XO-7 b: A Transiting Hot Jupiter with a Massive Companion on a Wide Orbit

The planet host star has a nearby star at a separation of 8'' that is five magnitudes fainter (G = 15.8407, where G is the Gaia G band magnitude). This neighbor is a K star (T_eff = 4038 K, ${BP}\,-{RP}=1.87$, where BP and RP are magnitudes from the Gaia blue and red photometers), it has a parallax of 2.9315 ± 0.0354 mas, a proper motion of 0.947 ± 0.066 and 17.027 ± 0.082 mas yr⁻¹ in R.A. and decl., respectively, and an estimated distance of 338 ± 4 pc as inferred from Gaia DR2 (Bailer-Jones et al. 2018; Gaia Collaboration et al. 2018). It is located about 140 pixels in the scan direction on the Gaia detectors and does not affect the astrometry of the main target (which could be the case for separations of 10–20 pixels). Its Gaia DR2 astrometric data do not show anything suspicious. The parallax, proper motion, and distance of the main target are reported in Table 2. These measurements show that both stars are unbound and the K star is in the background. They are not resolved by the XO instruments but the K star is faint enough to be negligible at the level of precision of the XO data. Both stars are well resolved in the follow-up observations (Figure 2). No other companion with a magnitude difference less than five is present within 1 arcmin in the bands used for the detection and follow-up (from B to i').

Zoom In Zoom Out Reset image size

Extensive photometric follow-up was conducted by a team of amateur and professional astronomers. We observed 22 transit events between 2017 June 12 and 2018 December 14 with facilities reported in Table 1 using different filters (B, V, g', R, r', i') or without a filter (labeled C for "Clear"). We obtained 32 good-quality light curves that we used in the analysis (Figures 3 and 4).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 1.  Facilities used for the Photometric Follow-up

Observatory	Telescope	Label
Observatoire de Nice, France	Schaumasse, 16 inches (40 cm)	NICE
(a) 2017 Jun 12, (b) 2017 Jul 2, (c) 2017 Jul 25, (d) 2017 Aug 14	FOV: 31' × 23'; Pixel size: 056 px−1
(e) 2017 Aug 17, (f) 2017 Sep 6, (g) 2018 Aug 7, (h) 2018 Aug 27
Observatori Astronòmic del Montsec, Catalonia, Spain	Joan Oró Telescope, 31 inches (80 cm)	TJO
(a) 2017 Jun 12, (b) 2017 Jul 2, (c) 2017 Sep 6, (d) 2017 Oct 14	FOV: 123 × 123; Pixel size: 036 px−1
(e) 2017 Oct 16, (f) 2017 Nov 26, (g) 2018 Aug 5, (h) 2018 Aug 7
(i) 2018 Dec 14
Elgin Observatory, Elgin, Oregon, USA	12 inches (30 cm)	ELGIN
(a) 2017 Jun 24, (b) 2017 Jul 17, (c) 2017 Aug 6, (d) 2017 Aug 9	FOV: 157 × 105; Pixel size: 123 px−1
Las Cumbres Observatory, McDonald Observatory, TX, USA	16 inches (40 cm)	LCOGT-MDO
(a) 2018 Jun 20	FOV: 29' × 19'; Pixel size: 057 px−1
Las Cumbres Observatory, Observatorio del Teide, Tenerife, Spain	16 inches (40 cm)	LCOGT-OT
(a) 2018 Jun 25, (b) 2018 Aug 7, (c) 2018 Aug 30, (d) 2018 Sep 19	FOV: 29' × 19'; Pixel size: 057 px−1

Note. Letters beneath each observatory indicate the dates of observation (see Figure 3). The field of view (FOV) in arcminute and pixel scale in arcsecond are also indicated.

