HII 2407: A Low-Mass Eclipsing Binary Revealed by K2 Observations of the Pleiades
Trevor J. David, John Stauffer, Lynne A. Hillenbrand, Ann Marie Cody, Kyle Conroy, Keivan G. Stassun, Benjamin Pope, Suzanne Aigrain, Ed Gillen, Andrew Collier Cameron, David Barrado, L.M. Rebull, Howard Isaacson, Geoffrey W. Marcy, Celia Zhang, Reed L. Riddle, Carl Ziegler, Nicholas M. Law, Christoph Baranec
DDraft version August 22, 2018
Preprint typeset using L A TEX style emulateapj v. 5/2/11
HII 2407: A LOW-MASS ECLIPSING BINARY REVEALED BY K2 OBSERVATIONS OF THE PLEIADES
Trevor J. David , John Stauffer , Lynne A. Hillenbrand , Ann Marie Cody , Kyle Conroy , Keivan G.Stassun , Benjamin Pope , Suzanne Aigrain , Ed Gillen , Andrew Collier Cameron , David Barrado ,L.M. Rebull , Howard Isaacson , Geoffrey W. Marcy , Celia Zhang , Reed L. Riddle , Carl Ziegler ,Nicholas M. Law , Christoph Baranec Draft version August 22, 2018
ABSTRACTThe star HII 2407 is a member of the relatively young Pleiades star cluster and was previouslydiscovered to be a single-lined spectroscopic binary. It is newly identified here within
Kepler / K M/M (cid:12) and
R/R (cid:12) values with stellar evolutionary models. We alsodemonstrate that the system is likely to be tidally synchronized. Follow-up infrared spectroscopy islikely to reveal the lines of the secondary, allowing for dynamically measured masses and elevating thesystem to benchmark eclipsing binary status. INTRODUCTION
Binary stars, notably double-lined eclipsing binaries,are fundamental astrophysical systems whose study iskey to obtaining accurate empirical measurements of stel-lar radii, masses, and temperatures. These precisely de-rived quantities are necessary for calibrating theoreti-cal models of stars, and understanding stellar evolution.Particularly valuable are well-characterized systems in ei-ther the pre-main sequence or post-main sequence phaseswhere stellar evolution is more rapid, and fundamen-tal calibrators correspondingly more rare relative to themain sequence. Among pre-main-sequence stars, fewerthan 10 systems with masses below 1.5 M (cid:12) have pub-lished orbital solutions and fundamentally derived stellarparameters; see Stassun et al. (2014) and Ismailov et al.(2014) for reviews.The Pleiades cluster (d = 136 . ± . ± [email protected] Department of Astronomy, California Institute of Technol-ogy, Pasadena, CA 91125, USA NSF Graduate Research Fellow Spitzer Science Center, California Institute of Technology,Pasadena, CA 91125, USA NASA Ames Research Center, Mountain View, CA 94035,USA Department of Physics & Astronomy, Vanderbilt University,Nashville, TN 37235, USA Department of Physics, Fisk University, Nashville, TN37208, USA. Department of Physics, University of Oxford, Keble Road,Oxford OX1 3RH, UK SUPA, School of Physics and Astronomy, University of StAndrews, North Haugh, St Andrews, Fife KY16 9SS, UK Centro de Astrobiolog´ıa, INTA-CSIC, Dpto. Astrof´ısica,ESAC Campus, P.O. Box 78, 28691 Villanueva de la Ca˜nada,Madrid, Spain Department of Astronomy, University of California, Berke-ley CA 94720, USA Department of Physics and Astronomy, University of NorthCarolina at Chapel Hill, Chapel Hill, NC 27599-3255, USA Institute for Astronomy, University of Hawai‘i at M¯anoa,Hilo, HI 96720-2700, USA evolved, with a well-populated main sequence between ∼ ∼ M (cid:12) (or M0 through B8 spectral types), andat lower masses the stars are still contracting as pre-mainsequence objects. The K ∼
800 bona fide andcandidate Pleiads.We report here the detection of Pleiades member HII2407 as an eclipsing binary system, and make use of K K2 OBSERVATIONS AND ANALYSIS
