Characterizing Extrasolar Planets with Multi-Color Photometry

MISSING IMAGE

Material Information

Title:
Characterizing Extrasolar Planets with Multi-Color Photometry
Physical Description:
1 online resource (202 p.)
Language:
english
Creator:
Colon, Knicole D
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Astronomy
Committee Chair:
Ford, Eric B
Committee Members:
Telesco, Charles M
Matcheva, Katia Ivandva
Ge, Jian
Fry, James N

Subjects

Subjects / Keywords:
astronomy -- photometry -- planets -- telescope
Astronomy -- Dissertations, Academic -- UF
Genre:
Astronomy thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Over the past twenty years, nearly 800 planets have been discovered orbiting stars other than the Sun. The discovery of these extrasolar planets (or simply, exoplanets) has led to a renewed interest in planet formation and evolution, as many exoplanets have properties that are nothing like those of the planets found in the Solar System. A subset of exoplanets are known to transit, or pass in front of, their host star, which provides a unique opportunity to measure how their radius changes with wavelength. Such measurements can be used to study the atmospheres of exoplanets, since changes in the measured radius can indicate absorption of stellar photons by the exoplanet atmosphere. Finding a significant change in the radius with wavelength can also indicate that a planet candidate is not a planet at all, but is instead an eclipsing binary star composed of two stars with different temperatures and therefore colors. With over 200 confirmed transiting exoplanets and NASA’s Kepler mission’s recent discovery of over 2000 transiting exoplanet candidates, detailed investigations into the properties of exoplanetary atmospheres and false positive rates for planet search surveys can now be conducted. To aid these investigations, I developed a novel technique of using the Optical System for Imaging and low Resolution Integrated Spectroscopy (OSIRIS) installed on the 10.4 meter Gran Telescopio Canarias (GTC) to acquire near-simultaneous, multi-color, narrow-band photometry of exoplanet transits. I first used this technique to observe the transits of the hot-Jupiters TrES-2b and TrES-3b, from which I reached some of the best photometric precisions (0.343-0.470 mmag) achieved to date using a ground-based telescope. I subsequently used this technique to measure a ~4.2% change in the apparent planetary radius of the giant exoplanet HD 80606b during transit between wavelengths that probe potassium. I hypothesize that the excess absorption is due to potassium in a high-speed wind being driven from the exoplanet’s exosphere. This was one of the first detections of potassium in an exoplanet atmosphere. In a similar study, I compared the transit depths for the “super-Earth” GJ 1214b as measured in and out of a predicted methane absorption feature, but I was not able to confirm or refute the presence of methane in GJ 1214b’s atmosphere due to the significant impact that stellar variability had on the measurements. Finally, I used the measured color change during transit to identify three short-period Kepler planet candidates as false positives and validate two as planets. These results test recent predictions of the false positive rates for Kepler candidates and suggest that stellar eclipsing binaries significantly contaminate short-period planet candidates. These results demonstrate the capability of the GTC for constraining the properties of transiting planets, which in turn allows us to better understand how different types of planets form and evolve.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Knicole D Colon.
Thesis:
Thesis (Ph.D.)--University of Florida, 2012.
Local:
Adviser: Ford, Eric B.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2012
System ID:
UFE0044491:00001


This item is only available as the following downloads:


Full Text

PAGE 1

CHARACTERIZINGEXTRASOLARPLANETSWITHMULTI-COLORPHOTOMETRY By KNICOLEDAWNCOL ON ADISSERTATIONPRESENTEDTOTHEGRADUATESCHOOL OFTHEUNIVERSITYOFFLORIDAINPARTIALFULFILLMENT OFTHEREQUIREMENTSFORTHEDEGREEOF DOCTOROFPHILOSOPHY UNIVERSITYOFFLORIDA 2012

PAGE 2

c 2012KnicoleDawnCol on 2

PAGE 3

Formyparents,DaveandDebbie.Youalwaystoldmetoreachforthestars! 3

PAGE 4

ACKNOWLEDGMENTS Manypeoplehavesupportedmeindifferentwaysovertheyears,andthis dissertationwouldnothavebeenpossiblewithoutthatsupport!Ithankmyfamily (Mom,Dad,Dave,Kimandallherlittleones,andTeenee-maysherestinpeace)for theircontinuedsupportthroughoutallmyendeavors.Youhavetrulyhelpedmakeme whoIamtoday,andIloveyouallsomuch!Iwouldliketothankmybestfriends,Kelly WellsandKristinaPengler,foralwaysbelievinginmeandforalwaysknowinghowto saytherightthinginanysituation.AshleyBright(Maloney),ErinCrupi,MichelleBlakely (Vihonski)-youallhavebeengreatfriendstomeandkeptmesanethroughourcrazy collegeyears.MuchlovegoesouttoT8andSouth4!Ialsowanttoacknowledgethe physicscrew(n00bs!)atTCNJ-withoutyouIdenitelywouldnotbeheretoday!Ialso wouldliketothankalltheastronomygraduatestudentsandpostdocsatUFfortheir supportovertheyears,includingJustinCrepp,MichelleEdwards,PaolaRodriguez Hidalgo,DimitriVerasandDaveVollbach.Youallespeciallyhelpedmesurvivemyrst yearofgraduateschool(andIwillneverforgetthatIwonthe"LateNightCookieTime" awardthatyear!).IespeciallywouldliketoacknowledgeDanCapellupo,whohasbeen atrulygreatfriendeversinceIstartedatUF(andIcan'tbelievewearegoingtobe halfwayacrosstheworldfromeachother!). IwouldliketothankmyawesomephysicsprofessorsatTCNJfortheirsupport duringcollegeandforencouragingmetopursueastronomy.Iamsincerelygrateful forthesupportmyadvisor,Dr.EricFord,hasgivenmethroughtheyearsandfor hisseeminglyendlessknowledgeabouteverythinginastronomy.Iwouldalsoliketo acknowledgemymanycollaboratorsforalltheirhelpandadvice,andIwanttogive specialrecognitiontothestaffattheGTCfortheirhelpandpatienceaswelearnedhow touseanewtelescopeandinstrumenttogether. Finally,IwanttothankPhilipJacksonforalwaysbeingthereformeandforalways beingabletomakemelaugh.YouprovidedmewithmotivationwhenIneededtoget 4

PAGE 5

workdone,butyoualsoremindedmetohavefunwheneverIbecametoostressed aboutwork.Iamgratefuleverydayforyourencouragementandsupport.Iloveyou! 5

PAGE 6

TABLEOFCONTENTS page ACKNOWLEDGMENTS .................................. 4 LISTOFTABLES ...................................... 9 LISTOFFIGURES ..................................... 10 ABSTRACT ......................................... 12 CHAPTER 1INTRODUCTION ................................... 14 2CHARACTERIZINGTRANSITINGEXTRASOLARPLANETSWITH NARROW-BANDPHOTOMETRYANDGTC/OSIRIS ............... 20 2.1Observations .................................. 21 2.2DataReduction ................................. 23 2.3Light-CurveAnalysis .............................. 26 2.4Results ..................................... 30 2.4.1PlanetaryParameters ......................... 30 2.4.2Light-CurveResiduals ......................... 32 2.4.3TransitColour .............................. 33 2.5Discussion ................................... 33 3PROBINGPOTASSIUMINTHEATMOSPHEREOFHD80606BWITH TUNABLEFILTERTRANSITSPECTROPHOTOMETRYFROMTHEGRAN TELESCOPIOCANARIAS ............................. 43 3.1Observations .................................. 48 3.1.1In-TransitandOut-of-TransitObservations .............. 48 3.1.2DataReductionandAnalysis ..................... 51 3.2Results ..................................... 55 3.2.1EffectsofEarth'sAtmosphere ..................... 58 3.2.2Limb-DarkeningEffects ........................ 59 3.2.3TransitColour .............................. 60 3.3Discussion ................................... 61 3.3.1InterpretationofLight-CurveShape .................. 61 3.3.2ComparisontoPreviousObservations ................ 62 3.3.3LackofaK I LineCore ......................... 63 3.3.4PlanetaryAtmosphereModels .................... 64 3.3.5ChangeinApparentRadiuswithWavelength ............ 65 3.3.6AbsorptionbyanExosphere ...................... 66 3.3.7PossibilityofOtherAbsorbers ..................... 67 3.3.8AbsorptionbyaWind .......................... 67 3.3.9PotentialSystematics .......................... 69 6