Download table as:  ASCIITypeset image

Observations with the 40 cm Schaumasse telescope at the Observatoire de Nice (Nice, France) were conducted using a Johnson B or R filter, sometimes alternating between the two. The images were calibrated using bias, darks, and flat fields. We reduced the data with the IRIS astronomical image processing software (Buil 2005) and performed differential aperture photometry with four reference stars chosen from their brightness and photometric stability. Observations with the 80 cm Telescopi Joan Oró telescope at Montsec Astronomical Observatory (Lleida, Spain) were conducted with the MEIA2 instrument, a 2k × 2k Andor CCD camera with a pixel scale of 036 and a squared field of view of 123, using a Johnson V filter. The images were calibrated with darks, bias, and flat fields with the ICAT pipeline (Colome & Ribas 2006). Differential photometry was extracted with AstroImageJ (Collins et al. 2017) using the aperture size and the set of comparison stars that minimized the rms of the out-of-transit photometry. Observations with the 30 cm telescope at Elgin Observatory (Elgin, Oregon, USA) were conducted using a CCD without a filter. The images were calibrated using bias, darks, and flat fields. We reduced the data with the AIP4Win v2.4.8 Magnitude Measurement Tool (Berry & Burnell 2005) and performed differential aperture photometry with three or five reference stars depending on the image and seeing quality. These stars were selected for each data set based on lowest noise and lack of curvature in the lightcurve and were averaged. Observations with the 40 cm telescopes of the Las Cumbres Observatory Global Telescope network (LCOGT) were conducted using Sloan Digital Sky Survey (SDSS) g', r', and i' filters. The images were reduced with the LCOGT's BANZAI pipeline (McCully et al. 2018) including bad-pixel masking, bias and dark subtraction, flat-fielding, and image plate solving from astrometry.net (Lang et al. 2010). We performed aperture photometry using the Astropy Photutils package (Bradley et al. 2019). Changes in the target star's position on the detector between nights led us to use nine reference stars for the g' band and 10 for the r' and i' bands.

We gathered these light curves using a consistent format and analyzed them jointly. We converted the dates into Barycentric Julian Date (BJD) and performed a transit fit for each bandpass, during which each light curve was corrected for a linear trend and outliers more than three sigma away were removed. We used these consistent, corrected light curves for the combined fit in Section 4.2.

RV measurements were obtained between 2016 July 23 and 2018 July 4 with the SOPHIE spectrograph (Perruchot et al. 2008; Bouchy et al. 2009, 2013) at the 193 cm telescope of Observatoire de Haute-Provence, France (Figure 5, Appendix). We used its High-Resolution mode (resolving power R = 75,000). Exposure times were around 13 minutes allowing signal-to-noise ratios of around 27 per pixel at 550 nm to be reached on most of the exposures. We used the SOPHIE pipeline to extract the spectra from the detector images, cross-correlate them with a numerical mask to produce clear cross-correlation functions (CCFs), then fit the CCFs by Gaussian curves to derive the RVs (Baranne et al. 1996; Pepe et al. 2002).

Zoom In Zoom Out Reset image size

The resulting CCFs have a contrast that represents ∼31% of the continuum, and a FWHM of 11.0 km s⁻¹ showing some stellar rotation (we measured $v\,\sin \,i$ = 6 ± 1 km s⁻¹ from the CCF width; see Section 4.1). The RVs have typical uncertainties around ±13 m s⁻¹, whereas we removed from our final data set three exposures having uncertainties larger than ±30 m s⁻¹. Only five spectra were contaminated by moonlight. We estimated and corrected for that contamination by using the second SOPHIE fiber aperture, which was placed on the sky while the first aperture pointed toward the target (e.g., Hébrard et al. 2008; Bonomo et al. 2010); this resulted in RV corrections around 35 m s⁻¹ or smaller (whereas the dispersion of the residuals after the combined fit in Section 4.2 is 14.7 m s⁻¹). Excluding those five Moon-contaminated observations does not significantly change our results. The final RV data set shows significant variations in phase with the transit ephemeris and with a semi-amplitude around 80 m s⁻¹ implying a companion mass in the giant-planet regime, as shown in Figure 5.

Radial velocities measured using different stellar masks (G2, K0, or K5) produce variations with similar amplitudes, so it is unlikely that these variations are produced by blend scenarios composed of stars of different spectral types. We finally adopted the RVs obtained with the K0 mask as they provide the least dispersed residuals. Using the RVs obtained from the G2 mask does not significantly change our results. Similarly, the measured CCF bisector spans quantify possible shape variations of the spectral lines. They show no correlations with the RVs, and no significant variations: their dispersion is two times smaller than the RV dispersion, whereas each bisector span is roughly one-half as precise as the corresponding RV measurement. This reinforces the conclusion that the RV variations are due to a planetary companion, and not caused by spectral-line profile changes attributable to blends or stellar activity.

We begin our analysis of the data with a study of the host star. Stellar atmospheric parameters (T_eff, $\mathrm{log}g$, and [Fe/H]) and respective uncertainties were derived using the methodology described in Sousa et al. (2008) and Santos et al. (2013). In brief, we make use of the equivalent widths of tens of iron lines and we assume ionization and excitation equilibrium. The process makes use of a grid of Kurucz model atmospheres (Kurucz 1993) and the radiative transfer code MOOG (Sneden 1973).

