Transiting Extrasolar Planets: Detection And

Home , 450 Brigitta, ELODIE spectrograph, HD 20782 b, HD 4308 b, HD 82943 b, HD 82943 c

TRANSITING EXTRASOLAR PLANETS:

DETECTION AND FOLLOW-UP

Thesis submitted for the degree “doctor of philosophy”

Avi Shporer

Submitted to the senate of Tel Aviv University

October 2009

This work was carried out under the supervision of

Professor Tsevi Mazeh

To my parents

Contents

Abstract...... III

1 Introduction 1 1.1 OverviewofExtrasolarPlanets ...... 4 1.1.1 ExtrasolarPlanetsDetection...... 4 Barnard’sstar ...... 5 TheRadialVelocityMethod ...... 6 PulsarPlanets ...... 9 High-PrecisionRadial Velocity ...... 10 TheTransitMethod ...... 14 OtherTechniques ...... 23 Current Status and Future Prospects. . . . 25 1.1.2 The Role of Observations in Developing Plan- etaryTheory ...... 29 TheSolarSystem ...... 30 Planetformationindisks ...... 30 Thecoreaccretiontheory ...... 31 The gravitational instability theory. . . . . 33 PlanetaryMigration ...... 34 Planetaryradius-mass relation ...... 38 Planetaryatmospheres ...... 41 Spin-orbitalignment...... 43 Looking for a second planet in the system . 44 Themass-periodcorrelation...... 45 1.2 StructureoftheThesis ...... 46 1.2.1 Searching for Transiting Extrasolar Planets . . 47

I 1.2.2 Follow-up Observations of Known Transiting Planets...... 49 WASP-1b...... 50 HAT-P-2b ...... 50 GJ436b ...... 51

2 The Papers 53 PaperI ...... 53 PaperII...... 60 PaperIII ...... 64 PaperIV...... 72 PaperV...... 77 PaperVI...... 82

3 Discussion 91 3.1 Papers I – III — Detection of Three New Transiting ExtrasolarPlanets ...... 91 3.1.1 PaperI—HAT-P-2b...... 91 3.1.2 PaperII—HAT-P-5b ...... 96 3.1.3 PaperIII—HAT-P-9b...... 97 3.2 Papers IV – VI — Follow-up Studies of Known Tran- sitingExtrasolarPlanets ...... 99 3.2.1 Paper IV — WASP-1b photometric follow-up 100 3.2.2 Paper V — HAT-P-2b spectroscopic follow-up 100 3.2.3 Paper VI — GJ 436b photometric follow-up campaign ...... 103 3.3 Summary ...... 104

Bibliography 107

Acknowledgments 123

II Abstract

This thesis, consisting of six published papers, includes the discovery of three new transiting extrasolar planets and follow-up studies of three known transiting planets.

Papers I, II and III present the discovery of HAT-P-2b, HAT-P-5b and HAT-P-9b, respectively. The three planets were discovered as candidates by the Hungarian Automated Telescope Network (HAT-

Net), working in collaboration with the Israeli HAT telescope at the

Wise Observatory (Wise-HAT, or WHAT), for which I am part of the operational group.

The planetary nature of HAT-P-2b was conﬁrmed spectroscopically at the Keck telescope, with the HIRES spectrograph. This planet is extraordinary as it is both massive (Mp =8.80 ± 0.16 MJ ) and residing in an eccentric orbit (e =0.5156 ± 0.0018). For HAT-P-

5b and HAT-P-9b, spectroscopic conﬁrmation was done at the OHP

1.93-m telescope, with the SOPHIE spectrograph, in observations I carried out. I also carried out photometric observations of transits of these planets with the Wise Observatory 0.46-m and 1.0-m telescopes. HAT-P-5b is a close-in planet (P =2.788491±0.000025 days) similar to Jupiter in mass and radius (Mp = 1.06 ± 0.11 MJ , Rp =

III 1.257 ± 0.053 RJ ), and HAT-P-9b is a low-density planet (ρp =

−3 0.35 ± 0.06 g cm = 0.28 ± 0.05ρJ ), belonging to a small group of inﬂated planets.

Papers IV, V and VI present three follow-up studies of known transiting planets, WASP-1b, HAT-P-2b and GJ 436b, respectively.

Each of these planets has some unique characteristics and the three studies were carried out right after the announcement of the planet discovery.

WASP-1b was observed with the Wise Observatory 1.0-m telescope and using the two transit light curves obtained we were able to refine the system parameters, especially the planetary and stellar radii, confirming the planet’s inflated nature (Rp =1.40 ± 0.06 RJ ).

The massive and eccentric planet HAT-P-2b was followed-up spectroscopically at the OHP 1.93-m telescope, with the SOPHIE spectrograph, leading to an improved orbital solution and the measurement of the spin-orbit alignment (λ =0.2 ± 12.5◦). GJ 436b was the ﬁrst planet detected in the Neptune-mass range, and its eccentric orbit made it even more interesting. Immediately after the discovery of its transiting nature (in a work I was part of), it was followed-up photometrically from three observatories worldwide, allowing us to

IV obtain reﬁned light curve parameters, look for transit timing variation (TTV) and variation in the transit impact parameter. Although no signiﬁcant variation was found, we could not rule out a small TTV, of the order of a minute, and a long-term modulation of the impact parameter, of the order of +0.2 yr−1.

The contribution of this work, as a whole, to the exciting ﬁeld of transiting planets, is the study of transiting exoplanets positioned in several key areas of the planetary radius-mass parameter space.

Chapter 1

Introduction

The discovery of planets orbiting other stars, similar to the Sun, at the end of the second millennium (Mayor & Queloz 1995) marks an important breakthrough in modern astrophysics. Further detections made since, including short period planets (e.g., Butler et al. 1997,

1998; Mazeh et al. 2000; Konacki et al. 2003) and planets on eccentric orbits (e.g., Marcy & Butler 1996; Cochran et al. 1997; Naef et al.

2001), revealed that the planet phenomenon is more diverse than expected from observations of the Solar System. The current sample, which is growing fast, includes 350 exoplanets1. Of those, about 60 cross the line-of-sight to their host star every orbital revolution. This selected group of transiting planets presents an opportunity for an

1Throughout this thesis data regarding the known sample of exoplanets is taken from the on-line Extrasolar Planets Encyclopedia (http://exoplanet.eu/). It is updated to July 1st, 2009, and includes published planets only.

1 in-depth study of the planet phenomenon, much more than non- transiting planets (e.g., Burrows et al. 2007; Guillot 2008; Johnson

2009). As described in more details in the following, only transiting planets allow the measurement of both the radius and exact mass, the study of their atmospheres and measuring the alignment between the stellar spin and the planetary orbit. All this makes them highly important astrophysical objects.

In addition, some of the transiting planets discovered already show unusual characteristics, including (I) large planetary radius (e.g.,

WASP-1b, Cameron et al. 2007), larger than theoretical expecta- tions at the time of discovery (e.g., Burrows et al. 2007; Chabrier

& Baraﬀe 2007), (II) high planetary mass (e.g., HAT-P-2b, Bakos et al. 2007a), which is vital for studying the mass-radius relation at this mass range (e.g., Baraﬀe et al. 2003, Chabrier et al. 2009), where objects are rare, and (III) short-period eccentric orbits (e.g.,

GJ 436b, Butler et al. 2004; Gillon et al. 2007), which are unexpected as the tidal circularization process is more eﬃcient for close-in orbits

(e.g., Mazeh 2008). These interesting discoveries call for follow-up observations, leading to better understanding. For example, looking for a second planet in a transiting planet system (e.g., Ballard et al.

2 2009), where the gravitational interaction between the two planets maintains the non-zero eccentricity (e.g., Adams & Laughlin 2006;

Maness et al. 2007), or, a more accurate measurement of the orbital eccentricity for an inﬂated planet, to see if its orbit is circular or not, since tidal friction may be responsible for the large radius of a planet on an eccentric orbit (e.g., Bodenheimer et al. 2001; Miller et al. 2009; Ibgui & Burrows 2009).

Transiting planets were the focus of my Ph.D. study. It included searching for new transiting planets, identiﬁed as candidates by WHAT

(Shporer et al. 2006, 2007a, 2009b) and HATNet (Bakos et al. 2004), and follow-up observations of known transiting planets I carried out at the OHP 1.93-m telescope, with the SOPHIE spectrograph, and at the Wise Observatory 0.46-m (Brosch et al. 2008) and 1.0-m telescopes.

This thesis consists of six published papers and is divided into two parts. The ﬁrst part, including Papers I–III, presents the detection of three new transiting planets. In the second part, consisting of

Papers IV-VI, three follow-up studies of known transiting planets are described.

In Sec. 1.1 I give an historical overview of the ﬁeld, from the early

3 discoveries to the modern ones. I also brieﬂy describe the primary methods used and summarize the current theories. In Sec. 1.2 I describe the contents of the thesis.

1.1 Overview of Extrasolar Planets

1.1.1 Extrasolar Planets Detection

The existence of planets, or worlds, orbiting other stars was discussed already in Ancient Greece. In the 4th century B.C.E. Aristotle postulated his geocentric cosmology, which views the Earth as unique

(Fraser 2006). As in other ﬁelds of science, the Aristotelean approach was well accepted for almost two millennia. In the 16th century C.E.

Nicolaus Copernicus suggested that the Earth orbits the Sun, thereby starting the Copernican revolution (e.g., Blumenberg 1987). Soon after, in 1584, Giordano Bruno took the Copernican approach a huge leap forward when he postulated in his book de l’infinito universo et mondi (on the infinite universe and worlds) that “There are innu- merable suns and an infinite number of planets which circle around those suns” (Singer 1950).

Up until the 20th century the existence of planets orbiting other stars remained a theoretical question. That was mainly because the

4 detection of extrasolar planets was beyond the technological reach until then. The basic problem is identifying a very faint object — a planet — right next to a star, which can be up to 108 times more luminous. This calls for the use of indirect methods, as described in the following subsections.

Barnard’s star

The ﬁrst scientiﬁcally-based claim for a detection of a planetary-mass object orbiting a star other than the Sun was made by van de Kamp

(1963, 1969a). Using astrometric measurements of the nearby M- type star known as “Barnard’s star” (named after the discoverer of its high proper motion, of 10′′.3 per year) he identified a perturbation, interpreted as resulting from the orbital motion of a 1.6 MJ planet, where MJ is the mass of Jupiter. Moreover, upon further analysis he claimed the identification of a second planet in the system (van de Kamp 1969b, 1975). However, other astrometric studies could not confirm van de Kamp’s detection (e.g., Gatewood & Eichhorn

1973). A modern astrometric study, using HST, did not detect a companion with a period longer than 150 days, down to the mass of

Jupiter (Benedict et al. 1999). Therefore, Van de Kamp’s claim was rejected.

5 The Radial Velocity Method

For a planet-star system, only the parent star spectrum is observed as the planet is much fainter, so this system can be considered a single- lined spectroscopic binary (SB1). However, as a gas giant planet, similar to Jupiter in mass, is three orders of magnitude lighter than a Sun-like star, the RV variation induced by the planet on its parent star is small. The following equation gives the RV semi-amplitude,

K, of a star of mass M1, orbited by a lighter body, of mass M2, when

M1 ≫ M2:

1 2 − 3 − 3 P M1 − 1 M2 sin i − K ≃ 203 (1 − e2) 2 m s 1, (1.1) day M⊙ MJ where P is the orbital period, e the orbital eccentricity and i the inclination angle, meaning the angle between the orbital angular momentum axis and the line-of-sight. A Jupiter-mass planet orbiting a

Sun-like star in a week long, edge-on, circular orbit, induces an amplitude of K ≈ 100 ms−1. For comparison, RV amplitudes of stellar binaries, in similar orbits, are larger by a few orders of magnitude.

Figure 1.1 gives a visual illustration of Eq. 1.1, presenting the RV amplitude induced by a planet on its parent star, as a function of

6 the minimum mass, orbital period and semi-major axis.

Semi−major Axis [AU]

2 0.01 0.1 1 10 10

1 10 HD 114762b −1 0 1000 m s 10 ] J

−1 51 Peg b −1 100 m s 10 sin(i) [M 2 −1 M −2 10 m s 10

−1 −3 1 m s 10

−4 10 0 2 4 10 10 10 Period [days]

Figure 1.1: Diagram illustrating the capabilities of the RV method, according to Eq. 1.1. The minimum mass, M2 sin i, in MJ , is plotted against orbital period, in days (bottom X-axis), and orbital semi-major axis, in AU (top X-axis). The dashed lines represent constant RV amplitude (assuming a circular orbit and a Solar-mass primary). The positions of the Solar System planets are marked with ﬁlled circles, and diamonds mark the positions of HD 114762b and 51 Peg b, described below.

The values of K, P and e are measured by ﬁtting the orbit, while

M1 is estimated using stellar atmospheric and evolution models.

Therefore, using Eq. 1.1, the value of M2 sin i is derived, which is a lower limit on the companion’s mass, depending on the inclination angle. Since this angle can not be determined spectroscopically, the exact mass can not be measured by the RV method alone. However, assuming the orbital plane is randomly oriented, sin i expectation

π 4 value is 4 , so the exact mass is expected to be close to π (= 1.27) times M2 sin i.

7 Looking for exoplanets with the RV method was suggested already by Struve (1952). The ﬁrst discovery of a planet candidate using this method was done in 1989 by Latham et al., who discovered HD 114762b. Further observations were done by Mazeh et al.

(1996), improving the system parameters. The planet orbits its parent star every 84 days in an eccentric orbit (e =0.35), inducing a RV amplitude of 0.6 kms−1. The star was measured by the CfA Digital

Speedometers (Latham 1992), where the typical RV uncertainty is only slightly smaller than the observed RV amplitude. Hence, the

RV periodic modulation was identiﬁed only with a power spectrum analysis, presented in Figure 1.2, illustrating how close this discovery was to the detection threshold.

Figure 1.2: Left: Power Spectrum of HD 114762 RVs, showing a clear peak at the orbital period. Right: RV curve folded on the orbital period. The solid line is the orbital ﬁt and the dashed line is the RV zero point, i.e., the center-of-mass velocity. The RV amplitude is only slightly larger than the measurements’ scatter, showing how this discovery was close to the detection limit and was possible due to the large number of measurements (ﬁgure taken from Latham et al. 1989).

8 The orbital fit, combined with the estimate of the stellar mass implied M2 sin i of 9 ± 2 MJ (Mazeh et al. 1996). Therefore, the discovery of HD 114762b was the first detection of a secondary with a minimum mass within the planetary mass range (i.e., below the traditional 13 MJ threshold for deuterium burning, although the definition of a planet is probably more complicated, see Basri &

Brown 2006), for an object orbiting a Sun-like star.

Pulsar Planets

The spinning rate of pulsars is known to be highly accurate, thus allowing to measure minute deviations in their radio pulses arrival time. A planetary body orbiting the spinning neutron star will induce periodic deviations, as the two bodies orbit the system’s center of mass. In 1992, Wolszczan & Frail announced the discovery of a planetary system around the pulsar PSR B1257+12. Further measurements done by Wolszczan (1994) conﬁrmed the detection, revealing a 3-planet system with masses of 0.02, 3.9 and 4.3 ME (Earth mass) and periods of 25, 66, and 98 days, respectively. Interestingly, this planetary system is similar to a down-size of the inner Solar

System (Mazeh & Goldman 1995).

If the PSR B1257+12 planetary system existed before the star

9 became a supernova, it is unclear how the system could have survived the violent explosion. Therefore, the more accepted theory is that the system evolved after the pulsar was formed, i.e., post-supernova evolution, probably from a debris disk of a binary companion (e.g.,

Miller & Hamilton 2001; Podsiadlowski 1993; Phinney & Hansen

1993).

The discovery of planets around PSR B1257+12 was the ﬁrst detection of planetary-mass bodies outside the Solar System. However, it was still not a convincing detection of planets around a main sequence, Sun-like star.

High-Precision Radial Velocity

In 1995, working with the ELODIE spectrograph (which was mounted on the OHP 1.93-m telescope), Mayor & Queloz discovered the exoplanet 51 Peg b, orbiting a G5V star. From the measured RV amplitude of 59 ms−1, shown in Fig. 1.3, they inferred a minimum mass of

0.44 MJ . Therefore, this was the ﬁrst discovery of a planetary-mass companion to a Sun-like star.

As illustrated in Figure 1.1, detecting Jupiter-mass objects within

1 AU requires RV accuracy of the order of 10 m s−1.

Several groups worldwide have been trying to reach this high-

10 Figure 1.3: RV curve of 51 Peg, folded on the orbital period, of 4.23 days, showing a clear sinusoidal modulation with an amplitude of 59 m s−1 (figure taken from Mayor & Queloz 1995). precision. Two groups were the first to conquer this challenge, in the 1990’s, and they have discovered most of the extrasolar planets known today: The Geneva and the California groups. The Geneva group was the first to detect an exoplanet using this method, with the discovery of 51 Peg b in 1995, by Mayor & Queloz. The main difference between the teams is in the spectroscopic technique they use nowadays, although they both report reaching a precision of the order of 1 ms−1 (Butler et al. 1996, 2006; Lovis & Pepe 2007; Pepe

& Lovis 2008). The Geneva group uses the simultaneous thorium technique, while the California group uses the iodine cell technique, as described below.

The Geneva group, led by M. Mayor, uses the SOPHIE spectrograph (successor of ELODIE), mounted on the 1.93-m telescope

11 at the Observatoire de Haute-Provence (OHP), the HARPS spectrograph, on the 3.6-m ESO telescope at La Silla, and CORALIE, mounted on the 1.2-m Euler telescope, also at La Silla. These are high-resolution (up to R ≈ 100,000 with HARPS; Pepe et al. 2002) echelle spectrographs, feb by two optical fibers. Light coming from both fibers is simultaneously recorded on the detector. One fiber is centered on the target, and the other is exposed to a thorium-argon

(ThAr) lamp, acting as a wavelength reference source. The RV is obtained by cross correlating the observed spectra with a box-shaped, binary template (Baranne et al. 1996).

The California group, led by J. Marcy, uses primarily the HIRES spectrograph (Vogt et al. 1994), mounted on the Keck I telescope, and the high-resolution spectrograph mounted on the Lick observatory Shane 3-m telescope (Marcy & Butler 1992; based on the

HAMILTON spectrograph, described in Vogt 1987). Both are high- resolution (up to R ≈ 70,000) echelle spectrographs, with a spectral coverage from the optical to the near-IR, about 3,500 – 10,000 A.˚

Obtaining the precise RV is done by placing a gaseous iodine (I2) cell in the optical path, between the slit and the detector, so the observed stellar spectra is superimposed with the well known I2 lines.

12 This way, both the wavelength calibration and the instrumental point spread function (PSF) are derived. Doppler analysis is done by modeling the composite spectrum (star and iodine) after breaking it into several hundred ∼2 A˚ segments. The stellar spectrum is derived by taking a high S/N spectroscopic observation of the star without the iodine cell (Butler et al. 1996; Eggenberger & Udry 2009). This high

S/N measurement, also referred to as a stellar template, deﬁnes an arbitrary velocity zero point.

From an historical point of view, it is interesting to note that a search for extrasolar planets using a similar technique was already attempted in the 1980’s. A group led by the Canadian astronomer

B. Campbell used an echelle spectrograph equipped with a hydrogen-

ﬂuoride (HF ) cell, mounted on the CFHT (Campbell & Walker 1979;

Campbell et al. 1988). This group monitored a few dozen Sun-like stars for ∼10 years (Campbell et al. 1991), and although their search was capable of detecting extrasolar planets, no planets were found.

Following the discovery of 51 Peg b, many more exoplanets were detected using the RV technique. In fact, most of the exoplanets known today were discovered this way. However, as noted earlier, this method lacks an intrinsic characterization of the planet, which

13 is essential for the study of planetary structure and composition.

The Transit Method

When the planetary orbit is edge on, meaning i = 90◦ or close, it in- tersects the observer’s line-of-sight to the star. Therefore, once every orbital revolution the planet moves across the stellar disk. During this crossing, called a transit, the planet blocks part of the light coming from the stellar surface, inducing a periodic dimming in the observed ﬂux. The relative decrease in stellar ﬂux is:

∆F R 2 = p , (1.2) F Rs where Rp and Rs are the planetary and stellar radii, respectively.

Eq. 1.2 ignores the limb darkening eﬀect and assumes the planetary disk is completely engulfed in the stellar disk, i.e., a complete transit, as apposed to partial, or grazing transit. For a Jupiter-size planet transiting a Solar-size star, the transit depth is about 1 %.

Transiting planets are not common. Even if a star is orbited by a planet the orbital inclination needs to be close to 90◦ for a transit to occur. More quantitatively, given a planet-star system with a randomly oriented orbital plane, the probability it will show transits

14 is (Sackett 1999):

R + R R Prob.(transit) = s p ≈ s , (1.3) a a where a is the orbital semi-major axis. In case the orbit is non- circular, then a is replaced by the planet-star distance when the planet moves through the plane deﬁned by the line-of-sight to the host star and the planetary orbital angular momentum axis. For a planet orbiting a Solar-size star at a close-in orbit, similar to that of

51 Peg b, the transit probability is about 10 %. Eqs. 1.2 & 1.3 show how the transit method is biased towards large planets, in close-in orbits.

Although they are rare, their scientiﬁc return is high. The transit conﬁguration allows a wealth of measurements, primarily the planetary radius and exact mass. In addition, follow-up studies, done after the planet is known, can study the planet in detail (see also

Section 1.1.2).

The photometric signature of a transiting planet was discussed already several decades ago (Struve 1952; Resonblatt 1971; Borucki

& Summers 1984). However, most of the surveys operating today were launched only after the ﬁrst detections of exoplanets by the RV

15 method, as they proved the existence of Jupiter-like close-in planets

(e.g., Mayor & Queloz 1995).

At ﬁrst, planets detected spectroscopically were immediately searched for transits, and indeed, the ﬁrst transiting exoplanet was detected this way. In 1999 the 3.5 day period planet HD 209458b was detected spectroscopically by Mazeh et al. (2000). Immediately after the initial detection transits were observed by Charbonneau et al. (2000) using the 10-cm STARE telescope (Brown & Charbonneau 2000), and by Henry et al. (2000; partial transit detection) using the 0.8- m APT at Fairborn Observatory, Arizona. Interestingly, this planet was the 10th close-in planet discovered spectroscopically. Due to the importance of this discovery, four transit events were observed in the year 2000 by HST, obtaining a highly accurate transit light curve

(Brown et al. 2001). Figure 1.4 presents the discovery light curve of

Charbonneau et al. (2000) in the left panel, and in the right panel the space-based follow-up light curve of Brown et al. (2001). The drop in ﬂux during transit, of 1.6 %, along with the RV amplitude

−1 of 86 ms (Mazeh et al. 2000), infers a planetary radius of 1.3 RJ

(Jupiter radius) and mass of 0.69 MJ .

Following the detections at the end of the 20th century, sev-

16 Figure 1.4: Transit light curves of HD 209458. Left: Discovery light curve, from Charbonneau et al. (2000). Right: HST follow-up light curve (Brown et al. 2001). eral groups initiated photometric surveys to search for planets, i.e., through the transit technique. Such an endeavor is highly involved and consists of several steps, described in the following paragraphs.

The photometric survey: The ﬁrst part of the hunt for exoplanets through the transit technique consists of a photometric survey, searching for targets showing a transit-like light curve. Such light curves, showing a shallow periodic dimming are listed as transiting planet candidates. Since there are several scenarios producing transit-like light curves, other than a transiting planet-star system, most of these candidates, up to about 90%, are eventually identiﬁed as false alarms. The common false alarm scenarios are:

• Astrophysical: Astrophysical systems whose light curves mimic

a transit light curve (e.g., Charbonneau et al. 2004), such as

grazing eclipsing binaries and low-mass eclipsing stellar objects

17 (e.g., Pont et al. 2005). Another astrophysical scenario consists

of an eclipsing binary in a triple stellar system, where the ﬂux

decrease during the deep stellar eclipses is seen as a small vari-

ation relative to the total ﬂux, from the binary and third star,

making it similar to a planetary transit (e.g., O’Donovan et al.

2006).

• Blending: When small-aperture wide-ﬁeld telescopes are used

for carrying out the photometric survey, their typical large pixel

scale may cause stars to become blended together on the CCD.

In such a case the light of a variable star is diluted, resultingin a

an apparent small amplitude variability which can be similar to

a transit-likesignal. This may also be the result of a variable star

positioned near the target, so the target’s photometric aperture

includes the variable star’s PSF limb (e.g., O’Donovan et al.

2007, see their Figs. 6 & 7).

• Statistical: Systematic trends in the light curve may appear as

a transit signal when folded on some period (Pont et al. 2006).

Light curves obtained by the photometric survey can not diﬀer between the false alarm scenarios and a true planet-star system. There- fore, follow-up observations, spectroscopic and photometric, are re-

18 quired.

Spectroscopic follow-up: Spectroscopic follow-up observations are aimed at closely examining the system, and measuring the companion’s mass. These observations are commonly done using high- resolution spectrographs. For transiting candidates, since the transit ephemeris, including both period and phase, is known from the light curve, only a few RV measurements are needed for conﬁrming or rejecting the planet-star hypothesis. The orbit is usually ﬁtted with a

2-parameter model: the RV amplitude, K, and the center of mass velocity, γ. The period, P , and mid-transit time, Tc, are held ﬁxed, and the eccentricity, e, is assumed to be zero as most transiting planets have short periods, where the tidal circularization process is eﬃcient.

If the orbit is clearly non-circular, the model becomes 4-parameter, including also the eccentricity and angle of periastron, ω.

The spectrum of the host star is needed also for determining its physical parameters, its mass in particular. This is done by modeling the spectrum and using stellar evolution models (e.g., Yi et al.

2001). In case a high quality photometric follow-up light curve is available (see next paragraph), a somewhat diﬀerent approach can be used (Sozzetti et al. 2007). Subsequently, the planetary mass is

19 calculated using Eq. 1.1, and the orbital semi-major axis is derived using Newton’s version of Kepler’s third law.

Photometric follow-up: Photometric follow-up of transiting candidates is usually done with 1m-class telescopes, or larger, where the photometric accuracy and resolution is higher than that of the telescopes used to identify the candidate. A better resolution is needed for identifying instrumental false alarms, described above (an extreme case is CoRoT, where the instrumental PSF is very large, see

Deeg et al. 2009). The high photometric accuracy allows to exam- ine the light curve shape, thus identifying V-shaped stellar eclipses.

Finally, if no false alarm scenario is identiﬁed the light curve parameters are derived by ﬁtting a planetary transit model (Mandel & Agol

2002; Gimenez 2006). Assuming a circular orbit, the model includes the orbital period, mid-transit time, ratio between the planetary and stellar radii, Rp/Rs, semi-major axis normalized by the stellar radius, a/Rs, and the impact parameter, b = a cos(i)/Rs. Since a is known from the orbital ﬁt, both the planetary and stellar radii, and the inclination angle are derived. Therefore, the stellar mass is a single parameter needed for obtaining the planetary radius and mass, from the independent observational measurements of the planet-star mass

20 and radius ratios. A transit light curve can, in principle, constrain e and ω, in addition to the ﬁve parameters listed above, thereby giving a photometric estimate of these two parameters, independent of the spectroscopic one (e.g., Barnes 2007; Ford et al. 2008). For an eccentric orbit the transit duration will be diﬀerent than for a circular orbit with the same semi-major axis (see Fig. 2 of Ford et al. 2008). For an e=0.5 orbit, similar to that of HAT-P-2b, this difference is close to a factor of 2 if the transit occurs near apsis, and for a HD 80606b-like orbit (e=0.93) this factor is about 5 (Barnes

2007). In addition, the transit light curve of a planet in an eccentric orbit becomes asymmetric, due to the varying planet-star relative sky-projected velocity during transit, leading to a diﬀerence between the durations of ingress and egress, depending on ω. For example, this diﬀerence is about 2 minutes for HAT-P-2b (Barnes 2007).

The ﬁrst successful photometric campaign was carried out by the

OGLE project (e.g., Udalski et al. 1992, 1997), initially devised to look for gravitational microlensing events (see a short description below). Using the 1.3-m Warsaw telescope at Las Campanas Ob- servatory, Chile, they monitored a dense ﬁeld of stars towards the

21 Galactic bulge. A few dozen transit-like light curves were identiﬁed and made available to the community (Udalski et al. 2002a, 2002b).

Shortly after, Konacki et al. (2003) conﬁrmed the planetary nature of OGLE-TR-56b, with the HIRES spectrograph.

A somewhat different approach was taken by several other groups, utilizing small-aperture, wide-field, dedicated telescopes for looking for transits. The first success of this approach was announced in

2004 by the TrES group (Dunham et al. 2004; O’Donovan et al.

2004; Alonso et al. 2004a), with the discovery of TrES-1 (Alonso et al. 2004b). In 2006, three other teams (HAT, Bakos et al. 2004;

WASP, Pollacco et al. 2006; XO, McCullough et al. 2005), also using small-aperture telescopes, announced their ﬁrst discoveries (HAT-P-

1b, Bakos et al. 2007b; WASP-1b & WASP-2b, Cameron et al. 2007;

XO-1b, McCullough et al. 2006). Since then, the discovery rate has increased dramatically (Charbonneau 2009).

While ground-based surveys are aﬀected by local weather conditions, the blurring of the atmosphere, and can operate only during local night time, space-based telescopes are free of these limitations.

The ﬁrst space-based search for transiting planets was done already in the year 2000. Gilliland et al. (2000) observed the globular cluster

22 47 Tucanae (NGC 104) for 8.3 days with HST, but no transits were identiﬁed among the 34,000 stars observed. Another search using

HST, for 7 days continuously, was performed by Sahu et al. (2006; the SWEEPS project) towards the Galactic bulge. They identiﬁed

16 planetary candidates, where for two of them RV follow-up rules out stellar companions.

The ﬁrst dedictaed space-based telescope is CoRoT (Deleuil et al.

2000), where I am part of the ground-based photometric follow-up campaign (Deeg et al. 2009). This 27-cm telescope was launched in December 2006, and started detecting planets in 2007 (Barge et al. 2008; Alonso et al. 2008), including the smallest planet yet with a measured radius (Leger et al. 2009). A more ambitious mission is Kepler (Borucki et al. 2003), launched in March 2009. Using its

95-cm telescope, it will continuously monitor a single ﬁeld for 3.5 years, with the goal of detecting terrestrial planets, similar to Earth in radius, mass, orbit and parent star.

Other Techniques

Gravitational microlensing: A microlensing event occurs when the relative motion of a distant luminous background star (the lensed object), a foreground star (the lens) and the observer brings them to

23 alignment (e.g., Sackett 1997; Bennett 2009). During such an event, lasting an order of 10 days, the relative motion makes the light from the background star path through different regions of the gravitational field of the foreground star, and any planets it may host. This results in a magnification light curve, where the existence of a planet has a clear signature, in the form of a superimposed sharp peak, or peaks, typically lasting a few hours. The microlensing method is sensitive to planets orbiting distant stars (d ∼ 103 pc), in intermediate orbits (a ∼ 1 AU), down to Earth mass. It is a one-time event, lack- ing the capability of repeated measurements, and requires continuous monitoring during the event. The first microlensing planet was discovered in 2004 (OGLE 2003-BLG-235/MOA 2003-BLG-53, Bond et al. 2004) and several more were discovered since (e.g., Beaulieu et al.

2006; Gaudi et al. 2008).

Direct Imaging: Perhaps the most straightforward way to look for planets is to image the close vicinity of a star. Although ap- pealing, this method faces the challenge of resolving the faint planet from the nearby, bright host star. The steady advancement of imaging technologies continuously increases the observing capabilities, resulting so far in the direct imaging of planets around two nearby

24 stars, published recently: (I) Marois et al. (2008) used IR observations from Keck and Gemini to identify a 3-planet system around

HR 8799, and (II) Kalas et al. (2008) used optical HST observations to detect a planet orbiting Fomalhaut. Although currently limited, this method is sensitive to planets at large distances from their parent stars (a ∼ 102 AU) and is complementary in this way to other methods described above.

Current Status and Future Prospects

Since the ﬁrst discoveries at the end of the previous century, the discovery rate of RV planets and transiting planets has increased exponentially (Charbonneau 2009). However, it is only the transiting planets that allow a detailed characterization of their structure and composition (e.g., Guillot 2008).

Many of these discoveries pose challenges to the theoretical understanding of the planet phenomenon, such as the detection of planets in eccentric orbits (e.g., HD 80606b, Naef et al. 2001; HD 20782b,

Jones et al. 2006), planets on close-in orbits (dubbed Hot Jupiters, e.g., OGLE-TR-56b, Konacki et al. 2003; WASP-12b, Hebb et al.

2009), massive planets (e.g., XO-3b, Johns-Krull et al. 2008; WASP-

14b, Joshi et al. 2009), planets orbiting a component of a stellar

25 binary (e.g., 16 Cyg B b, Cochran et al. 1997; HD 41004 A b, Zucker et al. 2004) and inﬂated planets (e.g., TrES-4, Mandushev et al. 2007;

CoRoT-1b, Barge et al. 2008).

Figure 1.5 presents the planetary mass vs. mean orbital radius

(semi-major axis) for the currently known exoplanets, in log-log scale

(figure taken from Chambers 2009). The Mp sin i minimum mass is given for the non-transiting planets. A few observational biases are evident. The lack of planets beyond ∼5 AU, reflecting the time that passed since the major exoplanets RV search programs started operating, the deficiency of low-mass planets at large orbital distances, resulting from the difficulty in detecting such planets, and the inherent bias in the transiting planets sample towards close-in orbits. Other features seen in Fig. 1.5 are real, and are used to constrain modern planetary theory (see Sec. 1.1.2). Those include the increased frequency of planets beyond ∼1 AU (Cumming et al.

2008), the shortage of high-mass planets at short orbital distances

(Zucker & Mazeh 2002) and the correlation between planetary mass and orbital distance, or period, seen for transiting planets (Mazeh et al. 2005; described in Sec. 1.1.2).

Figure 1.6 shows the orbital period distribution, in log scale, of

26 Figure 1.5: Planetary mass vs. mean orbital radius of the currently known planets, in log-log scale. The Mp sin i minimum mass is given for the non-transiting planets. The figure shows a few observational biases, such as the lack of exoplanets beyond ∼5 AU and the deficiency of low-mass planets at large orbital distances. However, the increased frequency of planets beyond ∼1 AU (Cumming et al. 2008) and the shortage of high-mass planets at small orbital distances is probably real (Zucker & Mazeh 2002; figure taken from Chambers 2009). planets detected using the RV method (gray, solid line), and the dashed bins represent planets detected through the transit method.

The increased frequency of planets with long periods, beyond 100 days is real, as they are harder to detect. This is also seen in Fig. 1.5.

The decline beyond ∼ 500 days is, however, an observational bias, reﬂecting the decrease in the RV amplitude with increasing period

(see Eq. 1.1), and the current time span of the high-precision RV measurements. The short periods of the transiting planets is a clear observational bias, from several reasons. The transit probability de-

27 creases with increasing orbital distance, as shown in Eq. 1.3. In addition, the time the system spends in-transit relative to the entire orbital period decreases as a−1, and, detecting transiting planets with longer periods requires longer time span and more intensive photometric monitoring. However, when considering only the RV planets, there is a small increase in the frequency of planets with periods of a few days, relative to longer periods, of 10 to 100 days. Since the RV method is also biased towards short periods, although less than the transit method, it is hard to say whether this increase is real or not.

55 50

40 35

#30 25

20 15

10 5

0 0 0.5 1 1.5 2 2.5 3 3.5 4 Log (P [days]) 10

Figure 1.6: Histogram of orbital periods, in log scale, for planets detected through the RV method, shown in gray. The increased frequency of planets at long periods, of a few hundred days, is real (see also Fig. 1.5). The dashed bins show the eﬀect of adding also the planets detected through the transit method, which are biased towards short periods. The histogram shows there is a small increase in the frequency of RV planets with periods of a few days, relative to 10 to 100 days. As the RV method is also biased towards short periods, although less than the transit method, it is hard to say whether this increase is real or not.

28 As many teams world-wide are perfecting their detection capabilities, new campaigns are being launched, and new techniques being pursued, we can expect future detection rate to increase further, and probably witness more surprising discoveries.

Over the last few years the ﬁeld has been shifting from the exciting era of ﬁrst discoveries, to increasing the amount of known planets in order to characterize them as a sample, and study in detail interesting planets (as described in 1.1.2). A similar process took place in the study of other astrophysical phenomena in the past, such as asteroids and supernovae.

1.1.2 The Role of Observations in Developing Planetary

Theory

The study of planets is led by the available observational data. Up until the end of the 20th century only a single planetary system was known, the Solar System. Obviously, basing a theory on a sample of one bears the risk of reaching an incomplete understanding. However, sometimes, as in this case, there is no other choice. On the other hand, although the Solar System was a single example for a long time, it was thoroughly studied, much more than any other system known today, and probably also in the foreseeable future.

29 The Solar System

There are several key characteristics common to most, if not all planets in the Solar System. Their orbits are coplanar, nearly circular and well separated, and their motion is in the prograde direction, meaning the same direction as the Sun’s rotation. Six of the eight planets (apart from Venus and Uranus) have self rotation, or spin, in the same direction as they orbit the Sun, with obliquities smaller than 30◦. Although they contain only . 0.2% of the Solar System mass, their orbital motion is responsible for 98% of the system’s angular momentum (Lissauer 1993).

