A SPIN–ORBIT ALIGNMENT FOR THE HOT JUPITER HATS-3b^*

Recent measurements of the sky-projected spin–orbit angles⁴ for extra-solar planets (exoplanets) are revealing many surprises. These include planets on highly misaligned (e.g., Albrecht et al. 2012), polar (e.g., Addison et al. 2013), and even retrograde orbits (e.g., Bayliss et al. 2010). The dominant core-accretion model of planetary formation predicts that planets should form on nearly coplanar orbits with respect to their host star's equator (e.g., Pollack et al. 1996; Watson et al. 2011; Greaves et al. 2014) as is the case for the solar system (Lissauer 1993; Beck & Giles 2005). Moreover, plausible models for orbital migration (e.g., type I/II migration; Lin et al. 1996; Alibert et al. 2004; Chambers 2006) are expected to maintain this alignment.

The total number of confirmed planets has ballooned to over 1700 ⁵ including the 715 recently confirmed Kepler planets (Rowe et al. 2014). This rapid growth in the number of known exoplanetary systems will allow the finer details of planet formation to finally be better understood. Measurements of the obliquity of these systems will play a critical role in this endeavor, providing key empirical evidence that will elucidate the complex formation and orbital evolution mechanisms of extra-solar planets.

Ground-based transit surveys, such as the Wide Angle Search for Planets (WASP; Pollacco et al. 2006), the Hungarian Automated Telescope Network (Bakos et al. 2004), and its southern hemisphere counterpart HATSouth (Bakos et al. 2013), along with space-based transit surveys like CoRoT (Barge et al. 2008) and Kepler (Borucki et al. 2010; Batalha et al. 2013), are playing an important role in discovering and characterizing exoplanets. Measurements of spin–orbit alignments are increasingly becoming a critical part of the follow-up exoplanet characterization programs of the major exoplanet surveys.

The number of systems with spin–orbit alignment measurements has dramatically increased over the past few years (see, e.g., Albrecht et al. 2012). To date, 76 ⁶ planetary systems have measured obliquities, and of these 31 show substantial misalignments (>225). The vast majority of the sampled planets, however, are on short-period orbits with Jupiter-like masses (i.e., Hot Jupiters). This is because the amplitude of the radial velocity anomaly used to determine the spin–orbit alignments is dependent on the square of the planet to star radius ratio, (R_P/R_⋆)², and the rotational velocity of the host star (vsin i_⋆). Therefore, relatively few sub-Jovian, long-period, or multiple planet systems, have been studied and this parameter space remains largely unexplored. The large number of newly discovered Kepler planets has expanded the samples from which this parameter space can be explored,⁷ in particular, via obliquity measurements of Neptune-size planets (e.g., a nearly polar orbit for the Super-Neptune, HAT-P-11b; Winn et al. 2010b); long-period planets (e.g., stellar obliquity of the long-period planetary system HAT-P-17; Fulton et al. 2013); and multiple planet systems (see, e.g., Sanchis-Ojeda et al. 2012; Albrecht et al. 2013; Huber et al. 2013; Hirano et al. 2014).

The majority of spin–orbit alignments have been determined using spectroscopic measurements of the Rossiter–McLaughlin effect. The Rossiter–McLaughlin effect was first predicted for eclipsing binary stars over 100 yr ago by Holt (1893); however, it was not observed until 1924 for eclipsing binary stars (Rossiter 1924; McLaughlin 1924) and in 2000 it was observed for planets (Queloz et al. 2000). This effect is caused by the modification of the stellar spectrum as a transiting planet partially obscures the stellar disk of its host star, causing a radial velocity anomaly due to asymmetric distortions in the rotationally broadened stellar line profiles (Ohta et al. 2005).

Three additional techniques include: planetary starspot-crossings (e.g., the spin–orbit misalignment of HAT-P-11 was detected as a direct result of measured spot-crossing anomalies; Sanchis-Ojeda & Winn 2011), asteroseismology (e.g., inclination of the multi-planet hosting star HR 8799 from asteroseismology; Wright et al. 2011a, the large obliquity for the multiplanet system Kepler-56 from asteroseismology measurements; Huber et al. 2013), and gravity darkening (e.g., orbital obliquity of KOI368.01 from gravity darkening; Zhou & Huang 2013; Ahlers et al. 2014).

In this paper, we present spectroscopic measurements of the Rossiter–McLaughlin effect for HATS-3 showing that the system is in spin–orbit alignment. HATSouth is a global network of wide-field telescopes capable of continuous 24 hr monitoring of specific regions in the sky (Bakos et al. 2013). To date, HATSouth has discovered five transiting extra-solar planets (HATS-1b, Penev et al. 2013; HATS-2b, Mohler-Fischer et al. 2013; HATS-3b, Bayliss et al. 2013; HATS-4b, Jordán et al. 2014; HATS-5b, Zhou et al. 2014). The obliquity for HATS-3b, reported here, is the first that was measured for a system discovered by the HATSouth planet search.

HATS-3 is a mid-late F star with a mass of M_⋆ = 1.209 ± 0.036 M_☉, radius of R_⋆ = 1.404 ± 0.030 R_☉, effective temperature of T_eff = 6351 ± 76 K, age of $3.2^{+0.6}_{-0.4}$ Gyr, 453 ± 11 pc away from Earth, and has moderate rotation (vsin i_⋆ = 9.12 ± 1.31 km s⁻¹) as reported by Bayliss et al. (2013). The planet is a Hot Jupiter with a mass of M_P = 1.071 ± 0.136 M_J, slightly inflated with a radius of R_P = 1.381 ± 0.035 R_J, orbiting at distance of $a=0.0485^{+0.0004}_{-0.0006}$ AU, and an orbital period of P = 3.547851 ± 0.000005 (Bayliss et al. 2013). HATS-3b is the third planet discovered from HATSouth survey and a good candidate to have a measurable Rossiter–McLaughlin effect anomaly due to a high vsin i_⋆ and large R_P (Bayliss et al. 2013).