PAGE 7

3.3.9.1ExcludingTelluricAbsorption ................ 69 3.3.9.2ExcludingInstrumentalEffects ............... 73 3.3.9.3PossibleNon-PlanetaryAstrophysicalEffects ....... 75 3.4Conclusion ................................... 77 3.5FutureProspects ................................ 78 4ASEARCHFORMETHANEINTHEATMOSPHEREOFGJ1214BVIA NARROW-BANDTRANSMISSIONSPECTROPHOTOMETRY ......... 104 4.1Observations .................................. 105 4.1.12010July22Transit .......................... 106 4.1.22010August28and29TransitandBaselineObservations ..... 107 4.1.32011June11Transit .......................... 108 4.2DataReduction ................................. 109 4.3LightCurveAnalysis .............................. 110 4.4Results ..................................... 113 4.4.1Resultsfromthe2010July22Transit ................. 115 4.4.2Resultsfromthe2010August28Transit ............... 116 4.4.3Resultsfromthe2011June11Transit ................ 117 4.4.4ResultsfromAllTransits ........................ 117 4.5Discussion ................................... 118 4.5.1VariabilityduetoGJ1214b'sAtmosphere .............. 119 4.5.2VariabilityduetoStellarActivity .................... 120 4.5.3VariabilityduetoEarth'sAtmosphere ................. 120 4.6Conclusions ................................... 122 5VETTING KEPLER PLANETCANDIDATESWITHMULTICOLOR PHOTOMETRYFROMTHEGTC:IDENTIFICATIONOFANECLIPSING BINARYSTARNEARKOI565 ........................... 132 5.1Observations .................................. 134 5.2DataReductionandAnalysis ......................... 136 5.3Results ..................................... 137 5.4Discussion ................................... 140 6CONSTRAININGTHEFALSEPOSITIVERATEFOR KEPLER PLANET CANDIDATESWITHMULTI-COLORPHOTOMETRYFROMTHEGTC .... 151 6.1TargetSelection ................................ 153 6.2Observations .................................. 155 6.2.1KOI225.01 ............................... 156 6.2.2KOI420.01 ............................... 156 6.2.3KOI526.01 ............................... 157 6.2.4KOI1187.01 ............................... 157 6.3DataReduction ................................. 158 6.4LightCurveAnalysis .............................. 159 6.5Results ..................................... 162 7

PAGE 8

6.5.1KOI225.01 ............................... 163 6.5.2KOI420.01 ............................... 165 6.5.3KOI526.01 ............................... 166 6.5.4KOI1187.01 ............................... 167 6.6Discussion ................................... 168 6.6.1ComparisontoTheoreticalStudies .................. 170 6.6.2ComparisontoObservationalStudies ................ 173 6.7Conclusion ................................... 173 7SUMMARYANDCONCLUSIONS ......................... 190 REFERENCES ....................................... 195 BIOGRAPHICALSKETCH ................................ 202 8

PAGE 9

LISTOFTABLES Table page 2-1RelativetransitphotometryofTrES-2andTrES-3 ................. 38 2-2SystemparametersofTrES-2andTrES-3 ..................... 39 3-1AbsolutetransitphotometryofHD80606fromJanuary2010 .......... 81 3-2Absoluteout-of-transitphotometryofHD80606fromApril2010 ........ 82 3-3RelativetransitphotometryofHD80606fromJanuary2010 ........... 83 3-4NormalizedtransitphotometryofHD80606fromJanuary2010 ......... 84 3-5Relativeout-of-transitphotometryofHD80606fromApril2010 ......... 85 3-6Time-averageduxratiosandnoiseestimatesforHD80606 .......... 86 3-7Time-averageduxratiosandnoiseestimatesforHD80606(outlyingabsolute uxesexcluded) ................................... 87 3-8Absoluteout-of-transitphotometryofHD80606fromJanuary2010 ....... 88 3-9Relativeout-of-transitphotometryofHD80606fromJanuary2010 ....... 89 5-1NormalizedphotometryofKOI565 ......................... 143 5-2NormalizedphotometryofKIC7025851 ...................... 145 6-1Propertiesof Kepler targets ............................. 175 6-2PhotometricprecisionsachievedfromGTCobservationsof Kepler targets ... 176 6-3Best-tmodelparametersfor Kepler targets .................... 177 9

PAGE 10

LISTOFFIGURES Figure page 2-1TransitlightcurvesforTrES-2 ............................ 40 2-2TransitlightcurvesforTrES-3 ............................ 41 2-3Standarddeviationoftime-binnedresidualsforTrES-3 .............. 42 3-1AbsolutephotometryofHD80606andHD80607fromJanuary2010 ..... 90 3-2AbsolutephotometryofHD80606andHD80607fromApril2010 ....... 91 3-3TransitlightcurvesforHD80606 .......................... 92 3-4UncorrectedandcorrectedlightcurvesforHD80606fromaroundmid-transit 93 3-5CorrelationsbetweentransituxratiosandparametersforHD80606 ..... 94 3-6UncorrectedandcorrectedlightcurvesforHD80606fromApril2010 ..... 95 3-7ComparisonsofthecorrectedlightcurvesforHD80606fromaroundmid-transit 96 3-8HistogramsofnormalizeduxratiosforHD80606fromaroundmid-transit ... 97 3-9Standarddeviationoftime-binnedin-transitresidualsforHD80606 ....... 98 3-10Standarddeviationoftime-binnedout-of-transitresidualsforHD80606 .... 99 3-11ObservedandmodelspectraofHD80606baroundtheK I line ......... 100 3-12TransitcolorofHD80606 .............................. 101 3-13Standarddeviationoftime-binnedtransitcolorforHD80606 .......... 102 3-14MeantransitcolorsforHD80606 .......................... 103 4-1BaselineobservationsofGJ1214fromAugust2010 ............... 123 4-2TransitlightcurvesforGJ1214fromJuly2010 .................. 124 4-3TransitlightcurvesforGJ1214fromAugust2010 ................ 125 4-4TransitlightcurvesforGJ1214fromJune2011 .................. 126 4-5CombinedtransitlightcurvesforGJ1214 ..................... 127 4-6Planet-starradiusratiosmeasuredfromobservationsofGJ1214inJuly2010 128 4-7Planet-starradiusratiosmeasuredfromobservationsofGJ1214inAugust 2010 .......................................... 129 4-8Planet-starradiusratiosmeasuredfromobservationsofGJ1214inJune2011 130 10

PAGE 11

4-9Planet-starradiusratiosmeasuredfromcombinedobservationsofGJ1214 .. 131 5-1LightcurvesforKOI565 ............................... 147 5-2FieldofviewaroundKOI565 ............................ 148 5-3LightcurvesforKIC7025851 ............................ 149 5-4ColorsforKOI565,KIC7025851,andahypotheticalunresolvedsystem .... 150 6-1Numberofplanetcandidatesandeclipsingbinariesdiscoveredby Kepler ... 178 6-2Radiusversusorbitalperiodof Kepler planetcandidates ............. 179 6-3Signal-to-noiseratiopertransitasafunctionoforbitalperiodof Kepler planet candidates ...................................... 180 6-4 Kepler magnitudeversusstellareffectivetemperaturefor Kepler planet-hosting stars .......................................... 181 6-5FieldofviewaroundKOI1187 ........................... 182 6-6TransitlightcurvesforKOI225 ........................... 183 6-7TransitlightcurvesforKOI420 ........................... 184 6-8TransitlightcurvesforKOI526 ........................... 185 6-9TransitlightcurvesforKOI1187 .......................... 186 6-10CumulativedistributionfunctionsoftheGalacticlatitudeofplanet-hostingstars andeclipsingbinariesdiscoveredby Kepler .................... 187 6-11CumulativedistributionfunctionsoftheGalacticlatitudeofspecicsubsetsof planet-hostingstarsandeclipsingbinariesdiscoveredby Kepler ........ 188 6-12Numberofeclipsingbinariesdiscoveredby Kepler asafunctionoforbitalperiod andGalacticlatitude ................................. 189 11