The equivalent widths were measured on a SOPHIE spectrum built from the addition of the spectra used for the RV measurements, but excluding the five SOPHIE spectra presenting moonlight contamination. We obtained T_eff = 6220 ± 70 K, log g = 4.2 ± 0.1 (cgs), and [Fe/H] = +0.48 ± 0.05. Using the calibration of Torres et al. (2010) with a correction following Santos et al. (2013), we derive a mass and radius of 1.43 ± 0.09 M_⊙ and 1.47 ± 0.20 R_⊙, respectively. We also derived the projected rotational velocity $v\,\sin \,i$ = 6 ± 1 km s⁻¹ from the parameters of the CCF using the calibration of Boisse et al. (2010).

Proceeding to a comprehensive analysis of the system, we fit the photometric follow-up light curves and the radial velocities together using EXOFASTv2 (Eastman et al. 2019). In each light curve, we verified that the uncertainties of individual data points were of the same order as the standard deviation of the light curve (after subtracting the transit), to ensure that they were not under- or overestimated. In some cases, we rescaled the uncertainties accordingly. We used MIST stellar isochrones (Choi et al. 2016; Dotter 2016; Paxton et al. 2011, 2013, 2015), a spectral energy distribution (SED) constructed from Tycho (Høg et al. 2000), 2MASS (Cutri et al. 2003; Skrutskie et al. 2006), and WISE (Cutri et al. 2014) catalog magnitudes, and the Gaia DR2 parallax (Gaia Collaboration et al. 2018) for BD+85 317 to constrain the host star's parameters. We use a quadratic limb-darkening law:

$$\begin{eqnarray}&&\displaystyle \frac{{I}_{\mu }}{{I}_{0}}=1-{u}_{1}(1-\mu )-{u}_{2}{\left(1-\mu \right)}^{2}\end{eqnarray} \tag{ 1 }$$

where I is the intensity and μ is the angle between a line normal to the stellar surface and the line of sight of the observer. The limb-darkening coefficients u₁ and u₂ are free parameters with theoretical values interpolated from updated Claret (2017) tables. In the fit, we set four priors: we used the SOPHIE spectral analysis to constrain the effective temperature and metallicity for the stellar isochrone fitting. We also set a prior on the Gaia parallax with an updated uncertainty (see Section 4.3). Finally, we used Schlafly & Finkbeiner (2011) measurements to place an upper limit on V-band extinction (see also Green et al. 2019). We set unconstrained starting values (e.g., transit depth and duration) based on the posteriors of a brief fitting run. In the full Markov Chain Monte Carlo (MCMC) analysis, we simultaneously fit a model to the transits and RVs, including a linear fit to the long-term trend from the latter data. The chains were well mixed (Gelman–Rubin statistic <1.01) after ∼37,000 steps.

The final parameter values and their uncertainties are computed as the medians and 1σ values of their respective posterior distribution functions. The limb-darkening coefficients are free and are treated in the same way as other parameters. We also ran a zero-eccentricity model to the data and find no significant changes in the final parameters.

The residual RVs are consistent with zero after subtracting the best-fit hot Jupiter signal and the long-term trend containing the systemic velocity. We found an RV jitter of $8.1\pm 3.3$ m s⁻¹ during the MCMC. We list the results from the fit in Tables 2–4.

Table 2.  Parameters of the Planet Host Star

Quantity	Unit	Value	Notes
Name		BD+85 317	1
R.A.	J2000	18:29:54.929	2
decl.	J2000	+85:13:59.58	2
p	mas	4.2419 ± 0.0215	2
d	pc	234.1 ± 1.2	3
${\mu }_{\alpha }$	$\mathrm{mas}\,{\mathrm{yr}}^{-1}$	−15.354 ± 0.038	2
${\mu }_{\delta }$	$\mathrm{mas}\,{\mathrm{yr}}^{-1}$	24.461 ± 0.054	2
γ	$\mathrm{km}\,{{\rm{s}}}^{-1}$	−12.983 ± 0.015	7
B	mag	11.23 ± 0.06	4
V	mag	10.52 ± 0.04	4
G	mag	10.4575 ± 0.0004	2
J	mag	9.557 ± 0.024	5
H	mag	9.308 ± 0.030	5
K	mag	9.241 ± 0.024	5
BP	mag	10.7795 ± 0.0008	2
RP	mag	10.0087 ± 0.0010	2
BP − RP	mag	0.7707	2
AG	mag	$0.6980\,[0.5709\,0.8174]$	2, 8
$E({BP}-{RP})$	mag	$0.3450\,[0.2543\,0.4170]$	2, 8
Sp Type		G0V	6
${T}_{\mathrm{eff}}$	K	6250 ± 100	7
[Fe/H]	dex	$0.432\pm 0.057$	7
log g	cgs	$4.246\pm 0.023$	7
$v\,\sin \,i$	$\mathrm{km}\,{{\rm{s}}}^{-1}$	6 ± 1	7
M	${{\rm{M}}}_{\odot }$	$1.405\pm 0.059$	7
R	${{\rm{R}}}_{\odot }$	$1.480\pm 0.022$	7
Age	Gyr	$1.{18}_{-0.71}^{+0.98}$	7
RV slope	m s−1 ${\mathrm{day}}^{-1}$	$-0.148\pm 0.011$	7