2.1. Simultaneous Photometric Observations on the FTS 2 m

Previous studies have shown the value of acquiring simultaneous photometric observations of exoplanetary systems for which Rossiter–McLaughlin effect radial velocity measurements are being obtained (e.g., Sanchis-Ojeda & Winn 2011; Oshagh et al. 2013; Sanchis-Ojeda et al. 2013). Such observations are particularly useful in disentangling the signature of starspot-crossings from true Rossiter–McLaughlin induced radial velocity variations (e.g., Kepler-63b; Sanchis-Ojeda et al. 2013). Starspot-crossings have been observed in both radial velocity and transit photometry data of other transiting planetary systems (e.g., radial velocity and photometric spot-crossing anomalies of Kepler-63b, Sanchis-Ojeda et al. 2013; and photometric spot-crossing anomalies of HAT-P-11, Sanchis-Ojeda & Winn 2011). They can introduce systematic radial velocity anomalies that are comparable in size to, and superimposed with, the Rossiter–McLaughlin effect, making an accurate determination of λ more difficult. Spot-crossings are, however, easily observable with photometry for large R_P and large spots. Simultaneous photometry can, therefore, be used to effectively remove spot-crossing anomalies from radial velocity measurements of the Rossiter–McLaughlin effect (e.g., Kepler-63b; Sanchis-Ojeda et al. 2013). An additional benefit of photometry of spot-crossing events is that they can provide constraints on the true orbital obliquity (instead of just the projected obliquity, as described in Nutzman et al. 2011) of planets that transit spotty stars (e.g., HATS-2b; Mohler-Fischer et al. 2013).

We obtained simultaneous photometric transit observations of the HATS-3 transit on the night of 2013 August 27 using the Faulkes Telescope South (FTS) to provide an updated constraint on the time of transit ingress. The FTS is located at the same site as the Anglo-Australian Telescope (AAT; Siding Spring Observatory) and is part of the Las Cumbres Global Telescope (LCOGT) Network. Observations were obtained using the "Spectral" imaging camera in 2 × 2 binned readout mode. The telescope was moderately defocused to avoid saturating on longer exposures and to minimize flat-fielding errors. The i-band filter and 30 s integration times (cadence of 50 s including the 20 s readout) were used for our transit observation. The data was reduced to calibrated fits files using the LCOGT data reduction pipeline. Aperture photometry was then performed using Source Extractor (Bertin & Arnouts 1996). Unblended, non-variable reference stars from the images were selected to de-trend the photometry, although this was the limiting factor in the precision of the photometry as none of the reference stars were as bright as HATS-3 in the 10'× 10' field of view. Additionally the star passed near zenith during the observations, resulting in a very large systematic error in the photometry due to rapid pupil rotation. We did not attempt to derive photometry during this period of time. The final result was that we obtained simultaneous photometric follow-up for HATS-3 for a total of 38 minutes on the 2013 August 27 starting ∼15 minutes before transit ingress. The light curve is presented in Figure 1 and Table 1.

Zoom In Zoom Out Reset image size

Table 1. Differential Photometry of the 2013 August 27 HATS-3b Transit

Time	RMag	${\sigma _{R_{{\rm Mag}}}}$	Time	RMag	${\sigma _{R_{{\rm Mag}}}}$
BJD-2400000			BJD-2400000
56531.95360	−0.00315	0.00080	56531.96725	−0.00199	0.00080
56531.95421	0.00107	0.00080	56531.96787	−0.00087	0.00080
56531.95483	−0.00181	0.00080	56531.96848	0.00023	0.00080
56531.95545	0.00043	0.00080	56531.96911	−0.00439	0.00080
56531.95608	0.00019	0.00080	56531.96972	−0.00428	0.00080
56531.95669	0.00088	0.00080	56531.97034	−0.00342	0.00080
56531.95732	0.00042	0.00080	56531.97095	−0.00411	0.00080
56531.95794	−0.00119	0.00080	56531.97158	−0.00049	0.00080
56531.95855	−0.00023	0.00080	56531.97219	−0.00905	0.00080
56531.95918	−0.00009	0.00080	56531.97281	−0.00693	0.00080
56531.95980	0.00119	0.00080	56531.97343	−0.00804	0.00080
56531.96043	0.00005	0.00080	56531.97405	−0.00229	0.00080
56531.96105	0.00348	0.00080	56531.97467	−0.00692	0.00080
56531.96167	0.00143	0.00080	56531.97529	−0.00674	0.00080
56531.96229	0.00053	0.00080	56531.97590	−0.00610	0.00080
56531.96291	−0.00003	0.00080	56531.97652	−0.00711	0.00080
56531.96352	0.00111	0.00080	56531.97715	−0.00637	0.00080
56531.96414	−0.00116	0.00080	56531.97777	−0.01088	0.00080
56531.96477	0.00192	0.00080	56531.97838	−0.00845	0.00080
56531.96538	0.00219	0.00080	56531.97901	−0.00394	0.00080
56531.96601	−0.00070	0.00080	56531.97963	−0.00860	0.00080
56531.96663	0.00413	0.00080

Download table as:  ASCIITypeset image

2.2. Spectroscopic Observations with CYCLOPS2

The spectroscopic observations of HATS-3b were carried out using the CYCLOPS2 instrument on the AAT. Providing full details of the CYCLOPS2 instrument design and specifications are beyond the scope of this work, and we direct the interested reader to Horton et al. (2012) for that information. The instrumental set-up and observing strategy for HATS-3b transit observations substantially followed that employed by Addison et al. (2013). The observations were calibrated using both a thorium–argon (ThAr) calibration lamp, to illuminate all on-sky fibers, and a ThXe lamp to illuminate the simultaneous calibration fiber.

We observed a transit of HATS-3b on the night of 2013 August 20, starting ∼20 minutes before ingress and finishing ∼2 hr after egress. The observations are summarized in Table 2. A total of 20 spectra were obtained on that night (12 during the ∼3.5 hr transit) in reasonable observing conditions (seeing around 15 with some high-level cirrus). The airmass at which HATS-3 was observed ranged from 2.3 at the start of the night, 1.2 near mid-transit, close to 1.0 at egress, and 1.1 at the end of the observations. We obtained a signal-to-noise ratio (S/N) = 18 per 2.5 pixel resolution element at λ = 5490 Å (in total over all 16 fibers) at an airmass of 1.0 and in 15 seeing. HATS-3 on this night was at a distance of ∼15° from the full Moon. However, cross-correlation of our data with a solar-like spectral mask demonstrated no obvious signatures of solar spectrum contamination near the main cross-correlation peak and we conclude that the observations were not significantly impacted by lunar contamination.