The observations took place during K .Although we also produced our own light curve fromaperture photometry, in our final analysis, we use theSimple Aperture Photometry (SAP) light curve availablefrom the Mikulski Archive for Space Telescopes (MAST),which we corrected for systematics and intrinsic stellarvariability.The dominant characteristic of the K ∼
2% amplitude causedby rotational modulation of star spots. Primary eclipsesof ∼
5% depth were detected by inspection of the rawlight curve. Following removal of the spot modulationpattern, a Lomb-Scargle periodogram analysis of the cor-rected light curve yielded an orbital period of 7.05 days.Phase-folding the corrected light curve on this period Data release notes available at http://keplerscience.arc.nasa.gov/K2/C4drn.shtml a r X i v : . [ a s t r o - ph . S R ] O c t David et al.then revealed secondary eclipses of depth < . ∼
71 days, with ∼
30 minute ca-dence, yielding ten 2015 epochs.In addition to the eclipses and star spot variability pat-tern, the light curve displays a saw-tooth-like patterninduced by the roll angle variations of the satellite. Fol-lowing Aigrain et al. (2015), we model the spot- androll- induced variations jointly, using a Gaussian process(GP) model with three components: a time-dependentterm to represent the spot modulation, a term depend-ing on the star’s position on the CCD (as measured viathe centroid) to represent the systematics, and a whitenoise term. This enables us to subtract the systematicsand, where appropriate, the spot-induced variability inorder to study the eclipses.A detailed description of GP regression applied to K x and y , rather than on a 1-D estimate of the roll-anglevariations. The time component was modelled as quasi-periodic to reflect the periodic but evolving nature ofthe spot-induced variations (see § BJD − HII 2407
The star is a classical member of the Pleiades withTrumpler (1921), van Maanen (1945) and Hertzsprung(1947) designations; the last is the name by which it ismost well known: Hz or HII 2407. The K V magnitude of 12.19, and J − K color of0.572. An R ≈ ,
000 spectrum from ∼ − T eff = 5170 K ac-cording to the Pecaut & Mamajek (2013) temperaturescale.Independent analysis of an R ≈ ,
000 spectrum overthe range of 6450-6850 ˚A taken with WIYN/Hydra in De-cember 1999 produces an effective temperature estimate T eff =4970 ±
95 K. This result is based on measurementof 51 lines using the ARES program , and is consistentwith a K2 spectral type from the empirical relations ofPecaut & Mamajek (2013).The star is known as a variable, and was identified bySoderblom et al. (1993a,b) to have Li I α and Ca II triplet core emission,and as a weak x-ray emitter by Stauffer et al. (1994) –all signs of youth that are consistent with the propertiesof many other low mass Pleiades members. From theHIRES spectrum, we measure EW(Li) = 43 . ± . σ contrastlimits of 2 mag, 4.25 mag, and 5.5 mag fainter than thestar at separations from the star of 0.5”, 1.5” and 3.5”,respectively. We use this information below to rule outany “third light” contamination in the eclipse fitting partof our analysis.The star was classified as spotted by Norton et al.(2007) from analysis of its WASP light curve (designation1SWASP J034942.26+242746.8), but the eclipses werenot identified by those authors. Our reanalysis of theWASP light curve using the WASP transit-search algo-rithm confirmed the eclipses and yielded the followingorbital ephemeris: P = 7 . ± . = 2455302 . ± . , consistent with our K Its categorization in SIMBAD as an RS CVn star is not thecorrect interpretation of source properties. n Eclipsing Binary in the Pleiades 3 r e l a t i v e f l u x BJD - 2454833 r e l a t i v e f l u x Figure 1.