PAGE 12

AbstractofDissertationPresentedtotheGraduateSchool oftheUniversityofFloridainPartialFulllmentofthe RequirementsfortheDegreeofDoctorofPhilosophy CHARACTERIZINGEXTRASOLARPLANETSWITHMULTI-COLORPHOTOMETRY By KnicoleDawnCol on August2012 Chair:EricB.Ford Major:Astronomy Overthepasttwentyyears,nearly800planetshavebeendiscoveredorbitingstars otherthantheSun.Thediscoveryoftheseextrasolarplanets(orsimply,exoplanets)has ledtoarenewedinterestinplanetformationandevolution,asmanyexoplanetshave propertiesthatarenothinglikethoseoftheplanetsfoundintheSolarSystem.Asubset ofexoplanetsareknowntotransit,orpassinfrontof,theirhoststar,whichprovides auniqueopportunitytomeasurehowtheirradiuschangeswithwavelength.Such measurementscanbeusedtostudytheatmospheresofexoplanets,sincechanges inthemeasuredradiuscanindicateabsorptionofstellarphotonsbytheexoplanet atmosphere.Findingasignicantchangeintheradiuswithwavelengthcanalsoindicate thataplanetcandidateisnotaplanetatall,butisinsteadaneclipsingbinarystar composedoftwostarswithdifferenttemperaturesandthereforecolors.Withover 200conrmedtransitingexoplanetsandNASA's Kepler mission'srecentdiscoveryof over2000transitingexoplanetcandidates,detailedinvestigationsintotheproperties ofexoplanetaryatmospheresandfalsepositiveratesforplanetsearchsurveyscan nowbeconducted.Toaidtheseinvestigations,Idevelopedanoveltechniqueof usingtheOpticalSystemforImagingandlowResolutionIntegratedSpectroscopy (OSIRIS)installedonthe10.4meterGranTelescopioCanarias(GTC)toacquire near-simultaneous,multi-color,narrow-bandphotometryofexoplanettransits. 12

PAGE 13

Irstusedthistechniquetoobservethetransitsofthehot-JupitersTrES-2band TrES-3b,fromwhichIreachedsomeofthebestphotometricprecisions(0.343-0.470 mmag)achievedtodateusingaground-basedtelescope.Isubsequentlyusedthis techniquetomeasurea 4.2%changeintheapparentplanetaryradiusofthegiant exoplanetHD80606bduringtransitbetweenwavelengthsthatprobepotassium.I hypothesizethattheexcessabsorptionisduetopotassiuminahigh-speedwind beingdrivenfromtheexoplanet'sexosphere.Thiswasoneoftherstdetections ofpotassiuminanexoplanetatmosphere.Inasimilarstudy,Icomparedthetransit depthsforthe"super-Earth"GJ1214basmeasuredinandoutofapredictedmethane absorptionfeature,butIwasnotabletoconrmorrefutethepresenceofmethanein GJ1214b'satmosphereduetothesignicantimpactthatstellarvariabilityhadonthe measurements.Finally,Iusedthemeasuredcolorchangeduringtransittoidentifythree short-period Kepler planetcandidatesasfalsepositivesandvalidatetwoasplanets. Theseresultstestrecentpredictionsofthefalsepositiveratesfor Kepler candidates andsuggestthatstellareclipsingbinariessignicantlycontaminateshort-periodplanet candidates. TheseresultsdemonstratethecapabilityoftheGTCforconstrainingtheproperties oftransitingplanets,whichinturnallowsustobetterunderstandhowdifferenttypesof planetsformandevolve. 13

PAGE 14

CHAPTER1 INTRODUCTION Inthepastdecade,theeldofextrasolarplanet(orexoplanet)researchhasgrown extraordinarily.Ground-andspace-basedsurveyshavediscoverednearly800planets orbitingstarsotherthantheSun.Thisincludesover200planetsthattransit,orpass infrontof,theirhoststarandthereforecanhavetheirradiusmeasured.Theseplanets arethefocusofthisdissertation.Whileamajorityofthesehavebeendiscoveredby wide-eldground-basedsurveys,space-basedmissionslike COROT and Kepler are quicklycatchingup(together,theyhaveproducedover80discoveriesoftransiting planetstodate).Regardlessofthediscoverymethod,thetremendousincreasein thenumberofknownexoplanetshasmadeitincreasinglyimportanttomakethe bestpossibleuseoffollow-upresources.Asdemonstratedby Col on&Ford ( 2009 ), largeground-basedtelescopesarecapableofcontributingsignicantlytophotometric follow-upeffortsevenforplanetsdiscoveredby,e.g., Kepler ,whichhasachieved superiorphotometricprecisionscapableofdetectingtransitingplanetswithradiismaller thantheEarth's( Boruckietal. 2009 ; Fressinetal. 2012 ; Gautieretal. 2012 ; Muirhead etal. 2012 ).Thisisbecauserelativelyhigh-precisionphotometricfollow-upcanbe conductedatnear-infraredwavelengths(ratherthaninasingle"white"broadbandlter), wherethedegeneracybetweenthetransitimpactparameter(theminimumdistance fromthecenteroftheplanettothecenterofthestarwhenprojectedontheskyplane; alsoequivalentto a cos i / R )andstellarlimbdarkeningparameterscanbeminimized (duetominimizationoftheeffectsofstellarlimbdarkeningatredderwavelengths).As aresult,thestrongerconstraintsontheimpactparameterimprovemeasurementsof theplanet-starradiusratio(or,simplytheradiusratio,denedas R p / R andwhichis approximatelyequaltothesquarerootofthetransitdepth, F / F ,intheabsence ofstellarlimbdarkening)andtransitduration.Inturn,theimprovedconstraintsonthe radiusratioallowforprecisecomparisonsoftheradiusratioasmeasuredatdifferent 14

PAGE 15

wavelengths.Asdemonstratedbythestudiespresentedinthisdissertation,measuring preciseradiusratiosasafunctionofwavelengthisextremelyimportantfor,forexample, atmosphericstudiesofexoplanetsaswellasidentifyingthetruenatureofplanet candidates. Inthisdissertation,Ifocusonusinganoveltechniquetoachievehigh-precision observationsofexoplanettransits.Specically,Iusenear-simultaneousnarrow-band multi-colorphotometryfromthe10.4mGranTelescopioCanarias(GTC)tocharacterize bothknownandcandidatetransitingplanets.Whilesuchatechniqueisinefcientfor planetsearches,itprovidesanotheravenuetowardsreachingthehighphotometric precisionsnecessaryforsomestudies(e.g.,atmosphericstudies).InChapter2( Col on etal. 2010 ),IdescribesomeoftherstscienticobservationsconductedwiththeGTC, whichbeganscienticoperationsinMarch2009andiscurrentlytheworld'slargest, ground-based,fullysteerable,single-aperture,opticaltelescope.Specically,Iuse theOpticalSystemforImagingandlowResolutionIntegratedSpectroscopy(OSIRIS) ( Cepaetal. 2000 2003 )installedontheGTC.OSIRISincludestwo2048 # 4096 pixelE2V44-82CCDs,whichprovideamaximumunvignettedeldofviewof 7.8 # 7.8arcmin 2 .ThereareasuiteofltersavailablewithOSIRIS,includingbroadband "ordersorter"ltersthatwerecustommadeforOSIRISandthatarenarrowerthan Sloanlters(whichhaveafull-widthathalf-maximumofabout62-153nm,comparedto 28-59nmfortheOSIRISlters).Thereisalsoatunablelter(TF)imagingmode,which currentlyallowstheusertospecifycustombandpasseswithacentralwavelengthof 651-934.5nmandafull-widthathalf-maximumof1.2-2.0nm. 1 Thus,bandpassescan bespecicallychosentoavoidwatervaporabsorptionandskyglowsoastominimize 1 A"blue"endoftheTFisexpectedtobeavailableinthefuture,whichwould effectivelyincreasethescienticreturnfromtheGTC.Inparticular,theblueTFwould allowforadditionalatmosphericfeaturestobeobserved,includingtheNa I featurethat hasbeenpreviouslydetectedinsomeexoplanetatmospheres. 15