Table 2. Summary of HATS-3b Transit Spectroscopic Observations

	2013 August 20	2013 August 27
UT time of Obs	08:40–14:32 UT	09:02–16:50 UT
Cadence	1120 s	1375–1675 s
Readout times	120 s	175 s
Readout speed	Fast	Normal
Readout noise	5.35 e−	3.19 e−
S/N (/2.5 pixels at λ = 5490 Å)	3–5	14–19
Resolution (λ/Δλ)	70,000	70,000
Number of spectra	20	19
Seeing	15	07–11
Weather conditions	Some cirrus	Clear
Airmass range	1.0–2.3	1.0–1.8

Download table as:  ASCIITypeset image

A second transit was observed on the night of 2013 August 27, starting ∼2 hr before ingress and finishing ∼2 hr after egress (see Table 2). A total of 19 spectra were acquired—including nine during the ∼3.5 hr transit. On this night, observations were obtained in a slower "NORMAL" readout mode. A quick-look analysis of the August 20 data had indicated that the additional read-noise delivered by the FAST readout mode was not worth the improved cadence delivered by the shorter read time. The observing conditions were excellent for Siding Spring Observatory with seeing between 07–11 and clear skies for the whole night. The airmass at which HATS-3 was observed that night varied between 1.4 for the first exposure, ∼1.1 near mid-transit, and 1.8 for the last exposure. We obtained a S/N = 28 per 2.5 pixel resolution element at λ = 5490 Å (in total over all 16 fibers) on this night when the star was observed at an airmass of 1.0 with 07 seeing.

We used the Exoplanetary Orbital Simulation and Analysis Model (ExOSAM; Addison et al. 2013) to determine the best-fit parameters from both transit photometry and Rossiter–McLaughlin effect measurements.

3.1. Transit Modeling

The ExOSAM light curve analysis model is a significantly improved version of an earlier model called Exoplanetary Pixelization Transit Model (see Addison et al. 2010). ExOSAM utilizes the small planet approximation of Mandel & Agol (2002), which assumes that the surface brightness of the star directly underneath the disk of the planet is constant (an excellent approximation for the vast majority of transiting systems, including HATS-3b), to determine the limb-darkened stellar flux blocked by the planet. Limb-darkening is described in our model using either a linear or quadratic law. For HATS-3b, a quadratic limb-darkening law (see Equations (1) and (2)) was used:

$$\begin{eqnarray} I_{0} &=& \frac{6}{\pi R_{\star }^{2} \left(6 - 2q_{1} - q_{2}\right)},\\ \ms F_{b} &=& I_{0}A_{P}\left[1-q_{1}\left\lbrace 1-\sqrt{\left|1-\frac{d_{c}^{2}}{R_{\star }^{2}}\right|}\right\rbrace \right] \nonumber\\ && -I_{0}A_{P}\left[q_{2} \left\lbrace 1 - \sqrt{\left|1-\frac{d_{c}^{2}}{R_{\star }^{2}}\right|}\right\rbrace ^{2}\right], \end{eqnarray} \tag{ 1 }$$

where I₀ is the central intensity of the stellar surface such that F_b is normalized to one (see Appendix A of Hirano et al. 2010 for simple derivation of I₀), R_⋆ is the stellar radius (in m), and q₁ and q₂ are the limb-darkening coefficients. F_b is the total fractional (normalized) intensity blocked by the planet, A_P is the area of the planet in front of the stellar disk, and d_c is the apparent distance between the center of the planetary disk to the center of the stellar disk along the transit chord as viewed from the Earth. d_c in Equation (2) determines the uniform limb-darkened surface brightness of the star directly underneath the disk of the transiting planet.

ExOSAM uses a total of 10 input parameters to calculate the best-fit transit light curves. Of these, ExOSAM can solve for eight, namely, the planet-to-star radius ratio (R_p/R_⋆); the orbital inclination angle (i); the orbital period (P); the orbital eccentricity (e); argument of periastron (ϖ); the two coefficients (q₁ and q₂) in the quadratic limb-darkening equation; and the mid-transit time (T₀). The final two parameters, the stellar mass (M_⋆ estimated from spectroscopy) and the planet mass (M_p estimated from radial velocity data), are held fixed. The eight free parameters are derived using a well-sampled grid search and minimizing χ² between the observed transit photometry and modeled light curve. The 1σ confidence levels in the free parameters are determined through the Δχ² method (Press 1992) which is based on the normal probability distribution of χ² as a function of the confidence level and degrees of freedom.

3.2. Rossiter–McLaughlin Effect Modeling

The ExOSAM Rossiter–McLaughlin analysis model described in Addison et al. (2013) instead uses 16 input parameters, of which we fix 14 in order to allow us to accurately determine the projected orbital obliquities λ and projected stellar rotational velocities vsin i_⋆ by fitting radial velocity data taken during a transit event (when the Rossiter–McLaughlin effect is observable). The 14 fixed parameters are as follows: the planet-to-star radius ratio (R_p/R_⋆); the orbital inclination angle (i); the orbital period (P); the mid-transit time (T₀) at the epoch of observation; the radial velocity offset (V_d) between our data sets and published data sets; a velocity offset term (V_s) accounting for systematic effects between our data sets taken over multiple nights; planet-to-star mass ratio (M_p/M_⋆); orbital eccentricity (e); argument of periastron (ϖ); two adopted quadratic limb-darkening coefficients (q₁ and q₂); the micro-turbulence velocity (ξ_t); the macroturbulence velocity (v_mac); and the center-of-mass velocity ($V_{T_{P}}$) at published epoch T_P.