Note. (1): Argelander (1903), (2): Gaia Collaboration et al. (2018), (3): Bailer-Jones et al. (2018), (4): Høg et al. (2000), (5): Cutri et al. (2003), (6): Pickles & Depagne (2010), (7): This work, (8): The values in brackets are the 16th and 84th percentiles of the probability density function.

Download table as:  ASCIITypeset image

Gaia DR2 data can be used to constrain exoplanet system parameters. In our combined fit, we used the Gaia DR2 parallax as an input. As a case study, we discuss the validity of this measurement using Gaia quality indicators and other studies.

There are a number of indications that the Gaia DR2 astrometry for this star is reliable. Its proper motion and parallax from Gaia DR2, Gaia DR1, and Tycho-2 are in full agreement. There is no nearby bright source: the nearest Tycho-2 star is more than 999 away (prox = 999). There is no indication from Gaia DR1 of a long-term curvature of the proper motion that could be caused by a companion (astrometric_delta_q = 0.00). The ecliptic latitude of the star (β = 7128) is in general good for astrometry, as confirmed by the Gaia DR2 statistics: 219 of the 227 observations (called "transits") have been used in the astrometric solution, distributed over 18 visibility periods, far above the minimum number of five periods required. The mean parallax factor is normal (mean_varpi_factor_al = 0.040), indicating that the astrometric fit should be straightforward, and the astrometric excess noise is zero (astrometric_excess_noise = 0.000 mas), which confirms the high quality of the Gaia DR2 astrometry. One suspicious element is the duplicate source flag (duplicated_source = 1). However, the number of "transits" is similar to that of the nearby faint star (Section 3.1), which has a duplicate source flag of 0. This precludes that the second, duplicated, source identifier is hiding large amounts of data; thus the Gaia DR2 astrometry is based on most data. The astrometric goodness of fit is poor (astrometric_gof_al = 9.9804), but the Revised Unit Weight Error (which is a rescaled astrometric quality indicator) is 1.12, which is lower than the threshold of 1.4 that indicates suspicious astrometry. The duplicate source flag likely originates from the partial saturation of this object.

Lindegren et al. (2018a, 2018b) state that the DR2 error for Gaia parallaxes does not represent the total uncertainty. We increased the uncertainty of the XO-7 parallax using the calibration formula given in slide 17 of Lindegren et al. (2018b; see "Known issues with the Gaia DR2 data"²¹ ). We also considered a systematic offset in the Gaia DR2 parallax of XO-7. Hall et al. (2019) provide a compilation of literature values in addition to their own finding and report a systematic offset around −50 μas (see their Figure 8, Tables 9 and 10). We ran our analysis for two extreme cases: no offset and a −82 ± 33 μas offset as found by Stassun & Torres (2018), propagating the uncertainties accordingly. The largest discrepancy between best-fit parameters from the two analyses was a 1σ difference in the stellar radius. Because the posterior values do not significantly change after the systematic parallax correction, we used the original Gaia DR2 parallax value and the calibrated uncertainty in our final analysis.

The effective temperature given by Gaia DR2 (T_eff = 5877 [5706, 6031] K) is lower than the one we measure from high-resolution spectroscopy (the values in brackets are the 16th and 84th percentiles of the probability density function). The radius from Gaia DR2 based on the BP and RP magnitudes (R = 1.58 [1.51, 1.68] R_⊙) is slightly larger than inferred from our fit. Gaia values are useful for ensemble analysis but are not necessarily accurate for single objects. The system RV from Gaia DR2 (V_sys = −12.82 ± 0.44 km s⁻¹) is in excellent agreement with our measurement.

We do not detect the secondary eclipse in the XO light curve. We set an upper limit on its depth δ_e in the XO bandpass by calculating the noise in the folded light curve. We eliminate the in-transit points, fold the light curve at 10⁵ different periods ranging from 2 to 3.7 days, split each light curve into segments equal to the eclipse duration (assumed to be equal to the transit duration), and calculate the mean flux in each segment. This yields the distribution of flux variations over the timescale of the eclipse. We take the 3σ values of this distribution as the 3σ upper limit on δ_e. We find δ_e < 0.00142 at 3σ in the R band. This limit is too high to constrain the brightness temperature and albedo of the planet's day side.

We measure the central times t_c of individual transits that were observed during the follow-up campaign. We do not find any correlation pattern between the t_c and the period index of the transits, calculated as the number of orbital periods after the first transit. We put an upper limit on the presence of transit timing variations (TTVs) by measuring the standard deviation of the distribution of t_c: TTVs of XO-7b should be lower than 5 minutes at 1σ (15 minutes at 3σ) over the two years of observations. This is consistent with the fact that our RV measurements rule out the presence of companions massive and close enough to induce significant TTVs on shorter timescales. We note that measuring TTVs from ground-based observations that are affected by correlated noise is challenging, as studied by Carter & Winn (2009).

The host star is BD+85 317 and is classified as G0V (Pickles & Depagne 2010), sometimes G2 (Wright et al. 2003). Its parameters are reported in Table 2. It is bright and circumpolar (within 5° of the celestial north pole), which could facilitate follow-up observations from the northern hemisphere. It has a high metallicity that is among the highest for stars harboring a hot Jupiter. Its age estimated from the best-fit MIST isochrone in the ${T}_{\mathrm{eff}}-\mathrm{log}\,g$ space indicates that it is relatively young. However, several isochrones of disparate ages are in close proximity and provide a good fit as reflected by the large uncertainty (nearly 100%). Thus, the age should be taken with caution and further study such as activity indicator analysis would be necessary for a better estimate. If indeed the star is young, XO-7 b would be one of the very few hot Jupiters known around young stars.

The hot Jupiter has an orbital period of 2.864 days, a mass of 0.709 ± 0.034 ${{\rm{M}}}_{{\rm{J}}}$, and a radius of $1.373\pm 0.026\,{{\rm{R}}}_{{\rm{J}}}$, yielding a density of $0.340\pm 0.027\,{\rm{g}}\,{\mathrm{cm}}^{-3}$. At a distance of $0.04421\,\pm 0.00062$ au from its star, the planet has an equilibrium temperature of $1743\pm 23\,{\rm{K}}$ assuming a zero albedo. The orbit is consistent with being circular. The parameters of the hot Jupiter are reported in Table 3 and the best-fit limb-darkening coefficients for each band are reported in Table 4.

Table 3.  Median Values and 68% Confidence Intervals for XO-7 b

Quantity	Unit	Value
P	Period (day)	2.8641424 ± 0.0000043
RP	Radius (${{\rm{R}}}_{{\rm{J}}}$)	1.373 ± 0.026
TC	Time of conjunction (${\mathrm{BJD}}_{\mathrm{TDB}}$)	2457917.47503 ± 0.00045
a	Semimajor axis (au)	0.04421 ± 0.00062
i	Inclination (degrees)	83.45 ± 0.29
e	Eccentricity	0.038 ± 0.033
Teq	Equilibrium temperature (K)	1743 ± 23
MP	Mass (${{\rm{M}}}_{{\rm{J}}}$)	0.709 ± 0.034
K	RV semi-amplitude (m s−1)	80.5 ± 3.2
${R}_{P}/{R}_{* }$	Radius of planet in stellar radii	0.09532 ± 0.00093
$a/{R}_{* }$	Semimajor axis in stellar radii	6.43 ± 0.14
δ	Transit depth (fraction)	0.00909 ± 0.00018
τ	Ingress/egress transit duration (day)	0.0190 ± 0.0015
T14	Total transit duration (day)	0.1155 ± 0.0014
TFWHM	FWHM transit duration (day)	0.09655 ± 0.00074
b	Transit impact parameter	0.709 ± 0.023
${\delta }_{S,3.6}$	Blackbody eclipse depth at 3.6 μm (ppm)	898 ± 29
${\delta }_{S,4.5}$	Blackbody eclipse depth at 4.5 μm (ppm)	1150 ± 34
${\rho }_{P}$	Density (cgs)	0.340 ± 0.027
loggP	Surface gravity	2.970 ± 0.028
Θ	Safronov Number	0.0325 ± 0.0014
$\langle F\rangle$	Incident Flux (109 erg s−1 cm−2)	2.09 ± 0.11
TS	Time of eclipse (${\mathrm{BJD}}_{\mathrm{TDB}}$)	2457918.900 ± 0.029
${M}_{P}/{M}_{* }$	Mass ratio	0.000482 ± 0.000021
PT	A priori non-grazing transit prob	0.1457 ± 0.0073
${P}_{T,G}$	A priori transit prob	0.1764 ± 0.0087

Download table as:  ASCIITypeset image

The RV measurements show a linear trend in addition to the radial velocities induced by the hot Jupiter, indicating the presence of a wide-orbit companion. The secular change is 100 m s⁻¹ over two years and no curvature is apparent in the data. Thus, the minimum orbital period is four years in the case of a circular orbit and two years for a very eccentric orbit. Assuming a circular orbit, we derive a minimum mass of 4 ${M}_{{\rm{J}}}$ for the companion, which could be a planet, a brown dwarf, or a star. This system is still under monitoring to characterize the long-period companion.

XO-7 b is an inflated hot Jupiter and is moderately hot. Its large atmospheric scale height (H = 671 km), combined with the brightness of the host star, makes it well suited to atmospheric characterization. It is among the 25 known transiting hot Jupiters with an atmospheric scale height larger than 500 km and a host star brighter than magnitude 11 in the V band. Assuming an absorption spanning two scale heights and estimating the amplitude of the transmission signal by $2\,H\times 2\,{R}_{p}/{R}_{\star }^{2}$, we expect a signal of 250 ppm, which could be detected with Hubble Space Telescope (HST; Figure 6). Thus, XO-7 b is a valuable target to investigate the atmospheric properties of moderately irradiated close-in gas giant planets.

Zoom In Zoom Out Reset image size

Among known transiting hot Jupiters with a mass, radius, and equilibrium temperature within 20% of those of XO-7 b, two have been observed in spectroscopy with HST STIS and/or WFC3: HD 209458 b and HAT-P-13 b. Atomic and molecular species have been detected in the atmosphere of HD 209458 b, including water vapor signatures around 1.4 μm with an amplitude of 200 ppm, which is about twice smaller than expected for a clear, solar-composition atmosphere, and indicates extra absorption by haze and/or dust (Deming et al. 2013; Sing et al. 2016). This low amplitude might also be due to a depletion in oxygen compared to solar abundance (Madhusudhan et al. 2014), although this explanation is not favored at the moment. The HAT-P-13 b WFC3 observations have not been published but are available on the MAST archive. Observing XO-7 b in transit spectroscopy would test if the atmospheric properties measured for HD 209458 b are also valid for a planet with similar characteristics and would help constrain hot Jupiter atmosphere models.

The host star has a high ecliptic latitude (β = 7128) and will be visible during a continuous period of 212 days with JWST. In addition, it is bright but will not saturate the JWST detectors in time-series spectroscopic observation modes (except in the NIRSpec PRISM/CLEAR configuration) and there is no surrounding star that could contaminate the spectrum (the faint nearby star is far enough away). Thus, it is an excellent target for JWST. It will also be a good target for ARIEL for similar reasons: it will be visible continuously, it is bright, and the presence of the companion makes it valuable as part of a larger sample to investigate connections between formation and migration mechanisms and atmospheric compositions. Thus, it would be a good addition to the ARIEL target list.

The transiting extrasolar planet XO-7 b reported here is probably the last of the series discovered by the XO project (McCullough et al. 2005). This section addresses some of the ways in which data from a ground-based survey such as XO differ from data from the TESS mission.²² Comparison of XO data for XO-7 b and TESS data for a similar planet (WASP-126b) demonstrates that, for discovering transiting planets, TESS data will be far superior. TESS is observing XO-7 as TIC 268403451 in camera 3 during its Sectors 18–20, between 2019 November and 2020 January. The TESS light curve will consist of nearly continuous monitoring of XO-7 for three months at 2 minutes cadence. For comparison, we selected an exoplanet candidate (TIC 25155310; TOI-114.01) with similar properties to XO-7/XO-7 b from the many candidates already reported by TESS. Like XO-7 b, the example that we selected happens to be a rediscovery by TESS of a planet, WASP-126b, discovered by a ground-based survey. Whereas XO-7 b transits a V = 10.5 mag G0 V star every 2.9 days, WASP-126b transits a V = 10.8 mag G2 V star every 3.3 days (Maxted et al. 2016). We obtained a TESS data validation report for WASP-126b from MAST²³ that includes 21 transits observed during TESS sectors 1–3. WASP-126b's transit depth is reported as 7182 ± 39 ppm.