Top panel: Systematics corrected K a 7.05 day orbital period with zero eccentricity. This pe-riod is the same as the 7.05 day eclipse period reportedabove from the K K ∼ >
10 km s − betweenthe primary and any putative secondary.We evaluated the rotation period by modelling theout-of-eclipse light curve using a Gaussian process (GP)model with likelihood: L = 1 (cid:112) (2 π ) n | K | exp (cid:18) − y T K − y (cid:19) (1)where y is a vector of n (normalised) flux measure-ments, and the elements of the covariance matrix K aregiven by K ij = k ( t i , t j )= A exp (cid:26) − Γ sin (cid:104) πP | t i − t j | (cid:105) − ( t i − t j ) L (cid:27) + σ δ ( t i − t j ) (2)where A is an amplitude, Γ an inverse length scale, P a period, L an evolutionary time-scale, and σ repre-sents the white noise standard deviation, while δ ( x ) isthe Kronecker delta function. This covariance functiongives rise to a family of functions which display peri-odic but slowly evolving behaviour, and has previouslybeen used to model the light curve of active stars (e.g.Aigrain et al. 2012). The GP model was implementedin Python using the george package (Ambikasaran etal. 2014). To speed up the computation, the light curvewas sub-sampled by selecting 500 data points at ran-dom. The posterior distribution for P was then evaluated(while marginalizing over the other parameters) usingan affine-invariant Markov Chain Monte Carlo (MCMC)implemented in the emcee package (Foreman-Mackey etal. 2013). The priors used were uniform in natural log between −
10 and 10 for all parameters, and we ran 36parallel chains of 700 steps each, discarding the first 200as burn-in. The resulting estimate of the rotation periodis P rot = 7 . ± . K ± K v sin i values from theliterature. Mermilliod et al. (2009) measured 5 . ± . − while Queloz et al. (1998) tabulated 6 . ± . − . From these two values we calculate a primaryradius of R = 0.77 ± (cid:12) in the first case or R = 0.93 ± (cid:12) in the second. Notably, the radiuscalculated from the Stefan-Boltzmann law ( ∼ (cid:12) )assuming our measured T eff and a luminosity from theliterature is consistent within error with the smaller ra-dius estimate above, but not the larger value. We notethat the large uncertainty in the primary radius is dom-inated by the v sin i measurement error.For comparison, from the evolutionary models of Siesset al. (2000) and assumed values of T eff = 4764 K and L = 0.29 L (cid:12) , Wright et al. (2011) estimated a radius of0.74 R (cid:12) and a mass of 0.83 M (cid:12) for HII 2407. Hartmanet al. (2010) reported 0.717 R (cid:12) and 0.817 M (cid:12) from theK-band magnitude and Yi et al. (2001) isochrones whileBouvier (1998a,b) found 0.81 M (cid:12) from the I magnitude,an assumed age of 120 Myr, and Baraffe et al. (1998)models.Ultimately, we adopt the following as the final primaryparameters: spectral type of K2 ± T eff , =4970 ±
95 K,log ( L /L (cid:12) )=-0.54, M =0.81 ± (cid:12) , and R =0.77 ± (cid:12) . Our adopted values for HII 2407 are mostlyconsistent with the highly precise measurements of well-studied double-lined eclipsing binaries. Among compa-rable main sequence systems compiled by Torres et al. David et al. orbital phase r e l a t i v e f l u x rotational phase r e l a t i v e f l u x Figure 2.
Systematics-corrected K § (2010), K1-K3 types have masses in the range 0.764-0.934M (cid:12) , radii of 0.768-0.906 R (cid:12) , T eff in the range 4720-5220K, and luminosities of log ( L/L (cid:12) ) -0.515 to -0.303. Wenote our mass uncertainty of 10% is arbitrary and in-tended to be conservative. For comparison, there is a ∼
7% dispersion in the masses of K1-K3 benchmarks dis-cussed above. HII 2407 has a luminosity slightly lowerthan typical, perhaps owing to the presence of spots givenits relatively young age. ORBITAL PARAMETER FITTING