ExOSAM models both the radial velocity from the motion of the host star due to the orbiting planet and the velocity anomaly due to the Rossiter–McLaughlin effect, using the analytical approach of Hirano et al. (2010) as described in Addison et al. (2013). We have included a Monte Carlo simulation in our model to obtain more robust estimates of the uncertainties on the spin–orbit angle (λ) and the rotational velocity (vsin i_⋆) of the host star from the given uncertainties on other fixed input parameters. Confidence intervals for λ and vsin i_⋆ (Δλ, Δvsin i_⋆) are derived from Monte Carlo simulations as adopted from Press (1992) and given by Equations (3) and (4):

$$\begin{equation} \Delta \lambda =\sqrt{\frac{1}{N_{{\rm MC}}-1}*\sum _{i=1}^{N_{{\rm MC}}}\left[\left(\lambda _{o}-\lambda _{i}\right)^{2}+\Delta \lambda _{i}^{2}\right]},\quad \end{equation} \tag{ 3 }$$

$$\begin{eqnarray} \Delta v\sin i &=&\sqrt{\frac{1}{N_{{\rm MC}}-1}} \nonumber\\ && \times \sqrt{\sum _{i=1}^{N_{{\rm MC}}}\left[(v\sin i_{(o)}-v\sin i_{(i)})^{2}+\Delta v\sin i_{(i)}^{2}\right]}, \quad \end{eqnarray} \tag{ 4 }$$

where N_MC is the number of Monte Carlo simulations, λ_o and vsin i_(o) are the best overall λ and vsin i_⋆ from all Monte Carlo runs (as determined from the minimum χ²), λ_i and vsin i_(i) are the best λ and vsin i_⋆ from the i^th Monte Carlo run, and Δλ_i and Δvsin i_(i) are the 1σ confidence levels of λ and vsin i_⋆ determined through the Δχ² method for the i^th Monte Carlo run.

3.3. Transit Analysis

3.4. Rossiter–McLaughlin Analysis

The spectroscopic data were reduced using custom MATLAB routines, which trace each fiber and optimally extract each spectral order as outlined previously in Addison et al. (2013). Each of the 16 fibers, in each of the 18 useful orders, is used to estimate a radial velocity (and associated uncertainty) by cross-correlation with a synthetic spectrum of similar spectral type using the IRAF⁸ task, fxcor. We created the synthetic template spectrum of a mid-F star ($T_{{\rm eff}_{\star }}=6500$ K and log g_⋆ = 4.0) using SYNSPEC,⁹ a general spectrum synthesis program. Fxcor uses the standard cross-correlation technique developed by Tonry & Davis (1979). We observed several radial velocity standard stars, including HD 206395, HD 10700, and HD 6735. We carried out tests using both these radial velocity standard observations and synthetic spectra as cross-correlation templates, and found that the lowest inter-fider¹⁰ velocity scatter was obtained using the spectrum of the synthetic template, and we therefore adopted this for use in the subsequent analysis. The weighted average velocities for each observation were computed using the method described in Addison et al. (2013) and the uncertainties for each weighted velocity were estimated from the weighted standard deviation of the fider velocity scatter. Our weighted radial velocities for the two transit observations, including their uncertainties and total S/N, are shown in Table 3.

Table 3. Radial Velocities for HATS-3 (Fiber and Order Averaged) Taken on 2013 August 20 and 27

Time	RV	S/N at	In/Out	Time	RV	S/N at	In/Out
BJD-2400000	( ms−1)	λ = 5490 Å	Transit	BJD-2400000	( ms−1)	λ = 5490 Å	Transit
56524.85790a,b	−40317 ± 103	18.6	Out	56525.20517b	−40763 ± 82	25.9	Out
56524.87216a	−40741 ± 73	16.6	Out	56531.88589	−40761 ± 39	26.6	Out
56524.88526a	−40659 ± 74	16.3	In	56531.90552	−40794 ± 36	29.3	Out
56524.89835a	−40855 ± 67	16.7	In	56531.92201	−40731 ± 49	23.4	Out
56524.91145a	−40694 ± 60	16.9	In	56531.93561	−40785 ± 43	25.7	Out
56524.92455a	−40628 ± 70	17.2	In	56531.94921	−40802 ± 41	24.6	Out
56524.93765a	−40757 ± 73	17.3	In	56531.96477	−40759 ± 46	24.9	Out
56524.95077a	−40792 ± 66	18.6	In	56531.98069	−40720 ± 44	23.3	In
56524.96386a	−40894 ± 65	17.5	In	56531.99660	−40733 ± 32	25.9	In
56524.97697a	−40907 ± 71	17.6	In	56532.01425	−40752 ± 31	28.6	In
56524.99006a	−40895 ± 70	17.6	In	56532.03364	−40806 ± 31	28.8	In
56525.00316a	−40842 ± 63	17.6	In	56532.05304	−40821 ± 38	31.5	In
56525.01626a	−40837 ± 61	18.5	In	56532.07069	−40864 ± 44	22.8	In
56525.02936a	−40854 ± 66	18.4	In	56532.08834	−40875 ± 37	28.0	In
56525.04246a	−40787 ± 62	17.6	Out	56532.10773	−40836 ± 38	26.9	In
56525.05556a	−40619 ± 71	18.3	Out	56532.12712	−40797 ± 31	27.0	In
56525.06866a	−40829 ± 62	17.9	Out	56532.14651	−40843 ± 44	28.4	Out
56525.08176a	−40846 ± 45	20.1	Out	56532.16590	−40793 ± 39	30.9	Out
56525.09485a	−40835 ± 72	18.8	Out	56532.18529	−40836 ± 35	29.9	Out
56525.10796a	−40925 ± 63	18.6	Out	56532.20295	−40805 ± 46	26.4	Out

Notes. ^aFast readout mode. ^bNot used in analysis (high airmass).

Download table as:  ASCIITypeset image

λ and vsin i_⋆ were determined from the best-fit model of the Rossiter–McLaughlin effect using ExOSAM. The mid-transit time T₀ was fixed to the best derived value from our simultaneous photometry. The velocity offset, V_s, between our data sets taken on August 20 and 27 was determined by finding the different V_d offsets between the Bayliss et al. (2013) data set and our data sets for each night separately (August 20 $V_{d_{1}}$ and August 27 $V_{d_{2}}$) and applying the difference to our combined data set of both nights ($V_{s}=V_{d_{1}}-V_{d_{2}}$). The other seven input parameters (M_p/M_⋆, e, ϖ, q₁, q₂, ξ_t, and $V_{T_{P}}$) were adapted from Bayliss et al. (2013).

The confidence intervals for our λ and vsin i_⋆ were derived from running 5000 Monte Carlo iterations. For each iteration, a synthetic data set was generated by drawing from a normal distribution for each radial velocity datum and its 1σ uncertainty. We also drew from randomly generated Gaussian distributions for the model parameters R_p/R_⋆, i, P, T₀, and V_d about their mean and standard deviation as given in Bayliss et al. (2013). We determined that the remaining model parameters M_p/M_⋆, e, ϖ, q₁, q₂, ξ_t, and $V_{T_{P}}$ negligibly contribute to the overall uncertainty in λ and vsin i_⋆ and so held these fixed in our simulations. The input parameters and their uncertainties as adopted are given in Table 4.

Table 4. System Parameters for HATS-3

Parameter	Value
Parameters as given by Bayliss et al. (2013) and used as priors in model
Mid-transit epoch (2400000-HJD)a, T0	56532.03799 ± 0.00028
Orbital periodb, P	3.547851 ± 0.000005 d
Semi-major axisb, a	$0.0485^{+0.0004}_{-0.0006}$ AU
Orbital inclinationb, i	862 ± 03
Impact parameterb, b	$0.497^{+0.024}_{-0.027}$
Transit depthb, (RP/R⋆)2	0.01022 ± 0.00060
Orbital eccentricityc, e	0.0 (assumed)
Argument of periastronc, ϖ	N/A (e = 0)
Stellar reflex velocityc, K⋆	125.7 ± 15.7  ms−1
Stellar massc, M⋆	1.209 ± 0.036 M☉
Stellar radiusb, R⋆	1.404 ± 0.030 R☉
Planet massc, MP	1.071 ± 0.136 MJ
Planet radiusb, RP	1.381 ± 0.035 RJ
Stellar microturbulencec, ξt	N/A
Stellar macroturbulencec, vmac	N/A
Stellar limb-darkening coefficientc, q1	0.4135 (adopted)
Stellar limb-darkening coefficientc, q2	0.3301 (adopted)
Velocity at published epoch TPc, $V_{T_{P}}$	0.0  ms−1
RV offset between Bayliss and our complete data setb, Vd	40792 ± 16  ms−1
RV offset between Aug 20 and 27 data setsd, Vs	144 ± 20  ms−1
Parameters determined from model fit using our velocities from the complete data set
Projected obliquity angle, λT	3° ± 25°
Projected stellar rotation velocity, vsin i⋆(T)	5.75 ± 2.98  km s−1
Parameters determined from model fit using our velocities from the complete data set (errors
empirically adjusted)
Projected obliquity angle, $\lambda _{T_{e}}$	4° ± 16°
Projected stellar rotation velocity, ${v\sin i_{\star }}_{(T_{e})}$	5.75 ± 1.70  km s−1
Parameters determined from model fit using Aug 20 velocities
Projected obliquity angle, λAug20	−8° ± 55°
Projected stellar rotation velocity, vsin i⋆(Aug20)	7.50 ± 4.18  km s−1
Parameters determined from model fit using Aug 27 velocities
Projected obliquity angle, λAug27	8° ± 33°
Projected stellar rotation velocity, vsin i⋆(Aug27)	5.25 ± 3.01  km s−1
Independent measurements of vsin i⋆(Ind) and Bayliss et al. (2013) vsin i⋆(B) published value
Projected stellar rotation velocity, $v\sin i_{\star ({\rm Ind}_{1})}$	5.2 ± 0.6  km s−1
Projected stellar rotation velocity, $v\sin i_{\star ({\rm Ind}_{2})}$	5.3 ± 0.7  km s−1
Projected stellar rotation velocity, vsin i⋆(B)	9.12 ± 1.31  km s−1

Notes. ^aParameter fixed from the transit photometry at the indicated value for final fit, but allowed to vary (as described in Section 3) for uncertainty estimation. ^bParameters fixed to the indicated value for final fit, but allowed to vary (as described in Section 3) for uncertainty estimation. ^cParameters fixed at values given by Bayliss et al. (2013). ^dParameter fixed at value determined from the difference between $V_{d_{1}}$ and $V_{d_{2}}$ from Aug 20 and 27 data sets, respectively.

Download table as:  ASCIITypeset image

Best-fitting values for λ and vsin i_⋆ for each Monte Carlo run were determined from a grid search that minimized χ² on a uniformly distributed, randomly generated set of 120 λ values between −60° ⩽ λ ⩽ 60° and 75 vsin i_⋆ values between 1.0 ⩽ vsin i_⋆ ⩽ 12.0 km s⁻¹. The ranges for λ and vsin i_⋆ were chosen based on a quick inspection of the observed Rossiter–McLaughlin effect velocity anomaly. We then determined the best overall λ and vsin i_⋆ and these were used to compute their 1σ uncertainties (Δλ) and (Δvsin i_⋆) as given in Equations (3) and (4).

Table 4 shows the final best-fit parameters and their uncertainties for eccentricity fixed at zero. Bayliss et al. (2013) computed two sets of planetary parameters for HATS-3b: one based on a fixed circular orbit and another allowing the eccentricity to float. When they allowed the eccentricity to vary, they determined that the best-fit eccentricity was e = 0.25 ± 0.10. Unfortunately, the available radial velocity data does not decisively indicate whether the orbit is circular or eccentric. Given that the planet is in a ∼3.5 day orbit and the system is over 3 Gyr old, we consider an $e\not\approx 0$ as unlikely. Such a planet should have been tidally circularized long ago, thus our analysis is based on the assumption that e = 0.

Figure 2 shows the modeled Rossiter–McLaughlin anomaly with the observed velocities over-plotted. The density distribution for λ and vsin i_⋆ resulting from the Monte Carlo simulations is shown in Figure 3, along with the location of the χ² minimum as well as the 1σ and 2σ confidence contours. Normalized density functions collapsed into λ and vsin i_⋆ are also shown, along with fitted Gaussians.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We clearly detect the Rossiter–McLaughlin effect from the combined data sets of August 20 and 27 as a positive anomaly between ∼100 minutes prior to mid-transit (0 MJD) to ∼0 MJD and then negative anomaly between ∼0 MJD to ∼100 minutes after mid-transit. That is, the planet initially transits across the blue-shifted hemisphere during ingress, crosses the mid-transit point near the stellar rotation axis, and then transits across the red-shifted hemisphere during egress. This produces a nearly symmetrical velocity anomaly as seen in Figure 2. The lack of any asymmetry suggests that the planet is in an orbit well aligned to the rotational axis of its host star (i.e., that is in "spin–orbit alignment").

For the August 20 and 27 data sets, we obtained the projected obliquity as λ = 3° ± 25° and stellar rotation velocity as vsin i_⋆ = 5.75 ± 2.98 km s⁻¹. We also conducted the analysis on the data sets from the two nights separately and obtained λ = −8° ± 55° and vsin i_⋆ = 7.50 ± 4.18 km s⁻¹ for August 20; and λ = 8° ± 33° and vsin i_⋆ = 5.25 ± 3.01 km s⁻¹ for August 27. The two nights produce consistent results, though the August 20 data delivers significantly higher uncertainties for λ and vsin i_⋆ due to the lower S/N spectra obtained on that night.

Examination of the normalized residuals to the model (R_Norm as defined in Equation (5)) and reduced χ² ($\chi ^{2}_{{\rm red}}$) suggests that our estimated velocity uncertainties may have been overestimated, as we find R_Norm = 0.62 and $\chi ^{2}_{{\rm red}}=0.90$. We therefore experimented with an empirical adjustment of those uncertainties in a manner similar to Butler et al. (2004). An updated solution with these adjusted uncertainties produces values for λ and vsin i_⋆ consistent with those obtained previously, but with smaller uncertainties (λ = 4° ± 16° and vsin i_⋆ = 5.75 ± 1.70 km s⁻¹). However, in the absence of a plausible cause for our uncertainties being overestimated, we favor our original values, but we do quote both solution uncertainties in Table 4:

$$\begin{equation} R_{{\rm Norm}}=\frac{\sum _{i=1}^{N_{{\rm data}}}\left[\frac{\left|O-C\right|_{i}}{\sigma _{i}}\right]}{N_{{\rm data}}}. \end{equation} \tag{ 5 }$$

We therefore checked our vsin i_⋆ estimation in two additional ways. First, we fitted a rotationally broadened Gaussian to the cross-correlation peak produced by each of the HATS-3 spectra taken that night (summed over all echelle orders) to obtain vsin i_⋆ = 5.2 ± 0.6 km s⁻¹. Second, we fitted a rotationally broadened Gaussian to a least-squares deconvolution line profile for each spectral order (in a similar manner to that used in Addison et al. 2013) of the three best spectra of HATS-3 taken on August 27, giving vsin i_⋆ = 5.3 ± 0.7 km s⁻¹. Both of these estimates are consistent with the value determined from our Rossiter–McLaughlin fitting. Most critically, if the vsin i_⋆ in HATS-3 were as large as that presented in Bayliss et al. (2013) of 9.12 ± 1.31 km s⁻¹, we would have obtained a velocity anomaly 50% larger than we actually observed. We are therefore confident in adopting a vsin i_⋆ = 5.75 ± 2.98 km s⁻¹ for this system. One possible explanation for the discrepancy between the published vsin i_⋆ of Bayliss et al. (2013) and our measured vsin i_⋆ could be from not accounting for stellar macroturbulence, which can contribute significantly to the overall absorption line broadening observed in stars. There exists a degeneracy between rotational broadening and macroturbulence broadening, as determined by the width of the stellar absorption lines, which makes disentangling each of their contributions to the overall broadening difficult to measure (Jordán et al. 2014).

HATS-3 is the first exoplanetary system discovered from the HATSouth Transit Survey to have the Rossiter–McLaughlin effect measured. It is a relatively hot, T_eff = 6351 ± 76 K, mid-F primary star (Bayliss et al. 2013) hosting a planet in a well-aligned orbit. This system joins the rapidly growing list of planetary systems for which spin–orbit alignments have been measured. A substantial fraction (∼41%) of the 76 systems with measured obliquities show spin–orbit misalignments (λ > (π/8) = 225)¹¹ and the majority of planets on high-obliquity orbits are found around stars hotter than T ⩾ 6250 K (as noted by Winn et al. 2010a, Albrecht et al. 2012, and others). There are a few noteworthy exceptions to this general trend such as HAT-P-18b with λ = 132° ± 15° and T_eff = 4870 ± 50 K (Esposito et al. 2014) and Kepler-63b with $\psi = {104^{+9}_{-14}}^{\circ }$ (true orbital obliquity ψ as opposed to the sky-projected obliquity λ) and T_eff = 5576 ± 50 K (Sanchis-Ojeda et al. 2013).

Several mechanisms have been proposed to explain the high occurrence rate of exoplanetary systems observed to be in spin–orbit misalignment. These mechanisms include Kozai–Lidov cycles (Kozai 1962; Lidov 1962; Fabrycky & Tremaine 2007), stellar internal gravity waves (Rogers et al. 2013), chaotic star formation (e.g., Thies et al. 2011), primordial disk misalignments from interactions with a stellar binary (Batygin 2012; Lai 2014), planet–planet scatterings (Chatterjee et al. 2008), and secular chaos (Wu & Lithwick 2011). Misalignments produced through disk migration alone are disfavored because the disk from which planets form is expected to be well aligned with the stellar spin axis of their host star. This assertion is well supported by recent observations of debris disks around nearby stars (e.g., Kennedy et al. 2013; Greaves et al. 2014). Since debris disks are material left behind from the formation of planetary systems (e.g., Wyatt 2008), the growing number of well-aligned systems detected adds weight to the theory that most planetary systems form from protoplanetary disks that are aligned with their host star's equator. If a planet migrates solely through interaction with the disk, it is therefore expected to have its orbital plane remain aligned (Bate et al. 2010).

