Table Of Content

Draft version February 25, 2015 PreprinttypesetusingLATEXstyleemulateapjv.5/2/11 K2-3: A NEARBY M STAR WITH THREE TRANSITING SUPER-EARTHS DISCOVERED BY K2 Ian J. M. Crossfield1,12, Erik Petigura2, Joshua E. Schlieder3,13, Andrew W. Howard4, B.J. Fulton4, Kimberly M. Aller4, David R. Ciardi5, Se´bastien Le´pine6, Thomas Barclay3, Imke de Pater2, Katherine de Kleer2, Elisa V. Quintana3, Jessie L. Christiansen5, Eddie Schlafly7, Lisa Kaltenegger11, Justin R. Crepp8, Thomas Henning7, Christian Obermeier7, Niall Deacon9, Lauren M. Weiss2, Howard T. Isaacson2, Brad M. S. Hansen10, Michael C. Liu4, Tom Greene3, Steve B. Howell3, Travis Barman1, Christoph Mordasini7 Draft version February 25, 2015 ABSTRACT 5 Small, cool planets represent the typical end-products of planetary formation. Studying the archi- 1 tecturesofthesesystems,measuringplanetmassesandradii,andobservingtheseplanets’atmospheres 0 during transit directly informs theories of planet assembly, migration, and evolution. Here we report 2 thediscoveryofthreesmallplanetsorbitingabright(K =8.6mag)M0dwarf usingdatacollectedas s b part of K2, the new ecliptic survey using the re-purposed Kepler spacecraft. Stellar spectroscopy and e K2 photometry indicate that the system hosts three transiting planets with radii 1.5 – 2.1 R , strad- ⊕ F dling the transition region between rocky and increasingly volatile-dominated compositions. With orbital periods of 10–45 days the planets receive just 1.5–10× the flux incident on Earth, making 4 2 these some of the coolest small planets known orbiting a nearby star; planet d is located near the inner edge of the system’s habitable zone. The bright, low-mass star makes this system an excellent ] laboratory to determine the planets’ masses via Doppler spectroscopy and to constrain their atmo- P spheric compositions via transit spectroscopy. This discovery demonstrates the ability of K2 and E future space-based transit searches to find many fascinating objects of interest. . Subject headings: K2-3— techniques: photometric — techniques: spectroscopic — eclipses h p - o 1. INTRODUCTION quently (Howard et al. 2012), and their atmospheres are r easier to study when transiting (Stevenson et al. 2010; Surveys for new planets demonstrate that small, low- t s mass planets are common around FGK stars (Howard Kreidbergetal.2014). PlanetstransitingMdwarfsoffer a the best opportunity to study habitability and constrain et al. 2010, 2012). Petigura et al. (2013) used Kepler [ models of rocky planet assembly and migration (Swift data to measure the frequency of Earth-sized planets in et al. 2013; Hansen 2014) and of planetary atmospheres 2 Earth-like orbits to be 5–20%. Such small planets with (Kalteneggeretal.2011;Rodler&López-Morales2014). v moderateinsolationlevels(thestellarenergyreceivedby Multi-planet M dwarf systems are even more exciting, 8 the planet at the top of any atmosphere) are of consid- 9 erable interest for their ability to host Earth-like atmo- both because such candidates are extremely unlikely to 7 spheres that could potentially support life. result from astrophysical false positives (Lissauer et al. 3 M dwarfs offer a shortcut to observing rocky and po- 2012) and because they allow for studies of compara- 0 tentially habitable planets. Compared to nearby Sunlike tive planetology (Muirhead et al. 2012) with identical . initialconditions(i.e.,formationinthesamenataldisk). 1 stars, planets around M dwarfs are easier to find with However,relativelyfewconfirmedtransitingplanets(and 0 transits or radial velocities (RV), they occur more fre- fewer multiple systems) are known around M dwarfs, 5 and the (because Kepler’s prime mission targeted just 1 1Lunar & Planetary Laboratory, University of Arizona 3900 late-type dwarfs) the prevalence of planets around : v Lunar, 1629 E. University Blvd., Tucson, AZ, USA, M dwarfs is less well constrained than around Sunlike Xi [email protected], University of California, Berkeley, stars (Dressing & Charbonneau 2013). CA,USA We are using K2, the continuing mission of NASA’s r 3NASAAmesResearchCenter,MoffettField,CA,USA Kepler spacecraft (Howell et al. 2014), to target thou- a 4Institute for Astronomy, University of Hawaii, 2680 Wood- sands of M dwarfs in each K2 field to find new, small lawnDrive,Honolulu,HI,USA 5NASA Exoplanet Science Institute, California Institute of planets orbiting these stars. K2’s 80-day campaigns Technology,770S.WilsonAve.,Pasadena,CA,USA are ideally suited to finding large numbers of small, 6DepartmentofPhysics&Astronomy,GeorgiaStateUniver- cool planets around M dwarfs, out to semimajor axes sity,Atlanta,GA,USA in the stars’ habitable zones (e.g., Kopparapu et al. 7Max-PlanckInstitutfu¨rAstronomie,Ko¨nigstuhl17,Heidel- berg,Germany 2014). In addition, some of K2’s M-dwarf planets orbit 8Department of Physics, University of Notre Dame, 225 stars bright enough for atmospheric characterization via NieuwlandScienceHall,NotreDame,IN,USA JWST transmission or emission spectroscopy (Kalteneg- 9University of Hertfordshire, College Lane, AL10 9AB, Hat- ger & Traub 2009; Batalha et al. 2013; Beichman et al. field,UK 10DepartmentofPhysics&Astronomy,UniversityofCalifor- 2014). niaLosAngeles,LosAngeles,CA,USA Here, we present the discovery of a new multi-planet 11DepartmentofAstronomy,CornellUniversity,122Sciences systemorbitingabrightMdwarf(K2-3,PMI11293-0127, Drive,Ithaca,NY,USA 12NASASaganFellow UCAC4 443-054906, PPMX 112920.3-012717). We de- 13NASAPostdoctoralProgramFellow scribe our analysis of the K2 photometry and of supple- 2 Crossfield et al. around the telescope boresight. When the spacecraft TABLE 1 reaches a pre-determined limit the spacecraft corrects Stellar Parameters of K2-3 thisrollwithathrusterfire. Asthespacecraftrolls,stars move over different pixels having different sensitivities. Parameter Value Source Thus, motion of the star results in apparent changes in Identifyinginformation stellar brightness. αR.A.(hh:mm:ss) 11:29:20.388 Because a target star traces out similar paths during δ Dec. (dd:mm:ss) -01:27:17.23 EPICID 201367065 each roll of the spacecraft, it is possible to separate out 2MASSID 11292037-0127173 2MASS variations in stellar brightness that are roll angle depen- PhotometricProperties dent, and to remove these variations from the photome- B(mag).......... 13.52±0.06 APASS try. Ourextractionpipelinedrawsheavilyontheworkof V(mag).......... 12.17±0.01 APASS g(mag).......... 12.871±0.030 APASS Vanderburg & Johnson (2014). We begin by computing r(mag).......... 11.582±0.020 APASS the median flux for each frame and adopt this value as i(mag)........... 10.98±0.17 APASS the background flux level. The background flux is sub- J(mag).......... 9.421±0.027 2MASS tracted out on a frame by frame basis. We compute the H(mag)......... 8.805±0.044 2MASS Ks(mag)........ 8.561±0.023 2MASS raw photometry, FSAP, by summing the flux within a W1(mag)........ 8.443±0.022 AllWISE soft-edged circular aperture centered around the target W2(mag)........ 8.424±0.019 AllWISE star. We compute the row and column centroids within W3(mag)........ 8.322±0.021 AllWISE the aperture. SpectroscopicandDerivedProperties µα (masyr−1) 88.3±2.0 Zachariasetal.(2012) On short timescales, spacecraft roll is the dominant µδ (masyr−1) -73.6±2.7 Zachariasetal.(2012)motion term and can be described by a single variable. Barycentricrv(kms−1) 32.6±1 APF,thispaper Weidentifytherolldirectionbycomputingtheprinciple Distance(pc) 45±3 thispaper components of the row and column centroids, x(cid:48) and y(cid:48). EW(Hα)(˚A) 0.38±0.06 EFOSC,thispaper WefitforafunctionthatrelatesF tox(cid:48). Wedescribe Age(Gyr) (cid:38)1 EFOSC,thispaper SAP this trend by F = GP(x(cid:48)), where GP is a Gaussian SpectralType M0.0±0.5V Thispaper. SAP [Fe/H] -0.32±0.13 SpeX,thispaper process having a correlation matrix given by a squared Teff (K) 3896±189 SpeX,thispaper exponential kernel. Fitting the GP(x(cid:48)) is an iterative M∗ (M(cid:12)) 0.601±0.089 SpeX,thispaper process where outliers are identified and removed and R∗ (R(cid:12)) 0.561±0.068 SpeX,thispaper the hyperparameters associated with the squared expo- nentialkernelareadjustedtoyieldtheminimumresidual RMS. mentary imaging and spectroscopic data in Sec. 2. In The algorithm described in Vanderburg & Johnson Sec. 3 we present the results of our analysis of K2-3’s (2014) was developed for the K2 engineering campaign properties and discuss the potential for future observa- (C0), where the time baseline was short enough that tions of this and other systems discovered by K2. drifts in stellar position along the y(cid:48) direction could be 2. OBSERVATIONSANDANALYSIS ignored. During 80 day period of C1 observations, stars movedenoughalongthey(cid:48) directionthattheGP(x(cid:48))de- We identified the high proper motion star PMI11293- terminedusingdataearlyinthecampaignwasno-longer 0127asatargetforourCampaign1proposal(GO103614, anappropriatedescriptionoftheposition-dependentflux PI Crossfield) from the SUPERBLINK proper motion variations. Adopting an approach described in Vander- survey (Lépine & Shara 2005; Lépine & Gaidos 2011). burg (2014), we divided the C1 observations in to six We identified the star as a probable nearby M dwarf nearly equal segments and performed the 1D decorrela- basedonacolorandpropermotionselectionschemeand tion approach described above on each segment individ- selectingalltargetswith(V −J)>2.5,V +5logµ+5< ually. The entire procedure described above is repeated 10, and (6V −7J −3) < 5logµ, where µ is the proper for different aperture radii (2, 3, 4, 5, 6, and 7 pixels). motion. ThestarmatchedthesourceK2-3intheKepler Weselecttheaperturesizethatminimizesthecalibrated input catalog (Huber 2014). K2 then observed this tar- RMS.ForK2-3, acircularaperturewitha4pixelradius get in long-cadence mode during C1, covering 30 May yielded the best calibrated photometry (which is avail- to 21 Aug 2014. Target properties, including optical able as an electronic supplement to this paper). Our and NIR photometry from APASS (Henden et al. 2012), photometry suggests that K2-3 may exhibit photomet- 2MASS(Skrutskieetal.2006), andWISE(Wrightetal. ric variations of (cid:46)1% on week-to-month timescales, but 2010) are summarized in Table 1). K2’slong-termstabilityisconstrainedsufficientlypoorly that we cannot claim evidence for periodic modulation indicative stellar rotation. 2.1. K2 Photometry 2.1.1. Extracting the Photometry 2.1.2. Transit Detection WeextractedthephotometryK2-3fromthepixeldata, We searched through the calibrated and detrended which we downloaded from the MAST. Because K2 only photometry (shown in Fig. 1a) using the TERRA algo- has two functional reaction wheels, the telescope cannot rithm described in Petigura et al. (2013). TERRA iden- maintain the 50-millipixel pointing precision achieved tified a transit candidate having P = 10.056 days and during the prime mission. The dominant drift is roll SNR = 59. We fit this candidate with a Mandel & Agol (2002)modelandsubtractedthebestfitmodelfromthe 14 The star was also identified in programs GO1006, GO1050, photometry. We reran TERRA on the photometry with GO1052,GO1036,GO1075,GO1059,andGO1063. theP =10.056daycandidateremoved. Wefoundasec- A nearby M star with three transiting super-Earths from K2 3 ond candidate having P = 24.641 days and SNR = 30. et al. 2013a). A forthcoming paper will discuss these ef- Again we removed the best-fitting model. TERRA did forts; in brief, we bias-subtract and flat-field the data notfindanyadditionaltransits,buta∼45-daycandidate frames, extract spectra using IRAF, and wavelength- was identified by eye (TERRA currently requires 3 de- calibrate using EFOSC2’s internal HeAr lamps. We tected transits, and thus was not sensitive to the longest achieve a S/N per resolution element of ∼100 for K2- period candidate which only transits twice during C1). 