PAGE 16

effectsfromEarth'satmosphere.Also,withtheGTC/OSIRIS,itispossibletoobserve inmultipleltersnearlysimultaneously,asthereisminimaldeadtimeduetoswitching betweenlters.Therefore,theGTC/OSIRISallowsforfast,narrow-band,high-precision spectrophotometryofexoplanettransits,whichallowsusto(1)probethecomposition andotherpropertiesoftheatmospheresofexoplanetsand(2)measurethetransitcolor ofplanetcandidates.Furthermore,thistechniqueisextremelyefcient,considering thatmultiplebandpassescanbeobservednearlysimultaneouslywithinasingletransit event. Asdiscussedabove,thereisnowasignicantsampleofknowntransitingplanets. Thesediscoveriesallowforauniqueinvestigationoftheatmospheresofexoplanets, asthephysicalcharacteristicsofanexoplanetaryatmospherecanbeprobedby transmissionspectroscopyorspectrophotometryobservedagainstthespectrumof thehoststar(e.g.withtheGTC/OSIRIS).Ifthereisabsorptionofstellarphotons intheexoplanetaryatmosphere,thisleadstoalargerapparentsizeoftheplanet attheabsorbingwavelengths( Brown 2001 ).However,suchobservationsrequire extremelyhighprecisionobservations,astheexcessabsorptionistypicallyexpected tocauseadecreaseinthemeasuredin-transituxratioofmuchlessthan0.1%. Earlymodelsfocusedontheatmospheresofhot-Jupiters,astheseshort-period,giant planetswerethersttransitingplanetsdiscovered.Themodelspredictedabsorption, particularlyfromthealkalimetalssodium(Na I )andpotassium(K I )(e.g., Seager& Sasselov 2000 ; Brown 2001 ; Hubbardetal. 2001 ).Whilethestrongestabsorberis expectedtobeNa I ( !! 589.6,589.0nm),itisnotwithinthecurrentobservingmodeson GTC/OSIRIS.However,K I ( !! 769.9,766.5nm)isexpectedtobethesecondstrongest transmissionspectrumsignatureintheopticalwavelengthrangeandisreadilywithin thewavelengthrangethattheGTC/OSIRIScanprobe.Todate,therehavebeenseveral detectionsofabsorptionduetoNa I inexoplanetaryatmospheres(e.g. Moutouetal. 2001 ; Charbonneauetal. 2002 ; Winnetal. 2004 ; Naritaetal. 2005 ; Redeldetal. 16

PAGE 17

2008 ; Singetal. 2008a ; Snellenetal. 2008 ; Langland-Shulaetal. 2009 ),butthese detectionshavebeenprimarilymadeinjusttwoexoplanets,HD209458bandHD 189733b.IdiscussoneoftherstdetectionsofK I inadifferentexoplanet'satmosphere madeusingtheGTC/OSIRISTFinChapter3( Col onetal. 2012 ). 2 Whilehot-Jupitersorbitingbrightstarsareprimetargetsforatmosphericstudies, thediscoveryof"super-Earths"orbitingMdwarfstarsopenedanentirelynewareaof atmosphericresearch.Oneofthemorewell-studiedsuper-EarthsisGJ1214b,which wasdiscoveredinaground-basedsurveyby Charbonneauetal. ( 2009 ).Duetothe smallradiusofthehoststar,thetransitdepthisnearly1.5%,whichimmediatelylends itselftomakingGJ1214banexcellentcandidateforatmosphericstudies.Furthermore, Miller-Ricci&Fortney ( 2010 )modeleditsatmosphereandpredictedthatGJ1214b isrequiredtohaveasignicantatmospherebasedonitsobservedmassandradius. However,thecompositionoftheatmosphereisdebated(inparticular,whetherornot methaneispresent),somanyrecenteffortshavebeenmadetoobserveGJ1214b's atmosphereviatransmissionspectroscopy(e.g. Beanetal. 2010 2011 ; Crolletal. 2011 ; Crosseldetal. 2011 ; D esertetal. 2011 ; Bertaetal. 2012 ; deMooijetal. 2012 ).Inthisdissertation,ItakethesameapproachusedinChapter3( Col onetal. 2012 ),andIusetheGTC/OSIRISTFtoperformtransmissionspectrophotometryof GJ1214binordertosearchforabsorptionduetomethane,theresultsofwhichare presentedinChapter4. Atmosphericstudiesasdescribedabovearecriticalforunderstandinghow hot-Jupitersandsuper-Earthsformandevolve.However,itisadmittedlydifcultto grasptheoverallpictureofplanetformationandevolutionwhensuchstudiesfocuson oneplanetatatime.Ontheotherhand,the Kepler missionisprovidingunprecedented 2 AseconddetectionofK I wasreportedby Singetal. ( 2011 )aroundthesametime asthestudypresentedinChapter3( Col onetal. 2012 ). 17

PAGE 18

viewsintothebulkpropertiesofextrasolarplanetarysystemsduetoitsdiscovery ofover2000planetcandidates( Batalhaetal. 2012 ).Ofcourse,amajorityofthese discoveriesareunconrmed,meaningthattheydonothaveindependentconrmation suchasanassociatedmassmeasurementthatconrmstheirplanetarystatus.A recenttheoreticalstudyhaspredictedthatasmanyas 95%ofthesecandidates aretrueplanetsandnotfalsepositives(e.g.,ablendwithastellareclipsingbinary eitherinthebackground/foregroundorboundtothetargetstar)( Morton&Johnson 2011 ),butthisstudydoesnottakeintoaccountthatdifferentsubsetsof Kepler targets mayhavedifferentfalsepositiveratesassociatedwiththem.Inparticular,planet candidateswithshortorbitalperiods(P < 3d)maybesignicantlycontaminatedby eclipsingbinarystars,asthereisarapidriseinthenumberofeclipsingbinariesthat havebeendiscoveredby Kepler attheseperiods( Pr saetal. 2011 ; Slawsonetal. 2011 ).Whilepreliminaryresultsfromanobservationalstudythatiscurrentlybeing conductedwithwarmSpitzer ( D esertetal. 2012 )supportthendingsof Morton& Johnson ( 2011 ),inthisdissertationIpresentmulti-colorobservationsofveshort-period Kepler planetcandidatesthatsupporttheideathatthereisacorrelationbetweenthe falsepositiverateandorbitalperiod,particularlyattheshortestorbitalperiods.Since planetarytransitsshouldbelargelyachromaticwhenobservedatdifferentwavelengths (excludingthesmallcolorchangesduetostellarlimbdarkening),itispossibletousethe observedtransitcolortoidentifycandidatesaseitherfalsepositivesorvalidatedplanets. Specically,inChapter5( Col on&Ford 2011 ),IpresentGTC/OSIRISTFphotometry ofasuper-Earth-size Kepler planetcandidatewhichresultedintheidenticationof aneclipsingbinarystarasthesourceofthetransitsignal.InChapter6(Col onetal., accepted),Ipresentobservationsoffourmorecandidates,twoofwhichwereidentied asfalsepositivesandtwothatwerevalidatedasplanets.Thus,Iconcludethatthose candidateswithparticularlyshortperiodsaresignicantlycontaminatedbyapopulation 18

PAGE 19

ofeclipsingbinaries,whichinturnhasimplicationsregardingthepropertiesofthebulk populationofknowntransitingplanetsandcandidates. Tosummarize,inthisdissertationIpresentanoveltechniqueofusingthe GTC/OSIRIStoacquirehigh-precision,narrow-band,multi-colorphotometryof exoplanettransits(Chapter2).Ithenpresentobservationsthatusethistechniqueto searchforabsorptionduetopotassiumintheatmosphereofagiantexoplanet(Chapter 3).InChapter4,Ipresentsimilarobservationsthatprobethemethanecontentinthe atmosphereofasuper-Earth-sizeplanet.Next,inChapters5and6,Ipresentmulti-color observationsof Kepler planetcandidates,withthegoalofusingthetransitcolorto identifythetruenatureoftheplanetcandidates.Lastly,inChapter7,Isummarizethe resultsandconclusionsfromeachofthepreviouschapters.Additionaldetailsregarding allthetopicsdiscussedinthischaptercanbefoundinthecorrespondingchaptersin thisdissertation.Finally,InotethatChapters2,3,5and6inthisdissertationare,or aimtobe,self-containedjournalarticles.Chapters2and3arepublishedintheMonthly NoticesoftheRoyalAstronomicalSociety(MNRAS)journal( Col onetal. 2010 2012 ). Chapter5ispublishedinthePublicationsoftheAstronomicalSocietyofthePacic (PASP)journal( Col on&Ford 2011 ).Chapter6hasbeenacceptedtoMNRASin collaborationwithE.FordandR.Morehead. 19