Planet formation in disks

The structure of the Solar System, coplanarity and prograde motion in particular, were perceived already in the 18th century as evidence for formation in a protoplanetary disk — a ﬂattened disk of material revolving around the early Sun (Laplace 1796, Kant 1755). Models devised in the 20th century showed that such disks are a natural result of stellar formation (e.g., Hoyle 1960; Cameron 1962). By the end of the 20th century, observational support for formation through a protoplanetary disk was obtained, in the form of direct imaging of

30 circumstellar disks around nearby stars (e.g., Smith & Terrile 1984;

Holland et al. 1998; Schneider et al. 1999). A few of these disks show structure, believed to be sculpted by planets (e.g., Heap et al. 2000).

In addition, disks were also observed in the two systems detected recently by direct imaging (Zuckerman & Song 2004; Kalas et al.

2005, 2008; Marois et al. 2008).

Therefore, there is growing observational evidence indicating that formation of planets takes place within protoplanetary disks surrounding young stars. However, identifying their birth place is much easier than understanding how they are conceived and develop from a disk of gas and dust to the planets observed today (e.g., Boden- heimer & Lin 2002).

The core accretion theory

The core accretion theory (e.g., Pollack et al. 1996; Bodenheimer et al. 2000; Alibert et al. 2005; Mordasini et al. 2008) is the commonly accepted theory for planet formation.

Within the core accretion framework the process of planetary growth is usually divided into several stages (e.g., Lissauer 1993;

Perryman 2000; Chambers 2008; Fortier et al. 2009). First, micro- scopic grains coalesce and grow to become planetesimals — solid

31 bodies of about a kilometer-size. Next, the gravitational interaction between planetesimals dominates other eﬀects (e.g., electromagnetic forces and gas drag), and they grow via pairwise mergers to become protoplanets, or protoplanetary cores. If and when these protoplanets become massive enough they accrete gas from their immediate surrounding, and obtain a gaseous envelope.

This paradigm explains well many of the observations in the Solar

System, especially the circular, coplanar orbits, and the high mass and size of the outer planets (Jupiter, Saturn, Uranus and Neptune), relative to the smaller, inner planets (Mercury, Venus, Earth and

Mars). The primary diﬃculty with the core accretion theory is that it is not clear whether the gas depletion time is long enough to allow gas giants formation (e.g., Wuchterl et al. 2000; Haisch et al. 2001;

Wyatt et al. 2003; Matsuo et al. 2007). However, a possible solution to this problem was suggested by Alibert et al. (2005), involving the planetary migration mechanism (see below), since during migration the planetary feeding zone is continuously being replenished, leading to faster planet formation.

32 The gravitational instability theory

A competing theory for planet formation states that planets form through the gravitational collapse of dense fragments of the protoplanetary disk (e.g., Boss 1997, 2000; Durisen et al. 2007). In this top-down scenario the gaseous atmospheres are created ﬁrst, and only then the heavy elements sink in to create the dense core. Al- though it takes place in a disk, this process bears some resemblance to stellar formation. The main problems with this formation process are that it requires a relatively massive disk, over 10 times the minimum mass solar nebula model, and it is not clear whether the disk can cool fast enough to allow fragmentation (e.g., Armitage 2007;

Matsuo et al. 2007).

It is also possible that both formation processes, the core accretion and gravitational instability, occur, depending on the initial conditions in the protoplanetary disk (e.g., Ribas & Miralda-Escude 2007;

Armitage 2007). If true, we should expect the existence of two populations of planets. Planets formed through core accretion should be lower in mass and with small orbital radii relative to planets formed by disk instability (Armitage 2007). The latter may also have larger

33 orbital eccentricity (Ribas & Miralda-Escude 2007) and smaller host star metallicity (Rice et al. 2003). So far, studies attempting to detect these two distinct populations in the observational data gave inconclusive results (e.g., Rice et al. 2003; Marcy et al. 2005; Ribas

& Miralda-Escude 2007).

Planetary Migration

The discovery of close-in Jupiter-like planets was surprising since such orbits were not predicted theoretically, as explained below.

The presence of volatiles like water (H2O), ammonia (NH3) and methane (CH4) ices is critical for the eﬃcient growth of massive protoplanets, which, in turn, accrete a large gaseous envelope to become gas giants. As those materials require low temperatures, they are believed to exist in the protoplanetary disk only beyond a certain distance from the star, deﬁning the so-called “snow line”, which for the Sun is at bout 3 AU (Hayashi 1981). Therefore, gas giant planets were predicted to be found beyond this line, while small, rocky, gas-less planets, within it.

The prediction described above originates from a key component of the theory and is highly consistent with the Solar System. How- ever, it was contradicted already by the very ﬁrst exoplanet detection,

34 51 Peg b, in 1995. It’s Mp sin i, of 0.46 MJ , is intermediate between the mass of Jupiter and Saturn, while its orbital semi-major axis is only 0.05 AU, six times smaller than Mercury’s perihelion. Such a close-in orbit for a gas giant was inconsistent with the expected large orbits, beyond the snow line. The detection of gas giants in similar orbits during the following years (e.g., τ Boo b and υ And b; But- ler et al. 1997) made it clear that close-in massive planets are not a rare phenomenon, despite the selection eﬀect towards their detection.

Therefore, the importance of the ﬁrst exoplanet detections was not only in the detection of a new astrophysical object, but moreover in providing observational evidence for the incompleteness of the theoretical understanding at that time, triggering a revision of planetary theory.

It was suggested that during the last stage of its formation, or immediately after it forms, the planet goes through an orbital evolution phase where it interacts with the disk. This interaction goes on until the disk dissipates, shifting the planet from its formation site to the currently observed orbit. A number of orbital evolution mechanisms were proposed, commonly referred to as planetary migration (e.g., Chambers 2009). Interestingly, the migration process

35 was studied already in the 1980’s (Goldreich & Tremaine 1980; Lin

& Papaloizou 1986), but it received significant attention only after the first exoplanet discoveries (nicely reflected by the citation history of these papers).

Planetary migration is ideally divided into two kinds, where low- mass planets, such as the Earth, undergo type I migration, and massive planets, type II. In type I, the interaction between the planet and the disk causes the exterior part of the disk to exert a larger torque on the planet than the interior part, resulting in an inward motion of the planet. In type II, the planet is massive enough to gravitationally clear an annular gap in the disk. Once this happens the orbital evolution of the planet is coupled to that of the disk, as illustrated in Figs. 1.7 and 1.8 where the planet migrates inward, along with the disk itself.

When two planets are present in the disk, if the inner migrates slower than the outer then they may be captured in a mean motion resonance (MMR; e.g., Cresswell & Nelson 2008; Lee et al. 2009), where their orbital periods are related by a ratio of two small integers. This prediction is supported by several discoveries of systems in a MMR conﬁguration, such as GJ 876 (Marcy et al. 2001) and

36 Figure 1.7: Snapshots from a simulation of a type II migration process. Each panel shows the disk surface density, in grey-scale. The planet’s initial mass is 10−3 of that of the parent star and the simulation starts after the planet have already cleared a gap in the disk. The unit of length was taken to be the initial orbital radius. The planet is encircled in white, where the circle’s radius equals the Roche radius, and for each panel the time stamp is shown in the upper-right corner, in initial orbital period units. The four panels show the inward motion of the planet, along with the disk itself (ﬁgure taken from Nelson et al. 2000) .

HD 82943 (Lee et al. 2006), both in a 2:1 resonance.

Although widely accepted, the migration scenario is not complete.

One of the primary diﬃculties has to do with stopping migration before the planet falls into the star. On the other hand, as illustrated in Fig. 1.6, it is still not clear whether there is a pile-up of planets in close-in orbits (e.g., Rice et al. 2008), as the commonly used techniques are biased towards detecting such planets. The tendency of

37 Figure 1.8: A graphical presentation of the type II migration simulation shown in Fig. 1.7. Left panel shows the orbital distance evolution and the right panel shows the increase in planetary mass during migration, as the planet accretes material from the disk. The unit of time is the initial orbital period, planetary mass is given in 10−3 the host star mass, which is also the initial mass, and the unit of length is the initial orbital radius (ﬁgure taken from Nelson et al. 2000). planet-host stars to be of high metallicity (e.g., Gonzalez et al. 2001;

Gonzalez 2006; Neves et al. 2009) was suggested in the past to be the result of planetary material that migrated all the way into the star (e.g., Gonzalez 1997; Murray & Chaboyer 2002), enriching its photosphere with heavy elements and showing that migration has no eﬃcient stopping mechanism. However, several studies claim that the planet-metallicity correlation can not be explained by accretion of planetary, high-Z material (Santos et al. 2003; Gonzalez 2006;

Valenti & Fischer 2008).

Planetary radius-mass relation

The discovery of exoplanets in the second half of the 1990’s made it possible to test theories of planetary formation and evolution. The

38 discovery of transiting planets ten years ago allowed an even broader and more detailed study of planetary theory, as described in this and the next subsections.

Only for transiting planets can both the planetary radius and exact mass be derived, as almost independent measurements. Ob- taining these two fundamental planetary properties allows testing theories of planetary structure, and inferring planetary composition

(e.g., Guillot 2005; Chabrier et al. 2009). Comparing the measured radius and mass to predictions of planetary models can show whether the planet has a heavy-element core, or a large fraction of ice or rock

(e.g., Fortney et al. 2007). Moreover, the planet’s structure may hold some clues as to the way it was formed. For example, the existence of a massive heavy-element core favors formation through core accretion.

Some of the discoveries of transiting planets, including the very

ﬁrst one, HD 209458b, posed theoretical challenges since the measured radii were signiﬁcantly larger than theoretically predicted (Bur- rows et al. 2004; Guillot 2008). These discoveries triggered an intensive theoretical study (e.g., Hubbard et al. 2001; Fortney et al. 2007;

Burrows et al. 2007; Baraﬀe et al. 2008), which is still on-going, lead-

39 ing to improved understanding of planetary structure. The mechanisms suggested for explaining the enlarged radii include intense stellar irradiation (Guillot & Showman 2002), downward transport of kinetic energy which is transformed into thermal energy deep inside the planet (Showman & Guillot 2002), layered convection (Chabrier &

Baraﬀe 2007), the transit radius eﬀect (Hubbard et al. 2001; Burrows et al. 2007), and enhanced planetary atmospheric opacities (Burrows et al. 2007). Another intriguing possibility is that the enlarged radius results from the interaction with a second, unseen planet in the system. The interaction between the two planets keeps the eccentricity of the transiting planet above zero, leading to tidal dissipation as the source of heat responsible for its large radius (Bodenheimer et al. 2001).

On the other hand, the radii of a few other planets seems to be signiﬁcantly smaller than theoretical predictions (Burrows et al.

2007), which is especially true for HD 149026b (Sato et al. 2005).

These radii may be explained by the existence of a dense, heavy- element core (Fortney et al. 2006; Burrows et al. 2007), which, in turn, favors the core-accretion scenario (Sato et al. 2005).

The study of planetary structure through their measured mass

40 and radius is ongoing, where most of the currently known planets are fairly understood. Although, some exceptions still remain (e.g.,

TrES-4, Mandushev et al. 2007; WASP-17b, Anderson et al. 2009).

Those planets are key objects for understanding the incompleteness in current planetary theory, as their radius and mass (and mean density) can not be explained using the mechanisms mentioned above.

This incompleteness will be the target of future follow-up studies, attempting to explain why they diﬀer from other planets, with similar orbits and/or host stars.

Planetary atmospheres

Transiting systems render the opportunity to quantitatively study the planetary atmosphere in several ways.

The atmospheric temperature is obtained via the measurement in the IR, of the secondary eclipse depth, when the star occults the planet (e.g., Deming et al. 2005; Charbonneau et al. 2005). Such measurements in several bands give a characterization of the atmospheric emission spectrum, by which a thermal inversion layer was identiﬁed in the atmospheres of a few planets (e.g., HD 209458b,

Knutson et al. 2008; TrES-4, Knutson et al. 2009). A similar measurement in the optical is aimed at detecting the light reﬂected by

41 the planet and measuring the atmospheric albedo. Current results and upper boundaries (e.g., Winn et al. 2008a; Rowe et al. 2008;

Snellen et al. 2009) are consistent with the low albedos of the Solar

System gas giants (Cox 2000).

Observing the transit in several bands and measuring the planetary radius gives also a rough characterization of the atmospheric transmission spectrum. A more detailed analysis is done by binning spectra taken during transit or comparing spectra taken in and out of transit, to look for the signature of speciﬁc gases (e.g., Brown 2001;

Brown et al. 2002; Pont et al. 2009a). In this way Charbonneau et al. (2002) identiﬁed Sodium in the atmosphere of HD 209458b (con-

ﬁrmed from the ground by Snellen et al. 2008), and Pont et al. (2008) identiﬁed atmospheric haze in HD 189733b.

Some aspects of atmospheric dynamics can be studied through a continuous photometric observation along more than half the orbital phase. Knutson et al. (2007) observed HD 189733 from transit to secondary eclipse, and determined that energy from the irradiated dayside is efficiently redistributed throughout the atmosphere. They also identified an atmospheric hot spot shifted to the east, meaning that the maximum flux appears slightly before opposition.

42 So far, the atmosphere of only a handful of planets was studied, since most of these measurements can be done only for bright host stars, and they require complicated and expensive observations. The future study of planetary atmospheres will beneﬁt from detections of new transiting planet-host stars, especially bright ones, and the development of new techniques and instruments (e.g., the recently launched Herschel satellite).

Spin-orbit alignment

The angle between the host star’s spin axis and the planetary orbital angular momentum axis is a trace of formation and orbital evolution processes. For transiting systems, the sky-projection of this angle is measurable via the Rossiter-McLaughlin eﬀect (RM; Rossiter 1924;

McLaughlin 1924) — an anomalous RV signal observed during transit, originating from the rotation of the host star (e.g., Queloz et al.

2000; Winn et al. 2005; Gaudi & Winn 2007). This measurement was obtained so far only for a small sample of systems. If spin-orbit alignment will be shown to be common, it will support migration processes which preserve and increase spin-orbit alignment, such as migration through interaction with the protoplanetary disk (Gaudi

& Winn 2007). On the other hand, migration involving processes

43 such as planet-planet interactions or Kozai oscillations (Kozai 1962;

Mazeh & Shaham 1979) can lead to misalignment (e.g., Mazeh et al.

1997; Wu & Murray 2003; Winn 2007; Fabrycky & Tremaine 2007).

Looking for a second planet in the system

As noted above, systems in a MMR conﬁguration are one of the predictions of the migration theory. For a transiting system, Hol- man & Murray (2005) and Agol et al. (2005) have shown that the gravitational inﬂuence of an additional, unobserved planet, can induce measurable deviations of the mid-transit times, especially when it orbits in a MMR. These deviations are known as Transit Timing

Variations, or TTVs, and can reach an order of a minute for a system with a second, Earth-mass planet orbiting in a 2:1 resonance with a

Jupiter-mass planet in a 3 day orbit (Agol et al. 2005). Therefore, a second planet can, in principle, be identiﬁed using photometric observations with 1m-class telescopes. However, as of this writing there was no clear detection of TTVs (although Diaz et al. 2008 present an interesting case) and MMR systems were detected only through the RV technique (e.g., GJ 876, Marcy et al. 2001; HD 82943, Lee et al. 2006). Nevertheless, a multi-planet system including a transiting planet is likely to exist, as shown by Fabrycky (2009).

44 The mass-period correlation

The mass-period correlation, as identified by Mazeh et al. (2005; see also Gaudi et al. 2005), is an intriguing observed correlation, not predicted theoretically, between increasing period and decreasing mass. The first identification of this correlation was based on the 6 transiting planets known at that time. A related correlation was identified later by Southworth et al. (2007), between the planetary surface-gravity and the period.

These correlations are still supported by the current sample of transiting planets (Mazeh 2009; Southworth 2009), although their nature is not completely understood. A possible explanation for the lack of low-mass planets at short periods is that the strong stellar

X-ray and EUV irradiation, which increases with decreasing period, caused these planets to evaporate (Baraﬀe et al. 2004; Mazeh et al.

2005; Davis & Wheatley 2009). Pont (2009) suggested that the mass- period correlation originates from tidal interaction between the star and the planet. Future detections of new transiting planets will allow to further study this interesting correlation.

Another kind of mass-period correlation was identiﬁed by Zucker

& Mazeh (2002), based on the minimum-mass of RV planets. This

45 correlation, conﬁrmed by more recent studies (e.g., Jiang et al. 2007,

2009), is seen as a deﬁciency of massive RV planets with short orbital periods (see Fig. 1.5). It is apparent in wider ranges of period and mass than the correlation for the short period, transiting planets described above, so it probably results from other processes. Rice et al. (2008) suggest that massive planets which migrate close to the star undergo rapid eccentricity growth, in a time scale of 105 yr, leading to their destruction. Their results also predict a pile-up of low-mass planets, at the Neptune-mass range, in close-in orbits, just beyond the Roche limit. The current sample of low-mass planets is too small to test this prediction.

1.2 Structure of the Thesis

The thesis consists of two scientiﬁc eﬀorts, searching for new transiting exoplanets in order to extend the current sample, and follow-up observations of interesting known transiting exoplanets, aimed at a more detailed study.

The ﬁrst part includes three discovery papers, of the planets HAT-

P-2b, also named HD 147506b (Paper I, Bakos et al. 2007a), HAT-P-

5b (Paper II, Bakos et al. 2007c) and HAT-P-9b (Paper III, Shporer

46 et al. 2009a). The second part includes the photometric follow-up of

WASP-1b (Paper IV, Shporer et al. 2007b), the spectroscopic follow- up of HAT-P-2b (Paper V, Loeillet et al. 2008), and the photometric follow-up campaign of GJ 436b (Paper VI, Shporer et al. 2009c).

1.2.1 Searching for Transiting Extrasolar Planets

As part of the WHAT2 (Shporer et al. 2006, 2007a, 2009b) team I took part in searching for transiting exoplanets, identiﬁed by WHAT, where observations were carried out in collaboration with similar instruments of the HATNet3 (Bakos et al. 2004). WHAT is located at the Wise Observatory4,5, Israel, and it consists of a 200 mm f/1.8 lens and a 2K×2K CCD, resulting in an 8.2◦ × 8.2◦ ﬁeld of view.

The collaboration with HATNet instruments, located in Arizona and

Hawaii allow the WHAT-HATNet network almost continuous monitoring, increasing the detection capabilities compared to a single-site campaign.

Each WHAT-HATNet ﬁeld, including an order of a few 10,000 stars, is usually observed for one full observational season. After observations are done the data is photometrically processed, using

2http://wise-obs.tau.ac.il/∼what/ 3http://www.hatnet.hu/ 4http://wise-obs.tau.ac.il/ 5Longitude: 34◦45’48” E, Latitude: 30◦35’45” N, Altitude: 875 m

47 aperture photometry and the application of detrending algorithms such as TFA (Kovacs et al. 2005) and SysRem (Tamuz et al. 2005).

Next, the light curves are scanned using the BLS algorithm (Kovacs et al. 2002), looking for a transit-like signal, deriving a list of transit candidates. Those candidates are discussed within the group and the high-priority ones are moved to the follow-up observations phase, where their true nature is investigated.

Most of the WHAT-HATNet candidates go through two phases of spectroscopic follow-up. The ﬁrst is carried out using the CfA speedometer (Latham 1992), mounted on the 1.5-m telescope at the

FLWO, on mount Hopkins, Arizona, USA, with a typical RV accuracy of 500 ms−1. This moderate-resolution spectrograph is used for rejecting clear false alarms, such as stellar binaries where the RV amplitude is too large for a planet. Candidates which survive this

ﬁrst screening are moved to the next phase, where high-resolution spectrographs are used.

During the years 2007–2008 I carried out several observing runs with the SOPHIE spectrograph (Bouchy et al. 2006), mounted on the 1.93-m telescope at the Observatoire de Haute Provence (OHP),

France. I used the high eﬃciency (HE) mode, and an exposure time

48 which resulted in a RV precision of about 10 m s−1, depending on the target’s brightness and local observing conditions. This precision is suﬃcient to detect the expected RV amplitude of Jupiter-size planets, of the order of 100 ms−1. The planetary nature of HAT-P-5b and

HAT-P-9b was identiﬁed in these observing runs.

During my Ph.D. study I regularly carried out photometric follow- up observations of WHAT-HATNet candidates, using the Wise Ob- servatory 0.46-m and 1.0-m telescopes. These observations were part of the detection of the three new planets presented in Papers I–III.

The ﬂexible scheduling and remote mode operation of both telescopes were important for obtaining this time critical data.

The data I obtained at Wise was reduced with a self-written pipeline using aperture photometry and IRAF routines. The target light curve was calibrated using a set of reference stars in the

field, and was de-correlated from several parameters, including airmass, time and PSF FWHM. The final transit light curve was fitted with the Mandel & Agol (2002) model.

1.2.2 Follow-up Observations of Known Transiting Planets

Papers IV–VI present three observational follow-up campaigns of known transiting planets, carried out immediately after their dis-

49 covery.

WASP-1b

WASP-1b is the ﬁrst planet detected by the WASP collaboration

(Pollaco et al. 2006). The discovery paper, Cameron et al. (2007), was made public at the end of September 2006 and did not include high quality transit light curves. We set out to obtain such light curves with the Wise Observatory 1.0-m telescope, in order to better constrain system parameters. Our observations were carried out at the beginning of October 2006, and the paper was submitted soon after.

HAT-P-2b

The discovery of this planet is presented in Paper I. We observed it with SOPHIE right after the discovery, motivated by the possibility of a second planet in the system, as suggested in the discovery paper.

The relatively large mass, close to 9 MJ , and large orbital eccentricity, of about 0.5, make this planet unusual, probably the result of an unusual formation or orbital evolution process. Measuring the spin- orbit angle allows to further investigate this process. Therefore, we obtained spectroscopic observations also during two transit events,

50 showing the RM eﬀect.

GJ 436b

The Neptune-mass planet orbiting GJ 436 was discovered spectroscopically by Butler et al. (2004), and was the ﬁrst planet discovered at this mass range. Its M-type host star and non-circular orbit made it even more interesting (e=0.16, Maness et al. 2007). A transit search was carried out as part of the discovery paper, but with a null result. Surprisingly, transits were discovered a few years later, by Gillon et al. (2007), in a work I was part of. Immediately after the discovery of transits we launched a photometric follow-up campaign in order to reﬁne the light curve parameters and look for any variation between transit events.

Chapter 2

The Papers Paper I – HAT-P-2b

HD 147506b: A SUPERMASSIVE PLANET IN AN ECCENTRIC ORBIT TRANSITING A BRIGHT STAR1 G. A´ . Bakos,2, 3 G. Kova´cs,4 G. Torres,2 D. A. Fischer,5 D. W. Latham,2 R. W. Noyes,2 D. D. Sasselov,2 T. Mazeh,6 A. Shporer,6 R. P. Butler,7 R. P. Stefanik,2 J. M. Ferna´ndez,2 A. Sozzetti,2, 8 A. Pa´l,9, 2 J. Johnson,10 G. W. Marcy,10 J. N. Winn,11 B. SipIcz,9, 2 J. La´za´r,12 I. Papp,12 and P. Sa´ri12 Received 2007 April 29; accepted 2007 July 20

ABSTRACT

We report the discovery of a massive (Mp 9:04 0:50 MJ) planet transiting the bright (V 8:7) F8 star HD 147506, with an orbital period of 5:63341 0¼:00013Æ days and an eccentricity of e 0:520 0:010.¼ From the tran- Æ 0:038 ¼ Æ sit light curve we determine that the radius of the planet is Rp 0:982þ0:105 RJ. HD 147506b (also coined HAT-P-2b) ¼ À 3 has a mass about 9 times the average mass of previously known transiting exoplanets and a density of p 12 g cmÀ , greater than that of rocky planets like the Earth. Its mass and radius are marginally consistent with theories of structure of massivegiant planets composed of pureH and He,andaccounting forthemmay require alarge(k100 M )core. The high eccentricity causes a ninefold variation of insolation of the planet between peri- and apastron. Using follow-upÈ photometry, we ﬁnd that the center of transit is Tmid 2;454;212:8559 0:0007(HJD) and the transit duration is 0:177 0:002 days. ¼ Æ Æ Subject headinggs: planetary systems — stars: individual (HD 147506) On-line material: color ﬁgure, machine-readable table

1. INTRODUCTION 2002; Bodenheimer et al. 2003). The longest period and lowest density transiting exoplanet (TEP) detected so far is HAT-P-1b, To date 18 extrasolar planets have been found that transit their 13 with P 4 46 days (Bakos et al. 2007). All TEPs have orbits con- parent stars and thus yield values for their mass and radius. : sistent with¼ circular Keplerian motion. Masses range from 0.3 to about 1.9 M , and radii from 0.7 to about J From existing radial velocity (RV) data, it might be expected 1.4 R . The majority fit approximately what one expects from theory J that there are some close-in (semimajor axis P0.07 AU, or P P for irradiated gas giant planets (e.g., Fortney et al. 2007 and ref- 10 days) giant planets with masses considerably larger than any erences therein, hereafter FMB07), although there are exceptions: HD 149026b has a small radius for its mass (Sato et al. 2005), im- of the 18 transiting planets now known. A well-known example, considering only objects below the deuterium burning threshold plying that it has a large heavy core ( 70 M ; Laughlin et al. 2005), È ( 13; e.g., Burrows et al. 1997), is Boo b, which was detected and several (HD 209458b, HAT-P-1b, WASP-1) have unexpect- edly large radii for their masses, perhaps suggesting some presently from RV variations, has a minimum mass of Mp sin i 3:9 MJ, and orbits only 0.046 AU from its star ( Butler et al. 1997).¼ Another unknown source of extra internal heating (Guillot & Showman example is HIP14810b ( Wright et al. 2007) with similar mass an orbital period of 6.7 days, and semimajor axis of 0.069 AU. At this orbital distance the a priori probability of such a planet transiting 1 Some of the data presented herein were obtained at the W. M. Keck Obser- vatory, which is operated as a scientific partnership among the California Institute its star is about 10%. Thus, ‘‘supermassive’’ planets should some- of Technology, the University of California, and the National Aeronautics and times be found transiting their parent stars. We report here the de- Space Administration. The Observatory was made possible by the generous fi- tection of the first such TEP and our determination of its mass and nancial support of the W. M. Keck Foundation. The authors wish to recognize and radius. This is also the longest period TEP and the first one to ex- acknowledge the very significant cultural role and reverence that the summit of hibit highly eccentric orbit. Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. 2. OBSERVATIONS AND ANALYSIS Keck time has been in part granted by NASA. 2 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, 2.1. Detection of the Transit in the HATNet Data MA 02138; [email protected]. 3 Hubble Fellow. HD 147506 is an F8 star with visual magnitude 8.7 and 4 Konkoly Observatory, Budapest, P.O. Box 67, H-1125, Hungary. 5 Hipparcos parallax 7:39 0:88 mas (Perryman et al. 1997). It Department of Physics and Astronomy, San Francisco State University, San was initially identified asÆ a transit candidate in our internally la- Francisco, CA 94132. 14 6 Wise Observatory, Tel Aviv University, Tel Aviv, Israel 69978. beled field G193 in the data obtained by HATNet’s (Bakos et al. 7 Department of Terrestrial Magnetism, Carnegie Institute of Washington DC, 2002, 2004) HAT-6 telescope at the Fred Lawrence Whipple Ob- 5241 Broad Branch Road, NW, Washington DC, 20015-1305. servatory ( FLWO) of the Smithsonian Astrophysical Observatory 8 INAFY Osservatorio Astronomico di Torino, Strada Osservatorio 20, 10025 (SAO). The detection of a 5 mmag transit with a 5.63 day period Pino Torinese, Italy. 9 in the light curve consisting of 7000 data points (with a 5.5 minute Department of Astronomy, Eo¨tvo¨s Lorańd University, Pf. 32, H-1518 Budapest, Hungary. cadence) was marginal. Fortunately, the star was in the overlapping 10 Department of Astronomy, University of California, Berkeley, CA 94720. corner with another field (G192) that has been jointly observed 11 Department of Physics, and Kavli Institute for Astrophysics and Space by HATNet’s HAT-9 telescope at the Submillimeter Array (SMA) Research, Massachusetts Institute of Technology, Cambridge, MA 02139. 12 site atop Mauna Kea, Hawaii, and by the Wise HAT telescope Hungarian Astronomical Association, P. O. Box 219, 1461 Budapest, Hungary. 13 See The Extrasolar Planets Encyclopedia at http://exoplanet.eu. 14 At http://www.hatnet.hu. 826 Paper I – HAT-P-2b

SUPERMASSIVE TRANSITING PLANET 827

Fig. 1.—Top : Unbinned HATNet and WHAT joint light curve with 26,400 data points, phased with the P 5:63341 day period. The 5 mmag deep transit is detected with a S/ N of 26. Middle: Same HATNet and WHAT data zooming in on the transit and binned with a ¼ 0:0005 bin-size. Bottom: Sloanz -band photometry taken with the FLWO 1.2 m telescope. Overplotted is our best ﬁt obtained with the Mandel & Agol (2002) formalism.¼

(WHAT; Wise Observatory, Israel; Shporer et al. 2006) for an vations from FLWO, Wise Observatory, Konkoly Observatory, and extended period that yielded 6700 and 3900 additional data the Clay Center (Boston), we finally succeeded in observing a full points, respectively. The transit was independently detected and transit using the KeplerCam detector on the FLWO1.2 m telescope confirmed with these data sets. By chance, the candidate is in yet (see Holman et al. 2007) on UT 2007 April 22. The Sloan z-band another joint field (G149) of HATNet (HAT-7 at FLWO) and light curve is shown in the bottom panel of Figure 1. From the com- WHAT, contributing 6200 and 2200 additional data points, bined HATNet and KeplerCam photometry, spanning a baseline respectively. Altogether this resulted in a light curve with excep- of 839 days, we derive a period of 5:63341 0:00013 days and Æ tional time coverage (570 days), an unprecedented number of data an epoch of midtransit of Tmid 2;454;212:8559 0:0007 days points (26,461 measurements at 5.5 minute cadence), and an rms of (HJD). From the FLWO 1.2 m¼ data alone (and theÆ analytic light 5 mmag. It is noteworthy that the network coverage by WHAT curve fit as described below), the length of transit is 0:177 Æ (longitude E35 ), HATNet at FLWO (W111) and HATNet at 0:002 days (4 hr, 15 minutes), the length of ingress is 0:012 Æ Hawaii (W155) played an important role in detecting such a long 0:002 days (17.5 minutes), and the depth (at the middle of the period and shallow transit. Data were reduced using astrometry transit) is 0.0052 mag. from Pa´l & Bakos (2006) and with a highly fine tuned aperture 2.2. Early Spectroscopy Follow-up photometry. We applied our external parameter decorrelation (EPD) technique on the light curves, whereby deviations from the Initial follow-up observations were made with the CfA Digital median were cross-correlated with a number of ‘‘external param- Speedometer (DS; Latham 1992) in order to characterize the host eters,’’ such as the X and Y subpixel position, hour angle, and star and to reject obvious astrophysical false-positive scenarios zenith distance. We have also applied the trend-filtering algo- that mimic planetary transits. These observations yielded values of 1 rithm (TFA; Kovaćs et al. 2005, hereafter KBN05) along with the TeA 6250 K, log g 4:0, and v sin i 22 km sÀ , correspond- box least-squares (BLS; Kovaćs et al. 2002) transit-search algo- ing to¼ a moderately rotating¼ main-sequence¼ F star. The radial veloc- 1 rithm in our analysis. TFA and BLS were combined in signal- ity (RV) measurements showed an rms residual of 0.82 km sÀ , reconstruction mode, assuming general signal shape, as described slightly larger than the nominal DS precision for a star with this in KBN05. The detection of this relatively shallow transit is a good rotation, and suggested that they may be variable. With a few dozen demonstration of the strengths of TFA. The top panel of Figure 1 additional DS observations, it was found that the RVappeared 1 shows the unbinned light curve with all 26,400 data points, whereas periodic, with P 5:63 days, semiamplitude 1kmsÀ ,and the middle panel displays the transit binned to 1/2000 of the period phasing in agreement with predictions from the HATNet+WHAT (4 minutes). We note that due to the large amount of data, the light curve. This gave strong evidence that there really was an RV binned light curve is of similar precision as a single-transit obser- signal resulting from Keplerian motion, although the precision vation by a 1 m class telescope. was insufficient to establish the orbit with confidence. Altogether, After several failed attempts (due to bad weather and instrument we collected 53 individual spectra spanning a time base of more failure) to carry out high-precision photometric follow-up obser- than a year (Table 1). Paper I – HAT-P-2b

828 BAKOS ET AL. Vol. 670

TABLE 1 TABLE 2 Radial Velocities for HD 147506 Summary of Stellar Parameters for HD 147506

BJD 2,400,000 RV a Uncertainty Parameter Value Source À 1 1 b (days) (m sÀ ) (m sÀ ) Observatory T (K)...... 6290 110 SME eff Æ 53,981.7775...... 556.0 8.4 Keck log g ...... 4.22 0.14 SME À Æ 53,982.8717...... 864.1 8.5 Keck v sin i (km s 1)...... 19.8 1.6 SME À À Æ 53,983.8148...... 62.9 8.8 Keck [Fe/H] (dex)...... +0.12 0.08 SME À Æ 53,984.8950...... 280.6 8.6 Keck Distance (pc) ...... 135 16 HIP Æ 2 54,023.6915...... 157.8 9.9 Keck Distance (pc) ...... 110 15 Y isochrones, a/R? constraint Æ0:085 2 54,186.9982...... 120.2 5.5 Keck log g ...... 4:214þ0:015 Y isochrones, a/R? constraint À0:062 2 54,187.1041...... 104.6 5.7 Keck Mass (M ) ...... 1:298þ0:098 Y isochrones, a/R? constraint À0:042 2 54,187.1599...... 130.1 5.3 Keck Radius (R )...... 1:474þ0:167 Y isochrones, a/R? constraint À0:052 2 54,188.0169...... 168.5 5.3 Keck log L? (L )...... 0:485þ0:134 Y isochrones, a/R? constraint À0:36 2 54,188.1596...... 198.2 5.5 Keck MV ...... 3:54þ0:15 Y isochrones, a/R? constraint À0:8 2 54,189.0104...... 68.9 5.7 Keck Age (Gyr)...... 2:6þ1:4 Y isochrones, a/R? constraint 54,189.0889...... 69.7 6.2 Keck À 54,189.1577...... 25.2 6.1 Keck 54,168.9679...... 152.7 42.1 Lick À of 6400 100 K, somewhat higher than the spectroscopic value, 54,169.9519...... 542.4 41.3 Lick Æ 54,170.8619...... 556.8 42.6 Lick but consistent within the errors. 54,171.0365...... 719.1 49.6 Lick Based on the Hipparcos parallax ( 7:39 0:88 mas), the 54,218.8081...... 1165.2 88.3 Lick apparent magnitude V 8:71 0:01 (Droege¼ Æ et al. 2006), the À 54,218.9856...... 1492.6 90.8 Lick ¼ Æ À SME temperature, and a bolometric correction of BCV 0:011 54,219.9373...... 28.2 43.9 Lick ¼À Æ À 0:011 mag (Flower 1996), the application of the Stefan-Boltzmann 54,219.9600...... 14.8 43.9 Lick À law yields a stellar radius of R? 1:84 0:24 R . 54,220.9641...... 451.6 38.4 Lick A more sophisticated approach¼ to determineÆ the stellar param- 54,220.9934...... 590.7 37.1 Lick eters uses stellar evolution models along with the observational 2 Note.—Table 1 is published in its entirety in the electronic edition of the constraints from spectroscopy. For this we used the Y models Astrophysical Journal. A portion is shown here for guidance regarding its by Yi et al. (2001) and Demarque et al. (2004), and explored a form and content. wide range of ages to find all models consistent with TeA, MV , and a The RVs include the barycentric correction. b [Fe/H] within the observational errors. Here MV 3:05 0:26 Only the Keck and Lick data points are shown here. Consult the elec- ¼ Æ tronic edition for a full data set that includes the CfA DS measurements. is the absolute visual magnitude, as calculated from V and the Hipparcos parallax. This procedure resulted in a mass and ra- 0:10 0:31 dius for the star of M? 1:42þ0:12 M and R? 1:85þ0:28 R , ¼1:4 À ¼ À 2.3. High-Precision Spectroscopy Follow-up and a best-fit age of 2:7þ0:6 Gyr. Other methods that rely on the À 15 In order to confirm or refute the planetary nature of the tran- Hipparcos parallax, such as the Padova stellar model grids (Girardi et al. 2002), consistently yielded a stellar mass of 1.4 M siting object, we pursued follow-up observations with the HIRES instrument (Vogt et al. 1994) on the W. M. Keck telescope and and stellar radius of 1.8 R . If we do not rely on the Hipparcos parallax, and use log g as a with the Hamilton echelle spectrograph at the Lick Observatory 2 proxy for luminosity (instead of MV ), then the Y stellar evolution (Vogt 1987). The spectrometer slit used at Keck is 0:8600, yield- 0:17 models yield a smaller stellar mass of M? 1:29þ0:12 M , a radius ing a resolving power of about 55;000 with a spectral coverage 0:36 ¼ 0:8À 8 of R? 1:46þ0:27 R , and a best-fit age of 2: 6þ2:5 Gyr. The surface between about 320000 and 8800 . The Hamilton echelle spectro- ¼ À À graph at Lick has a similar resolution of about 50;000. These spec- gravity is a sensitive measure of the degree of evolution of the tra were used (1) to more fully characterize the stellar properties star, as is luminosity, and therefore has a very strong influence on of the system, (2) to obtain a radial velocity orbit, and (3) to check the radius. However, log g is a notoriously difficult quantity to mea- for spectral line bisector variations that may be indicativeofa sure spectroscopically and is often strongly correlated with other blend. We gathered 13 spectra at Keck (plus an iodine-free tem- spectroscopic parameters. plate) spanning 207 days and 10 spectra at Lick (plus template) It has been pointed out by Sozzetti et al. (2007) that the nor- spanning 50 days. The radial velocities measured from these spec- malized separation a/R? can provide a much better constraint for tra are shown in Table 1, along with those from the CfA DS. stellar parameter determination than log g. The a/R? quantity can be determined directly from the photometric observations, with- 3. STELLAR PARAMETERS out additional assumptions, and it is related to the density of the A spectral synthesis modeling of the iodine-free Keck template central star. As discussed below in 6, an analytic fit to the FLWO 1.2 m light curve, taking into accountx an eccentric orbit, yielded spectrum was carried out using the SME software (Valenti & 1:10 a/R? 9:77þ0:02. Using this as a constraint, along with TeA and Piskunov 1996), with the wavelength ranges and atomic line data ¼ À 0:06 0:04 [Fe/H], we obtained M? 1:30þ0:10 M , R? 1:47þ0:17 R , and described by Valenti & Fischer (2005). Results are shown in 0:8 ¼ À 0:085¼ À A an age of 2:6þ1:4 Gyr. The log g 4:214þ0:015 derived this way Table 2. The values obtained for the effective temperature (Te ), À ¼ À surface gravity (log g), and projected rotational velocity (v sin i) is consistent with former value from SME. are consistent with those found from the CfA DS spectra. As a As seen from the above discussion, there is an inconsistency between stellar parameters depending on whether the Hipparcos check on TeA, we collected all available photometry for HD 147506 in the Johnson, Cousins, 2MASS, and Tycho systems, and applied parallax is used or not. Methods relying on the parallax (Stefan- a number of color-temperature calibrations (Ramı´rez & Meleńdez Boltzmann law, stellar evolution models with MV constraint, etc.) 2005; Masana et al. 2006; Casagrande et al. 2006) using seven different color indices. These resulted in an average temperature 15 At http://pleiadi.pd.astro.it. Paper I – HAT-P-2b