We used the jktebop orbit-fitting code (Southworth2013, and references therein) to derive the orbital andstellar parameters for the HII 2407 system. The code isbased on the Eclipsing Binary Orbit Program (Popper& Etzel 1981; Etzel 1981), which relies on the Nelson-Davis-Etzel biaxial ellipsoidal model for well-detachedEBs (Nelson & Davis 1972; Etzel 1975). jktebop mod-els the two components as biaxial spheroids for the cal-culation of the reflection and ellipsoidal effects, and asspheres for the eclipse shapes.Our procedure of removing the out-of-eclipse variabil-ity also eliminates gravity darkening, reflected light, andellipsoidal effects from the light curves. As such, parame-ters related to these effects are not included in the jkte-bop modeling. Additionally, out-of-eclipse observationsare excluded in order to reduce the effect these obser-vations have on the χ calculation and to expedite thefitting process. The observational errors were iterativelyscaled by jktebop to find a χ close to 1. A singleoutlier towards the center of secondary eclipse, locatedmore than 3- σ above the eclipse minimum, was deemedsystematic in nature and was excluded from further anal- ysis.The integration times of Kepler long cadence data arecomparable to the eclipse durations, resulting in “phase-smearing” of the light curve. The long exposure timeswere accounted for in jktebop by numerically integrat-ing the model light curves at ten points in a total time in-terval of 1766 seconds, corresponding to the
Kepler longcadence duration.The code finds the best-fit model to a light curvethrough Levenberg-Marquardt (L-M) optimization. Theinitial L-M fitting procedure requires reasonable esti-mates of the orbital parameters to be determined. Pe-riod estimates were obtained using Lomb-Scargle (Lomb1976; Scargle 1982) periodogram analysis. Approxi-mations of the ephemeris timebase, T , were obtainedby manually phase-folding the light curves on the peri-odogram period.Holding the period and ephemeris timebase fixed, ini-tial L-M fits are performed in succession for the remain-ing orbital parameters: the central surface brightness ra-tio, J = ( T eff , /T eff , ) (which can be approximated bythe ratio of the eclipse depths for circular orbits), the sumof the relative radii, ( R + R ) /a , the ratio of the radii, k = R /R , the orbital inclination, i , and the quantities e cos ω and e sin ω , where e and ω are the eccentricityand periastron longitude, respectively. We find an initialestimate for J from the ratio of secondary to primaryeclipse depths, which is ≈ /
13. For circular orbits thisratio approximately corresponds to a temperature ratioof ∼ ∼ v r , and the systemicRV, γ .After successively increasing the number of free pa-rameters in the fit, a final L-M fit was performed al-lowing all relevant parameters to be free. In modelingeach system, we assumed a linear limb-darkening lawfor both components and held the limb-darkening coeffi-cients fixed at 0.7, corresponding to the mean value tab-ulated by Sing (2010) for the Kepler bandpass and solarmetallicity stellar atmospheres with 3500 ≤ T eff ≤ . ≤ log g ≤ . a , b = 0.70, 0.04for the primary (corresponding to the mean values tab-ulated by Claret et al. (2012) for solar metallicity atmo-spheres with 4400 K ≤ T eff ≤ g =4.5) and a , b = 0.41, 0.29 for the secondary (corresponding tothe mean values for 3000 K ≤ T eff ≤ g =5.0).Using a quadratic limb-darkening law in this case pro-vided essentially no improvement to the quality of thelight curve fit. We suggest that grazing eclipses, spotactivity, the quality of the K jktebop model light curve and radial ve-locity curve are presented in Figure 3 with details given inTable 1. The χ of the best fit is 1.04 for the light curvewith out of eclipse observations removed. We also presentin Table 1 the best-fit parameters in the case of an ec-centric orbit (where e cos ω and e sin ω are allowed free),which are completely consistent with the correspondingparameters in the circular orbit solution. The best-fit ec-centricity in this case was e =0.0044 ± χ is minimized byexcluding the three low flux outliers. Moreover, exclusionof the entire bottom of primary eclipse (defined here asthose observations with relative flux values lower than0.955) leads to a best-fit with parameters more similarto those found when excluding just the three low fluxoutliers. Finally, a higher χ is found by forcing the fitto pass through the low flux outliers through excludingonly the cluster of observations occurring just prior tothe primary eclipse minimum in phase.However, given that we know the primary exhibits sig-nificant spot activity with periodicity similar to that ofthe binary orbit, we consider the removal of these out-liers a contrived choice. Furthermore, given the youthof the system, it is likely that the low-mass secondaryis also spotted. In such instances, complicated patternsmay arise during eclipses with contributions from boththe background and foreground stars (e.g. Gillen et al.2014). As such, we choose to include all observationsfrom primary eclipse in our final fit and suggest that theincreased scatter is likely due to spots. We note thatexcluding these three observations changes the best-fittemperature ratio by < ∼ ∼