Detection of the secondary eclipse of XO-7 b would permit a measure of the superposition of reflected starlight and emission from the planet's illuminated side. It would also confirm whether the orbital eccentricity is indeed close to zero. Unfortunately, the prospects are very poor for detecting the secondary eclipse of XO-7 b with TESS data. Based on the reported uncertainty of the depth of WASP-126b's transit, a secondary eclipse of similar duration would have been marginally detected by TESS at 3σ if its depth were 117 ppm. However, if the geometric albedo of XO-7 b is typical of hot Jupiters, e.g., 0.1, then its secondary eclipse depth due solely to reflected light is expected to be 21 ppm (Sheets & Deming 2017, Equation (4)). While thermal emission from a hot Jupiter can contribute to the depth of its secondary eclipse, even within the 600–1000 nm bandpasses of TESS or Kepler, Sheets & Deming (2017, Table 4) report a median of 35 ppm and a maximum of 91 ppm for the depths of secondary eclipses of 14 hot Jupiters. Because those two estimates of secondary eclipse depths (21 ppm and 35 ppm) are much smaller than the 117 ppm estimate of a marginal (3σ) detection of such an eclipse with TESS data similar to that expected for XO-7, we expect that TESS will not detect XO-7 b's secondary eclipse.

Similarly, a search of TESS data of XO-7 could turn up transits of a planet smaller in radius than XO-7 b. However, such a search is likely to be fruitless because such companions to hot Jupiters are rare (Steffen et al. 2012). Likewise, measuring XO-7's mean density or age from asteroseismology will not be possible with TESS data: XO-7 is a few magnitudes too faint (see Campante et al. 2016, Figure 12(a)). TESS photometry may enable measuring the rotation period of the star XO-7 and comparing XO-7's rotation period and its radius with its spectroscopically determined $v\,\sin (i)$ may yield the star's spin-axis inclination (i) for comparison with the projected angle of orbital obliquity of XO-7 b inferred from the Rossiter–McLaughlin effect (Holt 1893; McLaughlin 1924; Rossiter 1924). For stars hotter than approximately 6250 K, the pole of the planet's orbit could be preferentially misaligned with respect to the stellar spin axis (Winn et al. 2010a); XO-7's effective temperature (6250 ± 100 K; Table 2) places it at the threshold.

The secular trend in XO-7's radial velocities indicates an unseen companion, either a planet, brown dwarf, or star. Ngo et al. (2016) estimate that the hosts of hot Jupiters have stellar companions with separations less than 50 au in ${3.9}_{-2.0}^{+4.6}$% of their sample. Knutson et al. (2014) estimate that 27% ± 6% of hot Jupiters have a planetary companion in the range of mass 1–13 M_J and semimajor axis 1–10 au. In absolute value, the slope of XO-7's radial velocities, $-0.148\pm 0.001\ {\rm{m}}\,{{\rm{s}}}^{-1}\,{\mathrm{day}}^{-1}$, is 1.5 times larger than the maximum slope ($-0.097\pm 0.023\ {\rm{m}}\,{{\rm{s}}}^{-1}\,{\mathrm{day}}^{-1}$) exhibited in a sample of 51 hot Jupiters observed by Knutson et al. (2014), typically for at least five years each.²⁴ The slope of XO-7's radial velocities implies the following relationship between the unseen companion's mass (M_c, in solar units) and its angular separation from XO-7 (θ, in arc seconds; Knutson et al. 2014, Equation (1)):

$$\begin{eqnarray}&&{M}_{c}=40\ {\theta }^{2},\end{eqnarray} \tag{ 2 }$$

which implies that the companion must be hidden within a nominal ground-based seeing disk (2'' FWHM), otherwise it would be much more massive than XO-7 and would dominate the light. If it is a low-mass star (0.1–0.4 M_⊙), then its separation is 005–01, or 12–24 au, and could be revealed, with adaptive optics as a main-sequence M star, respectively 7–4 mag fainter than XO-7 in K band (Delfosse et al. 2000, Figure 1).

Regardless of whether it is a planet, brown dwarf, or a star, if its orbit is not very elliptical, then its orbital period must be measured in years, otherwise the trend of radial velocities would show some degree of curvature. For example, a M sin(i) = 5 M_J companion in a five-year circular orbit is consistent with the radial velocities measured to date. In that case, Gaia astrometry would be able to detect at ∼5σ/sin(i) the ∼50/sin(i) micro-arcsecond astrometric wobble of XO-7 induced by such a companion (Perryman et al. 2014). Astrometric orbits from Gaia will be available in DR4. Combining the astrometric orbit of XO-7 with our RV measurements will allow us to reconstruct the orbits in three dimensions and measure the mass of the wide-orbit companion. If it is stellar, the companion may also be detected in photometry by Gaia (de Bruijne et al. 2015). Companions at such small separations are very incomplete in Gaia DR2 and need special ground processing (Arenou et al. 2018); they will be available in Gaia DR4.