As the number of planetary systems with measured spin–orbit angles has grown, a few correlations have become apparent between the properties of the host star and the orbit of its planet. One of the first trends noted is that, as the temperature of the host star increases, so to does the likelihood of it hosting a planet on a significantly misaligned orbit. In particular, Winn et al. (2010a) and Albrecht et al. (2012) observed that the measured obliquities fall into two distinct populations. Around the coolest stars (T < 6250 K), the great majority of planets are found to be well aligned. In contrast, around the hotter stars (T > 6250 K), the distribution of obliquities are far more random—as can be clearly seen in Figure 4. This dichotomy may be explained by the fact that stars T_eff ⩾ 6250 K have thin convective layers and are unable to realigned planets on high-obliquity orbits through planet–star tidal interactions (Albrecht et al. 2012). These tidal interactions are thought to dampen orbital obliquities over time and primarily occur in the outer convective envelope of stars. Stars with effective temperature of T_eff < 6250 K have a much thicker convective envelope and can dampen obliquity more effectively on shorter timescales.

Zoom In Zoom Out Reset image size

In Figure 5, we show an updated plot of the projected orbital obliquity as a function of the relative tidal-dissipation timescales¹² of exoplanetary systems as calibrated from binary studies and adopted from Albrecht et al. (2013) and Addison et al. (2014). We determined the tidal dissipation timescale, using the methods presented in Albrecht et al. (2012), for HATS-3b as τ_RA = 1.07 × 10¹⁵ yr (using the radiative timescale for alignment) or in relative terms τ_RA = 2.13 × 10⁶ yr/(5 × 10⁹ yr). Taking into consideration the mass of the convective envelope (the second approach of Albrecht et al. 2012), we obtain τ_mcz = 1.11 × 10¹¹ yr and τ_mcz = 35 yr/(3.2 × 10⁹ yr) if we normalize τ_mcz to the age of HATS-3. One can see from Figure 5 that ∼two-thirds of exoplanetary systems have dissipation timescales shorter than that of HATS-3b. If HATS-3b did become misaligned during or shortly after migration, there would not have been enough time to realign the orbit. These results replicate those obtained in our earlier analysis of the misaligned system, WASP-79b (Addison et al. 2013). For that system, we calculated the tidal dissipation timescales as τ_RA = 3.3 × 10¹⁵ yr (τ_RA = 6.6 × 10⁶ yr/(5 × 10⁹ yr)) and τ_mcz = 1.60 × 10¹¹ yr (τ_mcz = 320 yr/(0.5 × 10⁹ yr)). Just as in the case for HAT-3b, these timescales are also sufficiently long that there has been insufficient time for a misaligned planet to re-align since the formation of the system.

Zoom In Zoom Out Reset image size

In Figure 4, we present an updated version of the projected orbital obliquity verses stellar temperature plot¹³ as shown in Esposito et al. (2014), to which we have added our recently measured obliquities for WASP-79b and HATS-3b. At first glance, it appears that there is only a weak correlation between obliquity and effective temperature as there are several systems that are outliers. While this does indeed illustrate that cool stars can host planets in misaligned orbits, it also tells us that in order for them to host such planets, the tidal-dissipation timescale for realignment must be very long (as shown in Figure 5). Therefore a more relevant factor in determining whether a star can host a planet on a high-obliquity orbit is the tidal-dissipation timescale and not just the stellar effective temperature. A long realignment timescale is possible for planets orbiting cool stars if the orbital distance-to-stellar radius ratio (a/R_⋆) is sufficiently large and/or if the planet-to-star mass ratio (M_P/M_⋆) is sufficiently small as the timescale is proportional to a/R_⋆ and inversely proportional to M_P/M_⋆. This supports the hypothesis, as proposed by Winn et al. (2010a) and Albrecht et al. (2012), that whatever mechanism(s) are producing Hot Jupiters, are also randomly changing their spin–orbit angles and the systems with long realignment timescales will still have their initial (post-migration) spin–orbit angles while systems with short realignment timescales will have their spin–orbit angles realigned.

Giant planets are expected to form several AU away from their host star in the surrounding protoplanetary disk. This disk is expected to be well aligned with the star's spin axis due to conservation of angular momentum from the collapse of the protostellar cloud (e.g., Pollack et al. 1996; Perryman 2011; Armitage 2013). During or shortly after formation, giant planets can migrate inward (through various proposed migration mechanisms) to become Hot Jupiters where they have been observed to reside at separations as close as 0.01 AU from their host star. In addition, nearly half of all solar-type stars in the Milky Way have one or more stellar companions and planets have been found to be as likely to form around single stars as they are multiple stellar systems (e.g., Raghavan et al. 2010; Lillo-Box et al. 2012; Orosz et al. 2012; Schwamb et al. 2013; Wang et al. 2014). This suggests that the stellar companions may have a significant role in shaping the evolution and migration of planets.

Several recent studies have suggested that Kozai resonances may explain the high frequency of spin–orbit misaligned exoplanets (e.g., Fabrycky & Tremaine 2007; Wu et al. 2007; Nagasawa et al. 2008; Queloz et al. 2010; Naoz et al. 2011; Narita et al. 2012; Plavchan & Bilinski 2013). Stellar companions to planet host stars that are widely separated (up to several hundred AU) and highly inclined with respect to the orbital plane of a planet can induce Kozai oscillations on the planet. These oscillations occur through the gravitational interactions between the planet and stellar companion and can drive the planet to high orbital obliquities. Eccentricity and orbital inclination are anti-correlated as described by the Kozai integral ($I_{k}=\sqrt{1-e^{2}}\cos i$ from Murray & Dermott 1999). The planetary orbital inclination is initially driven to match that of the orbital plane of the stellar companion. Then the Kozai oscillations will either cause the planetary eccentricity to increase while its orbital inclination (relative to the stellar companion) decreases or its orbital inclination (relative to the stellar companion) to increase while its eccentricity decreases. If the eccentricity is driven high enough, the planet will pass within a few stellar radii from its host star during periastron passage. The end result will be tidal dissipation and circularization of the orbit while the orbital inclination will be misaligned with the host star (Nagasawa et al. 2008). This process may naturally explain the 3 day orbital period pile-up observed for Hot Jupiters produced through Kozai or secular chaos migration (Wu & Lithwick 2011). Dawson et al. (2012) suggest, however, that high-eccentricity migration from Kozai resonances due to a stellar companion cannot be responsible for the production of all the Hot Jupiters. This is due to the lack of super-eccentric protohot Jupiters (planets currently undergoing high-eccentricity migration) discovered by Kepler. They suggest other migration processes, such as from disk migration, planet–planet scattering, planetary Kozai, or secular chaos, might instead be the dominant channel for the origin of Hot Jupiters.

In addition, it is worth noting that such Kozai-driven eccentricity excitation will also increase the likelihood of the excited planet acting to destabilize the orbits of any other planets orbiting nearby. It is well established that, aside from certain resonant orbits (e.g., Robertson et al. 2012; Wittenmyer et al. 2012a), increases in the eccentricity of an exoplanet's orbit will act to decrease the stability of multi-planet systems (e.g., Wittenmyer et al. 2012b, 2013). This might explain why no additional planets have been found orbiting the hosts of high-obliquity planets—those planets were ejected as a part of the evolution of the high-obliquity planet to its current orbit. The one exception to this rule is Kepler-56, which hosts two planets in coplanar orbits that are misaligned with respect to the host's equator (Huber et al. 2013). A massive companion in a wide orbit has been detected in this system and it is believed to be generating torques on the inner planets, driving them into coplanar orbits that are misaligned with the spin axis of the host star (Huber et al. 2013).

We have recently proposed to search for stellar companions around systems that host Hot Jupiters with measured obliquities to test the hypothesis that Kozai–Lidov cycles are the primary driver for spin–orbit misalignments (as discussed in Addison et al. 2014). Our search is being conducted by directly imaging a sample of nearby stars within 250 pc using the Magellan Adaptive Optics and Clio2 infrared camera instruments on the 6.5 m Magellan Telescope at the Las Campanas Observatory in Chile. We will be able to conclusively confirm or reject the presence of stellar companions to within 150 AU in our sample and test if the Kozai mechanism is responsible for producing the majority of misaligned Hot Jupiters. A complementary survey to ours is being conducted by Knutson et al. (2014) and they are searching for massive (>1 M_J), long-period (>1 yr) companions to close-in giant planets using radial velocity and adaptive optics imaging measurements. The first results from the Knutson et al. (2014) program have found evidence for 15 planetary and/or brown dwarf companions in 14 systems (out of a total of 51 sampled systems) which suggests that the dynamical evolution of Hot Jupiters could be driven by distant, massive companions.

We have measured the spin–orbit alignment of the newly discovered Hot Jupiter HATS-3b, and find the planet's orbit to be well aligned to the projected rotational axis of its host star (λ = 3° ± 25°). We obtained three separate values for the vsin i_⋆ of the stellar spin, namely: vsin i_⋆ = 5.75 ± 2.98 km s⁻¹(from the Rossiter–McLaughlin effect measurements); vsin i_⋆ = 5.3 ± 0.7 km s⁻¹(from a Gaussian fit to the least-squares deconvolution line profile); and vsin i_⋆ = 5.2 ± 0.6 km s⁻¹ (from a Gaussian fit to the cross-correlation function). They are all in good agreement with each other but in disagreement with the value of vsin i_⋆ = 9.12 ± 1.31 km s⁻¹ from Bayliss et al. (2013), possibly due to extra broadening from macroturbulence. Nonetheless, we were able to robustly measure the spin–orbit angle of HATS-3b. Simultaneous photometry was obtained for a portion of the August 27 transit and was used to constrain the mid-transit time. Such photometry for future Rossiter–McLaughlin effect observations is vital to ensure proper monitoring of starspot-crossing events that can constrain the true obliquity of exoplanetary systems being studied.

One may expect HATS-3b to be in a misaligned orbit, given that its host star has T_eff ⩾ 6250 K and the realignment timescale for this system is very long (τ_mcz = 1.11 × 10¹¹ yr). Orbital obliquities likely are, however, initially distributed randomly from the migration processes that produce Hot Jupiters regardless of the value of T_eff (Winn et al. 2010a; Albrecht et al. 2012). Therefore, we expect the observed obliquities to be randomly distributed for systems with long τ_mcz and low obliquities for systems with short τ_mcz. This is true for almost all observed planetary systems for which spin–orbit angles have been measured, including HATS-3b, thus supporting the Albrecht et al. (2012) hypothesis. Alternatively, the low-obliquity and short-period orbit of HATS-3b could just as well be explained by type I/II disk driven migration (e.g., Lin et al. 1996). If this is indeed the case, then the orbit of HATS-3b was likely well aligned to its host star's equator since its formation (e.g., Bate et al. 2010).

We are now beginning to unravel why high orbital obliquities are generally observed around stars with T_eff ⩾ 6250 K and long realignment timescales and not around stars with T_eff < 6250 K or short realignment timescales as more and more systems support the Albrecht et al. (2012) hypothesis. A lingering question remains, however, namely, what are the mechanism(s) most responsible for producing misaligned Hot Jupiters in the first place? This important question will likely be resolved in the near future from an expansion of both the sample and the parameter space of spin–orbit alignment measurements. The parameter space least explored includes multiple planet and long-period transiting planet systems. Testing the various mechanism(s) thought to produce high obliquities, such as searching for stellar companions around stars hosting Hot Jupiters with obliquity measurements (e.g., Addison et al. 2014; Knutson et al. 2014) and searching for evidence of additional planets orbiting the hosts of high-obliquity planets (e.g., Knutson et al. 2014), will also be important avenues to pursue in resolving this mystery.

The research work presented in the paper at UNSW has been supported by ARC Australian Professorial Fellowship grant DP0774000, ARC LIEF grant LE0989347, ARC Super Science Fellowships FS100100046, and ARC Discovery grant DP130102695. Work at the Australian National University is supported by ARC Laureate Fellowship Grant FL0992131. We thank Jonathan Horner, Jonty Marshall, and Shaila Akhter for helpful comments and suggestions on this manuscript. We acknowledge the use of the SIMBAD database, operated at CDS, Strasbourg, France. This research has made use of NASA's Astrophysics Data System, the René Heller's Holt–Rossiter–McLaughlin Encyclopaedia (http://www.physics.mcmaster.ca/~rheller/index.htm), the Exoplanet Orbit Database and the Exoplanet Data Explorer at exoplanets.org, and the Extrasolar Planets Encyclopaedia at http://exoplanet.eu.