No. 1, 2007 SUPERMASSIVE TRANSITING PLANET 829 favor a larger mass and radius ( 1.4 M and 1.8 R , respec- riod and transit epoch fixed at the photometric values given earlier. tively), whereas methods that do not rely on the parallax (stellar The parameters adjusted are the velocity semiamplitude K, the evolution models with log g or a/R? constraint) point to smaller eccentricity e, the longitude of periastron !, the center-of-mass mass and radius ( 1.3 M and 1.46 R , respectively). We have velocity for the Keck relative velocities , and offsets Áv KL, be- chosen to rely on the a/R? method, which yields considerably tween Keck and Lick, and Áv KC, between Keck and CfA DS. smaller uncertainties and a calculated transit duration that matches The fitted parameters were found to be fairly insensitive to the 1 the observations. In addition, it implies an angular diameter for the level of jitter assumed. However, only for a jitter of 60 m sÀ 0:021 1 2 star ( 0:127þ0:014 mas) that is in agreement with the more di- (or 70 m sÀ when the CfA DS data are included) did the rect estimate¼ ofÀ 0:117 0:001 mas from the near-infrared approach values expected from the number of degrees of free- surface-brightness relation¼ byÆ Kervella et al. (2004). The latter es- dom. There are thus two possible conclusions: if we accept that 1 timate depends only on the measured V Ks color and apparent HD 147506 has stellar jitter at the 60 m s level, then a single- À À Ks magnitude (ignoring extinction) from 2MASS (Skrutskie et al. planet solution such as ours adequately describes our observations. 1 2006), properly converted to the homogenized Bessell & Brett If, on the other hand, the true jitter is much smaller (P20 m sÀ ), system for this application (following Carpenter 2001). We note then the extra scatter requires further explanation (see below). Our that our results from the a/R? method imply a somewhat smaller adopted orbital parameters for the simplest single-planet Keplerian distance to HD 147506 than the one based on the Hipparcos solution are based only on the more precise Keck and Lick data, 1 parallax. The final adopted stellar parameters are listed in Table 2. andassumethat the jitteris 60 m sÀ (Table 3). The orbital fit is shown graphically in the top panel of Figure 2. In this figure, the 3.1. Stellar Jitter zero point of phase is chosen to occur at the epoch of midtransit, Stars with significant rotation are known to exhibit excess Tmid 2;454;212:8559 (HJD). The most significant results are scatter (‘‘jitter’’) in their radial velocities (e.g., Wright 2005 and the large¼ eccentricity (e 0:520 0:010) and the large velocity ¼ Æ 1 references therein), due to enhanced chromospheric activity and semiamplitude (K 1011 38 m sÀ , indicating a very massive the associated surface inhomogeneities (spottedness). This jitter companion). As we¼ show inÆ the next section ( 5), the compan- comes in addition to the internal errors in the measured velocities ion is a planet, i.e., HD 147506b, which we hereafterx refer to as and could potentially be significant in our case. We note that after HAT-P-2b. prewhitening the light curve with the transit component, we found As a consistency check we also fitted the orbits by fixing only no significant sinusoidal signal above 0.3 mmag amplitude. From the period and leaving the transit epoch as a free parameter. We this we conclude that there is no very significant spot activity found that for all values of the stellar jitter the predicted time of on the star (in the observed 500 day window). In order to estimate transit as derived from the RV fit was consistent with the photo- the level of chromospheric activity in the star, we have derived an metric ephemeris within the uncertainties. We also found thatin activity index from the Ca ii H and K lines in our Keck spectra of these fits the orbital parameters were insensitive to the level of jitter log RHK0 4:72 0:05. For this value the calibration by Wright and to whether or not the CfA DS data were included. The eccen- ¼À Æ 1 (2005) predicts velocity jitter ranging from 8 to 16 m sÀ . An tricity values ranged from 0.51 to 0.53. earlier calibration by Saar et al. (1998), parameterized in terms of We note that the accuracy of the HAT-P-2 system parameters, the projected rotational velocity, predicts a jitter level of up to notably the K velocity semiamplitude (i.e., the planetary mass), 1 1 50 m sÀ for our measured v sin i of 20 km sÀ . A different cal- will profit from extensive high-precision radial velocity monitor- 1 ibration by the same authors in terms of RHK0 gives 20 m sÀ . An ing, since the current data set has limited phase coverage and the additional way to estimate the jitter is to compare HD 147506 to periastron passage is only covered by two Lick points that have stars of the Lick Planet Search program (Cumming et al. 1999) that precision inferior to the Keck data.16 v 1 have similar properties (0:4 < B V < 0:5, sin i >15 km sÀ ). v v There are four such stars (J. JohnsonÀ 2007, private communica- 4.1. Solutions In ol ing Two Planets 1 tion), and their average jitter is 45 m sÀ . A more direct measure If we assume that the true stellar jitter is small, then the excess for the particular case of HD 147506 may be obtained from the scatter in the RV fit could be explained by a third body in the sys- multiple exposures we collected during a three-night Keck run in tem, i.e., a hypothetical HAT-P-2c. In addition, such a body could 2007 March. Ignoring the small velocity variations due to orbital provide a natural dynamical explanation for the large eccentric- motion during any given night, the overall scatter of these eight ity of HAT-P-2b at this relatively short period orbit. Preliminary 1 exposures relative to the nightly means is 20 m s À . This may two-planet orbital fits using all the data yielded solutions only sig- be taken as an estimate of the jitter on short timescales, although nificant at the 2 level, not compelling enough to consider as evi- it could be somewhat larger over the entire span of our observa- dence for such a configuration. Additional RV measurements are tions. Altogether, it is reasonable to expect the jitter to be at least needed to firmly establish or refute the existence of HAT-P-2c. 1 1 10 m sÀ , and possibly around 30Y50 m sÀ for this star. We also exploited the fact that the HATNet light curve has a 4. SPECTROSCOPIC ORBITAL SOLUTION unique time coverage and precision, and searched for signs of a second transit that might be due to another orbiting body around We have three velocity data sets available for analysis (13 rel- the host star. Successive box prewhitening based on the BLS spec- ative radial velocity measurements from Keck, 10 from Lick, and trum and assuming trapezoidal-shape transits revealed no second- 53 measurements from the CfA DS), which are nominally on an ary transit deeper than the 0.1% level and period P10 days. absolute scale (Table 1). Given the potential effect of stellar jitter, we performed weighted Keplerian orbital solutions for a range of 5. EXCLUDING BLEND SCENARIOS 1 1 jitter values from 10 to 80 m sÀ , with 10 m sÀ steps. These jitter As an initial test to explore the possibility that the photometric values were added in quadrature to all individual internal errors. signal we detect is a false positive (blend) due to contamination We performed separate fits for the star orbited by a single planet, from an unresolved eclipsing binary, we modeled the light curve both with and without the CfA DS measurements, since these have 1 1 errors ( 600 m sÀ ) significantly larger than Keck (5Y9msÀ ) 16 Measurements made after the acceptance of this paper indeed point toward Y 1 or Lick (40 90 m sÀ ). In all of these solutions we held the pe- a somewhat smaller planetary mass, between 8 and 9 MJ. Paper I – HAT-P-2b

830 BAKOS ET AL. Vol. 670

TABLE 3 Orbital Fit and Planetary Parameters for the HAT-P-2 System

Parameter Value

Period (days)a...... 5.63341 0.00013 a Æ T (HJD) ...... 2,454,212.8559 0.0007 mid Æ Transit duration (days)...... 0.177 0.002 Æ Ingress duration (days) ...... 0.012 0.002 1 b Æ Stellar jitter (m sÀ ) ...... 60 (m s 1) c ...... 278 20 À À Æ K (m s 1) ...... 1011 38 À Æ ! (deg) ...... 179.3 3.6 Æ e...... 0.520 0.010 Æ Tperi (HJD) ...... 2,454,213.369 0.041 1 Æ Áv KL (m sÀ )...... 380 35 1 d À Æ Áv KC (km sÀ ) ...... 19.827 0.087 Æ 9 f (M )(M ) ...... (376 42) ; 10À Æ 2/3 Mp sin i (MJ)...... 7.56 0.28([M? + Mp]/M ]) Æ a sin i (km) ...... (0.0669 0.0025) ; 106 ? Æ a (AU) ...... 0.0677 0.0014 rel Æ ip (deg)...... >84.6 (95% confidence) Mp (MJ)...... 9.04 0.50 Æ0:038 Rp (RJ)...... 0:982þ0:105 3 À4:8 p (g cmÀ )...... 11:9þ1:6 2 À44 gp (m sÀ ) ...... 227þ16 À Fig. 2.—Top : RV measurements phased with the period of P 5:63341days. ¼ a Fixed in the orbital fit. The zero point in phase corresponds to the epoch of midtransit. Large filled cir- b Adopted (see text). cles indicate Keck and Lick points. Small open circles denote CfA DS data (not c used for the fit). Overlaid is the fit that was based only on the Keck and Lick data The velocity is not in an absolute reference frame. 1 d The offset between Keck and CfA DS is given for reference from a fit assuming 60 m sÀ stellar jitter. Middle: Residuals from the fit. Bottom: Line bi- that includes all data sets, but does not affect our solution. sector spans on the same scale as in the top panel. No variation in the line bisectors is seen concomitant with that in the RVs, essentially confirming the planetary nature of the transiting object. assuming that there are three coeval stars in the system, as described by Torres et al. (2004). We were indeed able to reproduce the observed light curve with a configuration in which the brighter speed of the planet and the reflex motion of the star. Outside of object is accompanied by a slightly smaller F star that is in turn transits, the model flux is unity. During transits, the model flux is being eclipsed by a late-type M dwarf. However, the predicted rel- computed using the formalism of Mandel & Agol (2002), which ative brightness of the two brighter objects at optical wavelengths provides an analytic approximation of the flux of a limb-darkened would be 0.58, and this would have been easily detected in our star that is being eclipsed. The free parameters were the midtransit spectra. This configuration can thus be ruled out. time Tmid, the radius ratio Rp /R? , the orbital inclination i, and the The reality of the velocity variations was tested by carefully ex- scale parameter a/R?, where a is the semimajor axis of the rel- amining the spectral line bisectors of the star in our more numer- ative orbit. The latter parameter is determined by the timescales ous Keck spectra. If the velocity changes measured are due only of the transit (the total duration and the partial-transit duration) and is related to the mean density of the star (see 3). The orbital pe- to distortions in the line profiles arising from contamination of the x spectrum by the presence of a binary with a period of 5.63 days, riod, eccentricity, and argument of pericenter were fixed at the we would expect the bisector spans (which measure line asym- values determined previously by fitting the radial velocity data. metry) to vary with this period and with an amplitude similar to The limb-darkening law was assumed to be quadratic, with coef- the velocities (see, e.g., Queloz et al. 2001; Torres et al. 2005). ficients taken from Claret (2004). The bisector spans were computed from the cross-correlation func- To solve for the parameters and their uncertainties, we used a tion averaged over 15 spectral orders blueward of 5000 8 and Markov chain Monte Carlo algorithm that has been used exten- unaffected by the iodine lines, which is representative of the av- sively for modeling other transits (see, e.g., Winn et al. 2007; erage spectral line profile of the star. The cross-correlations were Holman et al. 2007). This algorithm determines the a posteriori performed against a synthetic spectrum matching the effective tem- probability distribution for each parameter, assuming independent perature, surface gravity, and rotational broadening of the star as (‘‘white’’) Gaussian noise in the photometric data. However, we determined from the SME analysis. As shown in Figure 2, while found that there are indeed correlated errors. Following Gillon et al. the measured velocities exhibit significant variation as a function (2006), we estimated the red noise r via the equation of phase (top panel ), the bisector spans are essentially constant 2 2=N within the errors (bottom panel ). Therefore, this analysis rules out 2 N À 1 ; 1 a blend scenario and confirms that the orbiting body is indeed a r ¼ 1 1=N ð Þ planet. À where is the standard deviation of the out-of-transit flux of the 6. PLANETARY PARAMETERS 1 original (unbinned) light curve, N is the standard deviation of For a precise determination of the physical properties of HAT- the light curve after binning into groups of N data points, and P-2b we have modeled the FLWO 1.2 m Sloan z-band photomet- N 40 corresponds to a binning duration of 20 minutes, which ric data shown in Figure 1. The model is an eccentric Keplerian is the¼ ingress/egress timescale that is criticalffiffiffiffi for parameter estima- orbit of a star and planet, thus accounting for the nonuniform tion. With white noise only, N 1/pN and r 0. We added ¼ ¼ Paper I – HAT-P-2b

No. 1, 2007 SUPERMASSIVE TRANSITING PLANET 831

r in quadrature to the error bar of each point, effectively inflating the error bars by a factor of 1.25. The result for the radius ratio is Rp /R? 0:0684 0:0009, and 1:10 ¼ Æ for the scale parameter a/R? 9:77þ0:02. The a posteriori distribu- ¼ À tion for a/R? is very asymmetric because the transit is consistent with being equatorial: i > 84:6, with 95% confidence. We confirmed that these uncertainties are dominated by the photometric errors, rather than by the covariances with the orbital parameters e, !, and P, and hence we were justified in fixing those orbital parameters at constant values. Based on the inclination, the mass of the star (Table 2), and the orbital parameters (Table 3), the planet mass is then 9:04 0:50 MJ. Based on the radius of the star Æ (Table 2) and the above Rp /R? determination, the radius of the 0:038 planet is Rp 0:982þ0:105 RJ. These properties are summarized in Table 3. ¼ À

7. DISCUSSION Fig. 4.—Surface gravity of TEPs as a function of orbital period. Data taken In comparison with the other 18 previously known transiting from Southworth et al. (2007) with the exception of HAT-P-2b. This object is exoplanets, HAT-P-2b is quite remarkable (Figs. 3 and 4). Its mass clearly not obeying the suspected correlation of gp and P for ‘‘Jupiter-mass’’ objects. of 9:04 0:50 MJ is 5 times greater than any of these 18 other Æ 4:8 3 exoplanets. Its mean density 11:9þ1:6 g cmÀ is 9 times that ¼ À 3 of the densest known exoplanet (OGLE-TR-113b, 1:35 g cmÀ ) 1.159 and 1.22 R , respectively, for coreless models. Our observed and indeed greater than that of the rocky planets of the¼ solar system J 3 44 2 radius of 0.982 RJ is smaller than any of the above values (4.5, 1, ( 5:5gcmÀ ). Its surface gravity of 227þ16 msÀ is 7 times 0.3 Gyr, with or without the 100 M core). Since the 1 positive that¼ of any of the previously known TEPs andÀ 30 times that of È error bar on our radius determination is 0.038 RJ, the inconsis- HAT-P-1b (Fig. 4). tency is only marginal (at the 3Y5 level). Nevertheless, the ob- We may compare the mass and radius for HAT-P-2b with evo- served radius favors either greater age or bigger core size, or both. lutionary models, including irradiation, as recently presented by Given the age of the host star (2.6 Gyr; Table 2), the greater age is FMB07. Given the inferred stellar luminosity (Table 2) and the an unlikely explanation. The required core size for this mass and time-integral of the insolation over an entire period (taking into radius according to FMB07 would be 300 M , an amount of icy account the orbital parameters, notably e and arel), the equivalent and rocky material that may be hard to accountÈ for. We note that semimajor axis aeq for the same amount of irradiation if the cen- using the alternate (and not favored) stellar radius of R? 1:84 R tral star were solar is 0.036 AU. At that separation, FMB07 find for that relies on methods based on the Hipparcos parallax ( 3), the a pure hydrogen/helium planet of mass 9 MJ and age of 4.5 Gyr x planetary radius would be Rp 1:2 RJ. This is broadly consistent a planetary radius about 1.099 RJ. A 100 M core has a negligi- È with young and coreless models of FMB07. ble effect on the radius (yielding 1.068 RJ), which is not surpris- Figure 3 also shows a theoretical mass-radius relation for ob- ing, since the mass of such a core is only a few percent of the total jects ranging from gas giant planets to stars ( Baraffe et al. 1998, mass. For younger ages, of 1 and 0.3 Gyr, the radii are even larger: 2003). Note that HAT-P-2b falls on the relation connecting giant planets to brown dwarfs to stars. It thus appears to be intermediate in its properties, between Jupiter-like planets and more massive objects such as brown dwarfs or even low-mass stars. According to theories, stars with mass k0.2 M have a core, in which internal pressure is dominated by classical gas (ions and electrons), and the R M radiusYmass relation holds in hydrostatic equilibrium (for a/ review and details on the following relations, see, e.g., Chabrier et al. 2000). Below 0.075 M (80 MJ) mass, however, the equation of state in the core becomes dominated by degenerate 1=3 electron gas (R M À for full degeneracy), yielding an expected / minimum in the massYradius relationship (around 73 MJ). Below this mass, the partial degeneracy of the object and the classical (R M 1=3) Coulomb pressure together yield an almost constant / 1=8 radius (R M À ). HAT-P-2b is a demonstration of this well- known phenomenon./ (The approximate relation breaks below M 4 MJ, where the degeneracy saturates, and a classical mass- radius behavior is recovered). Compared to the other 18 known transiting planets, HAT-P-2b Fig. 3.—Mass-radius diagram of known TEPs (from http://www.exoplanet.eu is also unique in having an orbit with remarkably high eccentric- and references therein), Jupiter and Saturn (large filled circles), and low-mass stars ity. The primary question is how such an eccentricity was created from Beatty et al. (2007). HAT-P-2b is an intermediate-mass object that is still in the in the first place. One possible explanation could be that the planet planetary regime (well below 13 MJ). Overlaid are equidensity lines (labeled), was scattered inward from a larger orbit, acquiring a high eccen- Baraffe et al. (1998) (stellar) and Baraffe et al. (2003) (zero insolation planetary) isochrones for ages of 0.5 Gyr (thick dotted line)and5Gyr(thick dashed-line), tricity in the process (Ford & Rasio 2007; Chatterjee et al. 2007). respectively. [See the electronic edition of the Journal for a color version of this If so, then the scattering event might have caused its new orbital figure.] plane to be inclined relative to the plane of the original disk, and Paper I – HAT-P-2b

832 BAKOS ET AL. hence out of the equatorial plane of the parent star (e.g., Fabrycky Boo b’s instead is) to the presence of a second planet in the & Tremaine 2007). This angle between these two planes should be HAT-P-2 system, or to rather different formation/migration sce- readily measurable from the Rossiter-McLaughlin effect (Winn narios altogether. et al. 2005). Indeed, the star HD 147506 is an ideal subject for HD 147506, with visual magnitude 8.71, is the fourth brightest studying this effect, because its rapid rotation should lead to a rel- among the known stars harboring transiting planets. Therefore, it atively large Rossiter-McLaughlin signal. has special interest because of the possibilities for follow-up with There are a number of other interesting issues related to the high large space- or ground-based telescopes. eccentricity of HAT-P-2b. During its 5.63 day orbit, the insolation reaching the planet’s surface varies by a factor of 9. Assuming an albedo of 0.1 (Rowe etal. 2006) and complete redistribution of insolation energy over the surface of the planet, the equilibrium Operation of the HATNet project is funded in part by NASA temperature varies from about 2150 K at periastron to 1240 K at grant NNG04GN74G. Work by G. A´ . B. was supported by NASA apastron. This would have a major influence on atmospheric dy- through Hubble Fellowship Grant HST-HF-01170.01-A. G. K. namics and photochemistry. wishes to offer thanks for support from the Hungarian Scientific It is interesting to compare the properties of the HAT-P-2 sys- Research Foundation (OTKA) grant K-60750. We acknowledge tem with the Boo system, which—as already noted—harbors partial support from the Kepler Mission under NASA Cooperative a close-in planet with minimum mass Mp sin i 3:9 MJ. Similar- Agreement NCC2-1390 (PI: D. W. L.). G. T. acknowledges par- ities of the two parent stars include the nearly¼ identical masses, tial support from NASA Origins grant NNG04LG89G. T. M. effective temperature, and rapid rotation, although Boo, with thanks the Israel Science Foundation for a support through grant Fe/H 0:28, is somewhat more metal-rich than HD 147506, 03/233. D. A. F. is a Cottrell Science Scholar of Research Corpora- ½with Fe/H¼þ 0:12. A striking difference is that, while the or- tion and acknowledges support from NASA grant NNG05G164G. bital eccentricity½ ¼þ of HAT-P-2b is 0.5, the eccentricity of Boo b We would like to thank Joel Hartman (CfA), Gil Esquerdo (CfA), is not measurably different from zero. However, Boo b’s orbital and Ron Dantowitz and Marek Kozubal (Clay Center) for their period, 3.3 days, is almost half that of HAT-P-2b. A large fraction efforts to observe HAT-P-2b in transit, and Howard Isaacson (San of close-in planets with 5 < P < 10 days have significant eccen- Francisco State University) for obtaining spectra at Lick Obser- tricities (0:1 < e < 0:3), although not as large as HAT-P-2b. For vatory. We wish to thank Amit Moran for his help in the obser- discussion on the eccentricity distribution, see Juric & Tremaine vations with the Wise HAT telescope. We owe special thanks to (2007). As circularization timescales are thought to be very steep the directors and staff of FLWO, SMA, and the Wise Observatory functions of the orbital semimajor axis (Terquem et al. 1998), for supporting the operation of HATNet and WHAT. We would one could then argue that HAT-P-2b’s large value of e is due to also like to thank the anonymous referee for the useful sugges- either the fact that the planet’s orbit is not yet circularized (while tions that improved this paper.

REFERENCES Bakos, G. A´ ., La´za´r, J., Papp, I., Sa´ri, P., & Green, E. M. 2002, PASP, 114, 974 Laughlin, G., Wolf, A., Vanmunster, T., Bodenheimer, P., Fischer, D., Marcy, G., Bakos, G. A´ ., Noyes, R. W., Kovaćs, G., Stanek, K. Z., Sasselov, D. D., & Butler, P., & Vogt, S. 2005, ApJ, 621, 1072 Domsa, I. 2004, PASP, 116, 266 Mandel, K., & Agol, E. 2002, ApJ, 580, L171 Bakos, G. A´ ., et al. 2007, ApJ, 656, 552 Masana, E., Jordi, C., & Ribas, I. 2006, A&A, 450, 735 Baraffe, I., Chabrier, G., Allard, F., & Hauschildt, P. H. 1998, A&A, 337, 403 Pa´l, A., & Bakos, G. A´ . 2006, PASP, 118, 1474 Baraffe, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt, P. H. 2003, Perryman, M. A. C., et al. 1997, A&A, 323, L49 A&A, 402, 701 Queloz, D., et al. 2001, A&A, 379, 279 Beatty, T. G., et al. 2007, ApJ, 663, 573 Ramı´rez, I., & Meleńdez, J. 2005, ApJ, 626, 465 Bodenheimer, P., Laughlin, G., & Lin, D. N. C. 2003, ApJ, 592, 555 Rowe, J. F., et al. 2006, ApJ, 646, 1241 Burrows, A., et al. 1997, ApJ, 491, 856 Saar, S. H., Butler, R. P., & Marcy, G. W. 1998, ApJ, 498, L153 Butler, R. P., Marcy, G. W., Williams, E., Hauser, H., & Shirts, P. 1997, ApJ, Sato, B., et al. 2005, ApJ, 633, 465 474, L115 Shporer, A., Mazeh, T., Moran, A., Bakos, G. A´ ., Kovacs, G., & Mashal, E. Carpenter, J. M. 2001, AJ, 121, 2851 2006, Tenth Anniversary of 51 Peg-b: Status of and Prospects for Hot Jupiter Casagrande, L., Portinari, L., & Flynn, C. 2006, MNRAS, 373, 13 Studies, ed. L. Arnold, F. Bouchy, & C. Moutou (Paris: Frontier Group), 196 Chabrier, G., Baraffe, I., Allard, F., & Hauschildt, P. 2000, ApJ, 542, L119 Skrutskie, M. F., et al. 2006, AJ, 131, 1163 Chatterjee, S., Ford, E. B., & Rasio, F. A. 2007, ApJ, submitted (astro-ph/0703166) Southworth, J., Wheatley, P. J., & Sams, G. 2007, MNRAS, 379, L11 Claret, A. 2004, A&A, 428, 1001 Sozzetti, A., Torres, G., Charbonneau, D., Latham, D. W., Holman, M. J., Cumming, A., Marcy, G. W., & Butler, R. P. 1999, ApJ, 526, 890 Winn, J. N., Laird, J. B., & O’Donovan, F. T. 2007, ApJ, 664, 1190 Demarque, P., Woo, J.-H., Kim, Y.-C., & Yi, S. K. 2004, ApJS, 155, 667 Terquem, C., Papaloizou, J. C. B., Nelson, R. P., & Lin, D. N. C. 1998, ApJ, Droege, T. F., Richmond, M. W., & Sallman, M. 2006, PASP, 118, 1666 502, 788 Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298 Torres, G., Konacki, M., Sasselov, D. D., & Jha, S. 2004, ApJ, 609, 1071 Flower, P. J. 1996, ApJ, 469, 355 ———. 2005, ApJ, 619, 558 Ford, E. B., & Rasio, F. A. 2007, ApJ, submitted (astro-ph/0703163) Valenti, J. A., & Fischer, D. A. 2005, ApJS, 159, 141 Fortney, J. J., Marley, M. S., & Barnes, J. W. 2007, ApJ, 659, 1661 Valenti, J. A., & Piskunov, N. 1996, A&AS, 118, 595 Gillon, M., Pont, F., Moutou, C., Bouchy, F., Courbin, F., Sohy, S., & Magain, P. Vogt, S. S. 1987, PASP, 99, 1214 2006, A&A, 459, 249 Vogt, S. S., et al. 1994, Proc. SPIE, 2198, 362 Girardi, L., et al. 2002, A&A, 391, 195 Winn, J. N., Holman, M. J., & Fuentes, C. I. 2007, AJ, 133, 11 Guillot, T., & Showman, A. P. 2002, A&A, 385, 156 Winn, J. N., et al. 2005, ApJ, 631, 1215 Holman, M. J., et al. 2007, ApJ, 664, 1185 Wright, J. T. 2005, PASP, 117, 657 Juric, M., & Tremaine, S. 2007, ApJ, submitted (astro-ph/0703160) Wright, J. T., et al. 2007, ApJ, 657, 533 Kervella, P., The´venin, Di Folco, E., & Se´gransan, D. 2004, A&A, 426, 297 Yi, S. K., Demarque, P., Kim, Y.-C., Lee, Y.-W., Ree, C. H., Lejeune, T., & Kovaćs, G., Bakos, G. A´ ., & Noyes, R. W. 2005, MNRAS, 356, 557 Barnes, S. 2001, ApJS, 136, 417 Kovaćs, G., Zucker, S., & Mazeh, T. 2002, A&A, 391, 369 Latham, D. W. 1992, in IAU Colloq. 135, Complementary Approaches to Double and Multiple Star Research, ed. H. A. McAlister & W. I. Hartkopf (ASP Conf. Ser. 32; San Francisco: ASP), 110 Paper II – HAT-P-5b

HAT-P-5b: A JUPITER-LIKE HOT JUPITER TRANSITING A BRIGHT STAR1 G. A´ . Bakos,2,3 A. Shporer,4 A. Pa´l,2,5 G. Torres,2 Ge´za Kovaćs,6 D. W. Latham,2 T. Mazeh,4 A. Ofir,4 R. W. Noyes,2 D. D. Sasselov,2 F. Bouchy,7 F. Pont,8 D. Queloz,8 S. Udry,8 G. Esquerdo,2 B. Sipo˝cz,2,5 Ga´bor Kovaćs,2 R. Stefanik,2 J. La´za´r,9 I. Papp,9 and P. Sa´ri9 Received 2007 October 6; accepted 2007 October 31; published 2007 November 30

ABSTRACT We report the discovery of a planet transiting a moderately bright (V p 12.00 ) G star, with an orbital period p ע of2.788491 0.000025 days. From the transit light curve we determine that the radius of the planet is R p ע p ע 1.257 0.053 RMJ. HAT-P-5b has a mass ofp 1.06 0.11 MJ , similar to the average mass of previously Ϫ3 ע p known transiting exoplanets, and a density ofrp 0.66 0.11 g cm . We ﬁnd that the center of transit is ע ע p Tc 2,454,241.77663 0.00022 days (HJD), and the total transit duration is0.1217 0.0012 days. Subject headings: planetary systems — stars: individual (GSC 02634Ϫ01087, HAT-P-5)

1. INTRODUCTION 2. OBSERVATIONS AND ANALYSIS 2.1. Detection of the Transit in the HATNet Data To date about 20 extrasolar planets have been found that GSC 02634Ϫ01087, also known as 2MASS transit their parent stars and thus yield values for their mass J18173731ϩ3637170, is a G star withI ≈ 11.3 and V ≈ and radius.10 Masses range from 0.07M (GJ 436; Gillon et al. J 12.00. It was initially identified as a transit candidate in our 2007) to about 9 (HAT-P-2b; Bakos et al. 2007), and radii hm MJ internally labeled field G196, centered ata p 18 08 , d p ′ from 0.4RRJ (GJ 436) to about 1.7J (TRES-4; Mandushev et 37Њ30 . The data were acquired by HATNet’s HAT-7 telescope al. 2007). These data provide an opportunity to compare ob- at the Fred Lawrence Whipple Observatory (FLWO) of the servations with theoretical models of planetary structure across Smithsonian Astrophysical Observatory (SAO) and HAT-9 a wide range of parameters, including those of the host star telescope at the Submillimeter Array (SMA) site atop Mauna (e.g., Burrows et al. 2007; Fortney et al. 2007 and references Kea, Hawaii. Following a standard calibration procedure of the therein). Transits also yield precise determination of other phys- frames (meaning bias, dark, and flat-field corrections), data ical parameters of the extrasolar planets, for instance the surface were reduced using the astrometry code of Pa´l & Bakos (2006) gravity. Interesting correlations between these parameters were and a highly fine-tuned aperture photometry. We applied our noted early on, such as that between masses and periods (Mazeh external parameter decorrelation technique on the light curves, et al. 2005) or periods and surface gravities (Southworth et al. whereby deviations from the median were cross-correlated with 2007). Classes of these close-in planets have also been sug- a number of “external parameters,” such as the X and Y subpixel gested, such as very hot Jupiters (VHJs;P p 1–3 days) and position, FWHM, hour angle, and zenith distance. We have hot Jupiters (HJs;P p 3–9 days; Gaudi et al. 2005), or a also applied the trend filtering algorithm (Kovaćs et al. 2005), possible dichotomy based on Safronov numbers (Hansen & along with the box least squares (BLS; Kovaćs et al. 2002) Barman 2007). However, the small ensemble of transiting ex- transit-search algorithm in our analysis. For field G196 we ∼ ∼ oplanets (TEPs) does not allow robust conclusions; thus the gathered 3750 (HAT-7) plus 890 (HAT-9) data points at 5.5 addition of new discoveries is valuable. minute cadence between 2005 June 8 and 2005 December 5 11 (UT). In the light curve of star GSC 02634Ϫ01087 we de- Over the past year the HATNet project (Bakos et al. 2002, ∼ 2004), a wide-angle photometric survey, has announced four tected a 13 mmag transit with a 2.7881 day period, a signal- TEPs. In this Letter we report on the detection of a new tran- to-noise ratio of 12 in the BLS frequency spectrum, and a dip significance of 18 (Kovaćs & Bakos 2005). The top panel siting exoplanet, which we label HAT-P-5b, and our determi- ∼ nation of its parameters, such as mass, radius, density, and of Figure 1 shows the unbinned light curve with all 4640 surface gravity. data points, folded with the period that we derived subsequently, based on high-precision follow-up photometry, as 1 Based in part on observations obtained with the SOPHIE spectrograph described in § 2.4 below. mounted on the 1.93 m telescope at the Haute Provance Observatory. 2 Harvard-Smithsonian Center for Astrophysics (CfA), 60 Garden Street, 2.2. Early Spectroscopy Follow-Up Cambridge, MA 02138; [email protected]. 3 Hubble Fellow. Initial follow-up observations were made with the CfA Dig- 4 Wise Observatory, Tel Aviv University, Tel Aviv, Israel 69978. ital Speedometer (DS; Latham 1992) in order to characterize 5 Department of Astronomy, Eo¨tvo¨s Lorańd University, Pf. 32, H-1518 Bu- the host star and to reject obvious astrophysical false-positive dapest, Hungary. scenarios that mimic planetary transits. The four radial velocity 6 Konkoly Observatory, P.O. Box 67, H-1125 Budapest, Hungary. 7 Institut d’Astrophysique de Paris, 98bis Boulevard Arago, 75014 Paris, (RV) measurements obtained over an interval of 33 days Ϫ1 France. showed an rms residual of 0.41km s , consistent with no 8 Observatoire Astronomique de l’Universite´ de Gene`ve, 51 chemin des detectable RV variation. Atmospheric parameters for the star Maillettes, CH-1290 Sauverny, Switzerland. (effective temperatureT , surface gravitylog g , metallicity 9 Hungarian Astronomical Association, P. O. Box 219, H-1461 Budapest, eff v Hungary. [Fe/H], and projected rotational velocitysin i ) were derived 10 Extrasolar Planets Encyclopedia, http://exoplanet.eu. as described by Torres et al. (2002). The first three quantities 11 See http://www.hatnet.hu. are strongly correlated and difficult to determine simulta- L173 Paper II – HAT-P-5b

L174 BAKOS ET AL. Vol. 671

TABLE 1 Radial Velocities for HAT-P-5

Ϫ a b BJD 2,400,000 RV jRV BS (days) (m sϪ1) (m sϪ1) (m sϪ1) 54227.5199 ...... 7721.4 12.2 22.5 54228.5949 ...... 7457.4 22.3 18.2 54229.6098 ...... 7710.4 17.2 Ϫ7.0 54230.4900 ...... 7603.8 22.4 Ϫ20.0 54231.6088 ...... 7521.3 14.3 15.2 54233.6057 ...... 7579.4 15.1 Ϫ31.0 54234.5210 ...... 7510.5 21.3 Ϫ54.8 54255.5171 ...... 7680.5 9.8 4.2 a The RVs include the barycentric correction. b Bisector spans.

of Cameron et al. (2007) we estimated the RV photon-noise uncertainties to be 10–25msϪ1 . We present here eight radial velocity measurements taken when the planet was out of transit, listed in Table 1. Measurements taken during transit, revealing the Rossiter-McLaughlin effect (e.g., Winn et al. 2005), will be presented in a separate paper.