3 and somewhat higher for our reference sample. We We fit each of these two transits individually and find flux-calibrate the extracted spectrum using observations consistenttransitparameters, supportingthehypothesis of spectrophotometric standards. that they result from a single planet. At half of this pe- We observed K2-3 on 2015 January 11 UT using the riod a third transit would occur in C1’s data gap (see recently refurbished SpeX spectrograph (Rayner et al. Fig. 1), but this would give the outer two planets a pe- 2003) on the 3.0m NASA Infrared Telescope Facility riod ratio of just 1.1. The previous record-holder for a (IRTF). The data were taken under clear skies with an close period ratio is the Kepler-36 system (Carter et al. averageseeingof∼0(cid:48).(cid:48)7. Weobservedwiththeinstrument 2012;Winn&Fabrycky2014),whosetwoplanetsexhibit inshortcrossdispersedmode(SXD)usingthe0.3X15” aconsiderablylargerperiodratioof1.17andtransittim- slit. Thissetupprovidessimultaneouswavelengthcover- ing variations of many hours. It is unlikely that such an age from 0.7 to 2.5 µm at a resolution of R≈2000. The unusualsystemwouldliejust45pcaway;inaddition,our extended blue wavelength coverage is a result of the re- dynamical analysis (described below) indicates that this centchipupgradeSpeXreceivedinJuly2014. Thetarget period ratio would be dynamically unstable. We there- wasplacedattwopositionsalongtheslitandobservedin fore conclude that the third planet’s period is ∼45 d. an ABBA pattern for subsequent sky subtraction. The observing sequence consisted of 8 × 40s exposures for a 2.2. Target Validation and Stellar Spectroscopy total integration time of 320s. Once the exposures were stacked, this integration time led to a signal-to-noise of We conducted a number of pixel-level diagnostics and > 140 per resolution element. We obtained standard observed K2-3 using several spectrographs to constrain SpeX calibration frames consisting of flats and arclamp the stellar properties. These observations are described exposures immediately before observing K2-3. below. Thereducedspectraareattachedasanelectronic The SpeX spectrum was reduced using the SpeX- supplementtothispaper,andthederivedparametersare Tool software package (Cushing et al. 2004). SpeXTool listed in Table 1. performs flat-field correction and wavelength calibration from the calibration frames followed by sky subtraction 2.2.1. Pixel-Level and Photometric Data Validation and extraction of the one-dimensional spectrum. Indi- Experienceoverthelastdecadeshowsthattransit-like vidual exposures of the target were combined using the signals must be validated to ensure that they arise from xcombspec routine within SpeXTool. We corrected for true planets, not “false-positive” configurations such as atmospheric absorption and performed flux calibration background eclipsing binaries blended with foreground using the A0V-type star HD 97585 which was observed stars (e.g., Torres et al. 2004, 2011). We therefore im- within 20 minutes and 0.015 airmass of the target. A plement a large number of tests on the pixel-level data telluric correction spectrum was constructed from the andextractedphotometrytoidentifyandweedoutthese spectrum of the A0V using the xtellcor package (Vacca false positives. et al. 2003) and applied to the spectrum of K2-3. This Once transit-like events are identified, TERRA runs a packagealsoperformsfluxcalibration. Separate,telluric- suite of diagnostics to distinguish planets from phenom- corrected SpeX orders were combined and flux matched ena like eclipsing binaries, starspots, and other periodic into a continuous spectrum using the xmergeorders rou- stellar variability. We subject targets passing this first tine. To minimize errors in the spectral slope due to steptoanextensivebatteryoffurthertestswhichsearch changesinseeing,guiding,anddifferentialrefraction,we for blends using an examination of centroid motions in alignedtheslitwiththeparallacticangleandminimized and out of transit, difference imaging analyses, and con- thetimebetweenobservationsofthetargetandstandard structionofpixelcorrelationimages(Brysonetal.2013). star. Prior to performing any spectroscopic analyses, we Thoughwearestilllearninghowtooptimallytunethese also applied corrections for the barycentric velocity of tests to account for K2’s few-pixel pointing variations, the observatory and the measured radial velocity. The validation results for large numbers of targets indicates final, calibrated spectrum is shown in Fig. 2. that the transit-like events identified with K2-3 occur within roughly one pixel (4”) of the target star. When 2.2.3. Stellar Parameters combined with our seeing-limited and adaptive optics Mannetal.