4% (or < σ ), and theratio of radii by ∼
12% (or < σ ).Independent of the jktebop analysis, we also modeledthe light curve and radial velocities with PHOEBE (Prˇsa& Zwitter 2005). Based on the fact that we could notdetect the secondary component in the HIRES spectrum,we can place an upper limit on the optical flux ratio of ∼ K R + R ) /a = 0.0506 ± a sin i = 2.69 ± (cid:12) , i = 88.09 ± T eff , /T eff , = 0.612 ± T eff , = 3040 ±
60 K. These values are close tothose found by jktebop , though there is a ∼
15% differ-ence in the sum of fractional radii and a ∼
20% discrep-ancy between the temperature ratios favored by the twodifferent codes. These differences suggest the statisticaluncertainties we report in Table 1 may not reflect thetrue uncertainties. A possible etiology of this behavior isthe reliance of PHOEBE on stellar atmospheres to con-vert surface brightness to T eff at cool temperatures, incontrast to jktebop which does not rely on such mod-els. We consider the results of both modeling efforts inthe analysis that follows. DISCUSSION
Simultaneous fitting of the light curve and primary ra-dial velocities yield an RV semi-amplitude of the pri-mary, systemic RV, and binary mass function that areentirely consistent with the values reported in Mermil-liod et al. (1992). We find a binary mass function f ( M )= 0.005521 ± (cid:12) , providing an absolute lowerlimit of ∼ Jup to the mass of the secondary.Since HII 2407 is an SB1 binary, the radial velocitiescontain information only about the projected orbit of theprimary component (i.e. a sin i ) and fail to provide usinformation about the separation between the two starsor the mass ratio, as would be the case in an SB2 binary.Without significant ellipsoidal variations, the light curvecan not constrain the mass ratio and, since the eclipsesare merely grazing, does not provide a strong constrainton the radius ratio either. Instead, the light curve con-tains robust information only about the sum of fractionalradii ( R + R ) /a , the inclination, and the temperatureratio.Nevertheless, auxiliary information about the systemallows for coarse characterization of the secondary. Thebroad-band photometry places HII 2407 close to thePleiades single star locus, which provides a rough con-straint on the mass ratio of q (cid:46) . − .
4. This upperlimit, combined with the lower limit from the preciselymeasured mass function, places the companion firmly inthe ∼ (cid:12) mass range. This, of course, assumingthe primary mass from photometry, which again is con-sistent with benchmark K dwarfs. Moreover, the incli-nation is robustly constrained and given the well-definedrange of dynamical masses for K1-K3 type benchmarkdouble-lined EBs, one can use the radial velocity equa-tion (Lehmann-Filh´es 1894) to obtain a reasonable, andmore precise, approximation for the secondary mass.As noted in Table 1, the upper limit on the flux ra- David et al. r e l a t i v e f l u x primary eclipse -0.015 -0.010 -0.005 0.0 0.005 0.010 0.015 phase -4.0-2.00.02.04.0 O - C ( pp t ) secondary eclipse phase r ad i a l v e l o c i t y ( k m / s ) phase -2.0-1.00.01.02.0 O - C ( k m / s ) Figure 3.
Best-fit jktebop model to the K Kepler data reduction pipeline. The horizontal dashed line in the bottom panel indicates the best-fit systemic radial velocity. tio from the HIRES spectrum can be used to place anupper limit on the radius ratio. However, significantlydifferent limits arise from the different temperature ra-tios favored by jktebop and PHOEBE. An upper limitof R /R < jktebop best-fit J value, while PHOEBE favors a higher temperature ratiothat implies R /R < R =0.77 ± (cid:12) and M = 0 . ± .
08 M (cid:12) ( § R ≈ ± (cid:12) , M ≈ ± (cid:12) (given the primarymass, and the best-fit radial velocity semi-amplitude andinclination). For these parameters and an assumed age of120 Myr, interpolation of (Baraffe et al. 2015, hereafterBHAC15) models predicts temperatures of T = 4975 K and T = 3120 K. The predicted flux ratio of this config-uration is thus F /F ∼ ∼ µ m(L. Prato, private communication). Detection of spectrallines from the secondary is likely possible in the infrared. jktebop modeling suggests a luminosity ratio L /L ≈ k J ≈ M = 0 .