Hot Jupiters are thought to form beyond the water ice line and migrate inward to reach close-in orbits. Two mechanisms have been proposed: disk or high-eccentricity migration. Disk migration should yield orbits that are circular and contained in the plane perpendicular to the star rotation axis, and can go on until the planet reaches the Roche limit a_Roche. In contrast, high-eccentricity migration should yield a wide range of obliquities and the planet is expected to reach a circular orbit at a distance almost exactly 2 a_Roche (Faber et al. 2005). The orbit may also retain some eccentricity depending on the circularization timescale. In this process, the initial eccentricity can originate from scattering interactions with another massive companion. Although no correlations have been found between hot Jupiter orbital parameters and the presence of distant companions (Knutson et al. 2014), we place XO-7 in the context of these mechanisms. XO-7 has an outer companion of at least 4 M_J and the hot Jupiter appears to have a nearly circular orbit. We calculate the Roche limit of the star–hot Jupiter system as ${a}_{\mathrm{Roche}}=2.7\,{R}_{p}\,{({M}_{\star }/{M}_{p})}^{1/3}$ and find a_Roche = 0.023 au. Interestingly, the semimajor axis is almost twice the Roche limit ($a/{a}_{\mathrm{Roche}}=1.95$). This is consistent with expectations for high-eccentricity migration (although disk migration is not ruled out). As illustrated in Figure 8 of Sarkis et al. (2018), almost all transiting hot Jupiters with a massive outer companion have a/a_Roche > 2. One exception has a/a_Roche just below 2, as for XO-7 b. Overall, this supports high-eccentricity migration for hot Jupiters in these systems.

If the planet–planet scattering mechanism played a role in the formation of this system, then we can expect a non-zero obliquity for the hot Jupiter. The wide-orbit companion may also affect the hot Jupiter's orbit via the Lidov–Kozai mechanism (Kozai 1962; Lidov 1962), which has been proposed to explain the observed sky-projected obliquities of hot Jupiters (Schlaufman 2010; Winn et al. 2010a; Albrecht et al. 2012; Dawson & Chiang 2014; Winn & Fabrycky 2015). Our attempts to observe the Rossiter–McLaughlin effect during a transit of XO-7 b to measure its obliquity have been unsuccessful so far, but we plan to make this observation in the near future. This will bring another clue to understand the formation, migration, and architecture of the XO-7 system.

The XO-7 system has striking similarities with HAT-P-13 (Bakos et al. 2009; Winn et al. 2010b). The two hot Jupiters XO-7 b and HAT-P-13 b have, respectively, similar periods (2.86 and 2.92 days), masses (0.71 and 0.85 ${{\rm{M}}}_{{\rm{J}}}$), radii (1.37 and 1.27 ${{\rm{R}}}_{{\rm{J}}}$), semimajor axes (0.044 and 0.043 au), and equilibrium temperatures (1747 and 1653 K); the host stars XO-7 and HAT-P-13 are metal-rich G dwarfs with similar metallicities (0.43 and 0.41 dex), and both have a second companion on a wide orbit. HAT-P-13 c is well characterized ($P=446.27\pm 0.22\,\mathrm{days}$, $M\,\sin \,i=14.28\pm 0.28\,{{\rm{M}}}_{{\rm{J}}}$, e = 0.6616 ± 0.0054) whereas XO-7 is still under monitoring to characterize that companion. HAT-P-13 has a third outer companion revealed by a linear trend in radial velocities (Winn et al. 2010b; Knutson et al. 2014). XO-7 b and HAT-P-13 b have radius ratios a/a_Roche of 1.96 and 2.27 and both have a nearly zero eccentricity (0.036 ± 0.032 and 0.0133 ± 0.0041). These systems have different ages (${1.12}_{-0.66}^{+0.94}$ and ${5.0}_{-0.7}^{+2.5}$ Gyr). It would be interesting to investigate if they could have formed in the same way. Also, comparing the XUV emission of the stars, for example, with the He i 1.08 μm absorption feature of the hot Jupiter's atmospheres with high-resolution spectroscopy could provide clues on evaporation scenarios of these atmospheres (e.g., Allart et al. 2018; Nortmann et al. 2018). Finally, both stars are metal-rich and host at least two massive companions, and it is known that giant planet formation is correlated with stellar metallicity (Fischer & Valenti 2005). On the other hand, no statistically significant correlation between the frequency of long-period companions and stellar metallicity has been found in hot Jupiter systems (Knutson et al. 2014). Thus, how to interpret these high metallicities in the context of the formation of these systems remains an open question.