(2013b)motivateasetoftemperaturesen- imagingdescribedbelow,asdescribedinSec.2.5wefind sitive spectral indices spanning the visible, J-, H-, and thatK2-3’stransitsarefarmorelikelytobeexplainedby K-bandsthatarecalibratedusingtheMdwarfsampleof amulti-planetsystemthanbynonplanetaryphenomena. Boyajian et al. (2012) with interferometrically measured radii. Weusedtheseindicestoestimatethetemperature 2.2.2. Optical and Infrared Spectroscopy ofK2-3. Wecalculatethemeanofthetemperaturesfrom WeobtainedR∼1500spectrafrom0.6–1.0µmofK2-3 eachofthethreeNIRbandindicesandtheirrmsscatter andanumberofcalibrationobjectsusingNTT/EFOSC2 and find T = 3896±117 K (±148 K systematic error, eff (Buzzoni et al. 1984) on UT 11 Jan 2015 as part of 70- ±189Ktotalerror). Thisrangeofeffectivetemperatures night K2 followup program (PID 194.C-0443, PI Cross- isconsistentwiththatmain-sequenceM0dwarfsofspec- field). Wedrawourcalibratorsfromseveralrecentworks tral type K8V to M0V (Pecaut & Mamajek 2013), using (Boyajian et al. 2012; Pecaut & Mamajek 2013; Mann the modified system which incorporates subtypes K8V 4 Crossfield et al. 1.0010 x u 1.0005 Fl d 1.0000 e z0.9995 ali m0.9990 r o0.9985 N 0.9980 1980 1990 2000 2010 2020 2030 2040 2050 BJD_TBD - 2454833 x1.0010 u b c d Fl 1.0005 d 1.0000 e z0.9995 ali0.9990 m r0.9985 o N0.9980 4 3 2 1 0 1 2 3 4 4 3 2 1 0 1 2 3 4 4 3 2 1 0 1 2 3 4 Hours From Mid-Transit Hours From Mid-Transit Hours From Mid-Transit Fig. 1.— Top: CalibratedK2photometryforK2-3. Verticalticksindicatethelocationsofeachplanets’transits. Bottom: Phase-folded photometryandbest-fitlightcurvesforeachplanet. Fig. 2.— CalibratedIRTF/SpeXspectraofourtargetcomparedtospectralstandards. StellarparametersaretabulatedinTable1. and K9V between K7V and M0V. T of an M dwarf. We estimate these additional fun- eff We adopt the metallicity calibration of Mann et al. damental parameters again using IDL software written (2013a) to remain consistent with our methods for de- by A. Mann16 to calculate radius, mass, and luminosity termining T , and other parameters. We use custom andtheirassociatederrorsusingtherelationsdetailedin eff IDL software provided by A. Mann15 to calculate the Mann et al. (2013a). Using the most conservative T eff metallicity in in the visible, J-, H-, and K-bands follow- errors, we calculate R = 0.561±0.068 R and M = ∗ (cid:12) ∗ ing the calibrations of Mann et al. (2013a). Since our 0.601±0.089 M . These values, and the other funda- (cid:12) SpeXspectrumdoesnotextend<0.7µm,wedonotuse mental parameters of the star, are tabulated in Table 1 thevisiblebandcalibrations. Followingthesuggestionof and are used for subsequent estimates of the individual Mann et al. (2013a), we also discard the J-band metal- planet properties. licity, which is often an outlier. Our final metallicity Independentoftheseparameters,wealsoassignaspec- is the mean of those measured from the H- and K-band tral type to this star using molecular band heads in our relationsandtheerroristhequadraturesumofthemea- opticalandNIRspectra. Intheoptical,theTiO5,CaH2, surement error and systematic error in each band. We and CaH3 indices (Reid et al. 1995; Gizis 1997) are cali- find [Fe/H] = -0.32±0.13. Thus, K2-3’s metallicity is brated for the earliest M dwarfs (Lépine et al. 2003) and sub-solar, broadly consistent with many other nearby, avoidregionsofthespectrumwithheavytelluriccontam- field-age, M dwarfs. ination. Following the most recent spectral type calibra- Mann et al. (2013b) provide empirical calibrations to tionsoftheseindicesbyLépineetal.(2013),ourEFOSC calculate the radii, masses, and luminosities given the spectrumyieldsaspectraltypeofK7.5±0.5, determined 15 https://github.com/awmann/metal 16 https://github.com/awmann/Teff_rad_mass_lum A nearby M star with three transiting super-Earths from K2 5 to a half-subtype scale and assuming a sequence K5-K7- using Bayesian inference and the observed proper mo- M0 (i.e., without the K8 and K9 subdivisions of Pecaut tion, sky coordinates, radial velocity and distance. The & Mamajek 2013). In the NIR, the H O-K2 index mea- probability of K2-3 being a member of one of the known 2 sures water opacity in the K-band, and was calibrated nearby young moving groups is <0.1% given the sky co- to a spectral subtype by Rojas-Ayala et al. (2012). We ordinates, proper motion, and radial velocity. Inclusion calculate this index from our SpeX spectrum and esti- ofthephotometricdistanceestimate(andconservatively mate a spectral type of M0.5±0.5. Lépine et al. (2013) assuming a 20% distance uncertainty) does not change also provide a calibration of the V −J color to spectral the BANYAN II results. Thus we conclude that K2-3 is subtype. Our target has V −J = 2.75, consistent with unlikely to be a member of any of these young moving subtype K7.5 on the scale of Pecaut & Mamajek (2013). groups. The spectroscopic and photometric classifications 2.3. Archival and Adaptive Optics Imaging are all consistent, although the NIR classification is marginally later. Here we average the optical and in- Toruleoutthepresenceofabackgroundstarbeingthe frared results and adopt a spectral type of M0.0±0.5V. source of or diluting the transit events, we compare two Using the riJHK photometric calibrations of Kraus & epochs of imaging data from the Digitized Sky Survey Hillenbrand (2007), we estimate a distance to K2-3 of (DSS) and the Sloan Digital Sky Survey (SDSS; Ahn 45±3 pc. et al. 2012) separated by 45 years. The data shown We obtained high-resolution (2” slit width with the in Fig. 3 are the DSS-Red plates with a pixel scale of B decker) spectra of K2-3 with the Levy Spectrometer 1.7(cid:48)(cid:48)/pixel taken on 19 April 1955 and the SDSS r-band (Radovan et al. 2010) on the Automated Planet Finder image with a pixel scale of 0.396(cid:48)(cid:48)/pixel taken on 03 (APF)telescope(Vogtetal.2014). Thespectrawerere- March 2000. The images are 1 arcminute on a side and ducedusingstandardprocedures,asdescribedin(Fulton clearly show the proper motion of the primary target. etal.2015). Inspectionofthegravity-sensitivelinescon- Thenearbystarlocated27(cid:48)(cid:48) totheNEisconsistentwith firms that K2-3 is a high gravity target, consistent with zero motion within our astrometric uncertainties; this the medium resolution spectra described above. We do star lies outside the photometric aperture applied to the not see any evidence of a second set of spectral lines, K2 photometry. The primary target, in contrast, dis- ruling out companions ∼2.5 mag fainter than K2-3 at plays a clear proper motion of 6.2(cid:48)(cid:48) over 45 years, in visible wavelengths. reasonable agreement with the measured proper motion (Lépine & Gaidos 2011; Zacharias et al. 2012). In the DSS image there is no evidence of a background star, 2.2.4. Activity, Age, and Membership andweestimateifastarislocatedatthepositionofthe Lines in the Balmer series are associated with mag- primary target in the Kepler data, that star must be at netic activity in late-type stars. The strongest line in least6magnitudes(ormore)fainterthanthetargetstar. the series, Hα at 6563 ˚A, is classically used to asses the Near-infrared adaptive optics imaging of K2-3 was ob- activityofMdwarfsandasacrudeindicatorofage(West tained at Keck Observatory on the nights of 2015 Jan- et al. 2004, 2008). We therefore measure the Hα equiv- uary 12 UT and 2015 Janary 16 UT. Observations were alent width (EW) as defined by West et al. (2011) and obtainedwiththe1024×1024NIRC2arrayandthenat- Lépineetal.(2013)andfindconsistentresultsusingboth uralguidestarsystem;thetargetstarwasbrightenough approaches. We use two different integration regions to to be used as the guide star. The data were acquired in calculate this EW and apply Monte Carlo methods to the narrow-band K-band and J-band continuum filters estimate the uncertainty in the EW measurements. We (KcontandJcont)usingthenarrowcamerafieldofview find that EW = 0.38±0.06˚A, indicating that K2-3 is a with a pixel scale of 9.942 mas pix−1; the atmosphere relatively inactive star. We further investigate possible was less stable on night 1 and only Kcont was acquired chromosphericactivityinK2-3byanalyzingitsUVemis- on that night. A 3-point dither pattern was utilized to sionmeasuredbyGALEX(Martinetal.2005). Thestar avoid the noisier lower left quadrant of the NIRC2 ar- is a weak near-UV (NUV) emitter and is not detected in ray. For both nights, the 3-point dither pattern was ob- the far-UV (FUV). Its low NUV flux and non-detection served with 10 coadds and a 1.5 integration time, but in the FUV is consistent with quiescent emission, simi- on night 1 only 4 frames were acquired for a total of 60 lartoothernearbyfieldMdwarfs(Shkolniketal.2011). seconds of on-source exposure time. For night 2, three TheHαabsorption,UVfluxes,andlackofchromospheric full dither patterns were acquired for a total on-source activityinanM0dwarfallindicateanold, field-agestar exposure time of 135 seconds in both Kcont and Jcont andtranslatestoaloweragelimitof∼1Gyr(Westetal. filters. The data from each night were flatfielded and 2008). sky subtracted and the dither positions were shifted and WefurtherexaminedthepossibilitythatK2-3isyoung coadded into a single final image. The final images from by comparing its space position (XYZ) and kinematics night 2 are shown in Fig. 3c and d. (UVW)withthoseofknownyoungmovinggroups. Its6- For night 1, the target star was measured with a reso- dimensionalUVWXYZ positionisinconsistentwiththe lution of 0.07(cid:48)(cid:48) (FWHM), but the atmosphere was much well-knownnearbyyoung(≈10–100Myr)groupssumma- morestableduringnight2andtheseimageshaveareso- rized by Gagné et al. (2014), as well as other sparser or lution of 0.05(cid:48)(cid:48) in the Kcont filter and 0.04(cid:48)(cid:48) in the Jcont slightly older groups (Shkolnik et al. 2009; Zuckerman filter. No other stars were detected within the 10(cid:48)(cid:48) field etal.2013). Toprovideaquantitativeestimate, weused of view of the camera; speckles seen in the Kcont images theBANYANIIwebtool(Maloetal.2013;Gagnéetal. arenotco-spatialwiththespecklesseenintheJcontim- 2014). BANYAN II calculates the probability of an ob- age, indicating that the speckles are not faint compan- ject being a member of a nearby young moving group ions. The night 2 data were much more sensitive than 6 Crossfield et al. thenight1dataandwereporttheanalysisofthosedata ble1)islargeenoughthatopticalDSSsurveyimagesre- inthiswork. IntheKcontfilter,thedataaresensitiveto veal no objects at the star’s current location (see Fig. 3a stars that have K-band brightness of ∆mag = 2.4 mag and b). Our Keck/NIRC2 images also show no compan- at a separation of 0.05(cid:48)(cid:48) and ∆mag = 8.0 mag at a sep- ions at separations down to a fraction of an arc second aration of 0.5(cid:48)(cid:48) from the central star; in the Jcont filter, (see Fig. 3c and d), and so our data validation tests in- the data are sensitive to stars that have J-band bright- dicatethatthetransitsmustoccuraroundK2-3andnot ness of ∆mag = 2.0 mag at a separation of 0.05(cid:48)(cid:48) and around some other nearby star. Blends involving back- ∆mag = 7.5 mag at a separation of 0.5(cid:48)(cid:48) from the cen- ground eclipsing binaries are thus strongly disfavored. tralstar(seeFig.3candd). Weestimatethesensitivities The most likely remaining false positive configuration by injecting fake sources with a signal-to-noise of 5 into involvesaheirarchicaltriplesystem, withalater-typeM thefinalcombinedimagesatdistancesofN*FWHMfrom dwarfclosetoK2-3andwithitsowntransitingplanet(s) the central source. The 5σ sensitivities, as a function of — but this too is extremely unlikely. An M4 dwarf radius from the star are shown in Fig. 3c and d. would have ∆K ≈ 2.7 and so might be missed in p our APF and EFOSC spectra, but the M4 would have 2.4. Light Curve Fitting ∆Ks ≈2.0 (Kraus & Hillenbrand 2007) and so to avoid detection in our Keck/NIRC2 image it would need to lie We analyze the photometry using standard Python- at a(cid:46)2.3 AU — while still needing to host its own 2R basedminimizers,theemceeMarkovChainMonte-Carlo ⊕ transiting planet. The likelihood that K2-3 has such a (MCMC) package (Foreman-Mackey et al. 2013), and low-massstellarcompanionis∼0.4andthatsuchacom- the JKTEBOP lightcurve code (Southworth et al. 2004; panion would lie at a projected separation < 2.3 AU is Southworth2011)usingnumericalintegrationtoaccount ∼0.5 (Duchêne & Kraus 2013). For planet b, the like- for our ∼30-min cadence. We fit each planet’s transit lihood of an M dwarf hosting such a planet is (cid:46) 0.15 separately, after first masking out data taken during the (Dressing&Charbonneau2013); andthelikelihoodofit other planets’ transits. transiting is ∼0.02. Then the likelihood of such a con- Weusethebest-fitTERRAparameterstoinitializethe trived configuration is just ∼6×10−4 (1 in 1700), so we fits. Weassumedalinearlimb-darkeningrelationforthe eliminatethisscenarioaswell–inanycase,suchabinary star. Because the data are insufficient to break all de- would be quickly revealed by even crude radial velocity generaciesbetweenthelightcurveparameters(Muirhead measurements. We therefore conclude that K2-3 indeed et al. 2012), we impose Gaussian priors in our analysis. hosts a three-planet system. For the limb-darkening parameter u, we assume a distribution with center 0.560 and dispersion 0.044; these 2.6. System Stability values correspond to the mean and standard deviation, respectively,ofalllinearlimb-darkeningtermstabulated Here we investigate the dynamical stability of the by(Claretetal.2012)thatsatisfy3300≤T ≤3700K three-planetK2-3system. Theplanetmassesareuncon- eff and log g ≥ 4.5. Using the spectroscopic parameters strainedbytransitphotometry,soweadoptthefollowing 10 presented below (Table 1), we also impose a prior on mass-radius relationship: the stellar density to constrain R /a (Seager & Mallén- Ornelas 2003). This last point ass∗umes that the planets’ • M = 43πR3ρ, where ρ=(2.43+3.39∗(RP/R⊕)) g orbits are circular, an assumption that future RV mea- cm−3 for R <1.5R (Weiss & Marcy 2014) P ⊕ surements will test. We seed our 60 MCMC chains with values near the (cid:16) (cid:17)0.93 best-fit parameters. We assign our data points equal • M = 2.69M⊕ RR⊕P (Weiss & Marcy 2014) for weights, such that the best-fit likelihood equals −χ2/2. 1.5R⊕ <RP <4.0R⊕ After burn-in we run the MCMC sampler: after each set opfar2a0m00etsetresptso, wcheeocpktifmorizbeetthteerfifittssgtiovetnhebydaeatach. cWheainre’s- • M =M⊕(cid:16)RR⊕P(cid:17)2.06forRP >4.0R⊕(Lissaueretal. initialize the sampler and re-scale the data weights if 2012) we find an improved fit, repeating until all parameters’ Adopting the above mass-radius relationship we derive chains are well-mixed (as indicated by Gelman-Rubin masses of 5.3, 4.3, and 4.4 M for planets b, c, and d metrics ≤ 1.03; Gelman & Rubin 1992). As our final ⊕ respectively. We integrate the system forward in time confidence intervals, we use the 15.87% and 84.13% per- with the Mercury integration package (Chambers 1999) centiles of each parameters’ posterior distribution. The utilizing the hybrid integrator and found the system to finaldistributionsareunimodal. Fig.1showstheresult- be stable for the full 2×105 yr simulation. ing photometry and best-fit models, and Table 2 sum- We also evaluate analytically the system’s stability. marizes the final values and uncertainties. The relevant length scale for dynamical interactions between planets is the mutual Hill radius: 2.5. Ruling Out False Positives AlmostallcandidatesinKepler’smulti-planetsystems (cid:20)M +M (cid:21)1/3 a +a R = in out in out (1) are bona fide planets (Lissauer et al. 2011), but one per- H 3M 2 (cid:63) nicioussourceofconfusionisthepossibilityofmistaking blended stars each hosting their own planets for a sin- where M and a denote mass and semi-major axis, re- gle multi-planet system. We therefore investigated the spectively. The subscripts “in” and “out” correspond to possibility that K2-3 might be a blend of multiple stars. the inner an outer planets respectively. Following Fab- First, we note that K2-3’s proper motion (listed in Ta- rycky et al. (2012), for each pair of planets, we compute A nearby M star with three transiting super-Earths from K2 7 J band K band (c) (a) (b) (c) (d) Fig. 3.— Wedetectnoobjectswithin25”ofK2-3: inDSS(a);inSDSS(b);andwithKeck/NIRC2inJ-band(c)andK-band(d). TABLE 2 Planet Parameters Parameter Units b c d T0 BJDTDB−2454833 1980.4189+−00..00001111 1979.2786+−00..00002267 1993.2232+−00..00003473 P d 10.05403+0.00026 24.6454+0.0013 44.5631+0.0063 −0.00025 −0.0013 −0.0055 i deg 89.28+0.46 89.55+0.29 89.68+0.21 −0.60 −0.44 −0.26 RP/R∗ % 3.483+−00..102730 2.786+−00..104833 2.48+−00..1140 T14 hr 2.553+−00..004474 3.428+−00..100967 3.98+−00..1175 R∗/a – 0.0343+−00..00004290 0.0193+−00..00004114 0.0127+−00..00002150 b – 0.37+0.22 0.41+0.26 0.45+0.23 −0.23 −0.25 −0.28 u – 0.560+0.041 0.557+0.043 0.563+0.041 −0.042 −0.044 −0.042 a AU 0.0769+0.0036 0.1399+0.0066 0.2076+0.0098 −0.0040 −0.0073 −0.0108 RP R⊕ 2.14+−00..2276 1.72+−00..2232 1.52+−00..2210 Sinc S⊕ 11.0+−43..11 3.32+−10..2955 1.51+−00..5473 Teq (K) ∼800 ∼450 ∼300 ∆ = (a −a )/R , the separation between the plan- of modern RV spectrographs. These mass estimates out in H ets measured in units of their mutual Hill radii. If two assume that the planets fall on the mean mass-radius planets begin on circular orbits, they are stable indefi- relationship, characterized by high densities and rocky √ nitely if ∆>2 3≈3.5 Gladman (1993). In the case of compositions for planets smaller than ∼1.6 RE. How- K2-3, ∆ = 15.9 and ∆ = 11.0. Thus, the two pairs ever, most of the planets with measured masses and bc cd of adjacent planets do not violate the criterion of Hill Rp <1.6RE havehighincidentfluxes(e.g.,Batalhaetal. stability. 2011; Howard et al. 2013; Pepe et al. 2013). The mass- Thereisnoanalyticstabilitycriterionforsystemshav- radius relationship is poorly constrained for cool planets ing three or more planets Fabrycky et al. (2012). Fab- that are less likely to be sculpted by thermal evolution rycky et al. (2012) introduce ∆ + ∆ , as a heuris- and photo-evaporation (Lopez et al. 2012). Character- in out tic metric for assessing the stability of three planets izing the mass-radius relationship for these cool, small in triple or higher multiplicity systems. They adopt planets is an important step to learning whether Earth- ∆ +∆ >18 as a heuristic criterion for the stability size planets in the habitable zone also have Earth-like in out of three planets, motivated by suites of direct numerical atmospheres. integrations (e.g. Smith & Lissauer (2009)). This cri- Theplanets’receiveinsolationlevels(Sinc)roughly11, terion is empirically supported by the ensemble of sys- 3.3, and 1.5× that of the Earth for planets b, c, and d, tems with three or more transiting planets from the Ke- respectively. Planet d is located at the inner edge of the pler mission. Among the 413 such systems in Fabrycky system’s habitable zone, with S =1.51+0.57S – close inc −0.47 ⊕ et al. (2012), only six had ∆in +∆out < 18. For K2-3, to the limits of the empirical habitable zone (e.g., Kop- ∆bc +∆cd = 26.9, and thus has a similar architecture parapu et al. 2014)– making this planet a very interest- to the ensemble of triple and higher systems discovered ing potential super-Venus or super-Earth. Because this during the prime Kepler mission. system is so close the atmosphere of this planet can be explored in the near future; depending on atmospheric, 3. DISCUSSION cloud, and surface properties liquid water could poten- Ouranalysisindicatesthreesmallplanetsorbitingthis tiallypersistonplanetc(Zsometal.2013,butseeKast- bright, nearby M dwarf. The planets range in size from ing et al. 2014). 2R to 1.5R , indicating that they may span the gap The K2-3 system is a convenient system to measure ⊕ ⊕ between rock-dominated “Earths”/“super-Earths” and the atmospheric properties of small, cool planets. In- low-density “sub-Neptunes” with considerable volatile deed, the star is a full magnitude brighter than Kepler- content (Marcy et al. 2014; Rogers 2014; Dressing et al. 138 (Kipping et al. 2014), the previous best system for 2014). characterizingcool,nearlyEarth-sizeplanets. Forcloud- The planets’ radii imply masses of roughly 4–5 M free,hydrogen-dominatedatmospheres,weestimatethat E and Doppler amplitudes of 1.2–2.3 ms−1, within reach 8 Crossfield et al. theseplanetswillshowspectralfeatureswithamplitudes Bryson and our referee Don Pollacco for useful com- of10HR /R2 ontheorderof100–200ppm(Miller-Ricci ments that improved the quality of this manuscript. p (cid:63) et al. 2009), where H is the atmospheric scale height. A.W.H. acknowledges NASA grant NNX12AJ23G, and These features would be detectable with current instru- S.L. acknowledges NSF grant AST 09-08419. This work mentation on the Hubble Space Telescope (Kreidberg made use of the SIMBAD database (operated at CDS, et al. 2014). Transit features in a heavy atmosphere Strasbourg, France), NASA’s Astrophysics Data System (e.g., N , CO ) would be an order of magnitude smaller, Bibliographic Services, the Authorea collaborative writ- 2 2 and secondary eclipses will have depths on the order of ing website, the NASA Exoplanet Archive and Infrared (R /R )2T /T ∼ 50–150 ppm – either of these scenar- Science Archive, and data products from the Two Mi- p (cid:63) eq (cid:63) ios should be detectable with JWST. By allowing us to cron All Sky Survey (2MASS), the APASS database, measure masses and atmospheric conditions for 3 small the SDSS-III project, the Digitized Sky Survey, and the planets in a single system, K2-3 represents an exciting Wide-Field Infrared Survey Explorer. Portions of this opportunitytotesttheoriesofplanetformationandevo- work were performed at the California Institute of Tech- lution in a single extrasolar laboratory. nologyundercontractwiththeNationalAeronauticsand That K2 should reveal such a system in its first full SpaceAdministration. Someofthedatapresentedherein campaign demonstrates that the mission will extend were obtained at the W.M. Keck Observatory (which is Kepler’s compelling scientific legacy for years to come. operated as a scientific partnership among Caltech, UC, Along with HIP 116454 (Vanderburg et al. 2014), the andNASA)andattheInfraredTelescopeFacility(IRTF, discovery of K2-3 shows that K2 is already finding fasci- operated by UH under NASA contract NNH14CK55B). nating new targets for observation with JWST and her- Theauthorswishtorecognizeandacknowledgethevery alds an era of further unprecedented discoveries in the significantculturalroleandreverencethatthesummitof TESS era. MaunaKeahasalwayshadwithintheindigenousHawai- ian community. We are most fortunate to have the opportunity to conduct observations from this mountain. Acknowledgements We thank Geoff Marcy, Evan Facility: Kepler, K2, IRTF (SPEX), NTT (EFOSC2), Sinukoff, and Charles Beichman for helpful conversa- APF (Levy), Keck-II (NIRC2) tions, Vishnu Reddy for swapping SpeX time, and Steve REFERENCES Ahn,C.P.,Alexandroff,R.,AllendePrieto,C.,etal.2012,ApJS, Howard,A.W.,Marcy,G.W.,Bryson,S.T.,etal.2012,ApJS, 203,21 201,15 Batalha,N.,Kalirai,J.,Lunine,J.,Clampin,M.,&Lindler,D. Howard,A.W.,Sanchis-Ojeda,R.,Marcy,G.W.,etal.2013, 2013,TransitingExoplanetSimulationswiththeJamesWebb Nature,503,381 SpaceTelescope,Tech.rep. Howell,S.B.,Sobeck,C.,Haas,M.,etal.2014,PASP,126,398 Batalha,N.M.,Borucki,W.J.,Bryson,S.T.,etal.2011,ApJ, Huber,D.2014,K2:ExtendingKepler’sPowertotheEcliptic. 729,27 EclipticPlaneInputCatalog,Tech.rep. Beichman,C.,Benneke,B.,Knutson,H.,etal.2014, Kaltenegger,L.,Segura,A.,&Mohanty,S.2011,ApJ,733,35 arXiv:1411.1754 Kaltenegger,L.,&Traub,W.A.2009,ApJ,698,519 Boyajian,T.S.,vonBraun,K.,vanBelle,G.,etal.2012,ApJ, Kasting,J.F.,Kopparapu,R.,Ramirez,R.M.,&Harman,C.E. 757,112 2014,ProceedingsoftheNationalAcademyofScience,111, Bryson,S.T.,Jenkins,J.M.,Gilliland,R.L.,etal.2013,PASP, 12641 125,889 Kipping,D.M.,Nesvorny´,D.,Buchhave,L.A.,etal.2014,ApJ, Buzzoni,B.,Delabre,B.,Dekker,H.,etal.1984,TheMessenger, 784,28 38,9 Kopparapu,R.K.,Ramirez,R.M.,SchottelKotte,J.,etal.2014, Carter,J.A.,Agol,E.,Chaplin,W.J.,etal.2012,Science,337, ApJ,787,L29 556 Kraus,A.L.,&Hillenbrand,L.A.2007,AJ,134,2340 Chambers,J.E.1999,MNRAS,304,793 Kreidberg,L.,Bean,J.L.,Désert,J.-M.,etal.2014,Nature,505, Claret,A.,Hauschildt,P.H.,&Witte,S.2012,A&A,546,A14 69 Cushing,M.C.,Vacca,W.D.,&Rayner,J.T.2004,PASP,116, Lépine,S.,&Gaidos,E.2011,AJ,142,138 362 Lépine,S.,Hilton,E.J.,Mann,A.W.,etal.2013,AJ,145,102 Dressing,C.D.,&Charbonneau,D.2013,ApJ,767,95 Lépine,S.,&Shara,M.M.2005,AJ,129,1483 Dressing,C.D.,Charbonneau,D.,Dumusque,X.,etal.2014, Lépine,S.,Shara,M.M.,&Rich,R.M.2003,AJ,126,921 arXiv:1412.8687 Lissauer,J.J.,Fabrycky,D.C.,Ford,E.B.,etal.2011,Nature, Duchêne,G.,&Kraus,A.2013,ARA&A,51,269 470,53 Fabrycky,D.C.,Ford,E.B.,Steffen,J.H.,etal.2012,ApJ,750, Lissauer,J.J.,Marcy,G.W.,Rowe,J.F.,etal.2012,ApJ,750, 114 112 Foreman-Mackey,D.,Hogg,D.W.,Lang,D.,&Goodman,J. Lopez,E.D.,Fortney,J.J.,&Miller,N.2012,ApJ,761,59 2013,PASP,125,306 Malo,L.,Doyon,R.,Lafrenière,D.,etal.2013,ApJ,762,88 Fulton,Weiss,&Sinukoff,e.a.2015,ApJ,submitted Mandel,K.,&Agol,E.2002,ApJ,580,L171 Gagné,J.,Lafrenière,D.,Doyon,R.,Malo,L.,&Artigau,E´. Mann,A.W.,Brewer,J.M.,Gaidos,E.,Lépine,S.,&Hilton, 2014,ApJ,783,121 E.J.2013a,AJ,145,52 Gelman,A.,&Rubin,D.B.1992,StatisticalScience,7,457 Mann,A.W.,Gaidos,E.,&Ansdell,M.2013b,ApJ,779,188 Gizis,J.E.1997,AJ,113,806 Marcy,G.W.,Isaacson,H.,Howard,A.W.,etal.2014,ApJS, Gladman,B.1993,Icarus,106,247 210,20 Hansen,B.M.S.2014,arXiv:1403.6553 Martin,D.C.,Fanson,J.,Schiminovich,D.,etal.2005,ApJ, Henden,A.A.,Levine,S.E.,Terrell,D.,Smith,T.C.,&Welch, 619,L1 D.2012,JournaloftheAmericanAssociationofVariableStar Miller-Ricci,E.,Seager,S.,&Sasselov,D.2009,ApJ,690,1056 Observers(JAAVSO),40,430 Muirhead,P.S.,Johnson,J.A.,Apps,K.,etal.2012,ApJ,747, Howard,A.W.,Marcy,G.W.,Johnson,J.A.,etal.2010, 144 Science,330,653 Pecaut,M.J.,&Mamajek,E.E.2013,ApJS,208,9 A nearby M star with three transiting super-Earths from K2 9 Pepe,F.,Cameron,A.C.,Latham,D.W.,etal.2013,Nature, Torres,G.,Konacki,M.,Sasselov,D.D.,&Jha,S.2004,ApJ, 503,377 614,979 Petigura,E.A.,Howard,A.W.,&Marcy,G.W.2013, Torres,G.,Fressin,F.,Batalha,N.M.,etal.2011,ApJ,727,24 ProceedingsoftheNationalAcademyofScience,110,19273 Vacca,W.D.,Cushing,M.C.,&Rayner,J.T.2003,PASP,115, Radovan,M.V.,Cabak,G.F.,Laiterman,L.H.,Lockwood, 389 C.T.,&Vogt,S.S.2010,inSocietyofPhoto-Optical Vanderburg,A.2014,arXiv:1412.1827 InstrumentationEngineers(SPIE)ConferenceSeries,Vol.7735, Vanderburg,A.,&Johnson,J.A.2014,PASP,126,948 SocietyofPhoto-OpticalInstrumentationEngineers(SPIE) Vanderburg,A.,Montet,B.T.,Johnson,J.A.,etal.2014, ConferenceSeries,4 arXiv:1412.5674,arXiv:1412.5674 Rayner,J.T.,Toomey,D.W.,Onaka,P.M.,etal.2003,PASP, Vogt,S.S.,Radovan,M.,Kibrick,R.,etal.2014,PASP,126,359 115,362 Weiss,L.M.,&Marcy,G.W.2014,ApJ,783,L6 Reid,I.N.,Hawley,S.L.,&Gizis,J.E.1995,AJ,110,1838 West,A.A.,Hawley,S.L.,Bochanski,J.J.,etal.2008,AJ,135, Rodler,F.,&López-Morales,M.2014,ApJ,781,54 785 Rogers,L.A.2014,arXiv:1407.4457 West,A.A.,Hawley,S.L.,Walkowicz,L.M.,etal.2004,AJ, Rojas-Ayala,B.,Covey,K.R.,Muirhead,P.S.,&Lloyd,J.P. 128,426 2012,ApJ,748,93 West,A.A.,Morgan,D.P.,Bochanski,J.J.,etal.2011,AJ,141, Seager,S.,&Mallén-Ornelas,G.2003,ApJ,585,1038 97 Shkolnik,E.,Liu,M.C.,&Reid,I.N.2009,ApJ,699,649 Winn,J.N.,&Fabrycky,D.C.2014,ArXive-prints, Shkolnik,E.L.,Liu,M.C.,Reid,I.N.,Dupuy,T.,& arXiv:1410.4199 Weinberger,A.J.2011,ApJ,727,6 Wright,E.L.,Eisenhardt,P.R.M.,Mainzer,A.K.,etal.2010, Skrutskie,M.F.,Cutri,R.M.,Stiening,R.,etal.2006,AJ,131, AJ,140,1868 1163 Zacharias,N.,Finch,C.T.,Girard,T.M.,etal.2012,VizieR Smith,A.W.,&Lissauer,J.J.2009,Icarus,201,381 OnlineDataCatalog,1322,0 Southworth,J.2011,MNRAS,417,2166 Zsom,A.,Seager,S.,deWit,J.,&Stamenković,V.2013,ApJ, Southworth,J.,Maxted,P.F.L.,&Smalley,B.2004,MNRAS, 778,109 351,1277 Zuckerman,B.,Vican,L.,Song,I.,&Schneider,A.2013,ApJ, Stevenson,K.B.,Harrington,J.,Nymeyer,S.,etal.2010, 778,5 Nature,464,1161 Swift,J.J.,Johnson,J.A.,Morton,T.D.,etal.2013,ApJ,764, 105