2.4. Photometry Follow-Up

Fig. 1.—The top panel shows the unbinned HATNet light curve with 4940 In order to better characterize the transit parameters and also data points, phased with the periodP p 2.788491 days, and with the binned to improve the ephemerides, we performed follow-up photo- data overplotted. The 0.013 mag deep transit is detected with a dip significance metric observations with 1 m class telescopes. A partial transit of 18. The other panels show photometry follow-up in the following order: of HAT-P-5b was observed using the KeplerCam detector on Sloan z-band photometry taken with the FLWO 1.2 m telescope (on two separate dates: UT 2007 May 18 and May 21), Cousins R-band photometry the FLWO 1.2 m telescope (see Holman et al. 2007) on UT taken with the Wise 1 m telescope (on 4 nights: UT 2007 May 26, June 20, 2007 May 18. We refer to this event as having transit number p Ϫ p July 4, and July 18). Overplotted are our best analytic fits as described in the Ntr 1. Three days later a full transit,Ntr 0 , was observed text. Residuals from these fits are shown on the same scale at the bottom of with the same instrument. The two Sloan z-band light curves these panels. are shown in the middle panel of Figure 1. We also gathered p p data for four subsequent full transit events,Ntr2 , N tr p p p neously. For example, the unconstrained value log g 11,Ntr16 , andN tr 21 , using the Wise 1 m telescope in we obtained is somewhat lower than derived from the Cousins R band (bottom panel of Fig. 1). Data were reduced 0.2 ע 4.0 our stellar evolution modeling in § 3, which islog g p 4.37 . in a manner similar to the HATNet data, using aperture pho- Consequently, in a second iteration we heldlog g fixed at this tometry and an ensemble of ∼300 comparison stars in the field. p ∼ value and redetermined the other quantities, obtaining Teff Since the follow-up observations span 22 transit cycles ( 2 -and v sin i p month time span), we were able to obtain an accurate ephem ,0.15 ע K, [Fe/H] p ϩ0.24 100 ע 5960 km sϪ1. These correspond to a slowly rotating early eris. An analytic model was fit to these data, as described below 1.5 ע 2.6 days and 0.000025 ע G main-sequence star. in § 6, and yielded a period of2.788491 ע p reference epoch of midtransit Tc 2,454,241.77663 2.3. High-Precision Spectroscopy Follow-Up 0.00022 days (HJD). The length of the transit as determined ,(days (2 hr, 55 minutes 0.0012 ע from this joint fit is0.1217 ,(days (20.9 minutes 0.0007 ע High-resolution spectroscopic follow-up was carried out at the length of ingress is0.0145 the Haute Provence Observatory (OHP) 1.93 m telescope, with and the central transit depth is 0.0136 mag. the SOPHIE12 spectrograph (Bouchy et al. 2006). SOPHIE is a multiorder echelle spectrograph fed through two fibers, one 3. STELLAR PARAMETERS of which is used for starlight and the other for sky background or a wavelength calibration lamp. The instrument is entirely The mass (MRpp ) and radius ( ) of a transiting planet scale computer-controlled, and a standard data reduction pipeline au- with those of the parent star. In order to determine the stellar tomatically processes the data on CCD readout. RVs are cal- properties needed to placeMRpp and on an absolute scale, we culated by numerical cross-correlation with a high-resolution made use of stellar evolution models along with the observa- observed spectral template of a G2 star. Similar spectroscopic tional constraints from spectroscopy. Because of its relative follow-up with SOPHIE has already resulted in the confir- faintness, the host star does not have a parallax measurement mation of two TEPs: WASP-1b and WASP-2b (Cameron et al. from Hipparcos, and thus a direct estimate of the absolute 2007). HAT-P-5 was observed with SOPHIE in the high-effi- magnitude is not available for use as a constraint. An alternative ciency mode (R ∼ 39,000) during our 2007 May 2–13 ob- approach is to use the surface gravity of the star, which is a serving run, with an additional measurement taken on June 4. sensitive measure of the evolutionary state of the star and there- Depending on observing conditions, exposure times were in fore has a very strong influence on the radius. However, the range of 15–35 minutes, resulting in signal-to-noise ratios log g is a notoriously difficult quantity to measure spectro- of 20–55 pixelϪ1 atl p 5500 A˚ . Using the empirical relation scopically and is often strongly correlated with other spectroscopic parameters (see § 2.2). It has been pointed out by Soz- 12 See http://www.obs-hp.fr. zetti et al. (2007) that the normalized separation of the planet, Paper II – HAT-P-5b

No. 2, 2007 HAT-P-5b: JUPITER-LIKE TRANSITING HOT JUPITER L175

TABLE 2 Summary of Stellar Parameters for HAT-P-5 Parameter Value Source ע Teff (K)...... 5960 100 DS DS 1.5 ע v sin i (km sϪ1)...... 2.6 Yonsei-Yale 0.028 ע log g ...... 4.368 Yonsei-Yale 0.15 ע Fe/H] (dex) ...... ϩ0.24] Yonsei-Yale 30 ע Distancea (pc) ...... 340

Yonsei-Yale 0.062 ע Mass (M,) ...... 1.160

Yonsei-Yale 0.049 ע Radius (R,) ...... 1.167

Yonsei-Yale 0.064 ע log(L⋆/L,)...... 0.187 ע MV ...... 4.32 0.18 Yonsei-Yale Yonsei-Yale 1.8 ע Age (Gyr) ...... 2.6 a Assuming no extinction due to the proximity of the star. a/R⋆, can provide a much better constraint for stellar parameter determination thanlog ga . The/R⋆ quantity can be determined directly from the photometric observations with no additional assumptions (other than limb-darkening, which is a second- Fig. 2.—The top panel shows the RV measurements phased with a period order effect), and it is related to the density of the central star. ofP p 2.788491 days. The zero point in phase corresponds to the epoch of As discussed later in § 6, an analytic fit to the light curve yields midtransit. Overlaid is the best fit, assuming 14.4msϪ1 stellar jitter. The middle panel shows the residuals from the fit. The bottom panel displays the line .0.19 ע a/R⋆ p 7.50 This value, along withT and [Fe/H] from § 6, was com- bisector spans on the same scale as the top panel. No variation in the line eff bisectors is seen concomitant with that in the RVs, essentially confirming the pared with the Yonsei-Yale stellar evolution models of Yi et planetary nature of the transiting object. al. (2001) following Sozzetti et al. (2007). As described earlier, the initial temperature and metallicity from our DS spectros- shown in the bottom panel of Figure 2, the changes in the copy were subsequently improved by applying thelog g con- bisector spans are of the same order as the residual RV vari- straint from the models, and repeating the isochrone compar- ations, and much smaller than the radial velocity semiamplitude ison. This resulted in final values for the stellar mass and radius itself. This analysis shows that the orbiting body is a planet ע p ע p ofM⋆ 1.160 0.062 MR, and⋆ 1.167 0.049 R, , and rules out a possible blend scenario. Gyr. We summarize these 1.8 ע and an estimated age of2.6 and other properties in Table 2. 6. PLANETARY PARAMETERS

4. SPECTROSCOPIC ORBITAL SOLUTION The light-curve parameters of HAT-P-5b were determined from a joint fit based on the six distinct transit events, observed Our eight RV measurements from SOPHIE were fitted with with the FLWO 1.2 m and Wise 1 m telescopes. A circular a Keplerian orbit model solving for the velocity semiamplitude orbit was assumed, based on our analysis above. We adopted K and the center-of-mass velocity g, holding the period and a quadratic limb-darkening law for the star, and took the ap- transit epoch fixed at the well-determined values from photom- propriate coefficients from Claret (2004) for both the Sloan z etry. The eccentricity was initially set to zero. The resulting Ϫ and Cousins R bands. The drop in flux in the light curves was rms residual of ∼23.7 m s1 is somewhat larger than expected modeled with the formalism of Mandel & Agol (2002) using from the internal errors, and we find that a reducedx 2 value the equations for the general case (i.e., not the small-planet of unity necessitates the addition in quadrature of uncorrelated Ϫ approximation). The adjusted parameters in the fit were (1) the noise of 14.4 m s1 , which we attribute to “stellar jitter.” This midtransit times of the first full transit (N p 0 ,T ) and the level of jitter is consistent with the predictions of Saar et al. tr c0 last full transit (N p 21 ,T ) (this is equivalent to fitting for (1998) for a projected rotational velocity such as what we tr c21 an epoch E and a period P); (2) the relative planetary radius, measure for the parent star. The final fit, with the internal errors 2 ; Ϫ1 R p /R⋆; (3) the square of the normalized impact parameter,b ע p increased as described above, yieldsK 138 14ms and { Ϫ 2 1/2 p Ϫ1 and (4) the quantityz/R⋆⋆a/R (2p/P)/(1 b ) . From sim- m s . The observations and fitted RV curve 2 9.1 ע g 7613.8 ple geometric considerations,z/Rb⋆ and have an uncorrelated are displayed in the top panel of Figure 2, with the residuals a posteriori probability distribution in parameter space. This shown in the middle panel. amounts to an orthogonalization of the fitted parameters, similar As a test we allowed for the possibility of an eccentric orbit to but simpler than the one employed by Burke et al. (2007) and solved for the two additional quantitiese cos q and for the case of XO-2b. e sin q, but the results were insignificantly different from zero. We used the Markov Chain Monte Carlo algorithm (see, e.g., Holman et al. 2007) to derive the best-fit parameters. Uncer- 5. EXCLUDING BLEND SCENARIOS tainties were estimated using synthetic data sets, by adding We have tested the reality of the velocity variations by care- Gaussian noise to the fitted curve at the dates of our obser- fully examining the spectral line bisectors of our OHP data. If vations and resolving the light-curve fit. The magnitude of the the velocity changes measured are due only to distortions in noise was taken from the white- and red-noise estimations the line profiles arising from contamination of the spectrum by based on the measured residuals. This process was repeated the presence of a binary with a period of 2.79 days, we would 1.5 # 10 5 times, yielding a good representation of the a pos- expect the bisector spans (which measure line asymmetry) to teriori distribution of the best-fit parameter values. We found vary with this period and with an amplitude similar to the this method of error estimation to be robust, since it is not velocities (see, e.g., Queloz et al. 2001; Torres et al. 2005). As sensitive to the number of out-of-transit points. Paper II – HAT-P-5b

L176 BAKOS ET AL. Vol. 671

TABLE 3 ע p The result for the radius ratio is R p /R⋆ 0.1106 Orbital Fit and Planetary Parameters ע and the normalized separation is a/R⋆ p 7.50 ,0.0006 0.19. We found that the a posteriori distribution ofb 2 is con- for the HAT-P-5b System sistent with a symmetric Gaussian distribution and yields Parameter Value ע p 2 b 0.181 0.040; therefore the orbit is inclined. From the Ephemerides: 0.000025 ע inclination, the mass of the star (Table 2), and the orbital pa- Perioda (days) ...... 2.788491 ע a rameters (§ 4), the other planetary parameters (such as mass Tmid (HJD) ...... 2,454,241.77663 0.00022 0.0012 ע and radius) are derived in a straightforward way, and are sum- Transit duration (days) ...... 0.1217 0.0007 ע Ingress duration (days) ...... 0.0145 marized in Table 3. We note thata/R⋆ , as derived from the RV orbital fit: light-curve fit, is an important constraint in the stellar parameter Stellar jitterb (m sϪ1)...... 14.4 9.1 ע determination (§ 3), which in turn defines the limb-darkening gc (m sϪ1)...... 7613.8 14 ע coefficients that are used in the light-curve fit. Thus, after the K (m sϪ1) ...... 138 initial fit to the light curve and the stellar parameter determi- ea ...... 0 nation, we performed another iteration in the light-curve fit. Geometry and planetary parameters: 0.19 ע a/R⋆ ...... 7.50 ע .We found that the change in parameters was imperceptible Rp/R⋆ ...... 0.1106 0.0006 0.00076 ע The possibility of transit time variations (TTVs) was a (AU) ...... 0.04075 ע checked by fitting the center of the transit of the five full b/R⋆ ...... 0.425 0.048 ע ip (deg)...... 86.75 0.44 ע transit events independently. We found no sign of TTV, as d Mp (MJ) ...... 1.06 0.11 ע the transit times differ by less than 1 j from the expected Rp (RJ) ...... 1.26 0.05 ע Ϫ3 values (listed in Table 3). rp (g cm ) ...... 0.66 0.11 ע e Ϫ2 gp (m s ) ...... 16.5 1.9 Transit timing parameters: ע CONCLUSIONS p Ϫ5 .7 DTc,0, Ntr 0 (10 days)...... 6 27 ע p Ϫ5 HAT-P-5b is an ordinary hot Jupiter (P p 2.788 days) with DTc,0, Ntr 2 (10 days) ...... 79 58 ע p Ϫ5 p DTc,0, Ntr 11 (10 days) ...... 6 62 ע slightly inflated radius (R p 1.26 RJ ) for its mass of 1.06 p Ϫ5 Ϫ DTc,0, Ntr 16 (10 days) ...... 112 84 ע M , orbiting a slightly metal-rich solar-like star. The ∼20% p Ϫ5 J DTc,0, Ntr 21 (10 days) ...... 11 57 radius inflation is what current models predict for a planet with a ∼ Fixed in the orbital fit. equilibrium temperature of 1500 K (Burrows et al. 2007; b Adopted (see text). Fortney et al. 2007). c The g velocity is in an absolute reference frame. HAT-P-5b is more massive than any of the known TEPs with d Errors on the planetary parameters include the uncertainties in similar period (2.5 days Շ P Շ 3 days), such as XO-2b, WASP- the stellar parameters. e Based on only directly observable quantities; see Southworth et 1, HAT-P-3b, TRES-1, and HAT-P-4b, with the exception of al. (2007). TRES-2. The latter is fairly similar in mass, radius, orbital period, and stellar effective temperature. G. K. wishes to acknowledge support from Hungarian Scientific However, HAT-P-5b is interesting in that it falls between Research Foundation (OTKA) grant K-60750. We acknowl- Class I and II, as defined by the Safronov number andTeq of edge partial support from the Kepler mission under NASA the planet (Hansen & Barman 2007). HAT-P-5b has a Safronov cooperative agreement NCC2-1390 (D. W. L., PI). A. P. is while Class I is defined as grateful for the hospitality of the CfA, where this work was , 0.005 ע number of0.059 ∽ ע 0.070 0.01, especially atTeq 1500 K. It seems that the carried out. A. P. was also supported by the Doctoral School additional discovery and characterization of transiting planets of Eo¨tvo¨s University. G. T. acknowledges partial support from of Jupiter and higher masses would be very helpful in order NASA Origins grant NNG04LG89G. T. M. thanks the Israel to understand these new correlations and their significance. Science Foundation for support through grant 03/233. We owe special thanks to the directors and staff of FLWO and SMA Operation of the HATNet project is funded in part by NASA for supporting the operation of HATNet. We thank the OHP grant NNG04GN74G. Work by G. A´ . B. was supported by support team and the SOPHIE consortium for their help in NASA through Hubble Fellowship Grant HST-HF-01170.01-A. making the observations.

REFERENCES

Bakos, G. A´ ., La´za´r, J., Papp, I., Sa´ri, P., & Green, E. M. 2002, PASP, 114, Kovaćs, G., Zucker, S., & Mazeh, T. 2002, A&A, 391, 369 974 Latham, D. W. 1992, in IAU Colloq. 135, Complementary Approaches to Bakos, G. A´ ., Noyes, R. W., Kovaćs, G., Stanek, K. Z., Sasselov, D. D., & Double and Multiple Star Research, ed. H. A. McAlister & W. I. Hartkopf Domsa, I. 2004, PASP, 116, 266 (ASP Conf. Ser. 32; San Francisco: ASP), 110 Bakos, G. A´ ., et al. 2007, ApJ, 670, 826 Mandel, K., & Agol, E. 2002, ApJ, 580, L171 Bouchy, F., et al. 2006, in Tenth Anniversary of 51 Peg-b, Status of and Mandushev, G., et al. 2007, ApJ, 667, L195 Prospects for Hot Jupiter Studies (Paris: Frontier Group), 319 Mazeh, T., Zucker, S., & Pont, F. 2005, MNRAS, 356, 955 Burke, C. J., et al. 2007, ApJ, in press (arXiv:0705.0003) Pa´l, A., & Bakos, G. A´ . 2006, PASP, 118, 1474 Burrows, A., Hubeny, I., Budaj, J., & Hubbard, W. B. 2007, ApJ, 661, 502 Queloz, D., et al. 2001, A&A, 379, 279 Cameron, A. C., et al. 2007, MNRAS, 375, 951 Saar, S. H., Butler, R. P., & Marcy, G. W. 1998, ApJ, 498, L153 Claret, A. 2004, A&A, 428, 1001 Southworth, J., Wheatley, P. J., & Sams, G. 2007, MNRAS, 379, L11 Fortney, J. J., Marley, M. S., & Barnes, J. W. 2007, ApJ, 659, 1661 Sozzetti, A., Torres, G., Charbonneau, D., Latham, D. W., Holman, M. J., Gaudi, B. S., Seager, S., & Mallen-Ornelas, G. 2005, ApJ, 623, 472 Winn, J. N., Laird, J. B., & O’Donovan, F. T. 2007, ApJ, 664, 1190 Gillon, M., et al. 2007, A&A, 472, L13 Torres, G., Konacki, M., Sasselov, D. D., & Jha, S. 2005, ApJ, 619, 558 Hansen, B. M. S., & Barman, T. 2007, ApJ, in press (arXiv:0706.3052) Torres, G., Neuhaüser, R., & Guenther, E. W. 2002, AJ, 123, 1701 Holman, M. J., et al. 2007, ApJ, 664, 1185 Winn, J. N., et al. 2005, ApJ, 631, 1215 Kovaćs, G., & Bakos, G. A´ . 2005, preprint (astro-ph/0508081) Yi, S. K., Demarque, P., Kim, Y.-C., Lee, Y.-W., Ree, C. H., Lejeune, T., & Kovaćs, G., Bakos, G. A´ ., & Noyes, R. W. 2005, MNRAS, 356, 557 Barnes, S. 2001, ApJS, 136, 417 Paper III – HAT-P-9b

HAT-P-9b: A LOW-DENSITY PLANET TRANSITING A MODERATELY FAINT F STAR∗ Avi Shporer1,GAsp´ Ar´ A.´ Bakos2,9, Francois Bouchy3, Frederic Pont4,Geza´ Kovacs´ 5, Dave W. Latham2, Brigitta Sipocz¨ 2,6, Guillermo Torres2, Tsevi Mazeh1, Gilbert A. Esquerdo2,7, Andras´ Pal´ 2,6, Robert W. Noyes2, Dimitar D. Sasselov2,Jozsef´ Laz´ ar´ 8, Istvan´ Papp8,Pal´ Sari´ 8, and Gabor´ Kovacs´ 2 1 Wise Observatory, Tel Aviv University, Tel Aviv 69978, Israel; [email protected]. 2 Harvard-Smithsonian Center for Astrophysics (CfA), 60 Garden Street, Cambridge, MA 02138, USA 3 Institut d’Astrophysique de Paris, 98bis Bd Arago, 75014 Paris, France 4 School of Physics, University of Exeter, Stocker Road, Exeter EX4 4QL, UK 5 Konkoly Observatory, P.O. Box 67, H-1125 Budapest, Hungary 6 Department of Astronomy, Eotv¨ os¨ Lorand´ University, Pf. 32, H-1518 Budapest, Hungary 7 Planetary Science Institute, 620 N. 6th Avenue, Tucson, AZ 85705, USA 8 Hungarian Astronomical Association, P.O. Box 219, H-1461 Budapest, Hungary Received 2008 June 24; accepted 2008 September 11; published 2008 December 12

ABSTRACT We report the discovery of a planet transiting a moderately faint (V = 12.3 mag) late F star, with an orbital period of 3.92289 ± 0.00004 days. From the transit light curve and radial velocity measurements, we determine that the radius of the planet is Rp = 1.40 ± 0.06 RJup and that the mass is Mp = 0.78 ± 0.09 MJup. The density of the new −3 planet, ρp = 0.35 ± 0.06 g cm , fits to the low-density tail of the currently known transiting planets. We find that the center of transit is at Tc = 2454417.9077 ± 0.0003 (HJD), and the total transit duration is 0.143 ± 0.004 days. The host star has M⋆ = 1.28 ± 0.13 M⊙ and R⋆ = 1.32 ± 0.07 R⊙. Key words: planetary systems – stars: individual (GSC 02463 − 00281, HAT-P-9) Online-only material: color figures

1. INTRODUCTION 2. OBSERVATIONS AND ANALYSIS Transiting extra-solar planets are important astrophysical ob- 2.1. Detection of the Transit in the HATNet Data jects as they allow us to test planetary structure and evolution theory (e.g., Fortney 2008; Burrows et al. 2008; Baraffe et al. HAT-P-9 is positioned in HATNet’s internally labeled field 2008), especially because they yield a measurement of the G176, centered at α = 07h28m, δ = 37◦30′. This field was planetary mass and radius. Over the last few years, the sam- observed in network mode by the HAT-6 telescope, located ple of known transiting planets has grown substantially, lead- at the Fred Lawrence Whipple Observatory (FLWO) of the ing to an improved theoretical understanding of their phys- Smithsonian Astrophysical Observatory (SAO), and the HAT-9 ical nature (e.g., Guillot et al. 2006; Burrows et al. 2007; telescope at the Submillimeter Array (SMA) site atop Mauna Fortney et al. 2008; Chabrier & Baraffe 2007). In addition, the Kea, HI. In total, 6884 exposures were obtained at a 5.5 min increasing sample of transiting planets enabled the discovery cadence between 2004 November 26 and 2005 October 21 (UT). of some interesting correlations, such as the mass–period rela- Preliminary reduction included standard bias, dark and flat- tion (Mazeh et al. 2005; Gaudi et al. 2005). The astrophysics field corrections, followed by an astrometric solution using the behind these correlations is not fully understood, and a clear code of Pal´ & Bakos (2006). Photometry was applied using way toward progress is the discovery of many more transiting fine-tuned aperture photometry while the raw light curves were planets. processed by our external parameter decorrelation technique. We report here the discovery of another transiting planet Next, the trend filtering algorithm (TFA; Kovacs´ et al. 2005) was detected by the HATNet project10 (Bakos et al. 2002, 2004), applied to get the final light curves. To search for the signature labeled HAT-P-9b, and our determination of its parame- of a transiting planet in the light curves we used the box least ters, including mass, radius, density, and surface gravity. In squares (BLS; Kovacs´ et al. 2002) algorithm. Section 2, we describe our photometric and spectroscopic ob- A transit-like signal was identified in the light curve of servations, and in Section 3 we derive the stellar parameters. HAT-P-9 (GSC 02463 − 00281, 2MASS J07204044+3708263), The orbital solution is performed in Section 4, and a discussion which is a V = 12.3 mag star, fainter than most transit- of possible blend scenarios is brought in Section 5. The deter- ing planet host stars detected with small-aperture, wide-field, mination of the light curve parameters and the planet’s physical ground-based campaigns. The HATNet light curve is pre- parameters is described in Section 6, and we bring a discussion sented in the top panel of Figure 1, folded on the ephemeris in Section 7. obtained from analyzing the follow-up light curves (P = 3.92289 days and Tc = 2454417.9077, see Section 2.4), showing a flux decrement of ∼ 1% at phase zero. The figure shows a small shift between phase zero and transit center, sug- ∗ Based in part on radial velocities obtained with the SOPHIE spectrograph gesting a possible slight difference in the ephemeris for the mounted on the 1.93 m telescope at the Observatoire de Haute Provence, France (runs 07A.PNP.MAZE, 07B.PNP.MAZE, 08A.PNP.MAZE). HATNet data and the photometric follow-up data, obtained 9 NSF Fellow. ∼1000 days apart (see Section 7 for further discussion). This 10 http://www.hatnet.hu. transiting planet candidate, along with others from the same

1393 Paper III – HAT-P-9b

1394 SHPORER ET AL. Vol. 690 Relative Flux

Phase Relative Flux

Phase Figure 1. The top panel shows the unbinned HATNet light curve with 6884 data points, phased with the orbital period of P = 3.92289 d. The binned light curve is overplotted. The center of the transit is at phase zero. The bottom panel shows a zoom-in around the transit. The HATNet light curve is folded with the ephemeris obtained here, based on the follow-up light curves (see Section 2.4). The small offset between phase zero and transit center suggests there is a slight difference in the ephemeris for the HATNet data and the photometric follow-up data, obtained ∼ 1000 days apart. See Section 7 for further discussion.

field, was selected for follow-up observations to investigate Table 1 its nature. Radial Velocities for HAT-P-9 a b c BJD−2,400,000 RV σRV BS S/N 2.2. Early Spectroscopy Follow-Up (days) (m s−1) (ms−1) (ms−1) Initial follow-up observations were made with the CfA Digital 54233.3347 22667 24 −62 43 Speedometer (DS; Latham 1992) in order to characterize the 54379.6357 22586 20 5 44 host star and to reject obvious astrophysical false-positive 54380.6213 22672 24 −93 38 scenarios that mimic planetary transits. The five radial velocity 54381.5815 22749 23 −148 40 (RV) measurements obtained over an interval of 62 days showed 54413.6292 22727 21 3 43 −1 54414.6084 22610 24 −31 38 an rms residual of 1.1 km s , consistent with no detectable 54415.6457 22675 27 −45 34 RV variation. Atmospheric parameters for the star (effective 54416.6392 22740 21 −13 42 temperature Teff, surface gravity log g, and projected rotational 54421.6701 22659 31 −11 32 velocity v sin i) were derived as described by Torres et al. (2002), 54422.6651 22595 20 −59 46 initially assuming a fixed metallicity of [Fe/H] = 0.0 ± 0.2. 54587.3543 22536 21 10 44 We obtained log g = 4.05 ± 0.30, Teff = 6130 ± 150 K, and 54588.3571 22650 23 30 41 v sin i = 12.2 ± 1.0kms−1. As the DS results were consistent 54589.3525 22730 23 −2 40 with a planet orbiting a moderately rotating main sequence 54590.3425 22699 25 −45 38 star, this target was selected for high-precision spectroscopy 54591.3647 22564 30 10 32 follow-up. Notes. 2.3. High-Precision Spectroscopy Follow-Up a The RVs include the barycentric correction. b Bisector span. Observations were carried out at the Haute Provence Obser- c Signal-to-noise ratio per pixel at λ = 5500 Å. vatory (OHP) 1.93 m telescope, with the SOPHIE spectrograph (Bouchy & the Sophie Team 2006). SOPHIE is a multi-order were 32–46 per pixel at λ = 5500 Å. Using the empirical rela- echelle spectrograph fed through two fibers, one of which is used tion of Cameron et al. (2007), we estimated the RV photon-noise for starlight and the other for sky background or a wavelength uncertainties to be 20–31 m s−1. We made 17 RV measure- calibration lamp. The instrument is entirely computer-controlled ments in total. Two of those were highly contaminated by the and a standard data reduction pipeline automatically processes Moon and were ignored, leaving 15 RV measurements, listed in the data upon CCD readout. RVs are calculated by numerical Table 1. cross-correlation with a high-resolution observed spectral template of a G2 star. 2.4. Photometry Follow-Up HAT-P-9 was observed with SOPHIE in the high-efficiency mode (R ∼ 39, 000) during three observing runs, from In order to better characterize the transit parameters and to 2007 May until 2008 May. Due to the relative faintness of this derive a better ephemeris, we performed photometric follow- star, exposure times were in the range of 25–75 min, depend- up observations with 1 m class telescopes. We obtained a total ing on observing conditions. The resulting signal-to-noise ratios of four transit light curves of HAT-P-9b, shown in Figure 3. Paper III – HAT-P-9b

No. 2, 2009 HAT-P-9b 1395

Table 2 List of Follow-Up Light Curves of HAT-P-9

a Ntr Start Date Observatory + Filter u1 u2 Cadence β RMS UT Telescope min−1 % 0 2007 Nov 13 FLWO 1.2 m z 0.1313 0.3664 1.0 1.1 0.17 6 2007 Dec 6 Wise 0.46 m clear 0.2403 0.3816 1.1 1.0 0.21 7 2007 Dec 10 Wise 1.0 m R 0.2403 0.3816 0.4 1.2 0.22 15 2008 Jan 11 FLWO 1.2 m z 0.1313 0.3664 1.0 1.6 0.18

Note. a The correlated noise factor, by which the errors of each light curve are multiplied (see Section 2.4).

Two events were observed by the KeplerCam detector on the Table 3 FLWO 1.2 m telescope (see Holman et al. 2007) on UT 2007 Summary of Stellar Parameters for HAT-P-9 November 13 and UT 2008 January 11, in the Sloan z band. Parameter Value Source We refer to the 2007 November 13 event as having a transit R.A. 07h 20m 40.s44 2MASS number Ntr = 0, so the 2008 January 11 transit number is Decl. +37◦ 08′ 26.′3 2MASS N tr = 15. In addition, two light curves were obtained at the mV (mag) 12.297 ± 0.063 TASS Wise Observatory. On UT 2007 December 6 the Ntr = 6 transit Teff (K) 6350 ± 150 DS + Yonsei–Yale event was observed by the Wise 0.46 m telescope (Brosch et al. v sin i (km s−1) 11.9 ± 1.0 DS + Yonsei–Yale +0.03 2008), with no filter. The following event, with Ntr = 7, was log g 4.29−0.04 DS + Yonsei–Yale + light curve shape observed on UT 2007 December 10 by the Wise 1 m telescope in [Fe/H] (dex) 0.12 ± 0.20 DS + Yonsei–Yale the Cousins R band. An additional light curve, obtained with the Mass (M⊙)1.28 ± 0.13 Yonsei–Yale+lightcurveshape Radius (R⊙)1.32 ± 0.07 Yonsei–Yale+lightcurveshape FLWO 1.2 m telescope, was of poor quality and is not included +0.08 log(L⋆/L⊙)0.41−0.09 Yonsei–Yale here. Table 2 lists for each light curve its Ntr number, UT date, +0.3 MV (mag) 3.7−0.2 Yonsei–Yale observatory, telescope, filter used, limb-darkening coefficients +1.8 Age (Gyr) 1.6−1.4 Yonsei–Yale + light curve shape used in its analysis, mean cadence, the correlated noise β factor +0.06 B − V (mag) 0.50−0.05 Yonsei–Yale (see below), and the rms residuals from the fitted light curve Distance (pc)a 480 ± 60 Yonsei–Yale model. Data were reduced in a similar manner to the HATNet Note. a Assuming extinction of A(V ) = 0.15 mag (see Section 3). data, using aperture photometry and an ensemble of ∼100 comparison stars in the field. An analytic model was fitted to these data, as described below in Section 6, and yielded a than the values of the parameters describing limb darkening, period of 3.92289 ± 0.00004 d and a reference epoch of mid- which is a second-order effect), and it is related to the mean density of the host star. As discussed later in Section 6, an analytic transit Tc = 2454417.9077 ± 0.0003 d (HJD). The length of the transit as determined from this joint fit is 0.143 ± 0.004 d fit to the light curve yields a/R⋆ = 8.6 ± 0.2. This value, along (3 h 26 min), the length of ingress is 0.019 ± 0.003 d (27 min), with Teff from Section 2.2, and assuming [Fe/H] = 0.0 ± 0.2, and the central transit depth is 1.17 ± 0.01%. The latter value is was compared with the Yonsei–Yale stellar evolution models of simply the square of the radius ratio (see Table 4) if one ignores Yi et al. (2001). This resulted in values for the stellar mass and +0.12 the limb-darkening effect. This effect increases the depth by radius of M⋆ = 1.16−0.15 M⊙ and R⋆ = 1.28±0.08 R⊙, and an +3.3 +0.03 about 0.1% in the bands used here. estimated age of 3.6−2.2 Gyr. The result for log g was 4.29−0.04, consistent with the value derived from the DS spectra. Nevertheless, in order to verify the overall consistency of our 3. STELLAR PARAMETERS results and refine the stellar parameters, we carried out a new The mass (Mp) and radius (Rp) of a transiting planet, de- iteration. We imposed this latter value of log g (coming from termined from transit photometry and RV data, are dependent stellar evolution modeling), and analyzed the DS spectra by also on those of the parent star. In order to determine the stel- allowing the metallicity, v sin i, and Teff to vary. The new results −1 lar properties needed to place Mp and Rp on an absolute scale, were [Fe/H] = 0.12 ± 0.20, v sin i = 11.9 ± 1.0kms , and we made use of stellar evolution models along with the ob- Teff = 6350 ± 150 K. Repeating the stellar evolution modeling servational constraints from spectroscopy and photometry, as resulted in M⋆ = 1.28 ± 0.13 M⊙, R⋆ = 1.32 ± 0.07 R⊙, and +1.8 described in Torres et al. (2008). Because of its relative faint- an age of 1.6−1.4 Gyr. The stellar properties are summarized in ness, the host star does not have a parallax measurement from Table 3 and they correspond to a late F star. Hipparcos, and thus a direct estimate of the absolute magnitude We also used our SOPHIE high-resolution spectra to estimate is not available for use as a constraint. An alternative approach v sin i and [Fe/H]. Based on our result for B − V from the +0.06 is to use the surface gravity of the star, which is a measure of stellar evolution models, of B − V = 0.50−0.05 mag, we got the evolutionary state of the star and therefore has a very strong v sin i = 10.1 ± 1.0kms−1, and [Fe/H] = +0.18 ± 0.10 dex. influence on the radius. However, log g is a difficult quantity Those values are close to the results from the DS spectra and to measure spectroscopically and is often strongly correlated the stellar evolution model. However, the SOPHIE spectra were with other spectroscopic parameters (see Section 2.2). It has taken in high-efficiency mode, where there is a known problem been pointed out by Sozzetti et al. (2007) that the normalized with removing the echelle blaze function. Hence, these estimates separation of the planet, a/R⋆, can provide a much better con- are used only for comparison, and are not included in our final straint for stellar parameter determination than the spectroscopic result. log g. Thea/R ⋆ quantity can be determined directly from the To check our spectroscopically determined Teff we used photometric observations with no additional assumptions (other several publicly available color indices for this star, and the Paper III – HAT-P-9b

1396 SHPORER ET AL. Vol. 690

Figure 2. The top panel shows the RV measurements phased with the period of P = 3.92289 days and mid-transit time of Tc = 2454417.9077 (HJD). The zero point in phase corresponds to the epoch of mid-transit. Overlaid is the best sinusoidal fit. The bottom panel shows the residuals from the fit. (A color version of this figure is available in the online journal.) calibrations of Ram´ırez & Melendez´ (2005) and Casagrande Table 4 et al. (2006) to derive independent temperature estimates, Orbital Fit and Planetary Parameters for the HAT-P-9b System using the [Fe/H] value found above in these calibrations. Parameter Value We adjusted the reddening until the photometric temperature Ephemeris matched the spectroscopic temperature, yielding E(B − V ) = Period (day)a 3.92289 ± 0.00004 a 0.053 (Ram´ırez & Melendez´ 2005) and E(B − V ) = 0.043 Tc (HJD) 2454417.9077 ± 0.0003 (Casagrande et al. 2006). We adopted the mean value of 0.048 Transit duration (day) 0.143 ± 0.004 as reddening, implying an extinction of A(V ) = 0.15 mag. Ingressduration(day) 0.019 ± 0.003 Note that the Burstein & Heiles (1982) reddening maps for this Orbital parameters celestial position (l,b = 181.1, 21.4) give E(B − V ) = 0.072 γ (km s−1) 22.665 ± 0.006 and the Schlegel et al. (1998) maps yield E(B − V ) = 0.065, K (m s−1) 84.7 ± 7.9 broadly consistent with our findings, especially if we take into e a 0 account that these latter methods measure the total reddening Lightcurve parameters along the line of sight. a/R⋆ 8.6 ± 0.2 Using the V magnitude from the TASS survey (Droege et al. Rp/R⋆ 0.1083 ± 0.0005 2006), V = 12.297 ± 0.063 mag, the absolute V magnitude from b 0.52 ± 0.03 the stellar evolution model, and the A(V ) = 0.15 mag value Planetary parameters determined above, the distance to HAT-P-9 is 480 ± 60 pc. i (deg) 86.5 ± 0.2 a (AU) 0.053 ± 0.002 4. SPECTROSCOPIC ORBITAL SOLUTION Mp (MJ)0.78 ± 0.09 Rp (RJ)1.40 ± 0.06 −3 Our 15 RV measurements from SOPHIE were fitted with a ρp (g cm )0.35 ± 0.06 −2 b gp (m s ) 9.8 ± 1.0 Keplerian orbit model solving for the velocity semi-amplitude K c Teq (K) 1530 ± 40 and the center-of-mass velocity γ , holding the period and transit Θ d epoch fixed at the well-determined values from photometry 0.046 ± 0.007 f (109 erg cm−2 s−1)e 1.3+0.3 (see Table 4). The eccentricity was set to zero. The fit yields p −0.3 −1 −1 K = 84.7 ± 7.9ms and γ = 22.665 ± 0.006 km s . The Notes. observations and fitted RV curve are displayed in the top panel a Fixed in the orbital fit. of Figure 2. The residuals are presented in the bottom panel of b Based on only directly observable quantities; see the same figure. RMS residuals are 22.1 m s−1, consistent with Southworth et al. (2007, Equation (4)). the RV uncertainties. The value of χ 2 is 12.4 for 13 degrees of c Planetary thermal-equilibrium surface temperature. freedom. d Safronov number; see Hansen & Barman (2007, Equation (2)). e 5. EXCLUDING BLEND SCENARIOS Stellar flux at the planet. We tested the reality of the velocity variations by examining the spectral line bisector spans (BSs) of the star using our distortions in the line profiles arising from contamination of the SOPHIE data. If the measured velocity changes are due only to spectrum by the presence of a binary with a period of 3.92 days, Paper III – HAT-P-9b

