Qatar Exoplanet Survey: Qatar-8b, 9b, and 10b—A Hot Saturn and Two Hot Jupiters

The survey data were collected with QES, hosted by the New Mexico Skies Observatory¹⁹ located in Mayhill, NM, USA. A full description of QES can be found in our previous publications, e.g., Alsubai et al. (2013, 2017).

The discovery light curves of Qatar-8b and Qatar-9b contain 2959 and 2755 data points, respectively, obtained during observations from 2016 December 5 to 2017 May 9. For Qatar-10b, the discovery light curve has 2077 data points collected in the time period 2017 March 21–November 1. The survey images are run through the QES pipeline, which extracts the photometric measurements using the image subtraction algorithm by Bramich (2008). A full description of the pipeline can be found in Alsubai et al. (2013).

The output light curves are ingested into the QES archive and are detrended using a combination of the Trend Filtering Algorithm (TFA; Kovács et al. 2005), which constructs a filter function from a set of field stars considered to be a representative template for systematics in the field, and the DOHA algorithm (Mislis et al. 2017), a co-trending algorithm used to eliminate lingering, quasi-systematic patterns identified from groups of stars that are highly correlated to each other. The light curves are then further processed with the TSARDI algorithm (Mislis et al. 2018), a machine learning points rejection algorithm that deals with any residual data irregularities. Qatar-8b, 9b, and 10b were identified as strong candidates during a search for transit-like events using the Box Least Squares algorithm (BLS) of Kovács et al. (2002), following a procedure similar to that described in Collier Cameron et al. (2006). Note that, although the initial candidate selection is an automatic procedure, the final vetting is done by eye.

Figure 1 shows the discovery light curves for the three exoplanets discussed in this paper.

Zoom In Zoom Out Reset image size

Follow-up photometric observations of a number of transits of Qatar-8b, 9b, and 10b were collected at five different observatories with the following combination of telescopes and instruments:

FLWO: the 1.5 m telescope at the Fred L. Whipple Observatory (Mount Hopkins, Arizona, USA) in combination with KeplerCam, equipped with a single 4k × 4k Fairchild CCD with a 037 pixel⁻¹ and a 231 × 231 on-sky field of view (FOV).

QFT: the 0.5 m Qatar Follow-up Telescope (New Mexico Skies Observatory, Mayhill, New Mexico, USA), equipped with a 1k × 1k Andor iKon-M 934 CCD, yielding a FOV of 13' × 13'.

OBP: the 0.82 m telescope at the Observatoire des Baronnies Provençales²⁰ (Provence-Alpes-Côte d'Azur, France) equipped with a FLI ProLine PL230 camera with a 2k × 2k e2v CCD detector resulting in a 23' × 23' FOV.

CAHA: the 1.23 m Zeiss telescope at the Centro Astronómico Hispano-Alemán (Calar Alto, Spain) in combination with the DLR-MKIII camera with a 4k × 4k, e2v CCD resulting in a 215 × 215 FOV.

TCS: the 1.52 m Telescopio Carlos Sanchez at the Teide Observatory (Tenerife, Canary Islands, Spain) with the MuSCAT2 instrument, which takes images in four filters simultaneously. Each channel is equipped with a 1k × 1k CCD, resulting in a 74 × 74 on-sky FOV. For a detailed description of MuSCAT2 and its dedicated photometric pipeline see Narita et al. (2019).

TRAPPIST-North: the 0.6 m robotic Transiting Planets and Planetesimals Small Telescope²¹ is located at Oukaimeden Observatory, Morocco. It is equipped with a 2k × 2k deep-depletion Andor IKONL BEX2 DD CCD camera with a pixel scale of 060 and an on-sky FOV of 198 × 198.

All of the follow-up light curves were generated through differential aperture photometry performed on the sequence of images for each observing run. In each case a number of comparison stars were selected and those with excessive noise or suspected variability were excluded from the final analysis. For the CAHA observations, the telescope was defocused and data reduction was carried out using the DEFOT pipeline (Southworth et al. 2009, 2014). For observations taken at FLWO, OBP, and with QFT the telescope was kept only approximately in focus and we used the AstroImageJ (AIJ) software package (Choi et al. 2016) to extract the light curves. The MuSCAT2 instrument on TCS has a dedicated pipeline for extracting the light curves. The extraction of the fluxes for the TRAPPIST-North observations was done by aperture photometry on selected stars with the IRAF/DAOPHOT software package (Stetson 1987). Final transit light curves were produced by normalizing to a low order polynomial (maximum order 2 in one case, a straight line for all other cases) fitted to the flat part of the light curves that also removes small residual trends if present. The uncertainties in the CAHA and MuSCAT2 light curves are estimated from the point-to-point dispersion of the points out of transit and for all other cases are the combination of the photon noise and the background noise (see the AIJ package; Choi et al. 2016). Nevertheless, as the main source of uncertainties is residual systematics and not photon noise, when fitting the follow-up light curves EXOFASTv2 (see Section 3.2) adds a variance term to each transit to enforce reduced χ² ∼ 1.

A summary of our follow-up photometric observations is given in Table 1 where we list the date, telescope, filter, and cadence for each transit observation. In the last column we give the mean uncertainty in two-minute bins if the cadence is less than 2 minutes, otherwise for the original observations cadence. The resulting light curves, along with the best model fit and the corresponding residuals, are plotted in Figures 2–4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Follow-up spectroscopic observations to measure precision RVs for all three targets—Qatar-8b, 9b, and 10b—were obtained in the same manner as for all previous QES candidates (for details see Alsubai et al. 2011 and subsequent papers). In brief, we used the Tillinghast Reflector Echelle Spectrograph (TRES) on the 1.5 m Tillinghast Reflector at FLWO. All spectra were obtained using the medium fiber, which results in a resolving power of R ∼ 44,000 and a velocity resolution element of 6.8 km s⁻¹ FWHM. The wavelength calibration was established using a Th–Ar hollow-cathode lamp illuminating the science fiber with two exposures obtained immediately before and after each target spectrum.

For each one of the three target stars we obtained the following TRES spectra: (a) Qatar-8–15 spectra between 2017 March 6 and 2018 June 11 with exposure times in the range of 10–30 minutes and an average signal-to-noise ratio per resolution element (SNRe) ∼34 at the peak of the continuum in the echelle order centered on the Mg b triplet near 519 nm; (b) Qatar-9–9 spectra between 2018 February 23 and June 11 all with individual exposure times of 60 minutes and an average SNRe ∼14.9; (c) Qatar-10–16 spectra between 2018 May 20 and September 28, with exposure times in the range of 20–40 minutes and an average SNRe ∼30. For each of our target stars, we cross-correlated each observed spectrum against the reference spectrum (taken to be the strongest among the spectra of the given star) to obtain relative RVs. This was done for a set of echelle orders (in an order-by-order fashion) chosen so as to have both good SNRe and minimal telluric line contamination. These RVs are reported in Tables 2–4 (with the time stamps in Barycentric Julian Date in Barycentric Dynamical time, BJD_TDB) and plotted in Figures 5–7. The observation that was used for the template spectrum has, by definition, an RV of 0.0 km s⁻¹, and the error on the template RV is the median of the uncertainties of all the orders. In order to be sure that the periodic RV signal detected in all three cases is due to orbital motion and excludes other astrophysical phenomena that could potentially produce a similar periodic signal, we also derived the line profile bisector span values (BS; lower panels in Figures 5–7). Buchhave et al. (2010) describe the procedures used above for both the RV and the BS measurements in more detail. The absolute center-of-mass velocity of each system is determined in two steps as the precise RV of each spectrum is measured against the strongest observed spectrum of the same star. That reference spectrum could be anywhere on the RV curve, and so a relative systemic velocity (γ_rel) is determined when fitting the Keplerian model. The absolute systemic velocity is then the sum of γ_rel and the absolute RV offset of the reference spectrum which is determined by cross-correlating the Mg b order of the respective reference spectrum against the CfA library of synthetic templates. We also correct by −0.61 km s⁻¹, because the CfA library does not include the gravitational redshift. This offset has been determined empirically by many observations of IAU Radial Velocity Standard Stars system. We quote an uncertainty of ±0.1 km s⁻¹ in the resulting absolute velocity, which is an estimate of the residual systematic errors in the IAU Radial Velocity Standard Star system. Note that the error in determining the absolute center-of-mass RV of each system does not affect the determination of the planetary parameters. Although a dedicated paper on the assessment of the TRES absolute zero-point has not been published, we refer the interested reader to Quinn et al. (2014, 2015), which gives a detailed account of the instrument RV precision, stability, sources of error, and steps used to bring observed stellar velocities to the absolute scale of the IAU Radial Velocity Standard Star system.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For Qatar-8 we also obtained two RV measurements, on 2018 April 15 and 25, with the high-resolution Fiber-fed Echelle Spectrograph (FIES; Telting et al. 2014) on the 2.5 m Nordic Optical Telescope (NOT) at the Observatorio del Roque de los Muchachos (ORM) on the island of La Palma, Canary Islands, Spain. We used FIES in its high-resolution mode R ∼ 67,000 and a velocity resolution element of 4.8 km s⁻¹ FWHM. For the first observation we obtained a single 30-minute exposure spectrum, while for the second observation we obtained three consecutive 15-minute exposure spectra. Similar to the TRES observations, the wavelength calibration was established using exposures of a Th–Ar lamp illuminating the science fiber bracketing the target exposure spectra. Relative RV measurements were obtained through the cross-correlation technique described above and using the spectrum obtained on 2018 April 15 (T_exp = 30 minutes) as a template. Each one of the three spectra obtained on 2018 April 25 was measured separately against the template and the three measurements were averaged.

For all three targets, the observed transit light curves combined with the shape and amplitude of the RV variations, are well described by a planet orbiting a single star. Nevertheless, there are astrophysical scenarios not involving a planet that could mimic such a behavior. These include an eclipsing binary—either a background or in a hierarchical triple system—blended with the primary, much brighter star. Next we put forward arguments that allow us to exclude such scenarios.

It has been well established (e.g., Queloz et al. 2001; Torres et al. 2005) that if the observed RV pattern were a result of a blend with an eclipsing binary, the spectral line bisectors (see Tables 2–4) would follow a similar pattern, that is, they would vary in phase with the photometric period and with a similar amplitude. The measured line bisectors are shown in the bottom panels in Figures 5–7 on the same scale as the RV residuals from the Keplerian orbit fit. No obvious pattern is seen in all three cases. To quantify that we performed a significance test on the Pearson's correlation between the BS and RV for each star, and also the BS and the Keplerian fit model. In each of the three cases, the correlation between the BS and the RV (or between the BS and the Keplerian model) is insignificant, while the correlation between the RV and the Keplerian model is highly significant. Numerically, the probability of chance RS/RV correlation is higher than the typical significance level (5%): Qatar-8b—p-values, p = 0.061; Qatar-9b—p = 0.073; Qatar-10b—p = 0.067. This supports our argument that the observed RV pattern is a result of a gravitationally induced motion from a planet orbiting a single star.

A further argument in favor of the planet scenario can be drawn from the equal transit depths across all filters (accounting for limb-darkening effects). We do, however, note that an eclipse of a stellar companion with very similar colors would also lead to equal depths at different wavelengths. As such, we consider the equal depths as a supportive argument only, albeit in full agreement with our conclusion regarding the planetary nature of the transits.

Physical properties of each system were determined through a global model fit using the EXOFASTv2 package. A detailed description of EXOFASTv2 can be found in Eastman (2017) and Rodriguez et al. (2017). For each of the three exoplanetary systems the global fit includes the RV measurements listed in Tables 2–4, the follow-up transit light curves shown in Figures 2–4, respectively, the distance and broadband photometry for each star, and the priors on the stellar atmospheric parameters (T_eff and [Fe/H]) determined from the available spectra.

The effective temperature (T_eff), surface gravity (log g), metallicity ([m/H]), and projected rotational velocity (v_rot) for the host stars were determined using the stellar parameter classification tool (SPC; Buchhave et al. 2012). SPC works by cross-correlating an observed spectram against a grid of synthetic spectra based on the Kurucz atmospheric models. We used the ATLAS9 grid of models with the opacity distribution functions from Castelli & Kurucz (2004). In Table 5 we list the weighted mean values and the associated uncertainties of the stellar atmospheric parameters determined from the SPC analysis of each individual spectrum.

Table 5.  Basic Observational and Spectroscopic Parameters of Qatar-8b, 9b, 10b Host Stars and Photometry Used for the Spectral Energy Distribution Fit

Parameter	Description	Value			References
Names		Qatar-8b	Qatar-9b	Qatar-10b
	3UC	322-045544	302-100935	320-060170
	2MASS	J23540364+3701185	J23540364+3701185	J23540364+3701185
Astrometry
α2000	R.A. (J2000)	10h29m38962	10h42m59543	18h57m46537	(1)
δ2000	decl. (J2000)	+70°31'3750	+60°57'5083	+69°34'1501	(1)
Photometry
B	Johnson B, mag	12.132 ± 0.084	15.515 ± 0.088	13.437 ± 0.080	(2)
V	Johnson V, mag	11.526 ± 0.046	14.133 ± 0.048	12.879 ± 0.079	(2)
g	Sloan g, mag	11.806 ± 0.018	14.894 ± 0.065	13.147 ± 0.124	(2)
r	Sloan r, mag	11.355 ± 0.028	13.569 ± 0.061	12.712 ± 0.079	(2)
i	Sloan i, mag	11.174 ± 0.015	13.074 ± 0.078	12.569 ± 0.109	(2)
J	2MASS J, mag	10.299 ± 0.023	11.835 ± 0.019	11.754 ± 0.022	(3)
H	2MASS H, mag	10.005 ± 0.024	11.235 ± 0.017	11.474 ± 0.023	(3)
K	2MASS K, mag	9.937 ± 0.015	11.104 ± 0.023	11.382 ± 0.017	(3)
W1	WISE1, mag	9.890 ± 0.023	11.021 ± 0.024	11.375 ± 0.023	(4)
W2	WISE2, mag	9.932 ± 0.021	11.120 ± 0.021	11.413 ± 0.020	(4)
W3	WISE3, mag	9.902 ± 0.033	10.986 ± 0.090	11.348 ± 0.067	(4)
Spectroscopic Parameters
	Spectral type	G0V	K5V	F7V	this work
Teff	Effective temperature, K	5687 ± 50	4363 ± 51	6123 ± 50	this work
log g	Gravity, cgs	4.22 ± 0.10	4.65 ± 0.10	4.36 ± 0.10	this work
[m/H]	Metallicity	0.0 ± 0.08	0.25 ± 0.08	0.40 ± 0.08	this work
γabs	Systemic velocity, km s−1	5.57 ± 0.10	2.92 ± 0.10	−25.49 ± 0.10	this work
vrot	Rotational velocity, km s−1	2.7 ± 0.5	4.3 ± 0.5	5.9 ± 0.5	this work

References. (1) GAIA DR2 http://gea.esac.esa.int/archive/; (2) APASS9 http://www.aavso.org/apass; (3) 2MASS http://irsa.ipac.caltech.edu/Missions/2mass.html; (4) WISE http://irsa.ipac.caltech.edu/Missions/wise.html.

Download table as:  ASCIITypeset image

Broadband photometric surveys provide measurements across the electromagnetic spectrum for all three host stars, from the optical (APASS) to the mid-IR (WISE). These measurements are gathered in Table 5 and used to fit a model spectral energy distribution (SED) for each one of the stars as described below. The resulting SED fits are shown in Figure 8.

Zoom In Zoom Out Reset image size

In the global fit we apply Gaussian priors on the parallax from Gaia DR2, including the offset determined by Stassun & Torres (2016) and impose an upper limit on the V-band extinction from the Galactic dust reddening maps (Schlafly & Finkbeiner 2011). The limb-darkening coefficients were fit with a prior derived from an interpolation from the Claret & Bloemen (2011) tables for each band. We note that while all transit light curves were normalized as described in Section 2.2, in the global fit EXOFASTv2 treats the baseline flux (the flux outside the transit) as a free constant and adjusts its value to unity. Because all transit light curves were normalized in advance this adjustment is naturally very small.

All of the host stars properties were determined during the global fit. EXOFASTv2 simultaneously uses the SED, the stellar density and limb-darkening constraint from the transit, the MIST isochrones (Collier Cameron et al. 2006; Dotter 2016), priors from SPC (T_eff, [Fe/H]) and Gaia (parallax), and an upper limit on the reddening to simultaneously determine all the stellar properties. The stellar radius is predominantly constrained by the SED and Gaia parallax, while the stellar age and mass are predominantly constrained by the MIST isochrones and spectroscopic priors. In addition, consistency between the stellar mass and radius derived from these methods and the stellar density from the transit is strictly required.

In fitting Qatar-8b, 9b, and 10b we only considered circular orbits and kept the eccentricity fixed to zero. On the one hand, our RV data is not of high enough quality to allow investigation of potential small departures from circularity, and on the other hand we expect the planets orbits to have circularized. Following the equations from Jackson et al. (2008), and using the values from Table 6 of M_⋆, R_⋆, M_P, R_P, and a/R_⋆, we estimate orbit circularization timescales of τ_circ ∼ 0.04, 0.01, 0.01 Gyr, respectively, for the entire range of tidal quality factors Q_⋆ and Q_P considered by the authors (10⁴–10⁸ for each Q). This is much lower than the estimated age of the host stars and, thus, we expect the planet orbit to have circularized.

A precise orbital period for each system is also obtained during the global fit. The best ephemeris for each star is calculated by fitting all transits simultaneously to a linear ephemeris within EXOFASTv2:

$$\begin{eqnarray}&&{T}_{{\rm{C}}}=2458210.83980(85)+3.71495(100)E,\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{T}_{{\rm{C}}}=2458227.75643(27)+1.540731(38)E,\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{T}_{{\rm{C}}}=2458247.90746(36)+1.645321(10)E,\end{eqnarray} \tag{ 3 }$$

where E is the number of cycles after the reference epoch, which we take to be the transit time that minimizes the covariance between T_C and the period, and the numbers in parenthesis denote the uncertainty in the last two digits. Equations (1)–(3) correspond to Qatar-8b, 9b, and 10b, respectively.

Table 6 summarizes the physical parameters of each planetary system. We note that for Qatar-8b, for which we have RV measurements from two different telescopes and instruments, EXOFASTv2 fits the relative offset for each RV data set separately. These are the reported γ_rel values in the table. The Safronov number is not used in the current paper and is provided in Table 6 for completeness, as it may be useful for other studies.