PAGE 20

CHAPTER2 CHARACTERIZINGTRANSITINGEXTRASOLARPLANETSWITHNARROW-BAND PHOTOMETRYANDGTC/OSIRIS Space-basedmissionslike CoRoT and Kepler havethephotometriccapabilityto detecttransitingplanetswithradiinotmuchlargerthantheEarth's,commonlyreferred toas"super-Earths"( Boruckietal. 2009 ; L egeretal. 2009 ).Whilespace-based observatorieshaveprovidedthehighestphotometricprecisions,ground-basedfollow-up observationsplayanessentialroleinconrmingdetectionsoftransitingplanetsand characterizingtheplanets'orbits,interiors,compositionsandatmospheres( Torreset al. 2008 ; Johnsonetal. 2009 ).Inparticular,because CoRoT 'sand Kepler 'swhitelight observationsareaffectedbystellarlimbdarkening(LD),therecanbestrongcorrelations amongthestellarLDparameters,thetransitimpactparameterandthetransitduration (whichdependsontheeccentricityandpericentredirection).High-precisionphotometry conductedatfar-red,near-infrared(NIR)ormid-infraredwavelengthscanbreakthis degeneracy,providingmoreprecisemeasurementsofaplanet'sorbitalandphysical parameters( Col on&Ford 2009 ).Observingatthesewavelengthsalsoallowsfor reduceddifferentialatmosphericeffects,whichcanfurtherimprovethequalityofthe transitlightcurve(LC). ThedetectionoftransitingEarth-sizeplanetsaroundsolar-likestarsrequiressuch highphotometricprecisionthatastronomershadlongassumedthatcharacterizing suchplanetscouldonlybedonefromspacetoavoidthedeleteriouseffectsofEarth's atmosphere( Boruckietal. 1985 ).Narrow-bandphotometrywithlargetelescopes providesanalternativepathtowardshighphotometricprecision.Whileinefcientfor planetsearches,near-simultaneousnarrow-bandobservationsprovideadditional opportunitiesforthecharacterizationofknownplanets(orpreviouslyidentiedplanet candidates),suchasthemeasurementofatmosphericabsorption. Inthispaperwedescribeanovelobservationaltechniqueforhigh-precisiontransit photometry,usingnear-simultaneousobservationsinmultiplenarrowbandpasses 20

PAGE 21

(Section 2.1 ).WedescribeourdatareductioninSection 2.2 andLCanalysisinSection 2.3 .InSection 2.4 ,wepresentresultsforTrES-2andTrES-3anddemonstratethat thistechniquecanachieveaveryhighphotometricprecision.Finally,inSection 2.5 wecompareourresultstothosefoundintheliteratureanddiscusstheimplications forstudyingtheatmosphericcompositionofgiantplanetsandforcharacterizingthe bulkpropertiesofsuper-Earth-sizeplanets,includingthoseinthehabitablezoneof main-sequencestars. 2.1Observations The10.4-mGranTelescopioCanarias(GTC) 1 islocatedattheObservatoriodel RoquedelosMuchachos,ontheislandofLaPalma.WhiletheGTCbeganscientic operationsin2009March,commissioningofthetelescopeandrstlightinstrumentsis ongoing.WedescribesomeoftherstscienticobservationswiththeGTCusingthe OpticalSystemforImagingandlowResolutionIntegratedSpectroscopy(OSIRIS)inthe tunablelter(TF)imagingmode( Cepaetal. 2000 2003 ).OSIRISincludestwo2048 # 4096pixelE2V44-82CCDswhichprovideamaximumunvignettedeldofviewof 7.8 # 7.8arcmin 2 .OSIRISoffersasuiteoflters,includingaTFwhichallowstheuser tospecifycustombandpasseswithacentralwavelengthof651-934.5nmandafull widthathalf-maximum(FWHM)of1.2-2.0nm.NotethattheTFbandpassisnotuniform acrosstheeldofview,withtheeffectivewavelengthdecreasingradiallyoutwardsfrom thecentre.Therefore,forthewavelengthsusedintheseobservations,adifferenceof 10nmexistsbetweenthetunedwavelengthattheopticalcentreandthewavelength observedat4arcminfromtheopticalcentre(i.e.nearoneedgeoftheCCD). Weobservedonetransiteachoftwogiantextrasolarplanets,TrES-2bandTrES-3b, usingGTC/OSIRIS.Theeldofviewwaschosensothat(1)thetargetanda"primary" referencestarwereobservedatthesamewavelengthusingthesameCCDand(2) 1 http://www.gtc.iac.es/en/ 21

PAGE 22

severaladditional"secondary"referencestarswereobservedwiththesameCCDbutat differentdistancesfromtheopticalcentre(andthusatdifferentwavelengths).Duetothe restrictionsinCCDorientation,differentchipsintheOSIRISCCDmosaicwereusedfor theTrES-2andTrES-3observations. Duringeachtransit,observationsalternatedbetweentwobandpassescentred on790.2and794.4nmatthelocationofthetarget.Thesewavebandswerechosen tominimizeatmosphericeffectsbyminimizingwatervapourabsorptionandskyglow. Eachobservationwasfollowedby33.4sofdeadtimeforreadout.Weuse1 # 1 binningbututilizeafastreadoutmode(500kHz)inordertodecreasethedeadtime betweenexposures.RecentandfutureCCDcontrollerupgradeswillreducethedead timebetweenexposures.Thetelescopewasalsodefocusedtoincreaseefciency andtoreducetheimpactofpixel-to-pixelsensitivityvariations.Evenwithaslight defocus,theresultingpoint-spreadfunctionsofthestarswerefairlywelldened(i.e.not doughnut-shaped). TheTrES-2( V 11.4)observationstookplaceunderphotometricconditionson 2009June25(darktime)from1:58to5:43UT,duringwhichtheairmassrangedfrom 1.07to1.44.ThedefocusedFWHMofthetargetvariedfrom 1.6to2.8arcsec ( 12.5-22pixels)duringtheobservations,whiletheactualseeingwas 1.2arcsec.The autoguidersystemkepttheimagesalignedwithinafewpixels,withthetarget'scentroid coordinatesshiftingbylessthan4pixelsinthe x -directionandlessthan6pixelsinthe y -direction.Weused80-sexposures,resultinginanoverallcadenceof3.78minfor eachbandpass.Fouroftheimageswereexcludedfromouranalysisduetotracking problems.Thelast 16minofout-of-transit(OOT)datawereaffectedbytwilightand arenotincludedintheanalysis. TheTrES-3( V 12.4)observationstookplaceunderphotometricconditionsduring greytime,startingat23:53UTon2009August10andcontinuinguntil2:54UTthe nextmorning.Duringtheobservations,theairmassrangedfrom 1.13to2.22,the 22

PAGE 23

defocusedFWHMofthetargetvariedfrom 1.4to2.1arcsec( 11-16.5pixels)andthe target'scentroidcoordinatesshiftedbylessthan3pixelsinthe x -directionandlessthan 5pixelsinthe y -direction.Theexposuretimewas120s,yieldinga5.11-minobserving cadenceforeachbandpass.Problemswiththetelescopecausedourobservationsto beginshortlyafteringresshadalreadybegun. 2.2DataReduction StandardIRAFproceduresforbiassubtractionandat-eldcorrectionwereused fortheTrES-2observations.However,duringpreliminaryanalysis,wefoundthat thetotalnumberofcountsinthedomeatelds(withagivenexposuretime)was decreasingwithtime.Thus,foratelding,weincludeonlythosedomeatstaken aftertherst30minofthelampbeingturnedon,bywhichtimethelampintensity hadstabilized.ForTrES-2,wecombine40(outof60)domeatsforeachobserved wavelength(790.2and794.4nm).Aftersummingovereithersetofats,theaverage numberofcountsperpixelwas1.71 # 10 6 .FortheTrES-3observations,weperform similarreductionstoTrES-2,butweonlyuse45ofthe95totaldomeatstakenfor eachwavelength.Aftersummingovermultipleats,theaveragenumberofcountsper pixelwas5.28 # 10 5 (1.04 # 10 6 )for790.2(794.4)nm.FortheTrES-3datareduction,we replacethemedianbiasframewiththemedianofaseriesofdarkframestakenwiththe sameexposuretimeastheobservations(albeittakenseveralmonthsaftertheactual observationstookplace),sinceearlyobservationsrevealedahigherthanexpected darkcurrent.WenotethatsubtractingdarkswasbenecialonlyfortheTrES-3data, basedontherootmeansquare(rms)scatteroftheOOTLCforTrES-3.Theopposite istrueforTrES-2;i.e.thermsscatteroftheOOTLCwassmallerwhenbiasframes weresubtracted(insteadofdarkframes).Thisdifferenceislikelyduetothefactthat thedarkframesforbothtargetsweretakenin2009October,ratherthanintheexact sameconditionsasthetransitobservations(i.e.severalmonthsaftereachtransitwas observed).Therefore,eachdatasetwasaffecteddifferentlybyitsrespectivesetofdark 23

PAGE 24

frames,resultinginthedarksprovingtobeusefulonlyfortheTrES-3observations. Recenthardwareupgradeshavereducedthedarkcurrent. DuetotheTF'ssmallbandpassandposition-dependentwavelength,allobservations containsky(OH)emissionrings.Therefore,weperformedskysubtractiononallimages basedontheIRAFpackageTFred, 2 whichestimatestheskybackground,including ringsduetoskyemission.Weperformedaperturephotometryoneachtargetand referencestarusingthestandardIDLroutineAPER. 3 Werepeatedthisanalysisusing arangeofapertureradiiandmeasuredthermsscatteroftheuxratio(targetoversum ofreferences)outsideoftransitineachcolour.Afterconsideringtheresultsinboth bandpasses,weadoptedanapertureradiusof44pixels(5.6arcsec)forstarsintheeld ofTrES-2and38pixels(4.8arcsec)forstarsintheeldofTrES-3.Notethattheuseof TFredremovestheneedforaskyannulus,sowedidnotinputoneintheAPERroutine. Toaccountforanyatmosphericextinction,wetlinearairmasstrendstoeach referencestar'sLC(computedbydividingtheuxforagivenstarbythesumoftheux fromalltheotherreferencestars).Foreachreferencestar,wediscardedpointsthat resultedinauxratiogreaterthan3 fromthemean(typicallyoneandatmosttwo pointsperreferencestar).Wecomputedthereferenceuxateachtimeastheweighted sumoftheuxoftheremainingreferencestars.Theuxinthetargetaperturewas dividedbythereferencetocomputethenalLCs.Wefoundthatusinganensemble ofsixtoeightreferencestarsresultedinasmallerrmsOOTscatterthanusingjustthe primaryreference,despitethefactthatthesecondaryreferenceswereobservedat differentwavelengthsandthatthereferencestarsintheeldofTrES-2(TrES-3)ranged over 2(3)maginbrightness.Sincemostofthereferencestarswerefainterthanthe 2 WrittenbyD.H.JonesfortheTaurusTunableFilter,previouslyinstalledonthe Anglo-AustralianTelescope;http://www.aao.gov.au/local/www/jbh/ttf/adv reduc.html. 3 http://idlastro.gsfc.nasa.gov/ 24

PAGE 25

targets,weconcludethataddingadditionalfaintreferenceshelpstoreducethephoton noiseforalargerensembleandbalancesstellarvariabilitywithindifferentreference stars. Theuncertaintiesintheuxratiowereestimatedbycomputingthequadraturesum ofthephotonnoiseforthetargetandthereferenceensemble[medianvaluesare0.331 (0.339)and0.244(0.249)mmagforTrES-2and0.426(0.433)and0.210(0.214)mmag forTrES-3for790.2(794.4)nm],theuncertaintyinthesumofskybackgroundanddark currentforthetargetandthereferenceensemble[medianvaluesare0.0751(0.0737) and0.103(0.105)mmagforTrES-2and0.0998(0.0840)and0.0688(0.0667)mmag forTrES-3for790.2(794.4)nm]andthescintillationnoiseforthetargetandprimary reference[medianvaluesare0.0627(0.0630)mmagforTrES-2and0.0778(0.0766) mmagforTrES-3for790.2(794.4)nm].Theuncertaintyinthesumofskybackground anddarkcurrentwasestimatedbyperformingaperturephotometryontheskyframes producedbyTFredatthespeciclocationofeachtargetandreferencestar.The scintillationnoisewasestimatedusingtherelationgivenby Dravinsetal. ( 1998 ),based on Young ( 1967 ).Wecautionthatthisisanempiricalrelationandhasnotbeentested forlargetelescopesatexcellentsiteslikeLaPalma.Thus,weconsidertheexpression forscintillationnoisetobeonlyaroughestimate.Nevertheless,itdemonstratesthat scintillationisexpectedtobeonlyaverysmallcontributiontothetotalerrorbudget, thankstothenarrowlterbandpass.Readoutanddigitizationnoiseaswellasat-eld noiseisnegligiblecomparedtophotonnoiseandisnotincludedinthecalculationof theuncertaintiesintheuxratio,soweincludeestimatesforthesenoisesources(as basedonrelationsgivenby Southworthetal. 2009 )onlyforreference.Themedian readoutnoiseis7.44electronsperpixelor0.0116(0.0123)mmagforTrES-2and7.35 electronsperpixelor0.0150(0.0158)mmagforTrES-3for790.2(794.4)nm.The medianat-eldnoiseis6.07e-4(6.07e-4)electronsperpixelor0.00274(0.00269) mmagforTrES-2and1.07e-3(7.63e-4)electronsperpixelor0.00778(0.00535)mmag 25

PAGE 26

forTrES-3for790.2(794.4)nm.Basedontherelationgivenby Howell ( 2006 ),wend themediantotaluncertaintiesintheuxratioforeachexposuretobe0.434(0.443) mmagforTrES-2and0.495(0.499)mmagforTrES-3for790.2(794.4)nm.Wenotethat ourestimateduncertaintiesaresomewhatlargerthanthemeasuredresidualrmsgiven inSection 2.4.2 .Whiletheestimatedandmeasuredprecisionaremarginallyconsistent, itispossiblethatweoverestimatedtheuncertaintyinindividualmeasurements. Weperformtheaboveanalysisforbothbandpassesandeachtarget.Theresulting photometrictimeseriesisreportedandshown(withsomecorrections;Section 2.4 )in Table 2-1 (fullversionavailableonline $ SupportingInformation)andFigures 2-1 and 2-2 2.3Light-CurveAnalysis BeforettingmodelstoourLCs,werstappliedtheexternalparameterdecorrelation (EPD)technique(e.g. Bakosetal. 2007 2010 )toeachLCinordertoremoveany systematictrendsthatarecorrelatedwiththefollowingparameters:the x and y centroid coordinatesofthetargetontheimageframes,thesharpnessofthetarget'sprole [approximatelyequalto(2.35/FWHM) 2 ]andtheairmass.Wethenperformedthe followinganalysisonthedecorrelatedLCs. WettheuxratiowithastandardplanettransitmodelthatincludesaquadraticLD law( Mandel&Agol 2002 ).WeparametrizetheLCmodelusingthetimeofmid-transit ( t 0 ),impactparameter( b % a cos i / R ),transitduration(fromrsttofourthcontact; D ), planet-starradiusratio( p % R p / R ),averageOOTuxratio,alinearslopefortheOOT uxratio( # )andtwoLDcoefcients( c 1 % u 1 + u 2 and c 2 % u 1 $ u 2 ),where u 1 and u 2 are, respectively,thelinearandquadraticLDcoefcientsof Mandel&Agol ( 2002 ). Weusedthepubliclyavailablecode mptfun 4 toperformLevenberg-Marquardt minimizationof $ 2 toidentifyabest-ttingmodelforthetransitphotometry.Theinitial 4 http://www.physics.wisc.edu/ craigm/idl/idl.html 26

PAGE 27

guessesfor b D and p arebasedonestimates(andtheiruncertainties)from Holmanet al. ( 2007 )and Sozzettietal. ( 2007 )forTrES-2band Gibsonetal. ( 2009 )and Sozzettiet al. ( 2009 )forTrES-3b.Theinitialguessesfor t 0 werebasedontheephemerisof Rabus etal. ( 2009 )forTrES-2and Sozzettietal. ( 2009 )forTrES-3.Ineachcasedescribed below,werepeatthelocalminimizationusinganarrayofinitialguessesforthemodeled parametersbasedonthepublisheduncertaintiesandconcludedthattheresultsofthe non-linearmodelttingwerenotsensitivetoourinitialguesses. Similarly,wetestedseveralsetsofvaluesfor c 1 and c 2 basedonthetheoreticalLD coefcientsestimatedforeachstar.Specically,bothLDcoefcientswerecomputed forourspecicbandpassesforagridofstellarmodelsusingPHOEBE( Pr sa&Zwitter 2005 ).Weinterpolatedin( T e log g ,[Fe/H])toestimate c 1 and c 2 foreachofseveral setsofstellarparameters.Weadoptedstellarparametersfrom Sozzettietal. ( 2007 ) forTrES-2andfrom Sozzettietal. ( 2009 )forTrES-3,butwenotethattheseabundance measurementsaresystematicallylowerthanthosedeterminedby Ammler-vonEiffet al. ( 2009 ). Valenti&Fischer ( 2005 )emphasizethatdifferentstudiescanyielddifferent resultsdueto,e.g.,theanalysistechniquesand/orstellarmodelsused,butboth Sozzetti etal. ( 2007 2009 )and Ammler-vonEiffetal. ( 2009 )useverysimilartechniquesto estimatestellarparameters.Therefore,wechosetointerpolateLDcoefcientsover therangeofthestellarparametersplustheirmeasurementuncertaintiesasgivenby Sozzettietal. ( 2007 2009 ),sincetheiranalysisusedslightlyhigherresolutionspectra thanusedby Ammler-vonEiffetal. ( 2009 ).Asanextraprecaution,weconrmedthat theirmeasurementuncertaintieswererealisticbycomparingthemtotheadopted uncertaintiesgivenby Valenti&Fischer ( 2005 ),whichwerebasedontheanalysisof observationsofthesolarspectrumasreectedbytheasteroidVesta. Mostpreviousstudiesusetwo-parameterLDlawsandtforoneorbothLD parameters. Southworth ( 2008 )investigatedtheeffectsofvariousLDmodelsand assumptions.HeconcludedthatalinearLDlawwasofteninadequate,soanon-linear 27

PAGE 28

lawshouldbeused,butthattwo-parametertsweretypicallyhighlydegenerate.Inthe casesofTrES-2andTrES-3,two-parameterLDmodelsareparticularlydegenerate, sincebothhavealargeimpactparameterandtheplanetonlyprobesthestellarsurface brightnessnearthestellarlimb. Southworth ( 2008 )alsofoundthatincludinguncertainty inLDparameterswasimportanttoobtainrealisticuncertaintiesfortransitmodel parameters.Therefore, Southworth ( 2008 )recommendedholding c 2 xedandttingfor c 1 .Initially,wetriedthisapproachusingseparateLDparametersforeachstarandeach bandpassandttedfor c 1 .Wefoundthatthebest-ttingLDcoefcientsforonestarand ourtwobandpassescoulddiffersignicantly,eventhoughstellaratmospheremodels predictsimilarLDcoefcientsinthetwonearbybandpasses.Therefore,weadoptan alternativeapproach. ForbothTrES-2andTrES-3,wemodeltheLCsusingfourdifferentscenarios. First,wetseparatemodelstotheuxratiosatthetwowavelengths(Columns1 and2ofTable 2-2 ),xingboth c 1 and c 2 atapairofself-consistentvaluesbasedon PHOEBEmodelsusingasinglesetofspectroscopicparameters.Thetransitand LDparametersinthesersttwoscenariosneednotnecessarilybeself-consistent. Comparingthesersttwomodelsallowsustoevaluatethesensitivityofourresults tothetwodifferentbandpasses.Secondly,wecorrecteachLCbasedontheprevious resultsforthettedslopeandmeanOOTuxandtasinglemodeltotheLCsfor bothbandpassessimultaneously(Column3ofTable 2-2 ).Inthisscenario,thetransit andplanetaryparameters( t 0 b D p )areforcedtobethesameforbothbandpasses (self-consistentifweneglectthepossibilitiesofwavelength-dependentplanetradius andcontaminationfrombackgroundlight).Forthisandthesubsequentscenario,we xallfourLDparameters(twoforeachbandpass)atself-consistentvaluesbasedon PHOEBEmodels.Finally,wetasinglemodeltothetwo(corrected)LCsbutallowfor separatevaluesof D and p foreachLC(Column4ofTable 2-2 ).Forthenalscenario, wecontinuetoforcethettohavethesame t 0 and b forbothbandpassessincethese 28

PAGE 29

shouldbethesameregardlessofthebandpassusedforobservations,butallowthe modelsinColumn4tohavedifferentvaluesof p and D inthetwobandpassessoasto accountforapotentialdifferenceintheplanetradiusinthetwobandpassesorblending ofaputativebinarystar(Section 2.5 ).Sincewehaveobservationsofasingletransit inmultiplebandsandimposeadditionalconstraintsthatallfourLDparametersbe self-consistent,onewouldexpectourmethodtoresultinalargerrmsscatteraboutthe modelthanthettingproceduresusedbypreviousauthorsanddevelopedforanalyzing transitobservationsatasinglewaveband. Theprimarydifferencesinthemodelparametersamongthefourscenarioscanbe tracedtodifferentsetsofLDparameters.As Southworth ( 2008 )demonstrated,xing bothLDcoefcientsattheirtheoreticalvaluescouldproducemeasurementuncertainties thataretoosmall.ToaccountfortheuncertaintyintheLDmodel,werepeateachof theanalysesvaryingthespectroscopicparametersbythepublisheduncertaintiesand thusthecalculatedLDcoefcients.Wereportmodelparametersforthebest-tting modelwiththesmallest $ 2 value,butpresenttheirparameteruncertaintiesbased onthecompletesetofbest-ttingvaluesanduncertaintiesestimatedforallallowed LDparameters(Table 2-2 ).Forexample,weestimatetheuppererrorbarforagiven parameterastheresultofsubtractingitsbest-ttingvalue(asgivenbythemodel withthesmallest $ 2 value)fromthemaximumsumofatvalueanditsassociated measurementuncertainty,consideringthefullsetofmodelscomputedforthedifferent LDcoefcients. Tofurtherinvestigatethemeasurementerrors,weappliedthe"PrayerBead" method(e.g. D esertetal. 2009 ).Specically,weconstructsyntheticLCsbycalculating theresidualsfromtheinitialbest-ttingLCmodel,performingacircularshiftandadding theshiftedresidualsbacktothebest-ttingLCmodel.Then,weconductedtheabove analysisonthesyntheticdatasets.Weusethedispersionofthetparametersto estimatetheeffectsofanyothersystematicnoisesourcesnotremovedbytheEPD 29

PAGE 30

technique.Wediscusstheresultsfromthisanalysisaswellastheformal1 errors(as determinedfromthecovariancematrix)forthebest-ttingparametersinthefollowing sectionforthecasepresentedinColumn4. 2.4Results Weshowtheuxratiosandthebest-ttingmodels(aftercorrectingforslopes/trends basedonresultsgiveninColumn4ofTable 2-2 )inFigures 2-1 and 2-2 ,respectively. NotethatthetimeseriesgiveninTable 2-1 wascorrectedbasedonresultsfromColumn 4ofTable 2-2 aswell.Thebest-ttingLCparametersforTrES-2andTrES-3aregiven inTable 2-2 anddescribedinSection 2.4.1 .Weanalyzetheresidualsanddiscussthe photometricprecisioninSection 2.4.2 2.4.1PlanetaryParameters ForthejointanalysispresentedinColumn4ofTable 2-2 (whichweadoptasour baselinemodel),thebest-ttingparametervaluesandtheirformal1 uncertaintiesfor TrES-2are t 0 =2455007.64605 0.00013 (HJD), b =0.8642 0.0036 D (790.2nm) = 0.07365 0.00061 d, D (794.4nm) =0.07481 0.00063 d, p (790.2nm) =0.1253 0.0008 and p (794.4nm) =0.1245 0.0008 .Similarly,forthejointanalysisofTrES-3,the best-ttingparametervaluesandtheirformal1 uncertaintiesare t 0 =2455054.52447 0.00010 (HJD), b =0.8356 0.0073 D (790.2nm) =0.05817 0.00065 d, D (794.4nm) = 0.05652 0.00069 d, p (790.2nm) =0.1695 0.0025 and p (794.4nm) =0.1631 0.0022 WenotethatforbothTrES-2andTrES-3,theseuncertaintiesaretypicallysmaller thanthoseinTable 2-2 ,sincethosealsoaccountforuncertaintyintheLDmodel. Additionally,the1 errorsforthemedianbest-ttingparametersfromthePrayerBead analysisforbothTrES-2andTrES-3areslightlylargerthan,butstillcomparableto,the formal1 uncertainties,withthelargestdeviationsoccurringfortheuncertaintyinthe radiusratio.Wealsocomputedthermsofthebest-ttingparametervaluesfromthe PrayerBeadanalysisasanadditionalcheckonourmeasurementerrors.Wendthat 30

