁻¹; using the stellar radius of 4 R_⊙ in Table 5, this gives a rotation period of 169 days. Of course, since the measured v sin i is a lower bound, it remains possible that the true rotational velocity is smaller, arising from a nearly pole-on orientation. However, such an orientation would require an unreasonably high spot coverage to produce the observed variations in the radial velocities. In Section 4.1, we note that HD 29399 may host a debris disk; one might imagine quasi-periodic transits by debris (e.g., Croll et al. 2015; Vanderburg et al. 2015; Rappaport et al. 2016) to cause the photometric and radial velocity variations much as starspots would. However, we show in Section 4.1 that the required debris is best modeled at 1500 K, and hence must orbit far too close to the star to produce a 765 day period by Keplerian orbital motion. Indeed, for such a long periodicity, the most likely source is a stellar magnetic activity cycle. In Section 3.3, we give further evidence via an analysis of the Hα feature.

Zoom In Zoom Out Reset image size

For HD 86950, we find a clear signal with $P=1270\pm 57$ days and $K=49\pm 12$ m s⁻¹, corresponding to a planet with m sin i of 3.6 ± 0.7 ${M}_{\mathrm{Jup}}$ adopting a host star mass of 1.66 M_⊙. The data and model fit are shown in Figure 2. Examination of 8.9 years (515 epochs) of ASAS photometry shows no periodicities of significance near the planet's orbital period (Figure 3). The ASAS V band photometry has a mean value of 7.47 ± 0.02 mag.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For HD 222076, all three instruments clearly reveal a planet with $P=871\pm 19$ days and m sin $i=1.56\pm 0.11$ ${M}_{\mathrm{Jup}}$ assuming a host star mass of 1.07 M_⊙ (Wittenmyer et al. 2016d). The orbit is nearly circular, and this fit has residuals of 5.9 m s⁻¹ (Figure 4). Again, the ASAS photometry reveals no significant periodicities near the planet's orbital period (Figure 5). The ASAS V band photometry has a mean value of 7.46 ± 0.02 mag. For this star, we have 5.67 years of FEROS spectra, uncontaminated by iodine, from which we can derive several activity indices (bisector velocity span, CCF FWHM, and S-index). None of these indicators (Table 7) shows any periodicities or correlations with the radial velocities. We are thus confident that the observed velocity variations are due to an orbiting planet.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the light of recent debate over the detection of planet-induced stellar reflex motion amidst the confounding effect of stellar activity (e.g., Robertson et al. 2014; Anglada-Escudé & Tuomi 2015; Fischer et al. 2016), we discuss in this section the Hα activity indices for the three stars considered here.

Stellar activity can induce spurious velocity signatures that mimic the velocities produced by exoplanets. This can appear at the stellar rotation period or its harmonics (Boisse et al. 2011). Magnetic cycles can also produce radial velocity signals comparable in amplitude to orbiting planets (e.g., Santos et al. 2010; Dumusque et al. 2011; Robertson et al. 2015; Faria et al. 2016). Multiple mechanisms can produce these line profile variations. For example, variable levels of chromospheric activity can produce changes in the level of line profile reversal in some line cores, resulting in changes to both the line centroid and hence the measured radial velocity (Martínez-Arnáiz et al. 2010). These effects will also produce changes in the line's equivalent width (EW), and so measurement of the EW can provide an indicator of the presence of activity-induced radial velocity variations (Robertson et al. 2014). Stellar activity is also seen to be correlated with the presence of starspots (Berdyugina 2005; Strassmeier 2009) and suppression of convection (Haywood et al. 2016), which will produce changes in line profile shapes and so line centroids, resulting in velocity jitter (Reiners et al. 2010; Andersen & Korhonen 2015). So EW measurements for a line-profile sensitive feature would be expected to be correlated with this source of jitter as well.

We have therefore measured the EW of Hα absorption as an indicator of variability in chromospheric activity. Our analysis builds on that presented by Robertson et al. (2014), with the addition of an automated algorithm for continuum normalization and telluric contamination identification in the region of the Hα line. This method (detailed in the Appendix) has the advantage of being robust for arbitrary, slowly varying continua selection, without being parametric for the specific shape of the continuum or the location of specific absorption lines.

We show the stacked spectra of the three stars featured in this work, along with their RV–EW relations (Figures 6–8). For HD 29399, where a strong ∼2 yr periodicity was evident in the photometry and the radial velocities, we find a correlation between the Hα EW and the radial velocity, further evidence that the RV periodicity is intrinsic to the star and is most likely the result of a magnetic activity cycle as noted in Section 3.1. For the candidate planet hosts HD 86950 and HD 222076, no correlations are seen from the RV–EW plots, supporting our claim that the detected RV variations are not due to chromospheric stellar activity.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

HD 86950 has been identified by McDonald et al. (2012) as having a possible infrared excess based on the presence of excess emission at $9\,\mu {\rm{m}}$ in the AKARI/IRC All-Sky Survey (Ishihara et al. 2010), with a fractional luminosity (${L}_{\mathrm{dust}}/{L}_{\mathrm{star}}$) of $\sim 1.2\times \,{10}^{-3}$. The presence of both a planetary system and a debris disc around HD 86950 would make it a nearly unique object among sub-giant stars, joining κ CrB (HD 142091) as one of very few examples of an evolved star hosting both a debris disc and exoplanet (Bonsor et al. 2013). Confirmation of this excess (and a better determination of the disc's properties if confirmed) is therefore critical.

We compiled a spectral energy distribution from photometry spanning optical to mid-infrared wavelengths, including optical BV, near-infrared 2MASS ${{JHK}}_{s}$ (Skrutskie et al. 2006), and mid-infrared MSX (Egan et al. 2003) and WISE (Wright et al. 2010) measurements, in addition to the AKARI 9 μm datum. We fit the stellar photospheric emission with a model from the BT-SETTL/Nextgen stellar atmospheres grid appropriate for the spectral type (K0III; ${T}_{\mathrm{eff}}=4750$ K, log g = 2.0, [Fe/H] = 0.0), and scaled to the stellar radius and distance (van Leeuwen 2007; Wittenmyer & Marshall 2015). We color corrected the AKARI and WISE flux densities assuming blackbody emission from the star. The resulting spectral energy distribution is shown in the left panel of Figure 10. No evidence of significant excess (i.e., (${F}_{\mathrm{obs}}-{F}_{\star }/{\sigma }_{\mathrm{obs}})\gt 3$) from the system is observed out to wavelengths ∼22 μm, ruling out the warm, bright disc inferred from McDonald et al. (2012). The previous identification of infrared excess from HD 86950 can be attributed to the AKARI flux density not being color corrected.

Zoom In Zoom Out Reset image size

We performed a similar analysis of the available photometric data for HD 222076. We find no evidence for an excess, with the maximum at 9 μm and 1.8σ (Figure 10, right panel). McDonald et al. (2012) likewise find no excess in their analysis.

While we find that HD 29399 does not host a planet, we note that McDonald et al. (2012) also identify it as having an excess, with (${L}_{\mathrm{dust}}/{L}_{\mathrm{star}}$) $\sim 3.6\times \,{10}^{-3}$. Using the same approach as described above, we show the spectral energy distribution for HD 29399 in Figure 11. Photometry for the targets was compiled from Johnson BV, 2MASS JHKs (Cutri et al. 2003), WISE All-Sky Survey (Wright et al. 2010), and the Akari IRC All-Sky Survey (Ishihara et al. 2010). We avoided using WISE W1 and W2 photometry due to known saturation issues for bright ($V\lt 8$) stars. We fit the stellar photospheric emission with the following model atmosphere consistent with the physical parameters given in Table 5: ${T}_{\mathrm{eff}}=4800$ K, log g = 3.5, [Fe/H] = 0.0. Adopting a stellar radius of 3.97 R_⊙ and the distance as given in Table 5, we find an infrared excess at the 3–8σ level. This can be fitted with a star+dust model using 1500 K dust (${\chi }_{\nu }^{2}=1.085$). The star is not yet so evolved (nor is its photospheric temperature cool enough) that it would be expected to be surrounded by a dusty envelope, one potential origin of the excess. Likewise, the radial velocities rule out the presence of a cool binary stellar companion. If we adopt a 25% larger stellar radius of ∼5 R_⊙, then the excess disappears and would be consistent with the poorly constrained near-infrared photometry for this star. However, assuming we are confident in the derivation of our stellar parameters and that the circumstellar dust origin for the excess holds, such hot dust might therefore be produced by, e.g., delivery of comets from an outer, cool debris belt. Additional measurements of the spectral energy distribution at wavelengths $\gt 30\,\mu$ to search for such a debris belt would be of value in this case.

Zoom In Zoom Out Reset image size

Following recent work by Reffert et al. (2015) and Jones et al. (2016), we now examine the results of the PPPS in light of the planet–metallicity relation for evolved hosts as noted by those authors. Overall, our PPPS sample of 164 stars has 11 confirmed planet hosts, comparable to the 10 planet hosts in 166 stars reported by the EXPRESS survey (Jones et al. 2016). Reffert et al. (2015) reported 15 secure planet hosts (and a further 20 candidates) among 373 giant stars in their Lick Observatory sample. Figure 12 shows the host-star mass versus metallicity [Fe/H] for 164 PPPS stars; the 11 confirmed hosts are shown as red points.

Zoom In Zoom Out Reset image size

For ease of comparison, we cast our results into the same stellar mass–metallicity bins as used by Reffert et al. (2015). Table 8 shows the PPPS results in these bins; Jones et al. (2016) presented the EXPRESS results in the same format in their Table 3. Our results are broadly consistent; a key difference is that the PPPS sample has very few targets of super-solar metallicity compared to the EXPRESS and Lick surveys. This results in the highest-metallicity bins being sparsely populated and not quite amenable to a full analysis including the dimension of host-star mass (as was performed for the 373-star sample presented in Reffert et al. 2015).

Table 8.  Detection Fraction in Host Star Mass/Metallicity Bins

[Fe/H]	${M}_{* }$ (M⊙)	Nstars	Nhosts	f (%)
−0.36	1.4	10	0	${0.0}_{-0.0}^{+9.1}$
−0.36	2.2	1	0	${0.0}_{-0.0}^{+50.0}$
−0.36	3.0	0	0	⋯
−0.20	1.4	27	1	${3.7}_{-2.3}^{+5.6}$
−0.20	2.2	2	0	${0.0}_{-0.0}^{+33.4}$
−0.20	3.0	0	0	⋯
−0.04	1.4	60	4	${6.7}_{-2.6}^{+4.0}$
−0.04	2.2	4	1	${25.0}_{-15.0}^{+25.0}$
−0.04	3.0	0	0	⋯
+0.12	1.4	49	5	${10.2}_{-3.6}^{+5.2}$
+0.12	2.2	0	0	⋯
+0.12	3.0	0	0	⋯
+0.28	1.4	10	0	${0.0}_{-0.0}^{+9.1}$
+0.28	2.2	0	0	⋯
+0.28	3.0	0	0	⋯

Note. Bin sizes are the same as in Reffert et al. (2015), their Table 1.

Download table as:  ASCIITypeset image

We next investigate the occurrence rate only as a function of [Fe/H] as shown in Figure 13. For ease of comparison with Jones et al. (2016), we use the same bins as from their Figure 8. The 68.3% binomial confidence intervals are computed after Wilson (1927), which is noted by Brown et al. (2001) as a preferable method for small sample sizes.

Zoom In Zoom Out Reset image size

Consistent with Hekker & Meléndez (2007), Reffert et al. (2015), and Jones et al. (2016), we also show an increasing planet occurrence rate with stellar metallicity. The two samples, of planet hosts and the parent sample, are shown by the Kolmogorov–Smirnov (K-S) test to be drawn from the same distribution with a K-S probability of $P=0.708$, i.e., a 70.8% probability that they are from the same underlying distribution.

Our findings of a positive correlation between metallicity and planet occurrence are in contrast to the results of some other evolved star surveys (e.g., Pasquini et al. 2007; Takeda et al. 2008; Mortier et al. 2013). The disagreement most likely arises from different selection criteria for the samples of evolved stars observed by the various groups. Mortier et al. (2013) pointed out that since most surveys choose a color cutoff $(B-V)\leqslant 1.0$, the more metal-rich, low-gravity stars are excluded. To illustrate this, Figure 14 plots the distribution of log g versus [Fe/H] for the targets of four surveys, two of which find a planet–metallicity correlation (Jones et al. 2016; Wittenmyer et al. 2016d), and two of which do not (Takeda et al. 2008; Mortier et al. 2013). The surveys which did not find a correlation are shown as red open circles, generally lying at lower [Fe/H] and lower surface gravity than the others. Mortier et al. (2013) showed a similar result in their Figure 6. We also show the differences in $(B-V)$ colors from these four surveys in the right panel of Figure 14. Again, the surveys finding a planet–metallicity correlation tend to sample redder stars with $(B-V)\geqslant 1.0$.

Zoom In Zoom Out Reset image size

These results from the PPPS do not yet account for survey incompleteness induced by nonuniform detectability (e.g., Howard et al. 2010; Wittenmyer et al. 2011b, 2016a). A more comprehensive analysis of the overall survey completeness is the subject of a forthcoming paper.