No. 2, 2009 HAT-P-9b 1397 we would expect the BSs (which measure line asymmetry) to and their uncertainties. The MCMC algorithm results in a vary with this period, resulting in a correlation between BS and distribution of each of the fitted parameters. We took the RV (see, e.g., Queloz et al. 2001; Torres et al.2005 ). As shown distribution median to be the best-fit value and the values at in Figure 4, the BSs show no significant variations, except one 84.13 and 15.87 percentiles to be the +1σ and −1σ confidence or two outliers. The correlation coefficient between BS and RV limits, respectively. for all 15 measurements is −0.38. Ignoring the extreme point Our four follow-up light curves are presented in Figure 3, reduces the correlation to −0.18. where the fitted model is overplotted and the residuals are In order to estimate the statistical significance of a correlation also shown. The result for the radius ratio is Rp/R⋆ = between BS and RV we define the following statistics: 0.1083 ± 0.0005, and the normalized separation is a/R⋆ = 8.6 ± 0.2. From the mass of the star (Table 3), the orbital param- σBS,fit rσ = , (1) eters (Section 4), and the light curve parameters, the physical σBS planetary parameters (such as mass and radius) are calculated in where σBS is the standard deviation of the BS values and σBS,fit a straightforward way, and are summarized in Table 4. We note the standard deviation of the residuals of a BS fit to the RVs. that a/R⋆, as derived from the light curve fit, is an important Both standard deviations are the unbiased estimators. For pure constraint for the stellar parameter determination (Section 3), noise rσ will equal 1.0. When fitting the BSs to the RVs of which in turn defines the limb-darkening coefficients that are all 15 measurements, using polynomials of degrees 1–3 we used in the light curve analytic fit. Thus, after the initial analytic got rσ in the range 0.984–1.043, consistent with no significant fit to the light curves and the stellar parameter determination, correlation. we performed another iteration of the light curve fit. We found Another sign of a binary would be a dependence between the that the change in parameters was imperceptible. RV amplitude and the template used. This may happen when the components of the blended binary are of a different spectral 7. DISCUSSION type than the primary target. We re-calculated the RVs using We present here the discovery of a new transiting extra-solar F0 and K5 templates, and got an amplitude consistent with our planet, HAT − P − 9b, with a mass of 0.78 ± 0.09 M , radius original value, within 1σ (not shown). Jup of 1.40 ± 0.06 RJup, and orbital period of 3.92289 ± 0.00004 These analyses indicate that the orbiting body is a planet with days. The V = 12.3 mag host star is among the faintest planet high significance. host stars identified with small-aperture, wide-field, ground- 6. PLANETARY PARAMETERS based campaigns. In fact, together with WASP-5 (Anderson et al. 2008), they are currently the faintest planet host stars discovered To determine the light curve parameters of HAT-P-9b we with such campaigns, where the transit depth is below 2%. fitted the four available transit light curves simultaneously, as To show the main characteristics of the new planet relative described in Shporer et al. (2008). A circular orbit was assumed. to the currently known transiting planets, we plot in Figure 5 We adopted a quadratic limb-darkening law for the star, and de- the position of the latter on the radius–mass diagram on the top termined the appropriate coefficients, u1 and u2, by interpolating panel, and the mass–period diagram on the bottom panel. In on the Claret (2000, 2004) grids for the atmospheric model de- both panels HAT-P-9b position is marked by an open circle. scribed above. For the Wise 0.46 m light curve, observed with The radius–mass diagram, presented here in log–log scale, no filter, we adopted the Cousins R-band limb-darkening coeffi- visually shows that the new planet has a low density, of −3 cients as the CCD response resembles a “wide-R” filter (Brosch ρp = 0.35 ± 0.06 g cm , similar to that of HD 209458b (e.g., et al. 2008). The values of the limb-darkening coefficients used Knutson et al. 2007; Mazeh et al. 2000), WASP-1b (Cameron for each light curve are listed in Table 2. et al. 2007; Shporer et al. 2007; Charbonneau et al. 2007), HAT- The drop in flux in the light curves was modeled with P-1b (Bakos et al. 2007; Winn et al. 2007b), and CoRoT-Exo-1b the formalism of Mandel & Agol (2002). The five adjusted (Barge et al. 2008). parameters in the fit were (1) the period P, (2) the mid-transit Comparing the measured radius to the modeled radius of time of the Ntr = 0 transit event, Tc,0, denoted simply as Tc,(3) Burrows et al. (2007), for the same stellar luminosity, semi- the relative planetary radius, r ≡ Rp/R⋆, (4) the orbital semi- major axis, and (core-less) planet mass as the HAT-P-9 system, major axis scaled by the stellar radius, a/R⋆, and (5) the impact shows that the measured value is larger than the modeled one parameter, b ≡ a cos(i)/R⋆, where i is the orbital inclination by 1–2σ. In this comparison we assumed solar opacity for the angle. planetary atmosphere. As shown by Burrows et al. (2007), the Accounting for correlated noise in the photometric data (Pont transit radius effect together with enhanced planetary atmo- et al. 2006) was done similarly to the “time-averaging” method spheric opacities, relative to solar opacity, can be responsible of Winn et al. (2008; see also Shporer et al. 2008). After a for inflating the planetary radius. Since atmospheric opacities preliminary analysis we binned the residual light curves using are related to the host star metallicity, and HAT-P-9 metallicity bin sizes close to the duration of ingress and egress, which seems to be super-solar, it is likely that these two effects account is 27 min. The presence of correlated noise in the data was for the radius difference, as is the case for WASP-1b (Burrows quantified as the ratio between the standard deviation of the et al. 2007; see their Figure 7). Other mechanisms that may binned residual light curve and the expected standard deviation explain the increased radius involve downward transport and assuming pure white noise. This ratio is calculated separately dissipation of kinetic energy (Showman & Guillot 2002), and for each bin size and we defined β to be the largest ratio among layered convection (Chabrier & Baraffe 2007). the bin sizes we used. For each light curve we multiplied the According to its Teq and Safronov number, listed in Table 4, relative flux errors by β and repeated the analysis. The β value HAT-P-9b is likely a Class II planet, as defined by Hansen & of each light curve is listed in Table 2. Barman (2007). This classification is in accordance with its low We used the Markov Chain Monte Carlo algorithm (MCMC; density, since members of this class usually have smaller masses see, e.g., Holman et al. 2007) to derive the best-fit parameters and larger radii than those of Class I. Paper III – HAT-P-9b

1398 SHPORER ET AL. Vol. 690

Figure 3. Follow-up light curves of HAT-P-9. The four light curves are shown at the top of the figure, with the fitted model overplotted. The label next to each light curve gives the transit event number, observatory, telescope, and filter used. Residuals are presented at the bottom of the figure, with the same top to bottom order as the actual light curves. (A color version of this figure is available in the online journal.) BS [m/s]

Phase Figure 4. Line bisector spans folded on the orbital phase. Albeit one or two outliers, the bisector spans do not show a variation with the orbital phase.

The stellar incident ﬂux at the planet is fp = Fortney et al. (2008). Since it is positioned near the transition +0.3 9 −2 −1 region in f between these two classes of planets it may be used 1.3−0.310 erg cm s , which is at the lower end of the fp range p for pM planets, according to the pL/pM planet classiﬁcation of to further characterize them. Paper III – HAT-P-9b

No. 2, 2009 HAT-P-9b 1399

1.8 ρ = 0.3 ρ = 0.7 ρ = 1.5 1.6

1.4 ] J 1.2 [R p R 1

0.8 [ρ] = g cm

0.4 0.6 0.8 1 2 4 6 8 10 M [M ] p J 4

3.5

2.5 ] J

[M 2 p M 1.5

0.5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Period [days] Figure 5. Top panel: the radius–mass diagram for the known transiting planets, in log–log scale, with HAT-P-9b marked by an open circle. Constant density lines are overplotted. The diagram visually shows that the new planet has a relatively low density. The Neptune-mass planet GJ 436b, orbiting an M star, is positioned beyond the lower-left corner of this diagram. Bottom panel: the mass–period diagram for the known transiting planets, in linear scale, with HAT-P-9b marked by an open circle. Three planets are positioned outside the boundaries of this diagram: the long-period planet HD 17156b, and the massive planets HAT-P-2b (a.k.a. HD 147506b) and XO-3b. The planet with the lowest mass in this diagram is GJ 436b, orbiting an M star, and the one with the largest mass is CoRoT-Exo-2b. This figure is based on data taken from http://www.inscience.ch/transits/ on 2008 June 1. (A color version of this figure is available in the online journal.) In the mass–period diagram the new planet is positioned near ment of the Rossiter–Mclaughlin effect (e.g., Winn et al. 2005). XO-1b (McCullough et al. 2006; Holman et al.2006 ), OGLE- Equation (6) of Gaudi & Winn (2007) gives an expected am- TR-182b (Pont et al. 2008), OGLE-TR-111b (Pont et al. 2004; plitude for the effect of more than 100 m s−1, which is larger Winn et al. 2007a), and HAT-P-6b (Noyes et al. 2008). The than the orbital amplitude. However, since this is a relatively position of HAT-P-9b along with that of neighboring planets faint star, with V = 12.3 mag, such a measurement will be suggests that the mass–period relation (Mazeh et al. 2005; Gaudi challenging. et al. 2005), i.e., the decrease of planetary mass with increasing orbital period, levels off at periods 3 days. Operation of the HATNet project is funded by NASA The ephemeris obtained here is based on the four follow- grants NNG04GN74G and NNX08AF23G. These observations up light curves. We compared it to the mid-transit time of have been partially funded by the Optical Infrared Coordina- the HATNet light curve, obtained some 1000 days earlier, by tion network (OPTICON), a major international collaboration propagating the error on the period. The result shows a difference supported by the Research Infrastructures Programme of the of about 1.5σ between the mid-transit times of the follow-up European Commissions Sixth Framework Programme. T.M. ac- light curves and that of the HATNet light curve. Future follow- knowledges support from the Ministry of Science, Culture & up light curves will be able to study more thoroughly this Sport through a grant to encourage French–Israeli scientific col- possible discrepancy. laboration. G.B. is a National Science Foundation Fellow, under Finally, we note that due to the line-of-sight stellar rotation ve- grant AST-0702843. We acknowledge partial support from the locity of 11.9 ± 1.0kms−1 and the non-zero impact parameter Kepler Mission under NASA Cooperative Agreement NCC2- 0.52 ± 0.03, this planet is a good candidate for the measure- 1390 (DWL, PI). G.K. thanks the Hungarian Scientific Research Paper III – HAT-P-9b

1400 SHPORER ET AL. Vol. 690 Fund (OTKA) support through grant K-60750. G.T. acknowl- Hansen, B. M. S., & Barman, T. 2007, ApJ, 671, 861 edges partial support for this work from NASA Origins grant Holman, M. J., et al. 2006, ApJ, 652, 1715 Holman, M. J., et al. 2007, ApJ, 664, 1185 NNG04LG89b. Knutson, H. A., Charbonneau, D., Noyes, R. W., Brown, T. M., & Gilliland, R. L. 2007, ApJ, 655, 564 REFERENCES Kovacs,´ G., Bakos, G. A.,´ & Noyes, R. W. 2005, MNRAS, 356, 557 Kovacs,´ G., Zucker, S., & Mazeh, T. 2002, A&A, 391, 369 Anderson, D. R., et al. 2008, MNRAS, 387, L4 Latham, D. W. 1992, in ASP Conf. Ser. 32: IAU Colloq. 135: Complementary Bakos, G. A.,´ Laz´ ar,´ J., Papp, I., Sari,´ P., & Green, E. M. 2002, PASP, 114, 974 Approaches to Double and Multiple Star Research (San Francisco, CA: Bakos, G. A.,´ Noyes, R. W., Kovacs,´ G., Stanek, K. Z., Sasselov, D. D., & ASP), 110 Domsa, I. 2004, PASP, 116, 266 Mandel, K., & Agol, E. 2002, ApJ, 580, L171 Bakos, G. A.,´ et al. 2007, ApJ, 656, 552 Mazeh, T., Zucker, S., & Pont, F. 2005, MNRAS, 356, 955 Baraffe, I., Chabrier, G., & Barman, T. 2008, A&A, 482, 315 Mazeh, T., et al. 2000, ApJ, 532, L55 Barge, P., et al. 2008, A&A, 482, L17 McCullough, P. R., et al. 2006, ApJ, 648, 1228 Bouchy, F., & The Sophie Team 2006, in 10th Anniversary of 51 Peg-b: Status Noyes, R. W., et al. 2008, ApJ, 673, L79 of and Prospects for Hot Jupiter studies, ed. L. Arnold, F. Bouchy, & C. Pal,´ A., & Bakos, G. A.´ 2006, PASP, 118, 1474 Moutou (Paris: Frontier Group), 319 Pont, F., Bouchy, F., Queloz, D., Santos, N. C., Melo, C., Mayor, M., & Brosch, N., Polishook, D., Shporer, A., Kaspi, S., Berwald, A., & Manulis, I. Udry, S. 2004, A&A, 426, L15 2008, Ap&SS,, 314, 163 Pont, F., Zucker, S., & Queloz, D. 2006, MNRAS, 373, 231 Burrows, A., Budaj, J., & Hubeny, I. 2008, ApJ, 678, 1436 Pont, F., et al. 2008, A&A, 487, 749 Burrows, A., Hubeny, I., Budaj, J., & Hubbard, W. B. 2007, ApJ, 661, 502 Queloz, D., et al. 2001, A&A, 379, 279 Burstein, D., & Heiles, C. 1982, AJ, 87, 1165 Ram´ırez, I., & Melendez,´ J. 2005, ApJ, 626, 465 Cameron, A. C., et al. 2007, MNRAS, 375, 951 Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525 Casagrande, L., Portinari, L., & Flynn, C. 2006, MNRAS, 373, 13 Showman, A. P., & Guillot, T. 2002, A&A, 385, 166 Chabrier, G., & Baraffe, I. 2007, ApJ, 661, L81 Shporer, A., Mazeh, T., Winn, J. N., Holman, M. J., Latham, D. W., Pont, F., & Charbonneau, D., Winn, J. N., Everett, M. E., Latham, D. W., Holman, M. J., Esquerdo, G. A. 2008, ApJ, submitted (arXiv:0805.3915) Esquerdo, G. A., & O’Donovan, F. T. 2007, ApJ, 658, 1322 Shporer, A., Tamuz, O., Zucker, S., & Mazeh, T. 2007, MNRAS, 376, Claret, A. 2000, A&A, 363, 1081 1296 Claret, A. 2004, A&A, 428, 1001 Southworth, J., Wheatley, P. J., & Sams, G. 2007, MNRAS, 379, L11 Droege, T. F., Richmond, M. W., Sallman, M. P., & Creager, R. P. 2006, PASP, Sozzetti, A., et al. 2007, ApJ, 664, 1190 118, 1666 Torres, G., Konacki, M., Sasselov, D. D., & Jha, S. 2005, ApJ, 619, 558 Fortney, J. J. 2008, arXiv:0801.4943 Torres, G., Neuhauser,¨ R., & Guenther, E. W. 2002, AJ, 123, 1701 Fortney, J. J., Lodders, K., Marley, M. S., & Freedman, R. S. 2008, ApJ,, 678, Torres, G., Winn, J. N., & Holman, M. J. 2008, ApJ, 677, 1324 1419 Winn, J. N., Holman, M. J., & Fuentes, C. I. 2007a, AJ, 133, 11 Gaudi, B. S., Seager, S., & Mallen-Ornelas, G. 2005, ApJ, 623, 472 Winn, J. N., et al. 2007b, AJ, 134, 1707 Gaudi, B. S., & Winn, J. N. 2007, ApJ, 655, 550 Winn, J. N., et al. 2005, ApJ, 631, 1215 Guillot, T., Santos, N. C., Pont, F., Iro, N., Melo, C., & Ribas, I. 2006, A&A, Winn, J. N., et al. 2008, ApJ, 683, 1076 453, L21 Yi, S. K., et al. 2001, ApJS, 136, 417 Paper IV – WASP-1b Follow-Up

Mon. Not. R. Astron. Soc. 376, 1296–1300 (2007) doi:10.1111/j.1365-2966.2007.11537.x

Photometric follow-up of the transiting planet WASP-1b

A. Shporer,1⋆ O. Tamuz,1 S. Zucker2 and T. Mazeh1 1Wise Observatory, Raymond and Beverly Sackler Faculty of Exact Sciences, Tel Aviv University, Tel Aviv 69978, Israel 2Department of Geophysics & Planetary Sciences, Raymond and Beverly Sackler Faculty of Exact Sciences, Tel Aviv University, Tel Aviv 69978, Israel

Accepted 2007 January 22. Received 2007 January 16; in original form 2006 October 18

ABSTRACT We report on photometric follow-up of the recently discovered transiting planet WASP-1b. We observed two transits with the Wise Observatory 1-m telescope, and used a variant of the Eclipsing Binary Orbit Program (EBOP) code together with the Sys-Rem detrending approach to ﬁt the light curve. Assuming a stellar mass of 1.15 M⊙, we derived a planetary radius of Rp = 1.40 ± 0.06RJ and mass of Mp = 0.87 ± 0.07MJ . An uncertainty of 15 per cent in the stellar mass results in an additional systematic uncertainty of 5 per cent in the planetary radius and of 10 per cent in planetary mass. Our observations yielded a slightly better ephemeris for the centre of the transit: Tc [HJD] = (245 4013.3127 ± 0.0004) + Ntr(2.51996 ± 0.00002). The new planet is an inﬂated, low-density planet, similar to HAT-P-1b and HD 209458b. Key words: techniques: photometric – stars: individual: WASP-1 – planetary systems.

and no defocusing, using a Tektronix 1024 × 1024 pixel back- 1 INTRODUCTION illuminated CCD, with a pixel scale of 0.696 arcsec pixel−1 and an Wide-field small-aperture telescopes are currently used by a few 11.88 × 11.88 arcmin2 overall field of view (Kaspi et al. 1999). groups (e.g. Alonso et al. 2004; Bakos et al. 2004; McCullough On 2006 October 4 the exposure time was 60 s at the beginning of et al. 2005; Pollacco et al. 2006) to search for transiting planetary the transit, at airmass ∼2, decreasing to 45 s at lower airmass. Point candidates. However, light curves obtained using wide-field small- spread function (PSF) full width at half-maximum (FWHM) was aperture telescopes are usually not accurate enough to put useful about 2.9 arcsec that night. On 2006 October 9 exposure time varied constraints on the system parameters. Hence, photometric follow- between 20 and 45 s according to varying observing conditions, up using larger telescopes is essential. in particular the bright moon’s altitude. PSF FWHM was about The WASP1 consortium (Pollacco et al. 2006) has recently de- 2.1 arcsec that night. There was a short period of cloudiness on the tected two new transiting extrasolar planets, WASP-1b and WASP- October 9 night which prevented us from observing the egress of 2b (Collier Cameron et al. 2006). The discovery paper suggested that transit. Nevertheless, we were able to observe the star for a short that WASP-1b is probably an inflated, low-density planet, similar period after the transit, thus allowing calibration of the light curve to HAT-P-1b (Bakos et al. 2006) and HD 209458b (e.g. Knutson zero-point. et al. 2007). However, using the photometry of the SuperWASP We did not centre the field of view on the target star but instead small-aperture cameras and a single transit observed by a 35-cm positioned it at RA = 00h20m21s, Dec. =+32◦02′39′′ (J2000), in telescope, Collier Cameron et al. (2006) could not constrain the order to include a maximum number of comparison stars, which are system parameters of WASP-1 very well. In this work we have set essential for the transit light curve reduction. out to better constrain these parameters using the Wise Observatory 1-m telescope. We describe our observations in Section 2 and the data processing in Section 3. In Section 4 we briefly discuss our 3 DATA PROCESSING AND MODEL FITTING results. We used IRAF2 CCDPROC package for the bias subtraction and flat field correction, using calibration exposures taken nightly. Using the IRAF PHOT task we applied aperture photometry to all field stars 2 OBSERVATIONS in the reduced frames, with a few trial values for the aperture radius We observed two WASP-1b transits with the 1-m telescope at the and sky annulus size. For the 2006 October 4 frames, we obtained the Wise Observatory, on the nights of 2006 October 4 and 2006 October most satisfactory result with an aperture of 7 arcsec and an annulus 9. The observations were carried out in the I filter, with auto-guiding

2 IRAF (Image Reduction and Analysis Facility) is distributed by the National Optical Astronomy Observatories (NOAO), which are operated by the As- ⋆E-mail: [email protected] sociation of Universities for Research in Astronomy (AURA), Inc., under 1 http://www.superwasp.org cooperative agreement with the National Science Foundation.

C 2007 The Authors. Journal compilation C 2007 RAS Paper IV – WASP-1b Follow-Up

WASP-1 photometric follow-up 1297 inner and outer radii of 35 and 63 arcsec, respectively. For the 2006 Table 1. Table of photometric measurements. Only the first October 9 frames, an aperture radius of 6 arcsec and an annulus 10 measurements are listed here, the complete table is avail- inner and outer radii of 28 and 56 arcsec, respectively, yielded the able in the online version of this article in the Supplementary best result. Materials section. Following Winn, Holman & Roussanova (2006), we used nine HJD 245 4000 Rel. flux Err. reference stars and normalized their flux light curves to unit median. We combined these normalized light curves by a simple average and 13.18941 0.9971 0.0016 a3σ rejection, thus creating a normalized relative flux comparison 13.19324 0.9998 0.0016 light curve. We normalized the target star flux light curve by dividing 13.19432 0.9982 0.0016 it by the comparison light curve. Finally, we fitted a linear function 13.19537 1.0001 0.0016 to the out-of-transit measurements and divided the light curve by 13.19642 0.9997 0.0016 13.19749 1.0005 0.0016 this function. 13.19854 0.9979 0.0016 13.19960 0.9975 0.0016 3.1 Photometric parameters 13.20065 0.9987 0.0016 13.20171 0.9984 0.0016 We used the Eclipsing Binary Orbit Program (EBOP) code (Popper . . . & Etzel 1981) together with the Sys-Rem (Tamuz, Mazeh & Zucker . . . 2005; Mazeh, Tamuz & Zucker 2006) code in order to fit the transit light curve to our photometric measurements. EBOP is widely used for modelling eclipsing binary light curves, After fitting the above model to the data, the residual rms was and can be easily adapted to model transits (Gimenez 2006). It found to be 1.8 mmag in the first night and 2.4 mmag in the second. does not model proximity effects very well, but this is irrelevant for We therefore set the photometric errors of each measurement to be transits. EBOP consists of two modules: a light curve generator and a equal to the corresponding rms, and repeated the analysis. differential corrections module. Following Tamuz, Mazeh & North Table 1 lists all our Sys-Rem detrended measurements. Table 2 (2006), we used only the light curve generator and applied our own lists the best-fitting values for each of the two transits independently optimization program. and for both transits simultaneously. We estimated the errors of the Sys-Rem is an algorithm designed to remove systematic effects best-fitting values using Monte Carlo simulations. We also give the from photometric light curves without assuming any prior knowl- best-fitting values derived without using Sys-Rem, i.e. using only edge of the effects. These effects may result from varying observ- EBOP and no detrending. Accuracy of these parameters is up to seven ing conditions between observations, such as airmass and weather times better than that of Collier Cameron et al. (2006). We adopted conditions. Each systematic effect is a sequence of generalized ‘air- values derived by fitting both nights simultaneously and applying masses’ {aj ; j = 1, ... , M} assigned to each image, where the Sys-Rem. Using these parameters we estimate the orbital inclination index j refers to the image number and M is the number of images. to be 89.7◦ ± 1.8◦. Sys-Rem estimates both the effect and a set of coefficients, {ci ; i = Fig. 1 presents our light curves together with the fitted transit 1, ..., N}, which are generalized ‘colours’, assigned to each star i, model. Fig. 2 presents two χ 2 contour maps, from which we can where N is the number of stars. Essentially, Sys-Rem optimizes the learn about the relations among the best-fitting parameters. The left- effects and the coefficients such that subtracting the product c a 2 i j hand panel, presenting χ as a function of r∗ and b, shows the well- from rij, the jth measurement of star i, will minimize the rms of the known degeneracy of the impact parameter and the stellar radius, set {r − c a } (Tamuz et al. 2005). 2 ij i j while the right-hand panel, presenting χ as a function of r∗ and k, As a set of photometric measurements can include a few differ- shows that the two radii are practically uncorrelated. ent effects, in our analysis we first estimated six systematic effects We used our value of mid-transit time, Tc, together with the one for each night separately, using the nine reference stars. Then, we derived by Collier Cameron et al. (2006)3 to recalculate the orbital searched for a set of 12 generalized colour coefficients, six for each period, and obtained a value of 2.51996 ± 0.00002 d. There is no of the two nights, together with six transit parameters, giving a to- need to re-evaluate all the other parameters, as the errors of the tal of 18 parameters, that would best fit the target light curve. The best-fitting parameters are not dominated by the period error. six transit parameters included the mid-transit time Tc, the stellar Our new transit elements allowed us to estimate the transit total fractional radius r∗ = R∗/a, where a is the orbital semimajor axis, duration and duration of ingress and egress. Adopting equation (4) the ratio of planetary radius to stellar radius k = R /R∗, the impact p of Sackett (1999) for transit duration results in tT = 3.7 ± 0.2 h. As- parameter b = cos ia/(R∗ + Rp) and two magnitude zero-points for suming transit ingress and egress are symmetric, duration of ingress the two nights I1 and I2. is calculated by Transit light curves usually require two additional parameters 2 2 2 – the period and a limb-darkening coefficient. We did not opti- P (R∗ + Rp) − a cos(i) t = mize for the period and adopted the period published by Collier ingress π arcsin 2 a Cameron et al. (2006). Using the temperature and gravity given by 2 2 2 Collier Cameron et al. (2006), we adopted a value of u = 0.37 (Van (R∗ − Rp) − a cos(i) − arcsin , (1) Hamme 1993). When u is left as a free parameter, the best-fitting a value is u = 0.21 ± 0.08, which we considered an unphysical result. Given the four parameters Tc, r∗, k and b, all the other best-fitting 3 We used a mid-transit time of HJD = 245 3151.486 ± 0.006, from 2004, 14 parameters can be solved analytically. Thus we can perform a published in a preprint of Collier Cameron et al. (2006). This value was simple grid search over a four-dimensional space for the minimum changed in the final version of the discovery paper, which was published χ 2. only after we submitted this paper.

C 2007 The Authors. Journal compilation C 2007 RAS, MNRAS 376, 1296–1300 Paper IV – WASP-1b Follow-Up

1298 A. Shporer et al.

Table 2. Light curve ﬁtted parameters.

Without Sys-Rem With Sys-Rem

October 4 transit r∗ = R∗/a 0.180 ± 0.015 0.179 ± 0.019 k = rp/r∗ 0.101 ± 0.003 0.100 ± 0.003 b = cos ia/(R∗ + Rp) 0.24 ± 0.19 0.23 ± 0.20 Tc [HJD] 245 4013.3131 ± 0.0006 245 4013.3114 ± 0.0005 October 9 transit r∗ = R∗/a 0.187 ± 0.016 0.181 ± 0.015 k = rp/r∗ 0.107 ± 0.002 0.107 ± 0.002 b = cos ia/(R∗ + Rp) 0.36 ± 0.20 0.24 ± 0.19 Tc [HJD] 245 4013.3124 ± 0.0045 245 4013.3114 ± 0.0046 Both transits r∗ = R∗/a 0.186 ± 0.015 0.174 ± 0.007 k = rp/r∗ 0.104 ± 0.002 0.102 ± 0.001 b = cos ia/(R∗ + Rp) 0.34 ± 0.20 0.03 ± 0.17 Tc [HJD] 245 4013.3127 ± 0.0005 245 4013.3127 ± 0.0004

1.01 Oct 4 1.01 Oct 9 1.005 1.005

1 1

0.995 0.995

0.99 0.99

0.985 0.985

0.98 0.98 Relative Flux Relative RMS = 1.8 mmag Flux Relative RMS = 2.4 mmag 0.975 0.975

0.97 0.97

0.965 0.965

0.96 0.96 −0.06 −0.04 −0.02 0 0.02 0.04 0.06 −0.06 −0.04 −0.02 0 0.02 0.04 0.06 Phase Phase

Figure 1. Transit light curves of WASP-1, observed with the Wise Observatory 1-m telescope on the nights of 2006 October 4 (left) and 2006 October 9 (right). Relative ﬂux is plotted against orbital phase and the best-ﬁtting model is overplotted. Residuals, derived by subtracting the model from the measurements, are plotted at the bottom of each panel and their rms in each of the light curves is also given.

100 0.185 10 0.185 100 100 3

0.18 0.18 100 10 1 3 10 100 10 3 10 1 100 0.175 3 0.175 1 3 * 1 * r 3 r 1 10 100 3 0.17 10 100 0.17 10

100 0.165 100 0.165 100 100

0.16 0.16 0 0.1 0.2 0.3 0.4 0.09 0.095 0.1 0.105 0.11 b k

2 2 Figure 2. Contour plots of the χ function, plotted versus r∗ and b (left) and r∗ and k (right). The contour labels denote the difference between the χ of that contour and the minimal χ 2. which can be written also as P 3.2 Radial velocity elements 2 tingress = arcsin(r∗(1 + k) 1 − b ) 2π (2) Using the radial velocity measurements supplied by Collier 2 Cameron et al. (2006) and our improved photometric elements, we − arcsin( r∗(1 − k) 1 − b ) . were able to recalculate the radial velocity elements K1, the radial In our case tingress = 20.7 ±0.9 min. velocity amplitude and γ , the average radial velocity. The original

C 2007 The Authors. Journal compilation C 2007 RAS, MNRAS 376, 1296–1300 Paper IV – WASP-1b Follow-Up

WASP-1 photometric follow-up 1299

Table 3. System parameters derived from the ﬁtted parameters, assuming three different values for the stellar mass. We use the most likely value and 0.25 the upper and lower limits of Collier Cameron et al. (2006). ρ=0.3 0.2

M∗ aR∗ Rp Mp 0.15 ρ=0.7 (M ) (au) (R )(RJ )(MJ ) ⊙ ⊙ 0.1 ] ) J 1.06 0.037 1.38 ± 0.06 1.36 ± 0.06 0.80 ± 0.07 0.05 [R 1.15 0.038 1.42 ± 0.06 1.40 ± 0.06 0.87 ± 0.07 p ρ=1.5 1.39 0.041 1.51 ± 0.06 1.49 ± 0.06 1.05 ± 0.09 0 Log( R −0.05

−0.1

SuperWASP photometric measurements were obtained 2 yr before −0.15 the radial velocities, and therefore did not constrain usefully their −0.2 −3 [ρ] = g cm orbital phases. Using our ephemeris, we were able to determine −0.25 the phases of the radial velocities to a precision of about a minute. −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 Log( M [M ] ) Therefore, we have recalculated the orbital parameters using this p J −1 constraint. Our derived radial elements are K1 = 118 ± 10 m s Figure 3. Radius versus mass of the 14 known transiting extrasolar planets −1 and γ =−13.506 ± 0.008 km s . in log–log scale. WASP-1bis marked by a filled diamond. Data for this figure In order to derive the orbital separation, stellar and planetary radii were taken from http://exoplanet.eu/catalog-RV.php and the planetary mass, we used the stellar mass given by Collier Cameron et al. (2006). As there is a large uncertainty in the stellar mass, we repeated our calculation for the lower and upper limits planetary atmospheric metallicities, which increase atmospheric of the published mass range. The results are given in Table 3. For opacities, is the underlying mechanism responsible for inflating the most likely value of the stellar mass, the planetary radius and these extrasolar planets. This explanation does not require any ad- mass are Rp = 1.40 ± 0.06RJ , Mp = 0.87 ± 0.07MJ , and the stellar ditional heat source and is consistent with the increased probability radius is R∗ = 1.42 ± 0.06 R⊙. Since stellar and planetary radii of high metallicity stars to host planets (Fischer & Valenti 2005; 1/3 2/3 scale as M∗ and the planetary mass scales as M∗ , an uncertainty Santos et al. 2005). In addition, Burrows et al. (2006) also pointed of 15 per cent on the stellar mass results in an additional systematic out that the commonly used analysis of transiting light curves tend uncertainty of 5 per cent on the stellar and planetary radii and of 10 to derive radius larger than the one used in theory, which consider per cent in planetary mass. This systematic uncertainty should be a planet radius till the point where the optical depth in the planet’s added in quadrature to the errors in Table 3. atmosphere is τ = 2/3. Combined with all the currently available mass and radius of transiting extrasolar planets, our new values, plotted in Fig. 4, are 4 DISCUSSION consistent with the mass–period relation pointed out by Gaudi, Sea- We present here photometry of two transits of the planet WASP- ger & Mallen-Ornelas (2005) and by Mazeh, Zucker & Pont (2005). 1b, recently published by Collier Cameron et al. (2006). Our new The only outlier to this relation is HD 149026b (Santos et al. 2005) data better constrain the system parameters, mainly the stellar and which probably has a dense core (Fortney et al. 2006). We can gain planetary radii. Combined with previously published results, our a deeper understanding of this relation and its possible origin by new data provide a longer time-span, which we use in order to fix the phase of the radial velocity orbit. Our analysis confirms Collier 1.6 Cameron et al. (2006) suggestion that the new planet is an inflated, low-density planet. We put all our photometric measurements in the 1.4 public domain for any further study. In a simultaneous study, Charbonneau et al. (2006) conducted 1.2 follow-up observations of WASP-1 and WASP-2, and derived estimates for the stellar and planetary radii with the Mandel & Agol

] 1 (2002) formulae. Although being slightly larger, their derived radii J [M p

for the WASP-1 system, of R = 1.45 ± 0.03 R and R = 1.44 ± M ∗ ⊙ p 0.8 0.04 RJ , are consistent with ours. Fig. 3 presents the radii and masses of all currently known tran- 0.6 siting planets. The figure shows that WASP-1b radius is similar to HAT-P-1b (Bakos et al. 2006) and HD 209458b (e.g. Knutson et al. 2007) radii. This small but growing group of inflated extrasolar 0.4 planets, whose radii is larger than predicted by common theories of 0.2 planet formation and evolution (e.g. Laughlin et al. 2005), consists 1 1.5 2 2.5 3 3.5 4 4.5 a significant fraction of all currently known transiting planets. A Period [days] number of possible explanations were suggested to account for this Figure 4. Mass period relation for the 14 known transiting extrasolar plan- discrepancy between theory and observations by considering, for ex- ets. WASP-1b is marked by a filled diamond. The planet with the lowest ample, an internal heat source as the cause of these large radii (e.g. mass at the bottom of the figure is HD 149026b (Santos et al. 2005; Fortney Bodenheimer, Laughlin & Lin 2003; Winn & Holman 2005). How- et al. 2006). Data for this figure were taken from http://exoplanet.eu/catalog- ever, in a recent paper, Burrows et al. (2006) suggest that enhanced RV.php

C 2007 The Authors. Journal compilation C 2007 RAS, MNRAS 376, 1296–1300 Paper IV – WASP-1b Follow-Up

1300 A. Shporer et al. considering the energy diagram of Lecavelier des Etangs (2006, his Gimenez A., 2006, Ap&SS, 304, 21 fig. 1) presenting the surface potential energy of all extrasolar plan- Kaspi S., Ibbetson P. A., Mashal E., Brosch N., 1999, Wise Observatory ets versus the extreme ultraviolet (EUV) energy flux they receive Technical Report 95/6 from their host star. That diagram clearly shows a forbidden region, Knutson H., Charbonneau D., Noyes R. W., Brown T. M., Gilliland R. L., in which planets with masses too small evaporate because they ab- 2007, ApJ, 655, 564 Laughlin G., Wolf A., Vanmuster T., Bodenheimer P., Fischer D., Marcy G., sorb energy fluxes too large. Maybe the planets cannot populate the Butler P., Vogt S., 2005, ApJ, 621, 1072 left-hand bottom of our diagram because of evaporation. However, Lecavelier des Etangs A., 2006, A&A, 461, 1185 this scenario does not account for the paucity of transiting planets Mandel K., Agol E., 2002, ApJ, 580, L171 in the upper right of the diagram. Therefore, we have to find many Mazeh T., Zucker S., Pont F., 2005, MNRAS, 356, 955 more transiting planets, in order to verify the mass–period relation Mazeh T., Tamuz O., Zucker S., 2006, preprint (astro-ph/0612418) and understand its nature. McCullough P. R., Stys J. E., Valenti J. A., Fleming S. W., Janes K. A., Heasley J. N., 2005, PASP, 117, 783 Pollacco D. L. et al., 2006, PASP, 118, 1407 ACKNOWLEDGMENTS Popper D. M., Etzel P. B., 1981, AJ, 86, 102 Sackett P. D., 1999, in Mariotti J. M., Alloin D., eds, NATO ASIC Proc. 532, We would like to thank Efi Hoory for his dedicated work while Planets Outside the Solar System: Theory and Observations. Kluwer, observing WASP-1 on the night of 2006 October 9. We also wish Boston, p. 189 to thank Elia Leibowitz and Liliana Formiggini for allowing us to Santos N. C., Israelian G., Mayor M., Bento J. P., Almeida P. C., Sousa S. use their telescope time on the night of 2006 October 4. We thank G., Ecuvillon A., 2005, A&A, 437, 1127 the anonymous referee for his thorough reading of the paper and Tamuz O., Mazeh T., Zucker S., 2005, MNRAS, 356, 1466 his comments which allowed us to improve this paper. This work Tamuz O., Mazeh T., North P., 2006, MNRAS, 367, 1521 Van Hamme W., 1993, AJ, 106, 2096 was supported by the Israeli Science Foundation through Grant No. Winn J. N., Holman M. J., 2005, ApJ, 628, L159 03/323. This research has made use of NASA’s Astrophysics Data Winn J. N., Holman M. J., Roussanova A., 2006, ApJ, in press (astro- System Abstract Service and of the SIMBAD data base, operated at ph/0611404) CDS, Strasbourg, France.

SUPPLEMENTARYMATERIAL REFERENCES The following supplementary material is available for this article: Alonso R., Deeg H. J., Brown T. M., Belmonte J. A., 2004, Astron. Nachr., Table 1 Table of all 583 photometric measurements. 325, 594 This material is available as part of the online article Bakos G., Noyes R. W., Kov´acsG., Stanek K. Z., Sasselov D. D., Domsa I., 2004, PASP, 116, 266 from: http://www.blackwell-synergy.com/doi/abs/10.1111/j.1365- Bakos G. et al., 2006, ApJ, 656, 552 2966.2007.11537.x Bodenheimer P., Laughlin G., Lin D. N. C., 2003, ApJ, 592, 555 (this link will take you to the article abstract). Burrows A., Hubeny I., Budaj J., Hubbard W. B., 2006, ApJ, submitted Please note: Blackwell Publishing is not responsible for the content (astro-ph/0612703) or functionality of any supplementary materials supplied by the au- Charbonneau D. et al., 2006, ApJ, submitted (astro-ph/0610589) thors. Any queries (other than missing material) should be directed Collier Cameron A. et al., 2006, MNRAS, in press (astro-ph/0603688) Fischer D. A., Valenti J., 2005, ApJ, 622, 1102 to the corresponding author for the article. Fortney J. J., Saumon D., Marley M. S., Lodders K., Freedman R. S., 2006, ApJ, 642, 495 Gaudi B. S., Seager S., Mallen-Ornelas G., 2005, ApJ, 623, 472 This paper has been typeset from a TEX/LATEX ﬁle prepared by the author.

C 2007 The Authors. Journal compilation C 2007 RAS, MNRAS 376, 1296–1300 Paper V – HAT-P-2b Follow-Up

A&A 481, 529–533 (2008) Astronomy DOI: 10.1051/0004-6361:20078167 & c ESO 2008 Astrophysics

Reﬁned parameters and spectroscopic transit of the super-massive planet HD 147506b⋆

B. Loeillet1,2, A. Shporer3, F. Bouchy2, F. Pont4 , T. Mazeh3, J. L. Beuzit5 , I. Boisse2, X. Bonﬁls5 , R. Da Silva4, X. Delfosse5, M. Desort5 , A. Ecuvillon2, T. Forveille5, F. Galland5, A. Gallenne2 , G. Hébrard2 , A.-M. Lagrange5, C. Lovis4, M. Mayor4, C. Moutou1, F. Pepe4 , C. Perrier5, D. Queloz4 , D. Ségransan4, J. P. Sivan1 , N. C. Santos4 ,6, Y. Tsodikovich3,S.Udry4, and A. Vidal-Madjar2

1 Laboratoire d’Astrophysique de Marseille, BP 8, 13376 Marseille Cedex 12; Université de Provence, CNRS (UMR 6110) and CNES, France e-mail: [email protected] 2 Institut d’Astrophysique de Paris, UMR 7095 CNRS, Université Pierre et Marie Curie, 98bis Bd Arago, 75014 Paris, France 3 Wise Observatory, Tel Aviv University, 69978, Israel 4 Observatoire de Genève, Université de Genève, 51 Ch. des Maillettes, 1290 Sauverny, Switzerland 5 Laboratoire d’Astrophysique de Grenoble, Observatoire de Grenoble, UMR5571, Université J. Fourier et CNRS, BP 53, 38041 Grenoble, France 6 Centro de Astrofísica, Universidade do Porto, Rua das Estrelas, 4150-762 Porto, Portugal Received 26 June 2007 / Accepted 14 January 2008

ABSTRACT

In this paper, we report a reﬁned determination of the orbital parameters and the detection of the Rossiter-McLaughlin eﬀect of the recently discovered transiting exoplanet HD 147506b (HAT-P-2b). The large orbital eccentricity at the short orbital period of this exoplanet is unexpected and is distinguishing from other known transiting exoplanets. We performed high-precision radial velocity spectroscopic observations of HD 147506 (HAT-P-2) with the new spectrograph SOPHIE, mounted on the 1.93 m telescope at the Haute-Provence observatory (OHP). We obtained 63 new measurements, including 35 on May 14 and 20 on June 11, when the planet was transiting its parent star. The radial velocity (RV) anomaly observed illustrates that HAT-P-2b orbital motion is set in the same +12.2◦ direction as its parent star spin. The sky-projected angle between the normal of the orbital plane and the stellar spin axis, λ = 0.2−12.5 , +0.039 +0.040 is consistent with zero. The planetary and stellar radii were re-determined, yielding Rp = 0.951−0.053 RJup, Rs = 1.416−0.062 R⊙.The +0.39 +2.6 −3 mass (Mp = 8.62−0.55 MJup) and radius of HAT-P-2b indicate a density of 12.5−3.6 g cm , suggesting an object in between the known close-in planets with typical density of the order of 1 g cm−3, and the very low-mass stars, with density greater than 50 g cm−3. Key words. techniques: radial velocities – stars: individual: HD 147506 – stars: planetary systems

1. Introduction These properties suggest that HAT-P-2b might be an intermediate object between giant planets and low-mass stars, near the Thirty of the almost 270 known extrasolar planets have been brown dwarf population. Its density is indeed close to the upper 1 observed to transit their parent stars . This small subgroup of limit of the planetary models (Baraffe et al. 2003) and may put planets have the highest impact on our overall understanding of this object in the transition region between planets and brown close-in giant planets because we can estimate their mass and dwarfs. radius, and for some of them directly study their atmosphere. We report here new RV measurements of HD 147506 These transiting hot Jupiters have masses from 0.07 to about (HAT-P-2) obtained to provide additional information and con- 8 MJup and radii from 0.3 to about 1.4 RJup. This set of planets straints on this unusual planetary system. This was done by was recently extended to hot Neptune objects by the discoveryof 1) refining the orbital parameters, 2) refining the mass and ra- the transit of GJ436b (Gillon et al. 2007), and to super-massive dius of the companion and 3) measuring and modeling the planets by the detection of HD 147506b (HAT-P-2b) by Bakos Rossiter-McLaughlin effect (RV anomaly due to the partial et al. (2007) (hereafter B07). The discovery paper of the latter eclipse of the rotating stellar surface). We present here a more derived the key orbital and physical parameters of this excep- precise orbital solution and the measurement of the sky projec- tional object, which differs by its mass (9.04 MJup), orbital period tion of the inclination between the stellar spin axis and the nor- (5.63 days), and eccentricity (0.52) from other transiting plan- mal of the orbital plane. ets. The radius of HD 147506b implies an uncommon measured mean density (11.9 g cm−3) and surface gravity (227 ms−2). 2. Observations

⋆ Based on observations collected with the SOPHIE spectrograph on The parent star of HAT-P-2b, HD 147506, was observed in May the 1.93-m telescope at OHP, France (programs 07A.PNP.MAZE and and June 2007 with the new spectrograph SOPHIE (Bouchy 07A.PNP.CONS). & the Sophie team 2006) mounted on the 1.93-m telescope of 1 See http://obswww.unige.ch/~pont/TRANSITS.htm Haute Provence Observatory. Observations were conducted in

Article published by EDP Sciences Paper V – HAT-P-2b Follow-Up

530 B. Loeillet et al.: Refined parameters and spectroscopic transit of HD 147506b the high-efficiency mode (HE mode), which provides a spectral Table 1. SOPHIE radial velocities of HD 147506 (HAT-P-2). resolution of R ∼ 39000. The Thorium-Argon lamp was used to calibrate the wavelength scale. The simultaneous ThAr cali- BJD RV Photon-noise Signal-to-noise ratio bration mode was not used, as wavelength calibration was per- −2 400 000 [km s−1] uncertainties per pixel formed less than 1 h before and after the observations leading to [kms−1] atλ = 5500 Å an instrumental stability of less than 2 m s−1. 54 227.5016 −19.4014 0.0088 109 We obtained eight out-of-transit spectra on six nights be- 54 227.6000 −19.4082 0.0065 146 tween May 6 and 15, 2007, and a sequence of 6 h, comprised 54 228.5842 −19.5581 0.0188 54 of 35 spectra during and after the transit on May 14 (JD = 54 229.5993 −20.1874 0.0161 61 − 2454235). Unfortunately, we could not observe the ingress 54 230.4475 21.2249 0.0141 68 54 230.6029 −20.8536 0.0148 66 phase of the transit that occurredjust before twilight. Another se- 54 231.5987 −19.5311 0.0121 78 quence was obtained during transit on June 11 (JD = 2454263) 54 235.3466 −20.1916 0.0156 61 to get a full coverage of the transit. The typical exposure time 54 235.3538 −20.2318 0.0180 53 was 10 min, long enough to reach a sufficiently high signal-to- 54 235.3615 −20.3008 0.0167 57 noise ratio (SNR) and short enough to adequately sample the 54 235.3692 −20.2790 0.0173 55 observed transit. The average SNR per pixel, at λ = 5500 Å, was 54 235.3765 −20.3083 0.0172 56 about 70 during the first observed transit and about 90 during 54 235.3866 −20.3889 0.0209 46 the second one. Another limitation of high-precision RV mea- 54 235.3938 −20.4280 0.0171 56 54 235.4011 −20.4370 0.0173 55 surement in the mode used (HE) concerns the guidance centring 54 235.4088 −20.4450 0.0163 58 of the targetin the fiber. As the diameterof the fiber is quite large 54 235.4161 −20.4530 0.0180 53 (3 arcsec), a strong decentringunder very good seeing conditions 54 235.4234 −20.4987 0.0150 62 −1 (<1.5 arcsec) could indeed induce a RV shift of a few tens ms . 54 235.4310 −20.5197 0.0144 65 We determined RVs using a weighted cross-correlation 54 235.4383 −20.5032 0.0157 60 method, following the procedure of Baranne et al. (1996) and 54 235.4456 −20.5318 0.0178 53 Pepe et al. (2005), with a numerical mask constructed from the 54 235.4535 −20.5340 0.0182 52 solar spectrum atlas corresponding to a G2 dwarf star. This stan- 54 235.4608 −20.4990 0.0143 66 dard mask is well adaptedto the F8-typespectrum of the primary 54 235.4681 −20.5047 0.0124 75 star. We estimated the measurement uncertainties based on the 54 235.4759 −20.4747 0.0117 81 54 235.4831 −20.4974 0.0113 82 photon-noise empirical relation detailed by Bouchy et al. (2005) 54 235.4904 −20.5364 0.0111 84 and Cameron et al. (2007). For the spectra obtained during the 54 235.4981 −20.5384 0.0111 85 night of the first and the second transit, the typical photon-noise 54 235.5054 −20.5484 0.0108 87 −1 uncertainty is 14 and 10 ms , respectively, whereas for the 54 235.5126 −20.5785 0.0135 70 other 8 spectra this uncertainty ranges from 6 to 28 m s−1. The 54 235.5204 −20.6029 0.0142 66 journal of SOPHIE observations, including barycentric Julian 54 235.5277 −20.6066 0.0153 62 dates (BJD), RVs, photon-noiseuncertainties, and SNR per pixel 54 235.5350 −20.5946 0.0115 81 is given in Table 1. The phase-folded RVs are plotted in Fig. 1. 54 235.5434 −20.6155 0.0118 80 As illustrated by the higher SNR values for the data acquired 54 235.5507 −20.5987 0.0113 82 during the second nightof observations,we have strong evidence 54 235.5580 −20.6266 0.0110 85 54 235.5682 −20.6683 0.0119 80 that the observational conditions were very good. As explained 54 235.5755 −20.6712 0.0123 77 above, the data are thus affected by guiding noise and we as- − −1 54 235.5827 20.6585 0.0115 82 sumed an additional systematic error of 50 ms . We estimated 54 235.5905 −20.6819 0.0108 87 this systematic error thanks to later instrumental tests under the 54 235.5978 −20.7002 0.0128 74 same conditions. 54 235.6051 −20.7205 0.0117 81 54 236.5190 −20.2207 0.0056 99 54 263.4521 −20.2050 0.0157 71 3. Stellar properties 54 263.4594 −20.1880 0.0100 102 As described in B07, the spectroscopic determination of the 54 263.4666 −20.2048 0.0089 113 − radius of HD 147506 is very sensitive to the method used, 54 263.4739 20.1887 0.0092 110 54 263.4804 −20.1626 0.0114 90 as well as to the log g determination, which may be affected 54 263.4860 −20.1750 0.0120 86 by the uncertainty of the spectrum continuum and the large 54 263.4915 −20.1832 0.0128 81 projected rotational velocity, v sin Is of the star. The spectro- 54 263.4971 −20.1429 0.0096 104 scopic approach described in B07 provides a stellar radius of 54 263.5030 −20.1565 0.0102 99 +0.062 − 1.474−0.167 R⊙. Our combined SOPHIE spectrum provides an in- 54 263.5086 20.1615 0.0108 95 dependent spectroscopic determination of the stellar parameters 54 263.5141 −20.1610 0.0104 97 +0.040 54 263.5236 −20.1810 0.0103 99 (RS = 1.416−0.062 R⊙), consistent with B07. From the full width half maximum (FWHM) of the averaged 54 263.5291 −20.1955 0.0151 72 cross-correlation functions (CCF) of SOPHIE spectra, which 54 263.5347 −20.3046 0.0179 63 − were calibrated to yield stellar v sin I values (Santos et al. 2002), 54 263.5411 20.3391 0.0277 48 s 54 263.5561 −20.3457 0.0127 85 we determined the v sin Is of HD 147506 to be equal to 21.3 ± − −1 54 263.5634 20.4009 0.0133 82 1.3 kms , somewhat slightly larger than the velocity derived 54 263.5706 −20.4169 0.0173 67 −1 by B07 (19.8 ± 1.6 kms ). The metallicity index [Fe/H] = 54 263.5779 −20.4377 0.0107 99 0.11 ± 0.10 we obtained, using the method described in Santos 54 263.5852 −20.4763 0.0091 113 et al. (2002) is in full agreement with B07. The activity in- ′ dex of log RHK = −4.75 ± 0.02 was derived from the H and Paper V – HAT-P-2b Follow-Up

B. Loeillet et al.: Reﬁned parameters and spectroscopic transit of HD 147506b 531

observations. A linear limb-darkening coefficient ǫ = 0.71 was also used, considering the stellar Teff = 6250 K (Claret 2004). Hence, our model has 13 free parameters, where 8 are non-linear (T0, e, ω,R p/Rs, a/Rs, λ, v sin Is and i) and 5 linear (K, V0, ∆vKL, ∆vSK and ∆vS2). We used a Keplerian model for the orbit, and the analytic approach described by Ohta et al. (2005), to model the RM effect. Equations given by Ohta et al. (2005) for the RM RV anomaly were modified to make them dependent on Rp/Rs and a/Rs,instead of Rp, Rs and a. We searched the parameter space for the global minimum χ2 position, using the equation below. We used the linear least square method, along with the data, to calculate the linear parameters at each position in the parameter space of the non-linear parameters. We modified our χ2 function to account for external information, namely the line-of-sight stellar rotational velocity, derived here from the spectra, the radii ratio, and transit duration, from B07:

Fig. 1. Top: phase-folded Radial Velocity measurements of HAT-P-2 2 2 superimposed on the reﬁned Keplerian orbital solution. Open circles 2 RVo,i − RVc,i v sin Is − 21.3 χ =Σi + refer to the Keck and Lick measurements. Filled circles refer to the RVerr,i 1.3 SOPHIE measurements. The inset shows a zoom around the transit 2 2 Rp/Rs − 0.0684 T − 0.177 where radial velocities exhibit the Rossiter-McLaughlin eﬀect. Bottom: + + dur , residuals around the orbital solution. 0.0009 0.002

where RVo,i and RVc,i are the ith observed and calculated RVs K CaII lines and appears to be close to the value determined by and RV is its error. T is the transit duration and is related ′ = − ± err,i dur B07 (log RHK 4.72 0.05). As estimated in Santos et al. to a/R , R /R , i and the orbital parameters P, e,andω. The un- (2000), our activity index for a F8 dwarf star implies a stellar jit- s p s −1 −1 certainties were computed directly from the linear least squares terfromafewms toabout20ms , which is confirmedby the analysis for the linear parameters (Press et al. 1992, Sect. 15.4, calibration made by Wright (2005). Analysis of the line-bisector Eq. (15.4.15)). The procedure is a bit more complex concern- computed for all out-of-transit spectra does not show significant ing the non-linear parameters. For each specific parameter, we variations nor correlation with the RVs. increase and decrease it using small steps starting from the minimum χ2 solution value. At each step, the rest of the parameters 4. Determination of the planetary system are fitted while holding the specific parameter constant. Then, a 4th degree polynomial is fitted to the χ2 values obtained and the parameters 1-σ uncertainty is estimated by identifying the values of the fit In this section, we first describe the procedure used to fit the corresponding to the minimum χ2 value +1. observed velocities with the Keplerian orbit and the RM RV In the fit procedure, we adopted the B07 stellar jitter anomaly, and how we estimated the uncertainties. In the second of 60 ms−1 for the Lick and Keck measurements. For the and third subsections, we discuss our results. SOPHIE measurements, we adopted a jitter of 17 ms−1, resulting in a reduced χ2 = 1 for the out-of-transit SOPHIE measurements. This value is in a good agreement with the value esti- 4.1. Analysis of the RV data mated from our revised activity index and should be compared −1 We used all available high-precision spectroscopic data to model to the jitter of 60 m s mentioned in B07, considering the span the orbit and the RM effect simultaneously. This data includes covered by Lick and Keck observations was 240 nights. Over 10 Lick spectra and 13 Keck spectra obtained by B07, and the such a long span, one cannot exclude variations in the stellar ac- 63 SOPHIE spectra. Our model is comprised of 15 parameters: tivity level. We also note that the effect of this jitter may have a time scale comparable or longer than the time of a single expo- The period, P; periastron passage time, T0; orbital eccentricity, e; angle between ascending node and periastron, ω; RV semi- sure, inducing a correlated noise effect. amplitude, K; RV zero point, V0 (these first six are the classical orbital parameters); planetary to stellar radii ratio, Rp/Rs;or- 4.2. Results of the fitted orbital solution bital semi-major axis to stellar radius ratio, a/Rs; angle between sky projection of the orbital angular momentum axis and stel- Table 2 lists the result for the fitted parameters of our model. lar spin axis, λ; line of sight stellar rotational velocity, v sin Is; The solution depends on the RV data and the prior constraints orbital inclination angle, i; and the stellar linear limb darken- (v sin Is, Rp/Rs, transit duration), derived from the spectroscopic ing coefficient, ǫ. We have also determined a velocity shift be- analysis in this work and the photometry from B07. The re- tween Keck and Lick velocity zero points, ∆vKL, and SOPHIE fined orbital solution is plotted in the upper panel of Figs. 1, and Keck zero points, ∆vSK. We also estimated a velocity shift, and 2 presents the RM RV anomaly model after subtraction of ∆vS2, between the SOPHIE measurements taken on the second the Keplerian orbit. transit night (June 11) and the rest of the SOPHIE measure- The rms of the out-of-transit residuals of SOPHIE measurements, taken about a month earlier. The period was fixed on ments, spanning 36 days, is equal to 18 ms−1, whereas rms of the value given by B07 (P = 5.63341 days), considering its all out-of-transit measurements, spanning 282 days, is 50 m s−1. very high accuracy (11s) derived from extensive photometric The orbital parameters we derived are consistent with B07. Paper V – HAT-P-2b Follow-Up

532 B. Loeillet et al.: Reﬁned parameters and spectroscopic transit of HD 147506b

Table 2. Refined parameters of the Keplerian orbital solution and the parameters of the HAT-P-2 system, compared to those determined in (Bakos et al. 2007, B07). T0 refers to the periastron time. The fitted parameters are presented at the top of the table. The bottom part of the table lists the system parameters, derived from the fitted ones. We present in the table both the spectroscopic determination and the result from the fit of the star’s v sin Is.

Parameter Value Value from B07 Period [d] ⋆ 5.63341 (fixed) 5.63341±0.00013 +0.0053 T0 [HJD] 2 454 213.4794 −0.0030 2 454 213.369 ± 0.041 +0.0025 e 0.5163 −0.0023 0.520 ± 0.010 +1.06 ω [deg] 189.92 −1.20 179.3 ± 3.6 K [ms−1] 966.9 ± 8.3 1011 ± 38 −1 V0 [ms ] –19 855.1 ± 5.8 N.A. +0.00090 Rp/Rs 0.06891 −0.00086 0.0684 ± 0.0009 +0.12 +1.10 a/Rs 10.28 −0.19 9.77−0.02 +12.2 λ [deg] 0.2 −12.5 N.A. v sin I [kms−1] 22.9 +1.1 N.A. s RM −1.2 Fig. 2. Top: radial velocities of HAT-P-2 as a function of the orbital v sin I [kms−1] 21.3 ± 1.3 19.8 ± 1.6 s Spectro phase after subtraction of the Keplerian model and superimposed with i [deg] 90.0 +0.85 >84.6(2σ) −0.93 the best fit of the Rossiter-McLaughlin effect. The filled and open cir- −1 ∆vKL [ms ] –328 ± 41 –380 ± 35 cles represent the RV measurements obtained during the first and the −1 ∆vSK [ms ] –19584 ± 17 N.A. second sequence of observations of the transit with SOPHIE, respec- −1 ∆vS2 [ms ] –27 ± 12 N.A. tively. The open circle with dotted error bars represent one measure- Stellar jitter [ms−1] ⋆⋆ 17 60 ment from the Keck set which is a few minutes before the ingress [Fe/H] (dex) 0.11 ± 0.10 0.12 ± 0.08 phase. Bottom: RV residuals after subtracting the orbital solution and ′ the modeled RM effect. log RHK –4.75 ± 0.02 –4.72 ± 0.05 ⋆ +0.062 +0.062 Ms [M⊙] 1.298 −0.098 1.298−0.098 +0.0011 a [AU] 0.0677 −0.0017 0.0677 ± 0.0014 +0.04 +0.042 Rs [R⊙] 1.416 −0.062 1.474−0.167 +0.39 Mp [MJup] 8.62−0.55 9.04 ± 0.50 R [R ] 0.951 +0.039 0.982+0.038 p Jup −0.053 −0.105 larger velocities by approximately 10%. Moreover, as the planet ρ −3 +2.6 . +4.8 p [g cm ] 12.5 −3.6 11 9−1.6 crosses the star at the equatorial plane, assuming a differential g [m s−2] 237 +30 227+44 p −41 −16 rotational velocity of the star, the fitted v sin Is corresponds to the maximum value. Küker & Rüdiger (2005) show that a differen- ∗ Adopted from B07; ∗∗ short term jitter. tial rotationalrate can be as high as about10% for a F8-type star. The projected angle between the stellar spin axis and the normal of the orbital plane, λ = 0.2 ± 12.5◦, is consistent with complete However our uncertainties are smaller, due to a larger sample alignment of the stellar spin and orbital angular momentum. of high-precision RV data. We extended our analysis in the search for the global min- We searched for a second planetary signal in the RV residuals 2 imum χ , without any constraint on the v sin Is. The result ob- to look for hints of a third body in the system. No clear periodic tained gives a significantly larger v sin I value (29.5 +3.1 kms−1) signal appears in the RV residuals. We estimated that we can s −2.2 and a consistent lambda value (5.0+17.8◦). However, such a exclude the presence of a second planet of mass greater than 1.3, −6.0 v sin Is value is clearly incompatible with the spectroscopic de- 1.5 and 1.8 MJup for an orbital period shorter than 50, 100, and 200 days respectively. However, the increased RMS for all out- termination of the rotational velocity of the star. of-transit residuals suggests that a long-term RV follow-up of We also computed the line-bisector behavior following the this star is needed. procedure described in Santos et al. (2002). As shown in Fig. 3, a specific signature of the line-bisectors can be found during the transit and is anti-correlated with the RVs due to the fact that the 4.3. Measurement of the Rossiter-McLaughlin effect crossing planet mainly affects the bottom of the spectral lines. +0.062 As an important result, the sign of the RV anomaly shows that Assuming the B07 stellar mass and period, of 1.298−0.098 M⊙ the orbital motion is set in the same direction as the stellar and 5.63341 days, and using the fitted ratios Rp/Rs and a/Rs, spin, similar to the four previous observed RM effects on tran- we also provide a new determination of the system parameters. siting exoplanets (Queloz et al. 2000; Winn et al. 2005, for Those include: orbital semi-major axis, a, stellar radius, Rs,and HD 209458; Winn et al. 2006, for HD 189733; Wolf et al. 2007, the planetary radius, Rp, mass, Mp, density ρp, and surface grav- for HD 149026; and Narita et al. 2007, for TrES-1). The or- ity gp. Our results are similar to those of B07 with smaller errors. bital inclination angle i we derived is in full agreement with the The RVs describing the spectroscopic transit are quite noisy and value determined by B07. The projected rotation velocity of the some of the measured RVs (around orbital phase 0.885) present −1 star v sin Is determined from the RM fit is 22.9 ± 1.2 kms . an unexpected RV shift. This shift does not seem to be due to an This value is greater than our spectroscopic determination from instrumental deviation and may be explained by a guidance de- SOPHIE CCFs, and 2-σ greater than the determination of B07. centring of the telescope. However, the amplitude of the Rossiter However, Winn et al. (2005) showed that v sin Is measured with anomaly is large enough that we can still estimate the parameters the analytical formulae from Ohta et al. (2005) is biased toward of this system with quite good uncertainties. Paper V – HAT-P-2b Follow-Up

B. Loeillet et al.: Reﬁned parameters and spectroscopic transit of HD 147506b 533

velocity of the star is only about 21 kms−1, the star is deﬁnitely not pseudo-synchronized to the planetary orbit. Peale (1999)in- dicated that the alignment and the synchronizationtimescales are of the same order of magnitude. Therefore, the lack of pseudo- synchronization indicates that the system was formed with the stellar spin aligned to the orbital angular momentum.

Acknowledgements. Part of these observations have been funded by the Optical Infrared Coordination network (OPTICON), a major international collaboration supported by the Research Infrastructures Programme of the European Commission’s Sixth Framework Programme. N.C.S. would like to thank the support from Fundao para a Cincia e a Tecnologia, Portugal, in the form of a grant (reference POCI/CTE-AST/56453/2004). This work was supported in part Fig. 3. Line-bisectors signature during the spectroscopic transit as a by the EC’s FP6 and by FCT (with POCI2010 and FEDER funds), within the function of the orbital phase. As in Fig. 2 the ﬁlled and open circles HELAS international collaboration. A.E. would like to thank the support from represent the RV measurements obtained during the ﬁrst and second the Swiss National Science Foundation (SNSF), Switzerland, in the form of a sequences of transit observations with SOPHIE, respectively. grant (reference PBSK2–114688). We thank the technical team from OHP who worked on the instrument SOPHIE and for their exceptional work.

5. Discussion and conclusion References The value of HAT-P-2b radius (0.95 RJup) puts this object in Bakos, G. A., Kovacs, G., Torres, G., et al. 2007, ArXiv e-prints, 705 the mass-radius diagram as an intermediate case between Hot- ff −3 Bara e, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt, P. H. 2003, Jupiters and low-mass stars. Its mean density of 12.5 g cm A&A, 402, 701 is in between the Hot-Jupiter density (0.34−1.34 g cm−3) and Baranne, A., Queloz, D., Mayor, M., et al. 1996, A&AS, 119, 373 the density of the smallest transiting M dwarfs OGLE-TR-122 Bouchy, F., Pont, F., Melo, C., et al. 2005, A&A, 431, 1105 and OGLE-TR-123 (Pont et al. 2005, 2006), which are 75 and Bouchy, F., & The Sophie Team 2006, in Tenth Anniversary of 51 Peg-b: Status −3 of and prospects for hot Jupiter studies, ed. L. Arnold, F. Bouchy, & C. 51 gcm , respectively. A similar paper (Winn et al. 2007) pub- Moutou, 319 lished results that are consistent with our conclusions. Super- Cameron, A. C., Bouchy, F., Hébrard, G., et al. 2007, MNRAS, 375, 951 massive planets like HAT-P-2b may constitute a new class of Claret, A. 2004, A&A, 428, 1001 stellar companions, in between Hot-Jupiters and low-mass stars Gillon, M., Pont, F., Demory, B., et al. 2007, ArXiv e-prints, 705 Hut, P. 1981, A&A, 99, 126 and near the Brown dwarf population. HAT-P-2b is the first Ida, S., & Lin, D. N. C. 2005, ApJ, 626, 1045 super-massive object around a F8 star for which the exact mass Küker, M., & Rüdiger, G. 2005, A&A, 433, 1023 has been determined. Such a massive close-in planet is not in Narita, N., Enya, K., Sato, B., et al. 2007, ArXiv Astrophysics e-prints agreement with the type II migration mechanism, which appears Ohta, Y., Taruya, A., & Suto, Y. 2005, ApJ, 622, 1118 to bemoreefficient for planets around low-mass stars (Ida & Lin Peale, S. J. 1999, ARA&A, 37, 533 Pepe, F., Mayor, M., Queloz, D., et al. 2005, The Messenger, 120, 22 2005). This could suggest a different formation process for this Pont, F., Bouchy, F., Melo, C., et al. 2005, A&A, 438, 1123 object, such as fragmentation, or interactions between the planet Pont, F., Moutou, C., Bouchy, F., et al. 2006, A&A, 447, 1035 and another companion or between the planet and the disk in the Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992, evolution process. Numerical recipes in C. The art of scientific computing, 2nd Ed. (Cambridge: University Press) A system with an elliptic orbit is expected to move toward Queloz, D., Eggenberger, A., Mayor, M., et al. 2000, A&A, 359, L13 pseudo-synchronization, with the stellar angular rotation veloc- Santos, N. C., Mayor, M., Naef, D., et al. 2000, A&A, 361, 265 ity tuned to near the angular velocity of the companion at peri- Santos, N. C., Mayor, M., Naef, D., et al. 2002, A&A, 392, 215 astron passage (Zahn 1977). Given the Keplerian orbital param- Winn, J. N., Noyes, R. W., Holman, M. J., et al. 2005, ApJ, 631, 1215 eters, we computed the angular planetary speed at the periastron Winn, J. N., Johnson, J. A., Marcy, G. W., et al. 2006, ApJ, 653, L69 Winn, J. N., Johnson, J. A., Peek, K. M. G., et al. 2007, ApJ, 665, L167 position and found that a stellar radius of 1.42 R⊙ implies a Wolf, A. S., Laughlin, G., Henry, G. W., et al. 2007, ApJ, 667, 549 pseudo-synchronization rotational velocity v sin Is of about Wright, J. T. 2005, PASP, 117, 657 40 kms−1 (Hut 1981, Eq. (43)). Because the observed rotational Zahn, J.-P. 1977, A&A, 57, 383 Paper VI – GJ 436b Follow-Up

PHOTOMETRIC FOLLOW-UP OBSERVATIONS OF THE TRANSITING NEPTUNE-MASS PLANET GJ 436b Avi Shporer1, Tsevi Mazeh1, Frederic Pont2, Joshua N. Winn3, Matthew J. Holman4, David W. Latham4, and Gilbert A. Esquerdo4,5 1 Wise Observatory, Raymond and Beverly Sackler Faculty of Exact Sciences, Tel Aviv University, Tel Aviv 69978, Israel 2 School of Physics, University of Exeter, Stocker Road, Exeter EX4 4QL, United Kingdom 3 Department of Physics, and Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, Cambridge, MA, 02139, USA 4 Harvard-Smithsonian Center for Astrophysics, 60, Garden Street, Cambridge, MA, 02138, USA 5 Planetary Science Institute, 620 N. 6th Avenue, Tucson, Arizona 85705, USA Received 2008 May 24; accepted 2009 January 1; published 2009 March 25

ABSTRACT This paper presents multiband photometric follow-up observations of the Neptune-mass transiting planet GJ 436b, consisting of ﬁve new ground-based transit light curves obtained in 2007 May. Together with one already published light curve, we have at hand a total of six light curves, spanning 29 days. The analysis of the data yields an orbital period P = 2.64386 ± 0.00003 days, midtransit time Tc [HJD] = 2454235.8355 ± 0.0001, planet mass Mp = 23.1 ± 0.9 M⊕ = 0.073 ± 0.003 MJup, planet radius Rp = 4.2 ± 0.2 R⊕ = 0.37 ± 0.01 RJup, and stellar radius Rs = 0.45 ± 0.02 R⊙. Our typical precision for the midtransit timing for each transit is about 30 s. We searched the data for a possible signature of a second planet in the system through transit timing variations (TTV) and variation of the impact parameter. The analysis could not rule out a small, of the order of a minute, TTV and a long-term modulation of the impact parameter, of the order of +0.2 yr− 1. Key words: stars: individual (GJ 436) – planetary systems Online-only material: color ﬁgures, machine-readable table

1. INTRODUCTION Butler et al. (2004) obtained photometric observations of GJ 436 at the expected time of possible transits and concluded Of the almost 300 extrasolar planets discovered to date6,the that a transit was unlikely. Nevertheless, the transiting nature ∼35 transiting planets are the only ones allowing a measurement of GJ 436b was recently discovered by Gillon et al. (2007a) of their mass and radius and the study of their atmospheres (e.g., who measured a planetary radius of ≃ 4 R⊕ and mass of Charbonneau et al. 2007; Fortney 2008; Guillot 2008; Pont et al. 22.6 M⊕. Soon after the discovery, a transit and secondary 2008). Of those, GJ 436b serves as a unique opportunity to study eclipse events were observed by Spitzer Space Telescope at a planet with mass and radius as small as Neptune. Although 8 µm (Gillon et al. 2007b; Deming et al. 2007; Demory et al. several planets have already been detected with minimum 2007), further constraining the system parameters and reducing masses similar to the mass of GJ 436b, for example GJ 581b, the uncertainties by a factor of about 10 for the planet-to-star c, and d, (Bonfils et al. 2005; Udry et al.2007 ), HD 4308b radius ratio and midtransit timing. Those observations resulted (Udry et al. 2006), and Cnc 55e (Fischer et al. 2008), GJ 436b in a somewhat increased planetary radius of 4.2 ± 0.2 R⊕. is the only transiting Neptune-mass planet discovered so far. Torres (2007) was able to refine further the planet parameters by Moreover, it is the only known transiting planet orbiting an deriving more accurate constraints on the host star and obtained M-type star, presenting an opportunity to help constrain stellar a slightly larger planetary mass, 23.17 ± 0.79 M⊕, and a more +0.09 models in this mass range. precise planetary radius of 4.22−0.10 R⊕. GJ 436b was initially discovered by Butler et al. (2004) Recently, Ribas et al. (2008) restudied the system and through a radial velocity (RV) modulation of its host star. suggested that the observed radial velocities of the system are Maness et al. (2007) have refined its orbital elements, specif- consistent with an additional small, super-Earth planet in a close ically the eccentricity, e = 0.160 ± 0.019, and identified a orbit around GJ 436 in a 1:2 mean-motion resonance with the linear velocity trend of 1.36 ± 0.4ms−1 yr−1. Those authors known planet. Similar to Maness et al. (2007), they suggested suggested that the velocity trend might result from the pres- that this planet could pump eccentricity into the orbit. Ribas ence of a long-period second planet in the system, a planet that et al. (2008) further suggested that such an additional planet can could induce a periodic modulation of the orbital eccentricity of induce a precession of the orbital plane of GJ 436b, if the orbital GJ 436b. This type of effect was suggested already for triple planes of motion of the two planets are inclined relative to each stellar systems (Mazeh & Shaham 1979, see also Mazeh 1990), other. Such a precession should change the inclination relative and for a planet in a binary system for 16 Cygni B (Mazeh et al. to our line of sight, and therefore might explain why Butler et al. 1997; Holman et al.1997 ). Maness et al. (2007) pointed out that (2004) failed to observe a transit in 2004. this interpretation of the eccentricity looks especially attractive Immediately after the detection of the transits of GJ 436b if the circularization timescale of GJ 436b is shorter than the age by Gillon et al. (2007a) we launched a ground-based (GB) of the system, because then we need to explain why the orbit of observational campaign to obtain high-quality transit light the planet has not been circularized. curves in different filters. Our motivation was to obtain the first group of light curves which will be used by future studies to look for a possible variation of the impact parameter and to search 6 For an updated list, see: http://exoplanet.eu/. for transit timing variations (TTV; Agol et al. 2005; Holman &

1559 Paper VI – GJ 436b Follow-Up

1560 SHPORER ET AL. Vol. 694 Table 1 mmag per transit duration was in the range of 0.1–1.0 mmag. As Photometry of GJ 436 a final step, we divided the entire light curve by the OOT median Observatory + Filter HJD Relative Relative intensity and normalized the measurement uncertainties so the Telescope Flux Flux Error median uncertainty OOT is equal to the OOT RMS. Table 1 lists Wise 0.46 m clear 2454225.229864 1.0015 0.0020 all photometric measurements of the five light curves obtained Wise 1 m V 2454225.218116 0.9997 0.0010 in this campaign. FLWO 1.2 m z 2454243.658856 1.0002 0.0020 Table 2 lists for each light curve the UT start date and Wise 1 m R 2454246.309414 1.0024 0.0017 time, observatory and telescope, the filter used, limb-darkening FLWO 1.2 m z 2454251.641074 0.9987 0.0024 coefficients used in its analysis, average exposure time, average 1 cadence (in minute− ), duration of the entire observation, start Note. This table includes the five light curves obtained in this work. and end airmass, the β factor (see below), RMS residuals, and (This table is available in its entirety in a machine-readable form in the online the Photometric Noise Rate (PNR). The PNR is a quantity journal. A portion is shown here for guidance regarding its form and content.) which takes into account the RMS and the cadence, defined as RMS √cadence, in units of mmag per minute (see also Murray 2005). Our observations are described in Section 2.In Burke et× al. 2008, Section 2.2). our analysis, described in Section 3, we simultaneously analyze six complete GB transit light curves: five light curves obtained 3. DATA ANALYSIS AND RESULTS in our campaign and the light curve from Gillon et al. (2007a). In Section 4, we discuss our results. We decided not to include the publicly available Spitzer 8 µm light curve in our simultaneous fitting since the shape 2. OBSERVATIONS of that light curve is dependent on how the Spitzer “ramp” was modeled (Gillon et al. 2007b; Deming et al.2007 ). As In addition to the transit of UT 2007 May 2, which was the Spitzer measurements are much more precise than the GB reported by Gillon et al. (2007a) and was the first complete measurements, this ramp could have systematically influenced transit observation of GJ 436b, we obtained five complete our results. For comparison, the Spitzer light curve PNR is a transit light curves of four different transit events. The following factor of 3 smaller than that of our ground-based light curves. paragraphs briefly describe these observations. Accounting∼ for correlated noise (Pont et al. 2006) was done We observed the transits of UT 2007 May 4 and May 25 similarly to the “time-averaging” method of Winn et al. (2008). with the Wise Observatory 1 m telescope, located in Israel. After a preliminary analysis, we binned the residual light curves We used a Princeton Instruments (PI) VersArray camera and a using bin sizes close to the duration of ingress and egress. The 1340 1300 pixel back-illuminated CCD, with a pixel scale of presence of correlated noise in the data was quantified as the × 0′′.58 per pixel, giving a field of view of 13.′0 12.′6. On May 4 ratio between the binned residual light-curves standard deviation × we used a Johnson V filter and on May 25 a Bessel R filter. The and the expected standard deviation in the absence of correlated flexible scheduling of the Wise 1 m telescope was an important noise. For each light curve, we estimated β as the largest ratio factor in obtaining the transit light curve of May 4, less than among the bin sizes we used. We then multiplied the relative three days after the first observation of a complete transit. flux errors by β. Values of β are in the range of 1.0–2.1 and are The transit of UT 2007 May 4 was simultaneously observed listed in Table 2. with the Wise Observatory 0.46 m telescope operated remotely. The camera was a 2148 1472 pixel Santa Barbara Instrument 3.1. Simultaneous Fitting of All Parameters Group (SBIG) ST-10 XME× CCD detector with a field of view We fitted the system parameters to all six GB light curves of 40.′5 27.′3 and a scale of 1′′.1 per pixel. This camera has no filters.× For a detailed description of this telescope and simultaneously. Our transit light-curve model consisted of nine instrument, see Brosch et al. (2008). parameters: the orbital period, P; a particular midtransit time, On UT 2007 May 23 and May 31 we used the 1.2 m telescope Tc; planet-to-star radius ratio, Rp/Rs ; semimajor axis scaled by at the Fred L. Whipple Observatory (FLWO) on Mount Hopkins, the stellar radius, a/Rs ; impact parameter, b (see Equation (3) Arizona. The camera was the KeplerCam, which is a 40962 of Winn et al. 2007 for the formula of the impact parameter in an eccentric orbit); two limb-darkening coefficients u1 and Fairchild 486 CCD with a square field of view 23.′1 on a side (Szentgyorgyi et al. 2005). On both nights, the filter was Sloan u2, for a quadratic limb-darkening law; orbital eccentricity, e; and longitude of periastron, ω. The latter four parameters were z and a 2 2 binning was used, giving a pixel scale of 0′′.68 per binned pixel.× held fixed in the fitting process, hence our model had five free Basic data reduction procedures, including bias and flat-field parameters. correction, were applied to all images using standard IRAF7 For transiting planets on eccentric orbits, the midpoint, in routines. We then performed aperture photometry of GJ 436 time, between transit start and end is different than the time and several comparison stars of similar brightness showing no of sky-projected star–planet closest approach. For GJ 436b, significant variability. The light curve of GJ 436 was calibrated this difference is comparable to the typical uncertainty on the by dividing it by the summed flux of the comparison stars. midtransit times obtained here. In our model, the midtransit time Next, we fitted a polynomial of degree one or two to the out-of- is defined as the time of closest approach, which is also the time transit (OOT) points versus time, and divided all points by this of minimum light. polynomial. The amplitude of these polynomial corrections in For e and ω,we used the values given by Maness et al. (2007). Although other authors (e.g., Demory et al. 2007; Deming et al. 2007) report more precise values for e, their uncertainty on 7 The Image Reduction and Analysis Facility (IRAF) is distributed by the ω is either much higher than that of Maness et al. (2007) National Optical Astronomy Observatories, which is operated by the Association of Universities for Research in Astronomy, Inc., under cooperative or not reported. As the errors in two orbital parameters are agreement with the National Science Foundation. correlated, we chose to adopt the reference giving the smallest Paper VI – GJ 436b Follow-Up

No. 2, 2009 PHOTOMETRIC FOLLOW-UP OBSERVATIONS OF GJ 436b 1561

Table 2 List of Light Curves Analyzed in this Work

a Start Date Start Time Observatory + Filter u1 u2 Exp. Cadence Duration Start End β RMS PNR 1 1 UT (hh:mm) UT Telescope Time (s) (minute− ) (hr) AM AM (mmag) (mmagminute− ) 2007 May 2 00:08 Euler1.2m V 0.340 0.444 80 0.44 3.86 2.10 2.24 1.0 1.2 1.8 2007 May 4 17:39 Wise0.46m clear 0.343 0.398 20 1.23 3.19 1.03 1.14 1.6 2.0 1.8 2007 May 4 17:22 Wise1m V 0.340 0.444 60 0.65 2.90 1.05 1.07 1.5 0.9 1.1 2007 May 23 03:56 FLWO1.2m z 0.088 0.522 10 2.44 3.91 1.02 1.99 1.3 2.0 1.3 2007 May 25 19:33 Wise1m R 0.343 0.398 40 0.88 3.60 1.15 3.79 1.3 2.4 2.6 2007 May 31 03:30 FLWO1.2m z 0.088 0.522 10 2.38 3.20 1.02 1.64 2.1 2.1 1.4

Note. a Photometric noise rate RMS √cadence. = × uncertainty on both. To determine u1 and u2, we adopted a we considered only the long chain while ignoring the first 20% model of Teff 3500 K (Maness et al. 2007), log g 4.843 of its steps. (Torres 2007),= and [Fe/H] 0.03 (Bonfils et al. 2005= ) and We performed 10 MCMC runs as described above while interpolated on the Claret (2000=−, 2004) grids. For the light curve changing each time the initial guess and initial distribution obtained with the Wise Observatory 0.46 m telescope, which has widths. Results of those runs were highly consistent with each no filter, we used limb-darkening coefficients corresponding to other and our final result is based on all runs. the R filter, as its CCD response resembles a “wide-R” filter Before describing our results, we briefly describe how we (Brosch et al. 2008). The coefficients used for each light curve tested our analysis: are listed in Table 2. 1. To test the adaptive form of our MCMC algorithm, we reran We used the formulas of Mandel & Agol (2002) to determine it while fixing the distribution widths during the long chain the relative flux as a function of the planet–star sky-projected part. The differences in the results for the fitted parameters separation and the Monte Carlo Markov Chain (MCMC) al- did not exceed 0.15σ, and the differences in the estimated gorithm (e.g., Tegmark et al. 2004; Ford2005 ; Gregory 2005) uncertainties were no more than 10%. to determine the parameters that best fit the data, along with 2. We repeated our analysis while skipping the polynomial their uncertainties. MCMC algorithms have already been used corrections mentioned in Section 2. The results for the fitted extensively in the literature for fitting transit light curves (e.g., parameters were less than 0.2σ from the original ones, and Holman et al. 2006; Collier Cameron et al. 2007; Gillon et al. the estimated uncertainties were similar at the 20% level. 2008; Burke et al. 2007). Assuming a uniform prior for the five 3. We applied our analysis on the Spitzer 8 µm light curve fitted parameters, the algorithm can be viewed as a random walk alone, and derived results similar to Gillon et al. (2007b); in the five-dimensional parameter space of the model. Starting Deming et al. (2007); Southworth (2008), at the 1σ level. from an initial guess, at each step the algorithm examines a 4. To further test the validity of our results, we reran∼ MCMC new point in the parameter space which is reached by adding a while adopting a Gaussian prior probability for e and ω. The Gaussian random value to each of the parameters. The algorithm central values and standard deviations were those given decides whether to accept the new point in the parameter space by Maness et al. (2007). The results were not modified and move there, or repeat the current point in the chain depend- significantly. ing on the resulting posterior probability. If the new point has a higher relative posterior probability, the algorithm will move Figure 1 presents the six GB light curves, with our fitted there, and if not, it will move to the new point with a relative model overplotted in solid lines. The three light curves obtained probability of exp( ∆χ 2/2), where ∆χ 2 is the χ 2 difference in the beginning of 2007 May are of better quality, especially the between the two points− in parameter space. The distribution of Euler 1.2 m V and Wise 1 m V light curves, with RMS residuals parameter values in all points is used to derive the best-fit value of 1.2 and 0.9 mmag, respectively. The increased scatter in the and its uncertainties. Here, we took the distribution median three light curves obtained at the end of 2007 May is the result to be the best-fit value and the values at the 84.13 and 15.87 of the high airmass of these observations, done close to the end percentiles to be the +1σ and 1σ confidence uncertainties, of the observational season. respectively. − Our estimates for the fitted parameters, while fitting all The widths of the Gaussian distributions used to find a new six light curves simultaneously, are listed in Table 3. For point are a crucial part of the MCMC algorithm and they affect comparison, this table includes results from previous studies the fraction of accepted points, f. In order to have f close to whenever these parameters were fitted directly. In case they 25% (Gregory 2005; Holman et al.2006 ), we divided each long were not, but can be calculated from other fitted parameters, chain, of 500,000 steps, into smaller chains (minichains), of they are given without an error. Gillon et al. (2007a) based their 1000 steps, and re-evaluated the distribution widths after the parameter determinations on an analysis of the Euler 1.2 m V execution of each minichain. This re-evaluation was done by light curve from UT 2007 May 2, which is included in this work. scaling them according to the relative difference between f of The other three studies referred to in Table 3 (Gillon et al. 2007b; the previous minichain, and the target value of 25%. The same Deming et al. 2007; Southworth 2008) are based on the Spitzer scaling was applied to all parameters. 8 µm light curve. Results obtained by Torres et al. (2008) are To estimate the relative size of the distribution widths for each not presented since for the light-curve parameters of GJ 436b, one of the model parameters, we initially performed a sequence these authors gave the weighted average of previous works. of 10 minichains, where only that parameter was allowed to vary Table 4 lists our results for the physical parameters of and the target f was 50% (Gregory 2005). Thus, a single MCMC the system, along with the corresponding values obtained run was comprised of several sequences of 10 minichains and a previously. Assuming the mass of the star to be Ms 0.452 0.013 M (Torres 2007), and adopting the orbital parameters= of± long chain of 500 minichains. For determining the final result, ⊙ Paper VI – GJ 436b Follow-Up

1562 SHPORER ET AL. Vol. 694

1.01 1.01 V clear

1 1

0.99 0.99

Relative Flux Relative 0.98 0.98

22.5 22.55 22.6 22.65 25.25 25.3 25.35

1.01 1.01 V z

1 1

0.99 0.99

Relative Flux Relative 0.98 0.98

25.2 25.25 25.3 25.35 43.7 43.75 43.8

1.01 1.01 R z

1 1

0.99 0.99

Relative Flux Relative 0.98 0.98

46.3 46.35 46.4 46.45 51.65 51.7 51.75

Figure 1. Six light curves analyzed in this work. The overplotted solid line is our best-estimate model. For each light curve, residuals from the model are plotted, centered on a relative flux of 0.98 for clarity. Within each panel the UT date, observatory, telescope, and filter used are given. (A color version of this figure is available in the online journal.)

Table 3 close to a 2:1 mean-motion resonance with the transiting planet. Light-Curve Fitted Parameters The gravitational interaction between the two planets may cause a detected TTV signal (Holman & Murray 2005; Agol et al. Reference PTc-2454200 Rp/Rs a/Rs b (day) (HJD) 2005). Ribas et al. (2008) suggest that the second planet is ThisWork 2.64386 35.8355 0.085 13.6 0.85 responsible for changing the inclination angle i of the transiting 0.00003 0.0001 0.001 0.5 0.01 planet orbit. The presence of such a planet may be detected ± ± ± ± ± Gillon et al. (2007a)a 0.082 through measuring transit timing and the impact parameter of 0.005 many transit light curves. As the light curves obtained here are ± Gillon et al. (2007b)a 80.78148 0.0830 0.849 the first group of such light curves, we performed two additional +0.00015 +0.010 MCMC analyses described below. 0.00008 0.013 First, we fitted Rp/Rs , a/Rs , and b to all light curves and a − − Deming et al. (2007) 80.78149 0.0839 13.2 0.85 an independent Tc to each transit light curve to estimate the 0.00016 0.0005 0.6 +0.03 ± ± ± midtransit time of each event. In this analysis, the period does 0.02 not determine the time intervals between transit events. It is used a − Southworth (2008) 80.78174 0.08284 13.68 only for determining the true anomaly of each measurement 0.00011 0.00090 0.51 ± ± ± within each light curve, so it is only weakly constrained. Hence, in this analysis P was fixed at the value derived in our original Note. a Period was fixed at the Maness et al. (2007) value of P 2.64385 = ± 0.00009 days. analysis. The resulting six Tc values were later used to derive the transit ephemeris by a linear fit. The new period is 0.35σ ≈ Maness et al. (2007; specifically K, e, and ω), we derived the from the period fitted in our original MCMC run. Repeating the planet mass and semimajor axis. Using Newton’s version for process described in this paragraph with the new period results Kepler’s third law, we derive the star and planet radius. in a variation of only 0.02σ, so no further iterations were done. Table 5 contains the midtransit timings fitted independently 3.2. Fitting Transit Timing and Impact Parameter for Each to each light curve. Note that light curves with the smallest Light Curve RMS residuals do not necessarily have the smallest midtransit time uncertainty. For example, the FLWO 1.2 m light curve, As mentioned in Section 1, Ribas et al. (2008) suggested the at E 3, has relatively high RMS residuals although a small presence of a second planet orbiting GJ 436, in an outer orbit midtransit= time uncertainty. This is simply the effect of the short Paper VI – GJ 436b Follow-Up

No. 2, 2009 PHOTOMETRIC FOLLOW-UP OBSERVATIONS OF GJ 436b 1563

Table 4 Results for the Physical Parameters

Reference aMp Rs Rp (AU) (M⊕)(R⊙)(R⊕) ThisWork 0.02872 +0.00030 23.1 ±0.9 0.45 ±0.02 4.2 ±0.2 −0.00025 Gillon et al. (2007a) 22.6 ±1.9 0.44 ±0.04 3.95 +0.41 −0.28 Gillon et al. (2007b) 0.463 +0.022 4.19 +0.21 −0.017 −0.16 Deming et al. (2007) 0.0291 ±0.0004 22.24 ±0.95 0.47 ±0.02 4.33 ±0.18 Torres (2007) 0.02872 ±0.00027 23.17 ±0.79 0.464 +0.009 4.22 +0.09 −0.011 −0.10 Southworth (2008)a 0.452 ±0.017 4.08 ±0.16

a +0.014 Note. Assuming Ms = 0.452−0.012 M⊙ (Torres 2007), orbital parameters of Maness et al. (2007), and the period derived here.

Table 5 2 Midtransit Time and Impact Parameter Fitted Independently to Each Light a Curve 1.5 b c E Tc-2454200 ∆Tc b ∆b 1 (HJD) (minute) (σ )(σ) −5 22.61612 ± 0.00037 −0.15 −0.28 0.838 ± 0.017 −0.013 −0.61 0.5 −4 25.26002 ± 0.00038 −0.10 −0.18 0.828 ± 0.020 −0.023 −0.96 −4 25.26052 ± 0.00030 0.62 1.45 0.843 ± 0.016 −0.008 −0.37 0 3 43.76657 ± 0.00026 −0.71 −1.88 0.846 ± 0.018 −0.005 −0.23 4 46.40982 ± 0.00040 −1.58 −2.74 0.843 ± 0.014 −0.008 −0.41 6 51.69956 ± 0.00030 1.34 3.12 0.854 ± 0.016 0.003 0.13

Notes. a The table gives the result of two separate MCMC analyses. In each analysis, only a single parameter (Tc or b) was fitted separately to every light curve and all other parameters were fitted to all light curves simultaneously, except for P which was fixed when fitting an individual Tc to each light curve. b Difference between the measured Tc and the expected midtransit time 0 2 4 6 8 according to the fitted ephemeris (see Table 4). The difference is given in Epoch minutes and also in units of the uncertainty on each midtransit time. Figure 2. Observed minus calculated (O − C) transit timing of the six light c Difference between the measured impact parameter while fitting each light curves included in this study. The graph shows the residuals from a linear fit curve separately, and the result of the simultaneous fit (see Table 4). The to the transit timing as a function of the transit epoch. Table 5 lists the actual difference is also given in units of the uncertainty on each impact parameter. transit timings. (A color version of this figure is available in the online journal.) KeplerCam cadence, as the increased number of points act to better constrain the midtransit time. This effect is quantified by the impact parameter, b, was fitted independently to each light the PNR (see Table 2). curve. The derived impact parameters are listed in Table 5 and By fitting a linear function, plotted in Figure 3. As a reference, we overplotted in Figure 3 the value of our original run, in a solid line, and the upper Tc(E) = Tc(0) + P × E, (1) and lower 1σ confidence limits in dashed lines. No significant variation can be seen in the data, although a long-term trend of E P = . ± . to the transit epoch, , we derived 2 64386 0 00003 days the order of 0.2 yr−1 cannot be ruled out. T = . ± . and c(0) 2454235 8355 0 0001 [HJD]. The transit event The two separate analyses described in this section were re- E = with 0 was chosen to be right in the middle of our observed peated without applying the polynomial corrections mentioned events, although we did not observe that particular event. We in Section 2. The results were consistent with the original ones. take this ephemeris as our final result, listed in Table 3, and In addition, for both the O − C transit timing residuals and the P T note that it is consistent with and c derived from our original impact parameters derived here, we looked for a possible cor- MCMC fitting. relation with several external parameters, including midtransit The residuals from the linear fit are presented in Figure 2, airmass, filter central wavelength, and the light-curves polyno- O − C known as the diagram. The figure shows some indication mial correction amplitude. All correlations were below the 1.5σ χ 2 for a variability of the period, reflected by the high value confidence level. of the fit, of 23 for 4 degrees of freedom. However, the small number of points does not allow to assign any significance to the 4. DISCUSSION AND SUMMARY detection of this variability, especially because some systematic effects could shift some of the points. This paper presents an analysis of six complete transit light Second, we performed an MCMC analysis where we looked curves of GJ 436b, spanning 29 days, or 11 orbital periods. for a transit impact parameter variation. The four parameters Our period is consistent with the period given by Maness P, Tc, Rp/Rs , and a/Rs were fitted to the entire data while et al. (2007), of 2.64385 ± 0.00009 days, and is more precise. Paper VI – GJ 436b Follow-Up

1564 SHPORER ET AL. Vol. 694

1.1 0.87 1.05

0.86 1

0.95 0.85

0.9 b 0.84 b(ω) 0.85

0.83 0.8

0.75 0.82 0.7

0.81 0.65

0.8 0.6 0 2 4 6 8 0 50 100 150 200 250 300 350 ω Epoch [deg] Figure 3. Impact parameter b fitted independently to each light curve, as a Figure 4. Impact parameter, b, vs. ω, for the eccentricity and inclination of function of the transit epoch. The solid line marks the best estimate derived in GJ 436b (the solid line). The filled circle (green) marks the current position, our original MCMC run, while fitting the impact parameter to all light curves error bars are too small to be shown. The upper and lower dashed lines (red) simultaneously. The dashed lines mark the upper and lower 1σ confidence mark the impact parameter when the stellar radius is multiplied by a factor of 1 Rp/Rs and 1+ Rp/Rs , respectively. When multiplying the stellar radius by limits. Table 5 lists the actual values of the impact parameters presented here. − 1 Rp/Rs (the upper dashed line), the impact parameter will be smaller than (A color version of this figure is available in the online journal.) − one only for nongrazing transits. When multiplying by 1 + Rp/Rs (the lower dashed line), the impact parameter will equal one when the minimal star–planet Our value for Rp/Rs (see Table 3) is larger by 1σ –2σ , than the sky-projected distance equals the radii sum, i.e., for b>1 there will be no radii ratio derived for the Spitzer 8 µm light curve by Gillon transit at all. et al. (2007b), Deming et al. (2007), and Southworth (2008). (A color version of this figure is available in the online journal.) If this difference is real, it may be induced by stellar spots, or wavelength-dependent opacities in the planetary atmosphere (Barman 2007; Pont et al. 2008, 2009). Interestingly, a similar (Miralda-Escude´ 2002), where Mp,2 and P2 are the mass effect, i.e., an increased radius ratio in the optical relative to the and period of the second planet, respectively. For the planet IR, was already observed for HD 189733b (Pont et al. 2007). suggested by Ribas et al. (2008), this precession is of the order Our derived physical parameters, mainly the planetary ra- of 10 yr. dius and mass, are consistent with those derived previously For eccentric orbits, the impact parameter can also be mod- (see Table 4). ified by the precession of the periastron because the impact We examined the individual transit timing of each light curve parameter depends on the orientation of the line of apsides (see (TTV, see Figure 2) and derived the impact parameter of each Equation (3) of Winn et al. 2007). The apsidal motion can be light curve independently (see Figure 3). The transit timing driven by the stellar quadrupole moment, or by a second planet. residuals show a possible hint for variability, supported by the The precession driven by the stellar quadrupole moment could high χ 2 of the fit. However, many more light curves are needed be quite slow, and therefore probably does not contribute to the to establish the variability and to explore its nature. The impact variation of the impact parameter. The rate of the apsidal preces- parameter plot shows a possible trend, although a linear fit to sion induced by another planet is of the same order of magnitude the six data points gives a slope consistent with zero. as the rate of the nodal precession, estimated in Equation (2), Comparing our derived transit ephemeris to transit times and therefore can be of the order of years, depending on the published recently (Alonso et al. 2008; Bean et al. 2008) shows parameters of the unseen additional planet in the system. that the predicted transit times are somewhat earlier than the To estimate the possible implication of the apsidal preces- measured times. The difference is at the 1σ–2σ level, so it is not sion on the impact parameter, we plot in Figure 4 the im- highly significant. However, if real, it supports the claim made pact parameter of GJ 436b as a function of the argument of by Bean et al. (2008) for a long-term drift in the transit times. the periastron ω, at fixed orbital eccentricity and inclination. We can see that b is modulated between 1.0 and 0.7 over the Together with the long-term slope in GJ 436 RVs (Maness et al. ∼ 2007), it suggests the existence of another object in the system apsidal precession period. The upper dashed line of the figure with a long period. Future photometric and spectroscopic data indicates that at some phase of the precession period the transit will be able to study this possibility better. of GJ 436b will even lose its flat-bottom shape. The figure sug- As mentioned in the introduction, the impact parameter of gests that if the apsidal motion is of the order of 10 years, we GJ 436b can be changed by a precession of the line of nodes, should be able to observe soon a change of the impact parameter. induced by a second planet whose orbital plane is inclined In case this modulation is due to apsidal motion, we will then relative to the orbit of GJ 436b (e.g., Miralda-Escude´ 2002). be able to confirm the change of the line of apsides direction by Such an effect was already suggested by Soderhjelm (1975) timing the secondary eclipse or by precise RV measurements. and Mazeh & Shaham (1976) for stellar triple systems. The precession period is We are deeply thankful to Michael Gillon for his photometric processing of the data obtained on 2007 May 4 and May 25. 2 Ms P2 We thank the anonymous referee for his thorough reading of Unode P (2) ≈ Mp,2 P the paper and his useful comments which helped improve the Paper VI – GJ 436b Follow-Up

No. 2, 2009 PHOTOMETRIC FOLLOW-UP OBSERVATIONS OF GJ 436b 1565 paper. T. Mazeh and A. Shporer thank Elia Leibowitz and Ford, E. B. 2005, AJ, 129, 1706 Liliana Formiggini for allowing the use of their telescope time Fortney, J. J. 2008, in ASP Conf. Ser. 398, Extreme Solar Systems, ed. D. Fischer et al., 405 at the Wise Observatory 1 m telescope on 2007 May 4. This Gillon, M., Triaud, A. H. M. J., Mayor, M., Queloz, D., Udry, S., & North, P. work was partly supported by Grant No. 2006234 from the 2008, A&A, 485, 871 United States–Israel Binational Science Foundation (F), by the Gillon, M., et al. 2007, A&A, 472, L13 Kepler Mission under NASA Cooperative Agreement NCC Gillon, M., et al. 2007, A&A, 471, L51 2-1390 with the Smithsonian Astrophysical Observatory, and Gregory, P. C. 2005, ApJ, 631, 1198 Guillot, T. 2008, Physica Scripta, 130, 014023 by NASA Origins grant NNG06GH69G (to M. J. H.). Holman, M., Touma, J., & Tremaine, S. 1997, Nature, 386, 254 Holman, M. J., & Murray, N. W. 2005, Science, 307, 1288 REFERENCES Holman, M. J., et al. 2006, ApJ, 652, 1715 Mandel, K., & Agol, E. 2002, ApJ, 580, L171 Agol, E., Steffen, J., Sari, R., & Clarkson, W. 2005, MNRAS, 359, 567 Maness, H. L., Marcy, G. W., Ford, E. B., Hauschildt, P. H., Shreve, A. T., Basri, Alonso, R., Barbieri, M., Rabus, M., Deeg, H. J., Belmonte, J. A., & Almenara, G. B., Butler, R. P., & Vogt, S. S. 2007, PASP, 119, 90 J. M. 2008, A&A, 487, L5 Mazeh, T. 1990, AJ, 99, 675 Barman, T. 2007, ApJ, 661, L191 Mazeh, T., Krymolowski, Y., & Rosenfeld, G. 1997, ApJ, 477, L103 Bean, J. L., et al. 2008, A&A, 486, 1039 Mazeh, T., & Shaham, J. 1976, ApJ, 205, L147 Bonﬁls, X., Delfosse, X., Udry, S., Santos, N. C., Forveille, T., & Segransan,´ D. Mazeh, T., & Shaham, J. 1979, A&A, 77, 145 2005, A&A, 442, 635 Miralda Escude,´ J. 2002, ApJ, 564, 1019 Bonﬁls, X., et al. 2005, A&A, 443, L15 Pont, F., Knutson, H., Gilliland, R. L., Moutou, C., & Charbonneau, D. Brosch, N., Polishook, D., Shporer, A., Kaspi, S., Berwald, A., & Manulis, I. 2008, MNRAS, 385, 109 2008, Ap&SS, 314, 163 Pont, F., Gilliland, R. L., Knutson, H., Holman, M., & Charbonneau, D. 2009, Burke, C. J., et al. 2007, ApJ, 671, 2115 MNRAS, 393, 6 Burke, C. J., et al. 2008, ApJ, 686, 1331 Pont, F., Zucker, S., & Queloz, D. 2006, MNRAS, 373, 231 Butler, R. P., Vogt, S. S., Marcy, G. W., Fischer, D. A., Wright, J. T., Henry, Pont, F., et al. 2007, A&A, 476, 1347 G. W., Laughlin, G., & Lissauer, J. J. 2004, ApJ, 617, 580 Ribas, I., Font-Ribera, A., & Beaulieu, J.-P. 2008, ApJ, 677, L59 Charbonneau, D., Brown, T. M., Burrows, A., & Laughlin, G. 2007, in Protostars Soderhjelm, S. 1975, A&A, 42, 229 and Planets V, ed. B. Reipurth, D. Jewett, & K. Keil (Tucson, AZ: Univ. Southworth, J. 2008, MNRAS, 386, 1644 Arizona Press), 701 Szentgyorgyi, A. H., et al. 2005, BAAS, 37, 1339 Claret, A. 2000, A&A, 363, 1081 Tegmark, M., et al. 2004, Phys. Rev. D, 69, 103501 Claret, A. 2004, A&A, 428, 1001 Torres, G. 2007, ApJ, 671, L65 Collier Cameron, A., et al. 2007, MNRAS, 380, 1230 Torres, G., Winn, J. N., & Holman, M. J. 2008, ApJ, 677, 1324 Deming, D., Harrington, J., Laughlin, G., Seager, S., Navarro, S. B., Bowman, Udry, S., et al. 2006, A&A, 447, 361 W. C., & Horning, K. 2007, ApJ, 667, L199 Udry, S., et al. 2007, A&A, 469, L43 Demory, B.-O., et al. 2007, A&A, 475, 1125 Winn, J. N., et al. 2007, ApJ, 665, L167 Fischer, D. A., et al. 2008, ApJ, 675, 790 Winn, J. N., et al. 2008, ApJ, 683, 1076

Chapter 3

Discussion

In this chapter I summarize the ﬁndings of each of the six papers and brieﬂy discuss their astrophysical implications.

3.1 Papers I – III — Detection of Three New

Transiting Extrasolar Planets

This thesis includes the discovery of three new transiting planets, which I was an important part of.

3.1.1 Paper I — HAT-P-2b

The detection of HAT-P-2b (HD 147506b) as a transiting planet candidate was highly challenging. The transit is shallow, only 5 mmag deep, and the period is relatively long (5.63 days). In fact, it was the

91 longest period transiting planet at the time of discovery, and as of this writing it is the longest period planet discovered photometrically, using small-aperture ground-based instruments.

Undoubtedly, its discovery should be credited to the increased capabilities of the HATNet network of telescopes. HAT-P-2 happens to be located in an overlap corner of 3 ﬁelds observed by HATNet instruments, and was observed by 4 telescopes, one in Hawaii, two in Arizona and WHAT in Israel. This resulted in a large number of measurements (∼26,400), spanning two observational seasons.

High-resolution spectroscopic follow-up observations of this candidate were carried out at the Keck telescope, with HIRES. Photo- metric follow-up observations were done from several observatories, including Wise. Data obtained at Wise did not show transits but was important for constraining the transit ephemeris, leading to ﬁnally obtaining a full transit at FLWO on April 2007.

HAT-P-2b is an unusual planet from several aspects. With a radius similar to that of Jupiter, its mass is almost 9 MJ , rendering

HAT-P-2b as one of the most massive and dense transiting planets.

Its mass is relatively large also when compared to the Mp sin i of the RV planets, even though the RV method is biased towards the

92 detection of massive planets. In addition, it has an eccentric orbit, with e ∼= 0.5, larger than most of the known exoplanets.

In the years following HAT-P-2b discovery there were a few detections of transiting exoplanets which also have a relatively high mass (Mp & 3MJ ) and reside in non-circular orbits. Those include, in decreasing mass order, XO-3b (Johns-Krull et al. 2008; Winn et al. 2008b), WASP-14b (Joshi et al. 2009), HD 80606b (e.g., Naef et al. 2001; Laughlin et al. 2009; Moutou et al. 2009), HD 17156b (e.g.,

Fischer et al. 2007; Barbieri et al. 2007) and WASP-10b (Christian et al. 2009; Johnson et al. 2009). Table 3.1 lists the properties of these

ﬁve planets, and those of HAT-P-2b as well. Figure 3.1 presents intrinsic and orbital characteristics of the transiting planets (updated to July 2009). The left panel shows the region in the radius-mass diagram where most of the transiting planets are located. HAT-P-

2b is encircled in red and the ﬁve planets mentioned above in green.

The rightmost planet in this diagram is CoRoT-3b, which is a brown dwarf on a circular orbit (M2 = 21.7 MJ , Deleuil et al. 2008). The right panel of Fig. 3.1 shows the orbital eccentricity vs. semi-major axis for the transiting planets (ﬁlled circles) and RV planets (empty circles) with semi-major axis smaller than 0.7 AU. As in the left

93 panel, HAT-P-2b is encircled in red and the ﬁve planets mentioned above in green. These total of six planets show eccentricities larger than most, if not all, of the planets with a similar semi-major axis.

Figure 3.1 right panel also shows that close-in planets tend to be less eccentric. Actually, the ﬁgure shows that the planets are positioned below some envelope function, which increases from zero for the close-in plants, through HAT-P-2b (e=0.52, a=0.0677 AU), up to HD 80606b, with e=0.93 and a=0.449 AU. A similar envelope was pointed out for stellar binaries by Mazeh (2008), resulting from tidal circularization. This suggests that the same mechanism is at work in star-planet systems.

Name Mp e P MJ days XO-3b 11.8 0.26 4.26 HAT-P-2b 8.8 0.52 5.63 WASP-14b 7.7 0.09 2.24 HD80606b 3.9 0.93 111.44 HD17156b 3.2 0.68 21.22 WASP-10b 3.1 0.06 3.09

Table 3.1: List of six transiting exoplanets, including HAT-P-2b, with high mass (Mp > 3MJ ) and non-circular orbits. List is ordered in decreasing planetary mass, and includes planet name, mass, orbital eccentricity and period. See also Fig. 3.1. For HAT-P-2b, the values for the mass and eccentricity are the weighted average of values given in the literature (Bakos et al. 2007a, Winn et al. 2007, Loeillet et al. 2008, Torres et al. 2008 and Pal et al. 2009).

As illustrated in the left panel of Fig. 3.1, there are only a few high-mass transiting planets, with Mp >∼ 3 MJ . However, since

94 1

2 0.8

1.5

] 0.6 J e [R p R 0.4 1

0.2

0 0.5 1 5 10 50 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 M [M ] p J a [AU]

Figure 3.1: Comparing intrinsic and orbital properties of massive planets. Left: The region in the radius-mass diagram, in log-log scale, where most of the transiting planets are located. HAT- P-2b is encircled in red and five other massive planets in green (XO-3b, WASP-14b, HD 80606b, HD 17156b and WASP-10b). The rightmost planet in this diagram is a brown dwarf on a circular orbit (CoRoT-3b, M2 = 21.7 MJ , Deleuil et al. 2008). Right: Orbital eccentricity vs. semi-major axis for transiting planets (filled circles, blue) and RV planets (empty circles, blue), with a ≤ 0.7 AU. As in the left diagram, HAT-P-2b is encircled in red and the five massive planets in green (figure is updated to July 2009). their radii are similar to those of the other lower mass transiting planets, the small number of high-mass transiting planets is real, and not resulting from an observational bias. This leads to the conclusion that formation processes, of close-in planets, are less efficient in this mass range. This seems to be consistent with the mass-period correlation of Zucker & Mazeh (2002), although the latter refers to a wider range of orbital distances than that of the transiting planets.

The tendency of massive transiting planets to reside in eccentric orbits (e.g., Southworth 2009), noted also for non-transiting planets

(Jones et al. 2006, see their Fig. 7), might suggest that their migration could act to increase their eccentricity. An interesting possibility

95 is that the non-zero eccentricity results from the gravitational interaction with another body in the system, a second planet or a stellar binary companion. One such mechanism is a second planet orbiting in a mean motion resonance (MMR), where the orbital periods of the two planets are related by a ratio of two small integers. This scenario is consistent with the expectation that during the migration process massive planets capture other planets in MMR, and the interaction between the planets increases their orbital eccentricity.

(e.g., Papaloizou 2003; Cresswell & Nelson 2006).

3.1.2 Paper II — HAT-P-5b

HAT-P-5b was one of the candidates I observed spectroscopically with SOPHIE, during an observing run in May 2007. This candidate was conﬁrmed as a planet after a single run, and in the months that followed I obtained four photometric follow-up light curves with the Wise Observatory 1.0-m telescope. Our American collaborators obtained 1.5 light curves (one full and another of a partial transit) from the FLWO 1.2-m telescope, with KeplerCam, bringing the number of high-precision light curves presented in the discovery paper to

5.5. The mid-transit time of each of the ﬁve full transit light curves was measured but no TTV was detected.

96 The derived radius and mass of the planet (Rp =1.26 ± 0.05 RJ ,

Mp =1.06±0.11 MJ ), along with its orbital period (2.79 days) make it similar to other transiting Hot Jupiter planets.

3.1.3 Paper III — HAT-P-9b

HAT-P-9b was one of the candidates I followed-up spectroscopically at the OHP 1.93-m telescope, with SOPHIE, during three observing runs on 2007 and 2008. During that time I obtained two photometric follow-up light curves from the Wise Observatory, the ﬁrst with the

0.46-m and the second with the 1.0-m telescope. Two additional light curves were obtained by our American collaborators with the

FLWO 1.2-m telescope, with KeplerCam.

The position of HAT-P-9b in the radius-mass diagram is marked in the top panel of Figure 5 of Paper III, at the low-density limb of the transiting planets cluster. Its derived radius and mass (Rp =

1.40 ± 0.06 RJ , Mp =0.78 ± 0.09 MJ ) result in a relatively low mean

−3 density, of ρp =0.35 ± 0.06 g cm , or 0.28 ± 0.05 the mean density of Jupiter. Therefore, HAT-P-9b joins a small but growing group of planets with a density that is lower than expected by early planetary theory, and is the subject of intense theoretical work in recent years

(e.g., Burrows et al. 2007; Chabrier & Baraﬀe 2007).

97 With an observed magnitude of V =12.3 mag, HAT-P-9 is one of the faintest planet-host stars detected using small-aperture ground- based instruments. This is one of the reasons why several spectroscopic observing runs were required to determine that it is a planet- star system, and not a false alarm. In addition, the orbital period, of

3.92 days, is long relative to those of other systems with similar distance from the Sun, or similar host star brightness (Mazeh 2009). All this proves the challenging nature of HAT-P-9b detection, attributed at least partially to the increased capabilities of the HATNet network of instruments.

Figure 3.2 shows the mass-period diagram for transiting planets, where HAT-P-5b and HAT-P-9b are encircled in green. The positions of these two planets conﬁrm the mass-period correlation suggested by

Mazeh et al. (2005). Furthermore, the current sample suggests the correlation levels oﬀ at an orbital period of about 3 days, especially if we ignore the two low-mass planets, GJ 436b and HAT-P-11b (Bakos et al. 2009), both with Mp < 0.1 MJ , as their low mass may result in a diﬀerent mass-period relation than for the Jupiter-mass planets.

The leveling-oﬀ of the mass-period relation at ∼3 days is consistent with the prediction of Pont (2009), that below a period of 3–5 days

98 planetary orbits are shaped by tidal interaction with the host star.

2.5

1.5 ] J [M p M 1

0.5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 Period [days]

Figure 3.2: The mass-period diagram for transiting planets. The Y-axis shows the planetary mass, in Jupiter mass, and the X-axis shows the orbital period, in days. Both axes are in linear scale. The two planets with the lowest mass values, below 0.1 MJ , are Neptune-mass planets (GJ 436b and HAT-P-11b). The diagram shows the observed correlation between decreasing mass and increasing period, first pointed out by Mazeh et al. (2005). Further discoveries made since, including HAT-P-5b and HAT-P-9b, encircled in green, confirm this correlation and show it may be leveling off at an orbital period of about 3 days. Note that the massive planet HAT-P-2b is beyond the parameter space included in this figure (figure is updated to July 2009).

3.2 Papers IV – VI — Follow-up Studies of Known

Transiting Extrasolar Planets

This thesis includes follow-up studies of three interesting transiting planets, all carried out right after their discovery.

99 3.2.1 Paper IV — WASP-1b photometric follow-up

Two transits of WASP-1b were observed at the Wise Observatory 1.0- m telescope at the beginning of October 2006. The ﬂexible scheduling of this telescope was important for obtaining these observations within only two weeks following the planet publication. Both light curves we obtained were of better quality than the one presented in the discovery paper (Cameron et al. 2007), allowing to better constrain the system parameters, especially the planetary and stellar radii. The error bar of the stellar radius was reduced signiﬁcantly, from ∼0.4 to 0.06 R⊙, and by an order of magnitude for the planetary radius, from 0.6 to 0.06 RJ . Our results were consistent with those of a simultaneous paper by Charbonneau et al. (2007), based on data obtained by the FLWO 1.2-m telescope with KeplerCam, and con-

ﬁrmed the inﬂated nature of WASP-1b. With Rp = 1.40 ± 0.06 RJ

−3 and Mp = 0.87 ± 0.07 MJ , its mean density is only 0.39 g cm , which is 0.32 that of Jupiter.

3.2.2 Paper V — HAT-P-2b spectroscopic follow-up

On top of the unusual characteristics of the HAT-P-2 system, the host star is bright, with V = 8.7 mag, making it a good candidate

100 for follow-up observations. Our spectroscopic follow-up observations were obtained with the OHP 1.93-m telescope, with SOPHIE. These observations were done during May and June 2007, right after the announcement of the planet, and include 8 out-of-transit spectra and two in-transit sequences, from two diﬀerent events, consisting of 35 and 20 spectra each.

We ﬁtted simultaneously for the orbit and the RM eﬀect a 15- parameter model, using all available 86 spectra, including the 13

Keck and 10 Lick spectra from the discovery paper (Paper I). We were able to reﬁne the orbital solution, and search for the signature of a second planet in the RV residuals. We could not detect clear evidence for its existence; we estimated the data excludes a second planet of mass greater than 1.3, 1.5 and 1.8 MJ for an orbital period shorter than 50, 100 and 200 days, respectively. Our result for the alignment between the stellar spin and the orbital angular momentum is consistent with complete alignment, with λ = 0.2 ± 12.5◦.

This result is consistent with that of a simultaneous study, done at

Keck with HIRES (Winn et al. 2007).

This RM measurement is currently one of 11 such measurements done for transiting exoplanets, where a complete spectroscopic tran-

101 sit was observed. A clear misalignment was measured only for one system, XO-3 (Hebrard et al. 2008; Winn et al. 2009). Analysis of the current sample, done by Fabrycky & Winn (2009), suggests that

XO-3 misalignment indicates it evolved through a diﬀerent migration process than the other systems, leading to an isotropic distribution of the spin-orbit angle. This process probably includes multi-body interactions, such as planet-planet scattering or even collisions (e.g.,

Chatterjee et al. 2008; Nagasawa et al. 2008).

Recently, a partial spectroscopic transit was observed for the HD 80606 system (Moutou et al. 2009), and its analysis suggests that it is also misaligned (Moutou et al. 2009; Pont et al. 2009b; Gillon 2009). It is interesting to point out that both XO-3b and HD 80606b belong to the small group of massive planets on eccentric orbits (see Table 3.1 and Fig. 3.1 in Sec. 3.1.1). This contradicts the expected faster tidal damping of large misalignment for more massive planets (e.g., Fab- rycky et al. 2007; Jackson et al. 2008), suggesting that the eﬀect of tidal damping is small also in the 10 other systems, where λ is close or consistent with zero. Therefore, the alignment observed for these 10 systems is probably a relic of formation and orbital evolution processes (Fabrycky & Winn 2009).

102 3.2.3 Paper VI — GJ 436b photometric follow-up cam-

paign

This follow-up campaign was the ﬁrst photometric campaign for this target, carried out during May 2007, right after the discovery of the system’s transiting nature (Gillon et al. 2007). Four telescopes, from three observatories participated in this campaign (Wise in Israel,

FLWO in Arizona, USA, and ESO La-Silla in Chile), obtaining a total of six light curves of ﬁve diﬀerent transit events.

We searched for any variation in the light curve shape, including

TTV and a variation in the impact parameter, where the latter will change the transit duration and depth, especially since the transit in this system is almost grazing. A variation in the other light curve parameters (planetary to stellar radii ratio and the orbital semi-major axis divided by the stellar radius) is unlikely. Although we did not

ﬁnd any signiﬁcant variation, we could not rule out a small TTV, of the order of a minute, and a long-term modulation of the impact parameter, of the order of +0.2 yr−1.

In this campaign we have obtained the ﬁrst set of light curves, used later in the continuous study of this interesting system (e.g., Coughlin et al. 2008; Pont et al. 2009a). GJ 436 is currently monitored by our

103 group and probably by others as well. These observations will be part of future follow-up studies, with a longer time-span, allowing for a more detailed study of the system.

3.3 Summary

Fig. 3.3 presents the radius-mass diagram for transiting planets, marked in blue, and low-mass stars, for comparison, in red. The three planets whose discovery is presented in Papers I–III are encircled in green, and the three which are studied in Papers IV–VI are encircled in purple. The positions of the low-mass stars shows clearly the linear relation between radius and mass for these objects.

For the transiting planets, many interesting features of the current sample can be seen in the ﬁgure. Most of the planets are similar to

Jupiter in mass, and have a slightly larger radius, including HAT-P-

5b, located right at the center of the transiting planets cluster. Some of the planets, located at the upper left limb of the cluster are low- density, inﬂated planets, including HAT-P-9b and WASP-1b. These planets were a challenge to planetary theory at the time of discovery

(e.g., Burrows et al. 2007; Chabrier & Baraﬀe 2007), and some still are (e.g., TrES-4, Mandushev et al. 2007; WASP-17b, Anderson et

104 al. 2009). The sparse group of high-mass transiting planets is located to the right of the cluster. This interesting group of planets, which includes HAT-P-2b, is intermediate in mass between Jupiter- mass planets and low-mass stars. The radii of these planets show there is only a small variation in the radius of astrophysical objects from Jupiter-like planets through the lowest mass stars, across two orders of magnitude in mass, as expected theoretically (e.g., Baraﬀe et al. 2003, Chabrier et al. 2009). In addition, almost all of the high- mass planets have non-circular orbits, suggesting that the migration process does not decrease the initial eccentricity of high-mass planets, and maybe even increases it. The smallest and lightest planets currently known are located at the bottom left of the ﬁgure, close to Neptune, including GJ 436b and HAT-P-11b (Bakos et al. 2009).

Those planets are hard to detect with any method, so their very small number is an obvious observational bias. Once more Neptune-like planets are discovered, they will allow testing the theoretical radius- mass relation for these small planets (e.g., Baraﬀe et al. 2008), and pave the way to the study of even smaller, terrestrial planets (e.g.,

Chabrier et al. 2009).

The contribution of this work as a whole, as visually illustrated

105 in Fig. 3.3, is the study of transiting exoplanets positioned in several key areas of the radius-mass parameter space.

M [M ] P Sun 0.0001 0.001 0.01 0.1 1 8 ρ = 0.1 ρ = 1.0 ρ = 10.0 0.8

6 0.6

4 0.4 ]

] 2

J 0.2 Sun [R p [R P R R

1 0.1

0.5 0.05

[ρ] = g cm−3

0.1 0.5 1 5 10 50 100 500 M [M ] p J

Figure 3.3: The radius-mass diagram, presenting transiting planets (blue; updated to July 2009), and low-mass stars (red) from Beatty et al. (2007, Table 9). Solar System planets are marked by diamonds (Jupiter, Saturn and Neptune). The X-axis shows mass in Jupiter mass (bottom) and Solar mass (top), while the Y-axis shows radius in Jupiter radius (left) and Solar radius (right). Both axes are in logarithmic scale and the dashed lines are constant density lines. The three planets whose discovery is part of this thesis are encircled in green, and the three for which a follow-up study is included in this thesis are encircled in purple. The planets studied in this work are positioned at various key areas of the radius-mass parameter space.

106 Bibliography

Adams, F. C., & Laughlin, G. 2006, ApJ, 649, 992 Agol, E., Steffen, J., Sari, R., & Clarkson, W. 2005, MNRAS, 359, 567 Alibert, Y., Mordasini, C., Benz, W., & Winisdoerffer, C. 2005, A&A, 434, 343 Alonso, R., et al. 2008, A&A, 482, L21 Alonso, R., Deeg, H. J., Brown, T. M., & Belmonte, J. A. 2004a, Astronomische Nachrichten, 325, 594 Alonso, R., et al. 2004b, ApJ, 613, L153 Anderson, D. R., et al. 2009, arXiv:0908.1553 Armitage, P. J. 2007b, arXiv:astro-ph/0701485 Bakos, G., Noyes, R. W., Kovács, G., Stanek, K. Z., Sasselov, D. D., & Domsa, I. 2004, PASP, 116, 266 Bakos, G. A.,´ et al. 2007a, ApJ, 670, 826 (Paper I) Bakos, G. A.,´ et al. 2007b, ApJ, 656, 552 Bakos, G. A.,´ et al. 2007c, ApJ, 671, L173 (Paper II) Bakos, G. A.,´ et al. 2009, arXiv:0901.0282 Ballard, S., et al. 2009, arXiv:0909.2875 Baraffe, I., Chabrier, G., & Barman, T. 2008, A&A, 482, 315

107 Baraﬀe, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt, P. H. 2003, A&A, 402, 701 Baraﬀe, I., Selsis, F., Chabrier, G., Barman, T. S., Allard, F., Hauschildt, P. H., & Lammer, H. 2004, A&A, 419, L13 Barnes, J. W. 2007, PASP, 119, 986 Baranne, A., et al. 1996, A&AS, 119, 373 Barbieri, M., et al. 2007, A&A, 476, L13 Barge, P., et al. 2008, A&A, 482, L17 Basri, G., & Brown, M. E. 2006, Annual Review of Earth and Plan- etary Sciences, 34, 193 Beatty, T. G., et al. 2007, ApJ, 663, 573 Beaulieu, J.-P., et al. 2006, Nature, 439, 437 Benedict, G. F., et al. 1999, AJ, 118, 1086 Bennett, D. P. 2009, arXiv:0902.1761 Blumenberg, H. 1987, The Genesis of the Copernican World. Cam- bridge, Massachusetts: The MIT Press (translator: Wallace, R. M.) Bodenheimer, P., Hubickyj, O., & Lissauer, J. J. 2000, Icarus, 143, 2 Bodenheimer, P., & Lin, D. N. C. 2002, Annual Review of Earth and Planetary Sciences, 30, 113 Bodenheimer, P., Lin, D. N. C., & Mardling, R. A. 2001, ApJ, 548, 466 Bond, I. A., et al. 2004, ApJ, 606, L155

108 Borucki, W. J., et al. 2003, Scientiﬁc Frontiers in Research on Ex- trasolar Planets, 294, 427 Borucki, W. J., & Summers, A. L. 1984, Icarus, 58, 121 Boss, A. P. 1997, Science, 276, 1836 Boss, A. P. 2000, ApJ, 536, L101 Bouchy, F., & The Sophie Team 2006, Tenth Anniversary of 51 Peg- b: Status of and prospects for hot Jupiter studies, 319 Brosch, N., Polishook, D., Shporer, A., Kaspi, S., Berwald, A., & Manulis, I. 2008, Ap&SS, 314, 163 Brown, T. M. 2001, ApJ, 553, 1006 Brown, T. M., & Charbonneau, D. 2000, Disks, Planetesimals, and Planets, 219, 584 Brown, T. M., Charbonneau, D., Gilliland, R. L., Noyes, R. W., & Burrows, A. 2001, ApJ, 552, 699 Brown, T. M., Libbrecht, K. G., & Charbonneau, D. 2002, PASP, 114, 826 Burrows, A., Hubeny, I., Budaj, J., & Hubbard, W. B. 2007, ApJ, 661, 502 Burrows, A., Hubeny, I., Hubbard, W. B., Sudarsky, D., & Fortney, J. J. 2004, ApJ, 610, L53 Butler, R. P., Marcy, G. W., Williams, E., McCarthy, C., Dosanjh, P., & Vogt, S. S. 1996, PASP, 108, 500 Butler, R. P., et al. 2006, ApJ, 646, 505 Butler, R. P., Marcy, G. W., Vogt, S. S., & Apps, K. 1998, PASP, 110, 1389

109 Butler, R. P., Marcy, G. W., Williams, E., Hauser, H., & Shirts, P. 1997, ApJ, 474, L115 Butler, R. P., Vogt, S. S., Marcy, G. W., Fischer, D. A., Wright, J. T., Henry, G. W., Laughlin, G., & Lissauer, J. J. 2004, ApJ, 617, 580 Cameron, A. G. W. 1962, Icarus, 1, 13 Cameron, A. C., et al. 2007, MNRAS, 375, 951 Campbell, B., & Walker, G. A. H. 1979, PASP, 91, 540 Campbell, B., Walker, G. A. H., & Yang, S. 1988, The Impact of Very High S/N Spectroscopy on Stellar Physics, 132, 79 Campbell, B., Yang, S., Irwin, A. W., & Walker, G. A. H. 1991, Bioastronomy: The Search for Extraterrestial Life – The Explo- ration Broadens, 390, 19 Chabrier, G., & Baraﬀe, I. 2007, ApJ, 661, L81 Chabrier, G., Baraﬀe, I., Leconte, J., Gallardo, J., & Barman, T. 2009, American Institute of Physics Conference Series, 1094, 102 Chambers, J. 2008, Icarus, 198, 256 Chambers, J. E. 2009, Annual Review of Earth and Planetary Sci- ences, 37, 321 Charbonneau, D. 2009, IAU Symposium, 253, 1 (arXiv: 0808.3007) Charbonneau, D., et al. 2005, ApJ, 626, 523 Charbonneau, D., Brown, T. M., Dunham, E. W., Latham, D. W., Looper, D. L., & Mandushev, G. 2004, The Search for Other Worlds, 713, 151 Charbonneau, D., Brown, T. M., Latham, D. W., & Mayor, M. 2000, ApJ, 529, L45

110 Charbonneau, D., Brown, T. M., Noyes, R. W., & Gilliland, R. L. 2002, ApJ, 568, 377 Charbonneau, D., Winn, J. N., Everett, M. E., Latham, D. W., Holman, M. J., Esquerdo, G. A., & O’Donovan, F. T. 2007, ApJ, 658, 1322 Chatterjee, S., Ford, E. B., Matsumura, S., & Rasio, F. A. 2008, ApJ, 686, 580 Christian, D. J., et al. 2009, MNRAS, 392, 1585 Cochran, W. D., Hatzes, A. P., Butler, R. P., & Marcy, G. W. 1997, ApJ, 483, 457 Coughlin, J. L., Stringfellow, G. S., Becker, A. C., L´opez-Morales, M., Mezzalira, F., & Krajci, T. 2008, ApJ, 689, L149 Cox, A. N. 2000, Allen’s Astrophysical Quantities, Cresswell, P., & Nelson, R. P. 2006, A&A, 450, 833 Cresswell, P., & Nelson, R. P. 2008, A&A, 482, 677 Cumming, A., Butler, R. P., Marcy, G. W., Vogt, S. S., Wright, J. T., & Fischer, D. A. 2008, PASP, 120, 531 Davis, T. A., & Wheatley, P. J. 2009, MNRAS, 396, 1012 Deeg, H. J., et al. 2009, A&A, accepted (arXiv:0907.2653) Deleuil, M., Barge, P., Defay, C., L´eger, A., Rouan, D., & Schneider, J. 2000, Disks, Planetesimals, and Planets, 219, 656 Deleuil, M., et al. 2008, A&A, 491, 889 Deming, D., Seager, S., Richardson, L. J., & Harrington, J. 2005, Nature, 434, 740 D´ıaz, R. F., Rojo, P., Melita, M., Hoyer, S., Minniti, D., Mauas, P. J. D., & Ru´ız, M. T. 2008, ApJ, 682, L49

111 Dunham, E. W., Mandushev, G. I., Taylor, B. W., & Oetiker, B. 2004, PASP, 116, 1072 Durisen, R. H., Boss, A. P., Mayer, L., Nelson, A. F., Quinn, T., & Rice, W. K. M. 2007, Protostars and Planets V, 607 Eggenberger, A., & Udry, S. 2009, arXiv:0904.0415 Fabrycky, D., & Tremaine, S. 2007, ApJ, 669, 1298 Fabrycky, D. C., Johnson, E. T., & Goodman, J. 2007, ApJ, 665, 754 Fabrycky, D. C. 2009, IAU Symposium, 253, 173 Fabrycky, D. C., & Winn, J. N. 2009, ApJ, 696, 1230 Fischer, D. A., et al. 2007, ApJ, 669, 1336 Ford, E. B., Quinn, S. N., & Veras, D. 2008, ApJ, 678, 1407 Fortier, A., Benvenuto, O. G., & Brunini, A. 2009, A&A, 500, 1249 Fortney, J. J., Marley, M. S., & Barnes, J. W. 2007, ApJ, 659, 1661 Fortney, J. J., Saumon, D., Marley, M. S., Lodders, K., & Freedman, R. S. 2006, ApJ, 642, 495 Fraser, C. G. 2006, The cosmos: a historical perspective (Greenwood Guides to Great Ideas in Science), Greenwood Press, westport, Connecticut, London Gatewood, G., & Eichhorn, H. 1973, AJ, 78, 769 Gaudi, B. S., et al. 2008, Science, 319, 927 Gaudi, B. S., Seager, S., & Mallen-Ornelas, G. 2005, ApJ, 623, 472 Gaudi, B. S., & Winn, J. N. 2007, ApJ, 655, 550 Gilliland, R. L., et al. 2000, ApJ, 545, L47

112 Gillon, M. 2009, arXiv:0906.4904 Gillon, M., et al. 2007, A&A, 472, L13 Gim´enez, A. 2006, A&A, 450, 1231 Goldreich, P., & Tremaine, S. 1980, ApJ, 241, 425 Gonzalez, G. 1997, MNRAS, 285, 403 Gonzalez, G. 2006, PASP, 118, 1494 Gonzalez, G., Laws, C., Tyagi, S., & Reddy, B. E. 2001, AJ, 121, 432 Guillot, T. 2008, Physica Scripta Volume T, 130, 014023 Guillot, T., & Showman, A. P. 2002, A&A, 385, 156 Guillot, T. 2005, Annual Review of Earth and Planetary Sciences, 33, 493 Haisch, K. E., Jr., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, L153 Hayashi, C. 1981, Progress of Theoretical Physics Supplement, 70, 35 Heap, S. R., Lindler, D. J., Lanz, T. M., Cornett, R. H., Hubeny, I., Maran, S. P., & Woodgate, B. 2000, ApJ, 539, 435 Hebb, L., et al. 2009, ApJ, 693, 1920 H´ebrard, G., et al. 2008, A&A, 488, 763 Henry, G. W., Marcy, G. W., Butler, R. P., & Vogt, S. S. 2000, ApJ, 529, L41 Hoi Lee, M., & Thommes, E. W. 2009, arXiv:0907.4838 Holland, W. S., et al. 1998, Nature, 392, 788 Holman, M. J., & Murray, N. W. 2005, Science, 307, 1288

113 Hoyle, F. 1960, QJRAS, 1, 28 Hubbard, W. B., Fortney, J. J., Lunine, J. I., Burrows, A., Sudarsky, D., & Pinto, P. 2001, ApJ, 560, 413 Ibgui, L., & Burrows, A. 2009, ApJ, 700, 1921 Jackson, B., Greenberg, R., & Barnes, R. 2008, ApJ, 678, 1396 Jiang, I.-G., Yeh, L.-C., Chang, Y.-C., & Hung, W.-L. 2007, AJ, 134, 2061 Jiang, I.-G., Yeh, L.-C., Chang, Y.-C., & Hung, W.-L. 2009, AJ, 137, 329 Johns-Krull, C. M., et al. 2008, ApJ, 677, 657 Johnson, J. A. 2009, PASP, 121, 309 Johnson, J. A., Winn, J. N., Cabrera, N. E., & Carter, J. A. 2009, ApJ, 692, L100 Jones, H. R. A., Butler, R. P., Tinney, C. G., Marcy, G. W., Carter, B. D., Penny, A. J., McCarthy, C., & Bailey, J. 2006, MNRAS, 369, 249 Joshi, Y. C., et al. 2009, MNRAS, 392, 1532 Kalas, P., et al. 2008, Science, 322, 1345 Kalas, P., Graham, J. R., & Clampin, M. 2005, Nature, 435, 1067 Kant, I. 1755, Allegmeine Naturgeschichte und Theorie des Himmels. Koingsburg, Leipzig: J. F. Petersen. Engl. transl. by W. Haste 1968, “Universal Natural History and Theories of the Heavens.” In Kant’s Cosmology. New York: Greenwood Knutson, H. A., Charbonneau, D., Allen, L. E., Burrows, A., & Megeath, S. T. 2008, ApJ, 673, 526 Knutson, H. A., et al. 2007, Nature, 447, 183

114 Knutson, H. A., Charbonneau, D., Burrows, A., O’Donovan, F. T., & Mandushev, G. 2009, ApJ, 691, 866 Konacki, M., Torres, G., Jha, S., & Sasselov, D. D. 2003, Nature, 421, 507 Kov´acs, G., Bakos, G., & Noyes, R. W. 2005, MNRAS, 356, 557 Kov´acs, G., Zucker, S., & Mazeh, T. 2002, A&A, 391, 369 Kozai, Y. 1962, AJ, 67, 591 Laplace, P. S. 1796, Exposition du Systeme du Monde. Engl. transl. by H. H. Harte 1830, The System of the World. Dublin: Dublin Univ. Press Latham, D. W. 1992, IAU Colloq. 135: Complementary Approaches to Double and Multiple Star Research, 32, 110 Latham, D. W., Stefanik, R. P., Mazeh, T., Mayor, M., & Burki, G. 1989, Nature, 339, 38 Laughlin, G., Deming, D., Langton, J., Kasen, D., Vogt, S., Butler, P., Rivera, E., & Meschiari, S. 2009, Nature, 457, 562 Lee, M. H., Butler, R. P., Fischer, D. A., Marcy, G. W., & Vogt, S. S. 2006, ApJ, 641, 1178 Lee, A. T., Thommes, E. W., & Rasio, F. A. 2009, ApJ, 691, 1684 Leger, A., et al. 2009, arXiv:0908.0241 Lin, D. N. C., & Papaloizou, J. 1986, ApJ, 309, 846 Lissauer, J. J. 1993, ARA&A, 31, 129 Loeillet, B., et al. 2008, A&A, 481, 529 (Paper V) Longo, O. 1991, Memorie della Societa Astronomica Italiana, 62, 433 Lovis, C., & Pepe, F. 2007, A&A, 468, 1115

115 Mandel, K., & Agol, E. 2002, ApJ, 580, L171 Mandushev, G., et al. 2007, ApJ, 667, L195 Maness, H. L., Marcy, G. W., Ford, E. B., Hauschildt, P. H., Shreve, A. T., Basri, G. B., Butler, R. P., & Vogt, S. S. 2007, PASP, 119, 90 Marcy, G. W., & Butler, R. P. 1992, PASP, 104, 270 Marcy, G. W., & Butler, R. P. 1996, ApJ, 464, L147 Marcy, G. W., Butler, R. P., Fischer, D., Vogt, S. S., Lissauer, J. J., & Rivera, E. J. 2001, ApJ, 556, 296 Marcy, G., Butler, R. P., Fischer, D., Vogt, S., Wright, J. T., Tinney, C. G., & Jones, H. R. A. 2005, Progress of Theoretical Physics Supplement, 158, 24 Marois, C., Macintosh, B., Barman, T., Zuckerman, B., Song, I., Patience, J., Lafreni`ere, D., & Doyon, R. 2008, Science, 322, 1348 Matsuo, T., Shibai, H., Ootsubo, T., & Tamura, M. 2007, ApJ, 662, 1282 Mayor, M., & Queloz, D. 1995, Nature, 378, 355 Mazeh, T. 2008, EAS Publications Series, 29, 1 Mazeh, T., & Shaham, J. 1979, A&A, 77, 145 Mazeh, T. 2009, IAU Symposium, 253, 11 Mazeh, T., & Goldman, I. 1995, PASP, 107, 250 Mazeh, T., Krymolowski, Y., & Rosenfeld, G. 1997, ApJ, 477, L103 Mazeh, T., Latham, D. W., & Stefanik, R. P. 1996, ApJ, 466, 415 Mazeh, T., et al. 2000, ApJ, 532, L55 Mazeh, T., Zucker, S., & Pont, F. 2005, MNRAS, 356, 955

116 McCullough, P. R., Stys, J. E., Valenti, J. A., Fleming, S. W., Janes, K. A., & Heasley, J. N. 2005, PASP, 117, 783 McCullough, P. R., et al. 2006, ApJ, 648, 1228 McLaughlin, D. B. 1924, ApJ, 60, 22 Miller, N., Fortney, J. J., & Jackson, B. 2009, ApJ, 702, 1413 Miller, M. C., & Hamilton, D. P. 2001, ApJ, 550, 863 Mordasini, C., Alibert, Y., Benz, W., & Naef, D. 2008, Astronomical Society of the Paciﬁc Conference Series, 398, 235 Moutou, C., et al. 2009, A&A, 498, L5 Murray, N., & Chaboyer, B. 2002, ApJ, 566, 442 Naef, D., et al. 2001, A&A, 375, L27 Nagasawa, M., Ida, S., & Bessho, T. 2008, ApJ, 678, 498 Nelson, R. P., Papaloizou, J. C. B., Masset, F., & Kley, W. 2000, MNRAS, 318, 18 Neves, V., Santos, N. C., Sousa, S. G., Correia, A. C. M., & Israelian, G. 2009, A&A, 497, 563 O’Donovan, F. T., et al. 2007, ApJ, 662, 658 O’Donovan, F. T., Charbonneau, D., & Kotredes, L. 2004, The Search for Other Worlds, 713, 169 O’Donovan, F. T., et al. 2006, ApJ, 644, 1237 P´al, A., et al. 2009, arXiv:0908.1705 Papaloizou, J. C. B. 2003, Celestial Mechanics and Dynamical As- tronomy, 87, 53 Pepe, F., et al. 2002, The Messenger, 110, 9

117 Pepe, F. A., & Lovis, C. 2008, Physica Scripta Volume T, 130, 014007 Perryman, M. A. C. 2000, Reports on Progress in Physics, 63, 1209 Phinney, E. S., & Hansen, B. M. S. 1993, Planets Around Pulsars, 36, 371 Podsiadlowski, P. 1993, Planets Around Pulsars, 36, 149 Pollacco, D. L., et al. 2006, PASP, 118, 1407 Pollack, J. B., Hubickyj, O., Bodenheimer, P., Lissauer, J. J., Podolak, M., & Greenzweig, Y. 1996, Icarus, 124, 62 Pont, F. 2009, MNRAS, 396, 1789 Pont, F., Gilliland, R. L., Knutson, H., Holman, M., & Charbonneau, D. 2009a, MNRAS, 393, L6 Pont, F., et al. 2009b, A&A, 502, 695 Pont, F., Knutson, H., Gilliland, R. L., Moutou, C., & Charbonneau, D. 2008, MNRAS, 385, 109 Pont, F., Melo, C. H. F., Bouchy, F., Udry, S., Queloz, D., Mayor, M., & Santos, N. C. 2005, A&A, 433, L21 Pont, F., Zucker, S., & Queloz, D. 2006, MNRAS, 373, 231 Queloz, D., Eggenberger, A., Mayor, M., Perrier, C., Beuzit, J. L., Naef, D., Sivan, J. P., & Udry, S. 2000, A&A, 359, L13 Ribas, I., & Miralda-Escud´e, J. 2007, A&A, 464, 779 Rice, W. K. M., Armitage, P. J., Bonnell, I. A., Bate, M. R., Jeﬀers, S. V., & Vine, S. G. 2003, MNRAS, 346, L36 Rice, W. K. M., Armitage, P. J., & Hogg, D. F. 2008, MNRAS, 384, 1242 Rosenblatt, F. 1971, Icarus, 14, 71

118 Rossiter, R. A. 1924, ApJ, 60, 15 Rowe, J. F., et al. 2008, ApJ, 689, 1345 Sackett, P. D. 1999, NATO ASIC Proc. 532: Planets Outside the Solar System: Theory and Observations, 189 Sahu, K. C., et al. 2006, Nature, 443, 534 Santos, N. C., Israelian, G., Mayor, M., Rebolo, R., & Udry, S. 2003, A&A, 398, 363 Sato, B., et al. 2005, ApJ, 633, 465 Schneider, G., et al. 1999, ApJ, 513, L127 Showman, A. P., & Guillot, T. 2002, A&A, 385, 166 Shporer, A., et al. 2009a, ApJ, 690, 1393 (Paper III) Shporer, A., Bakos, G. A.,´ Mazeh, T., Kov´acs, G., & Sip˝ocz, B. 2009b, IAU Symposium, 253, 331 Shporer, A., Mazeh, T., Moran, A., Bakos, G., & Kovacs, G. 2007a, Transiting Extrapolar Planets Workshop, 366, 99 Shporer, A., Mazeh, T., Moran, A., Bakos, G., Kovacs, G., & Mashal, E. 2006, Tenth Anniversary of 51 Peg-b: Status of and prospects for hot Jupiter studies, 196 Shporer, A., Mazeh, T., Pont, F., Winn, J. N., Holman, M. J., Latham, D. W., & Esquerdo, G. A. 2009c, ApJ, 694, 1559 (Pa- per VI) Shporer, A., Tamuz, O., Zucker, S., & Mazeh, T. 2007b, MNRAS, 376, 1296 (Paper IV) Singer, D. W. 1950, Giordano Bruno His Life and Thought, with Annotated Translation of His Work: On the Inﬁnite Universe and Worlds, Henry Schuman, Inc., New York

119 Smith, B. A., & Terrile, R. J. 1984, Science, 226, 1421 Snellen, I. A. G., Albrecht, S., de Mooij, E. J. W., & Le Poole, R. S. 2008, A&A, 487, 357 Snellen, I. A. G., de Mooij, E. J. W., & Albrecht, S. 2009, Nature, 459, 543 Southworth, J. 2009, MNRAS, 394, 272 Southworth, J., Wheatley, P. J., & Sams, G. 2007, MNRAS, 379, L11 Sozzetti, A., Torres, G., Charbonneau, D., Latham, D. W., Holman, M. J., Winn, J. N., Laird, J. B., & O’Donovan, F. T. 2007, ApJ, 664, 1190 Struve, O. 1952, The Observatory, 72, 199 Tamuz, O., Mazeh, T., & Zucker, S. 2005, MNRAS, 356, 1466 Torres, G., Winn, J. N., & Holman, M. J. 2008, ApJ, 677, 1324 Udalski, A., Kubiak, M., & Szymanski, M. 1997, Acta Astronomica, 47, 319 Udalski, A., et al. 2002a, Acta Astronomica, 52, 1 Udalski, A., Szymanski, M., Kaluzny, J., Kubiak, M., & Mateo, M. 1992, Acta Astronomica, 42, 253 Udalski, A., Zebrun, K., Szymanski, M., Kubiak, M., Soszynski, I., Szewczyk, O., Wyrzykowski, L., & Pietrzynski, G. 2002b, Acta Astronomica, 52, 115 Valenti, J. A., & Fischer, D. A. 2008, Physica Scripta Volume T, 130, 014003 van de Kamp, P. 1963, AJ, 68, 515 van de Kamp, P. 1969a, AJ, 74, 238

120 van de Kamp, P. 1969b, AJ, 74, 757 van de Kamp, P. 1975, AJ, 80, 658 Vogt, S. S. 1987, PASP, 99, 1214 Vogt, S. S., et al. 1994, Proc. SPIE, 2198, 362 Winn, J. N. 2007, Transiting Extrapolar Planets Workshop, 366, 170 Winn, J. N., Holman, M. J., Shporer, A., Fern´andez, J., Mazeh, T., Latham, D. W., Charbonneau, D., & Everett, M. E. 2008a, AJ, 136, 267 Winn, J. N., et al. 2008b, ApJ, 683, 1076 Winn, J. N., et al. 2009, ApJ, 700, 302 Winn, J. N., et al. 2007, ApJ, 665, L167 Winn, J. N., et al. 2005, ApJ, 631, 1215 Wolszczan, A. 1994, Science, 264, 538 Wolszczan, A., & Frail, D. A. 1992, Nature, 355, 145 Wu, Y., & Murray, N. 2003, ApJ, 589, 605 Wuchterl, G., Guillot, T., & Lissauer, J. J. 2000, Protostars and Planets IV, 1081 Wyatt, M. C., Dent, W. R. F., & Greaves, J. S. 2003, MNRAS, 342, 876 Yi, S., Demarque, P., Kim, Y.-C., Lee, Y.-W., Ree, C. H., Lejeune, T., & Barnes, S. 2001, ApJS, 136, 417 Zucker, S., & Mazeh, T. 2002, ApJ, 568, L113 Zucker, S., Mazeh, T., Santos, N. C., Udry, S., & Mayor, M. 2004, A&A, 426, 695 Zuckerman, B., & Song, I. 2004, ApJ, 603, 738

121

Acknowledgments

First and foremost I wish to thank my thesis advisor, Professor Tsevi

Mazeh, for his guidance and support in the course of this work, and for allowing me numerous opportunities to expand my research.

During the years of my Ph.D. study I have beneﬁted from the collaboration and assistance of many faculty members, postdocs, students and other astronomers (in alphabetical order): Suzanne

Aigrain, Gaspar Bakos, Assaf Bervald, Francois Bouchy, Noah Brosch,

Eliel Cohen, Hans Deeg, Michael Gillon, Joel Hartmen, Matt Hol- man, Eﬁ Hoory, Shai Kaspi, Evgeny Gorbykov, Geza Kovacs, Dave

Latham, Benoit Loeillet, Yiftah Lipkin, Ilan Manulis, Dani Maoz,

Haim Mendelson, Philip Nutzman, Eran Ofek, Aviv Oﬁr, David Pol- ishook, Frederic Pont, Oded Spector, Omer Tamuz, Yevgeny Tsodikovich,

Josh Winn, Adi Zitrin & Shay Zucker.

I wish to warmly thank the Wise Observatory staﬀ: Ezra Mashal,

Friedel Loinger, Sammy Ben-Guigui and John Dan.

My work was supported by ISF grant no. 03/323.

123