81 M (cid:12) , the luminosity ratio suggests M ≈ (cid:12) , from interpolation among either BHAC15 or Siesset al. (2000) isochrones. We caution that this ratio isstrongly dependent on the poorly constrained ratio ofradii. The best-fit central surface brightness ratio cor-responds to a temperature ratio of T eff , /T eff , = 0.4953 ± Table 1
Best-fit Orbital ParametersParameter Symbol Value 1- σ Error UnitsCentral surface brightness ratio J ± R + R ) /a ± R /R ± a Inclination i ± P ± T ± v r ± − Systemic radial velocity γ ± − Eccentric Orbit Parameters
Central surface brightness ratio J ± R + R ) /a ± R /R ± i ± e cos ω ± e sin ω -0.0044 ± P ± T ± v r ± − Systemic radial velocity γ ± − Note . — Orbital parameters determined from a simultaneous fit of the corrected K σ errors determined from 10,000 Monte Carlo simulations with jktebop .Parameters in the eccentric orbit case were determined from 5,000 Monte Carlo simulations. All parameters in the eccentric caseare consistent within error with the circular orbit fit. a The statistical uncertainty in R /R is not reliable in the absence of a flux ratio measurement due to the intrinsic degeneraciesof EB lightcurves, particularly for circular orbits (see § ∼ ∼ jktebop temperature ratio or ∼ § of T eff , = 2460 ±
90, assuming a 3- σ error in the sur-face brightness ratio. We note this temperature is ∼ ∼ (cid:12) (Hartman et al. 2010).The ∼ ≈ × years for the primary’srotation. While this estimate is somewhat uncertain, itindicates that the synchronization timescale for such asystem should exceed the age of the cluster at rotationperiods longer than 10 days. This is consistent with thestudies of Meibom et al. (2006) and (Marilli et al. 2007),who found synchronized binaries in clusters of compara-ble age only at periods less than ten days.We conclude that there is good theoretical and obser-vational support for the interpretation that the similar-ity between the orbital and photometric periods is causalrather than coincidental, and that the primary’s rotationis tidally locked to the orbit. SUMMARY
We report the discovery of Pleiades member HII 2407as an eclipsing binary. The star was known previously asa spectroscopic binary, and we used the literature radialvelocities combined with new K -1 M/M fl -1 R / R fl Siess et al. (2000)
Torres et al. (2010) EBs 80 Myr120 Myr200 Myr 1 Gyr 10 Gyr -1 M/M fl -1 BHAC15 -1 M/M fl -1 PARSEC
Figure 4.
Isochrones in the mass-radius plane with the components of HII 2407 and benchmark EBs from Torres et al. (2010) overplotted.From left to right, the evolutionary models depicted are from Siess et al. (2000), Baraffe et al. (2015), and Bressan et al. (2012). Allmodels plotted are for solar metallicity (Z=0.02). Unlike the Torres et al. (2010) sample, the masses and radii of the HII 2407 componentsare model-dependent. troscopy, where the flux ratio is more favorable relativeto optical spectroscopy, is likely to reveal the lines ofthe secondary, allowing for dynamically measured massesand elevating the system to benchmark EB status.We thank the referee for suggestions that led to signif-icant improvements in this paper. We thank Lisa Pratofor her estimate of the infrared flux ratio and look for-ward to a direct detection of the secondary. The materialpresented herein is based upon work supported in 2015by the National Science Foundation Graduate ResearchFellowship under Grant No. DGE1144469. T.J.D. grate-fully acknowledges support from France C´ordova throughthe Neugebauer Scholarship. This research was partiallysupported by an appointment to the NASA Postdoc-toral Program at the Ames Research Center, adminis-tered by Oak Ridge Associated Universities through acontract with NASA. Support for this work was providedby NASA via grant NNX15AV62G. This paper includesdata collected by the Kepler mission. Funding for theKepler mission is provided by the NASA Science Mis-sion directorate. Some of the data presented in this pa-per were obtained from the Mikulski Archive for SpaceTelescopes (MAST). STScI is operated by the Associa-tion of Universities for Research in Astronomy, Inc., un-der NASA contract NAS5-26555. Support for MAST fornon-HST data is provided by the NASA Office of SpaceScience via grant NNX09AF08G and by other grants andcontracts. Some of the data presented herein were ob-tained at the W.M. Keck Observatory, which is operatedas a scientific partnership among the California Insti-tute of Technology, the University of California and theNational Aeronautics and Space Administration. TheObservatory was made possible by the generous finan-cial support of the W.M. Keck Foundation. The authorswish to recognize and acknowledge the very significantcultural role and reverence that the summit of MaunaKea has always had within the indigenous Hawaiian com-munity. We are most fortunate to have the opportunityto conduct observations from this mountain. The Robo-AO system was developed by collaborating partner in-stitutions, the California Institute of Technology and theInter-University Centre for Astronomy and Astrophysics,and supported by the National Science Foundation un- der Grant Nos. AST-0906060, AST-0960343 and AST-1207891, the Mt. Cuba Astronomical Foundation and bya gift from Samuel Oschin. C.B. acknowledges supportfrom the Alfred P. Sloan Foundation. A.C.C. acknowl-edges support from STFC grant ST/M001296/1. Fund-ing for WASP comes from consortium universities andfrom UKs Science and Technology Facilities Council.
Facilities: