HATS-5b: A TRANSITING HOT SATURN FROM THE HATSouth SURVEY^*

The transit signal around HATS-5 was first detected from photometric observations by the HATSouth survey (Bakos et al. 2013). HATSouth is a network of identical, fully robotic telescopes located at three sites spread around the Southern Hemisphere, allowing continuous coverage of the surveyed fields. Altogether 8066 observations of HATS-5 were obtained by the HATSouth units HS-1 in Chile, HS-3 in Namibia, and HS-5 in Australia from 2009 September to 2010 December. Each unit consists of four 0.18 m f/2.8 Takahasi astrographs and Apogee 4K × 4K U16M Alta CCD cameras. Each telescope has a field of view of 4° × 4°, with a pixel scale of 37 pixel⁻¹. The observations are performed with 4 minute exposures through the Sloan r' filter.

Discussions of the HATSouth photometric reduction and candidate identification process can be found in detail in Bakos et al. (2013) and Penev et al. (2013). Aperture photometry was performed and detrended using External Parameter Decorrelation (EPD; Bakos et al. 2007) and Trend Filtering Algorithm (TFA; Kovács et al. 2005). Transit signals were identified using the Box-fitting Least Squares analysis (Kovács et al. 2002). Table 1 summarizes the photometric observations for HATS-5. The HATSouth discovery light curve is plotted in Figure 1.

Zoom In Zoom Out Reset image size

Spectroscopic confirmation of HATS-5b consisted of separate reconnaissance observations to exclude most stellar binary false-positive scenarios that can mimic the transit signal of an exoplanet. High-resolution, high signal-to-noise (S/N) measurements of the radial velocity (RV) variation for HATS-5 were then obtained to confirm the planetary status of HATS-5b. The spectroscopic follow-up observations are presented in Table 2.

Low-resolution reconnaissance observations were performed using the Wide Field Spectrograph (WiFeS; Dopita et al. 2007) on the Australian National University (ANU) 2.3 m telescope at Siding Spring Observatory, Australia. A flux-calibrated spectrum was obtained at R ≡ λ/Δλ = 3000 to provide an initial spectral classification of HATS-5 as a G dwarf with T_eff = 5300 K, log g = 4.5, and [Fe/H] = 0. These stellar parameters are later refined by higher resolution observations (Section 3). Multi-epoch observations at R = 7000 confirmed that the candidate did not exhibit >1 km s⁻¹ RV variations. Such velocity variations are indicative of eclipsing stellar binaries, which have so far made up ${\sim }30{\rm \%}$ of HATSouth candidates. Details of the WiFeS follow-up procedure and stellar binary identification process can be found in Bayliss et al. (2013) and Zhou et al. (2014). Candidates that pass the WiFeS vetting process are passed on to higher resolution observations.

HATS-5 received nine high-resolution (R = 60, 000) reconnaissance RV observations with the CORALIE spectrograph on the Swiss Leonard Euler 1.2 m telescope at La Silla Observatory, Chile, and 14 R = 48, 000 observations with the FEROS spectrograph on the ESO/Max Planck Gesellschaft (MPG) 2.2 m telescope at La Silla. Detailed descriptions of the acquisition, reduction, and analyses of the CORALIE and FEROS observations can be found in the previous HATSouth discovery papers (Penev et al. 2013; Mohler-Fischer et al. 2013; Jordán et al. 2014). Velocities from these observations allowed us to constrain the RV orbit semi-amplitude to be <45 m s⁻¹.

The upper limit RV constraints from CORALIE, FEROS, and WiFeS indicated that HATS-5b is a low-density gas giant. High-S/N, high-resolution observations were required to determine the RV orbit of the system. Velocities of HATS-5 were obtained by the Planet Finding Spectrograph (PFS) on the 6.5 m Magellan Baade telescope at Las Campanas Observatory, Chile, and the High Dispersion Spectrograph (HDS) on the 8.2 m Subaru telescope at Manua Kea Observatory, Hawaii. The PFS and HDS velocities and bisector spans are presented in Table 3, and the RV orbit is plotted in Figure 2.

Zoom In Zoom Out Reset image size

The Subaru/HDS (Noguchi et al. 2002) observations were carried out on the nights of 2012 September 19–22 UT. Observations were made using an I₂ cell on four of the nights (Kambe et al. 2002), and without the I₂ cell on one of the nights. We used the KV370 filter, the 06 × 20 slit, and the StdI2b setup, yielding spectra with a resolution of R = 60, 000 and wavelength coverage of 3500–6200 Å. On each night we obtained three consecutive observations yielding a total S/N per resolution element of ∼100. The observations are split into three to reduce the impact of cosmic-ray contamination and changes in the barycentric velocity correction over the course of an exposure. The I₂-free observations were used to create a template spectrum needed to measure precise relative RV values from the observations made with the I₂ cell. The individual spectra were reduced to RV measurements using the procedure of Sato et al. (2002, 2012), which in turn is based on the method of Butler et al. (1996). Additionally, we measured spectral line bisectors following Bakos et al. (2007) for each observation. The rms scatter of the HDS velocities from the best-fit Keplerian curve is 4.8 m s⁻¹.

HATS-5 was also observed with the Carnegie Planet Finder Spectrograph (PFS; Crane et al. 2010) on Magellan II at Las Campanas Observatory, Chile, on the UT nights of 2012 December 28–31, 2013 February 21, and 2013 February 4. We obtained one iodine-free spectrum, and all other spectra were taken using the iodine cell and a slit width of 05. To increase the S/N of each spectrum, we read out with 2×2 binning and in slow readout mode. Consecutive pairs of 20 minute exposures were taken on each night. The RV for each spectrum was determined using the spectral synthesis technique detailed in Butler et al. (1996). The rms scatter of the PFS velocities from the best-fit Keplerian curve is 3.8 m s⁻¹.

High-precision photometric follow-ups of a partial and a full transit of HATS-5b were performed on 2012 October 10 and 2012 December 11, respectively, using GROND on the ESO/MPG 2.2 m telescope (Greiner et al. 2008). The GROND imager provides simultaneous photometric monitoring in four optical bands (g', r', i', z') over a 54 × 54 field of view at 0158 pixel⁻¹ sampling. Details of the GROND observation strategy, reduction, and photometry procedure can be found in Penev et al. (2013) and Mohler-Fischer et al. (2013). The GROND light curves are presented in Table 4 and plotted in Figure 3.

Zoom In Zoom Out Reset image size

Multi-band transit observations by GROND can provide constraints on the dependency between planet radius and wavelength (e.g., Nikolov et al. 2013; Mancini et al. 2013) and potentially probe for molecular absorption and Rayleigh scattering features in the transmission spectrum of a planet. We performed a separate fitting of the GROND full transit data from 2012 December 11, simultaneously fitting for the transit parameters T_c, a/R_⋆, and i, and the individual R_p/R_⋆ for each passband. The fitting is performed using the JKTEBOP eclipsing binary model (Nelson & Davis 1972; Southworth et al. 2004), with both quadratic limb-darkening coefficients fixed to that of Claret (2004) and freed and parameterized according to Kipping (2013). The best-fit parameters and uncertainties are explored by the emcee implementation of an MCMC routine (Foreman-Mackey et al. 2012) under Python. Simultaneous EPD is performed on the residuals for each iteration with a linear combination of the first-order terms for time, target star X position, Y position, FWHM, and airmass. The final R_p/R_⋆ values are consistent with each other to within errors for the fixed and free limb-darkening coefficient analyses.

The deviation from mean radius for each passband is plotted in Figure 5. For comparison, we also plot the wavelength–radius variation of HD 189733b, as measured using the Hubble Space Telescope (HST) by Pont et al. (2008) and Sing et al. (2011), and scaled to match the scale height (500 km) and R_p/R_⋆ of HATS-5b, assuming an H₂-dominated atmosphere (following Snellen et al. 2008). HD 189733b is a pL-class planet according to Fortney et al. (2008), in a similar temperature range and exhibiting similar spectra to L-type brown dwarfs. HATS-5b is mildly irradiated, similar to HD 189733b, and may be of the same spectral class. The results of the GROND observations are consistent with both a null detection of atmospheric features and that expected from the scaled measurements of HD 189733b. We do not see obvious starspot crossing events in the transit light curve, although unocculted spots can also cause a slope in the broadband R_p/R_⋆ measurements (e.g., Pont et al. 2008; Sing et al. 2011). While we do not detect any atmospheric features on HATS-5b, the large scale height makes HATS-5b an appealing target for future transmission spectroscopy observations. Future observations in the bluer U band may also reveal opacity variations in the atmosphere by H₂ Rayleigh scattering (e.g., Sing et al. 2011, 2013; Jordán et al. 2013; Nascimbeni et al. 2013).

Zoom In Zoom Out Reset image size

The mass and radius of HATS-5b are plotted in the context of existing close-in transiting gas giants in Figure 6. A number of previous studies have investigated the relationship between the planet radius distribution, host star metallicity, and levels of insolation (e.g., Guillot et al. 2006; Enoch et al. 2012, 2011; Béky et al. 2011). The factors that impact the radius of a gas giant should be mass dependent. For example, the level insolation should have a less significant impact on the radius of the denser, more massive gas giants and brown dwarfs than on the less dense Saturn-mass planets. Here we revisit the mass dependence of the planet radius on the host star metallicity and the planet equilibrium temperature.

Zoom In Zoom Out Reset image size

We bin the planet population into samples of 20 and perform a least-squares fit for a linear dependence between radius, mass, equilibrium temperature T_eq(K), and metallicity $\mathcal {Z}$:

$$\begin{eqnarray} \frac{R_{\rm fit}}{R_{\rm Jup}} &=& c_1 \mathcal {T} + c_2 \mathcal {Z} + c_3 \log \left(\frac{M_p}{M_{\rm Jup}} \right) + c_4\\ \mathcal {T} &\equiv & \frac{T_{\rm eq}-T_{\rm eq,mean}}{T_{\rm eq,max}-T_{\rm eq,min}} \nonumber \\ \mathcal {Z} &\equiv & \frac{\rm [Fe/H]-\rm [Fe/H]_{\rm mean}}{\rm [Fe/H]_{\rm max}-\rm [Fe/H]_{\rm min}}\protect, \protect \nonumber \end{eqnarray} \tag{ 1 }$$

where the magnitudes of c₁ and c₂ are used to judge the level of correlation for T_eq and [Fe/H], respectively. c₃ takes into account a linear dependence between mass and radius within the mass bin. c₄ is an arbitrary offset in the fit. We adopt normalized versions of effective temperature $(\mathcal {T})$ and [Fe/H] $(\mathcal {Z})$ such that the correlation coefficients are comparable between the two. The population means, maxima, and minima are T_{eq, mean} = 1490 K, T_{eq, max} = 2583 K, T_{eq, min} = 665 K, [Fe/H]_mean = 0.03, [Fe/H]_max = 0.45, [Fe/H]_min = −0.46. The errors in the coefficients are derived by bootstrapping the analysis within each mass bin. Since each mass bin covers a relatively small mass range, a linear dependence is sufficient (see Figure 7 for the sizes of each mass bin). We find a peak in the mass dependence of the T_eq correlation at M_p ∼ 1 M_J and a general lack of overall correlation between R_p and [Fe/H]. The correlation coefficients are plotted against their respective mass bins in Figure 7. We repeated the exercise using only planets with solar-mass hosts (0.8 < M_⋆ < 1.2 M_☉), to reduce any potential selection effects in the target selection and spectral classifications of the surveys.

Zoom In Zoom Out Reset image size

Smaller planets are found in longer periods (e.g., Mazeh et al. 2005; Davis & Wheatley 2009), biasing the T_eq–mass distribution. To reduce the effect of the bias, we re-perform the analysis using only mildly irradiated planets (T_eq < 1500 K). In addition, the radius–T_eq dependence is nonlinear over the general population (Demory & Seager 2011), and limiting the T_eq range has the added benefit of reducing the effect of the nonlinear dependence on the analysis. In all cases we find the peak dependence to T_eq to be ∼1 M_J, and a lack of dependence on [Fe/H]. We also note a minor correlation between c₁ and c₂, suggesting a correlation between the host star [Fe/H] and the irradiation received by the hot Jupiters, although the result is dependent on selection effects of the host star sample. For example, a correlation between the T_eff and [Fe/H] in the sampled host star population, potentially due to the age–metallicity relationship, can directly influence the relation between the planet T_eq and host star metallicity.

We also perform the same analysis for the entire population of hot gas giants, fitting for a second-order polynomial in mass–radius, and linear dependence to T_eq and [Fe/H]. We find a strong correlation in T_eq, with c₁ = 0.81 ± 0.17, and an insignificant correlation in [Fe/H], with c₂ = −0.25 ± 0.14. Increasing the order of the polynomial does not affect the coefficient values within errors. We find the overall dependence to [Fe/H] to be weak at best. Miller & Fortney (2011) suggests that the [Fe/H]–radius dependence is more prominent for the least irradiated planets; we limit the analysis to planets with T_eq < 1500 K, but still find a lack of correlation with [Fe/H], with c₁ = 0.37 ± 0.11 and c₂ = −0.15 ± 0.10.

We find that the radii of Saturn-mass planets are less affected by their equilibrium temperature than those of Jupiter-mass planets, in agreement with the singular value decomposition analysis performed by Enoch et al. (2012). In addition, we also find that the radii of planets with M_p > 1 M_J are less dependent on equilibrium temperature. This effect is reproduced by the isochrones from Fortney et al. (2007). The isochrones can also reproduce a drop in the correlation strength between irradiation and radius for the least massive gas giants (M_p < 0.3 M_J), but require the presence of a large core (M_c > 10 M_Earth). Interestingly, we find no statistically significant dependence of radius on the host star metallicity, contrary to previous examinations (e.g., Guillot et al. 2006; Béky et al. 2011; Enoch et al. 2011, 2012). It is not clear how the host star metallicity affects the metallicity and radius of the planet. A higher metallicity disk may produce planets with more massive cores, leading to a smaller overall radius (e.g., Guillot et al. 2006), but a higher opacity atmosphere is more efficient at retaining heat, reducing the rate of contraction, leading to a more inflated radius (e.g., Burrows et al. 2007, 2011).

We have presented the discovery of HATS-5b, a transiting hot Saturn with mass of 0.237 ± 0.012 M_J and radius of 0.912 ± 0.025 R_J. HATS-5b is the lowest mass and radius planet to date reported by the HATSouth survey. The host star is a quiet, slowly rotating G dwarf with a stellar mass of 0.936  ±  0.028 M_☉ and radius of 0.871  ±  0.023 R_☉. The large scale height of HATS-5b makes it favorable to further atmospheric characterizations via transmission spectroscopy.

The radius of HATS-5b is consistent with the model of an irradiated gas giant that formed via core accretion (Fortney et al. 2007). The radius is also consistent within 1σ to the empirical radius relationship for Saturn-mass planets from Enoch et al. (2012). We examined the correlation between the planet radius, equilibrium temperature, and host star metallicity within the known gas giant population and confirmed the strong correlation between planet radius and equilibrium temperature at ∼1 M_Jup. We report no robust correlation between metallicity and the planet radius.