PAGE 31

accountingfortheuncertaintyintheLDmodelsufcientlyaccountsforthedistributionof thebest-ttingparametersasderivedfromthePrayerBeadanalysis. ForbothTrES-2andTrES-3,thebest-ttingvaluesforeachparameterare consistentbetweenthedifferentmodelspresentedinTable 2-2 butonlyafteraccounting foruncertaintiesintheLDmodel.Themostnotabledifferenceinourresultswith differentmodelingproceduresisforTrES-3whencomparingtheimpactparameterin Columns2and3(analysisforthe794.4-nmLCandthejointanalysiswithcommon planetparameters)withtheothertwomodelsforTrES-3.Thisisatleastpartlydue totheneardegeneracybetweenimpactparameterandLDmodel.Indeed,theLD coefcientsforthemodelsinColumns2and3aresignicantlydifferentfromthebest valuesfortheothertwocases.Whilewewereabletoplacetightconstraintsonthe impactparameterforTrES-2b,theprecisionforTrES-3bissignicantlyreduced,since wearemodelinganincompleteLCandhavenodatapriortoingress. Thebest-ttingslopesandtheirformaluncertaintiesasmeasuredintheindividual LCsbeforeperformingthejointanalysesare 0.00012 0.00172 and $ 0.00020 0.00180 d 1 forthe790.2-and794.4-nmLCsforTrES-2,and 0.00256 0.00501 and 0.00092 0.00540 d 1 forTrES-3.WhiletheestimatedslopesforTrES-2arefairlyconsistent,the differenceintheestimatedslopesismuchlargerforTrES-3thanforTrES-2.Notethat themeasuredslopesaccountfortheeffectsofbothdifferentialatmosphericextinction (duetosomereferencestarsbeingobservedatadifferentwavelengththanthetarget andprimaryreference)andrealastrophysicalvariabilityinthecolourofthetarget and/orreferencestars.Ourobservationsarenotgenerallyabletodisentanglethetwo potentialcausesofaslope.Iftheslopewereprimarilyduetodifferentialextinction,then observationsofmultipletransitsshouldproduceconsistentresults.Ontheotherhand,if stellarvariabilityissignicant,theneachtransitwouldneedaseparateslopeparameter. InthecaseofTrES-3,thelargedifferenceinslopesbetweenthetwobandpasses suggeststhatdifferentialextinctionisnotresponsible.However,wecautionthatthe 31

PAGE 32

measurementoftheslopearoundtransitislikelyaffectedbythelackofpre-transit data.Thus,weemphasizetheimportanceofacquiringcompletetransitLCsaswellas extended,uninterruptedOOTbaselinedata. ForTrES-2,theuncertaintiesintheOOTux(asdeterminedfromtheindividual LCanalyses)were 0.000091 ( 0.000097 )forthe790.2(794.4)nmobservations, basedon26(25)exposurestotaling 35(33)minofintegrationtime.Theuncertainties fortheOOTuxforTrES-3areslightlylarger,perhapsduetothelackofpre-transit observationsandcovariancewiththeslope.However,theTrES-3observationswere morestableoverallandallowedforasinglebaselineofOOTdatathatwaslongerthan eitherthepre-orpost-transitbaselinesfortheTrES-2observations.Bothobservations demonstratethecapabilityforveryhigh-precisionmeasurementsthatcouldenablethe detectionofsuper-Earth-sizedplanetsaroundsolar-likestarsand/orthecharacterization oftheatmospheresofgiantplanetsorbitingbrightstars(Sections 2.4.2 and 2.5 ). 2.4.2Light-CurveResiduals Wecomputedtheresidualuxbysubtractingthebest-ttingmodelsgivenbythe jointanalysispresentedinColumn4ofTable 2-2 fromthedata(Figures 2-1 bandcand 2-2 bandc).Notethatsomeoftheresidualsshowevidenceofadditionalsystematic noisesourcesthatwerenotremovedwiththeEPDtechnique.Whilewedonotknowthe originofthesesystematics,wenotethattheuncertaintyestimatesfromthePrayerBead analysisareconsistentwithourbest-ttingparameters(asdiscussedinSection 2.4.1 ). Weestimatethephotometricprecisiontobe343 45and412 43partspermillion (ppm)forobservationsofTrES-2at790.2and794.4nm,basedonthermsdeviationof theresiduals.Despitehighairmass(upto 2.22),theTrES-3observationsproduceda precisionof470 64and424 57ppmat790.2and794.4nm,respectively.Because theestimatedprecisionsforbothtargetsaresmallerthan(orcomparableto,based ontheupperlimitscomputedfromtheexpectedstandarddeviations)theestimated 32

PAGE 33

measurementuncertainties(Section 2.2 ),wetakethemtobeconsistentwiththe theoreticallimit. Duetothelong,uninterruptedOOTbaselineobtainedforTrES-3,wetakethe TrES-3observationsasthebest-casescenariotoachievehighprecisionsconsistent withthetheoreticallimitandpresentthestandarddeviationofthetime-binnedresiduals fortheTrES-3LCsinFigure 2-3 .Theresidualsforthe790.2-nmLCareconsistent withthetrendexpectedforwhiteGaussiannoise,whiletheresidualsforthe794.4-nm LCdeviateatbinningfactorslargerthan5.Whenbinningtheuncorrelated790.2-nm residualsoverlongerobservingtimes( 40min),weestimateaprecisionof146ppm, whichissufcienttodetectthetransitofasuper-Earth-sizeplanet( & 1.3 R )orbiting asolar-likestaroranEarth-sizeplanetorbitingastarsmallerthan0.7 R # (assuming 120minofobservationsbetweensecondandthirdcontactand3 condencelevel). 2.4.3TransitColour InFigures 2-1 (d)and 2-2 (d),wepresentthecolouroftheresidualuxes.The colourwascomputedbyaveragingeachpairofpointsinthe790.2-nmLCresidualsand dividingbytheuxresidualsofthe794.4-nmLC.Wendnosignicantcolourdeviation ineithertheTrES-2orTrES-3system.Weestimatermsprecisionsof519and502 ppmforTrES-2andTrES-3,respectively,whicharebothslightlylargerthan,butstill consistentwith,theprecisionsestimatedfortheindividualLCs.Wefoundnosignicant differenceinthecoloursintransitandOOTforTrES-2.ForTrES-3,therewasaslight slopeinthecolourduringtransit,butwedonotconsiderthistobesignicantgiventhe largeuncertaintyintheslopesttedtotheindividualLCsduetothelackofpre-transit observations. 2.5Discussion TherstexoplanetobservationsfromtheGTCprovidedexcellentphotometric precision,despiteproblemswiththetelescope,highairmassandrelativelypoor atmosphericconditions. 33

PAGE 34

OurresultsforTrES-3areconsistentwithprevioushigh-precisionobservations ( Gibsonetal. 2009 ; Sozzettietal. 2009 ).Ourmeasurementsofplanet-starradiusratio, impactparameterandtransitdurationbracketthoseofpreviousstudies.Ourmeasured transittimehasanuncertaintyof 9.5sandoccurs44safterthatpredictedby Sozzetti etal. ( 2009 )and14saftertheupdatedephemerisof Gibsonetal. ( 2009 ). Gibsonet al. ( 2009 )conductedatransittiminganalysisofTrES-3b,andtheyndnoevidenceof transittimingvariations.Giventhedifferencesbetweenvariousephemeridesandthe possibilityforstellarvariabilitytocontributeadditionaltimingnoise,wedonotconsider suchadifferencesignicant. OurresultsforTrES-2barealsoingoodagreementwith Holmanetal. ( 2007 )and Sozzettietal. ( 2007 ).However,wenotethatallreportedvaluesfortheradiusratio couldbeaffectedbyabinarycompaniontoTrES-2,whichhasaseparationof1.089 0.008arcsecfromtheprimarystarand i $ 3.7magasreportedby Daemgenet al. ( 2009 ).Whilethecompanionstarisclearlyincludedwithinouraperture,ourresults donotaccountfortheuxofthecompanioninouranalysis.Inthiscase,ourmodel parameter p servesasadepthparameterwhichcanberelatedtotheactualplanet-star radiusratioforagivenamountofcontamination( Daemgenetal. 2009 ).Ifweassume thattheprimary-secondarystaruxratioinourbandpassesissimilartothatin i $ ,then theplanet-starradiusratiowouldincreaseby 1.6percentrelativetoourestimates inTable 2-2 .Inprinciple,thedifferentcolourofablendedstarcouldresultindifferent transitdepthswhenobservedinmultipledifferentbandpasses.Whileourresultsare suggestiveofsuchatransitdepthdifference,thedifferenceisnotstatisticallysignicant. Giventhehighprecisionofourobservationsandapossiblefeatureinthein-transit data(aroundmid-transit),wesuspectthattheapparentdifferencesin p maybedueto 34

PAGE 35

variationsinthestellarsurfacebrightness. 5 Inlightofthesepotentialcomplications, near-simultaneousphotometryinmultiplebandpassescouldbeparticularlyusefulfor recognizingwhenacompanionorbackgroundobjectisaffectingthephotometry. Severalrecentstudieshaveconsideredthepossibilityofvariationsinthetransit timesand/ordurationsofTrES-2b( Mislis&Schmitt 2009 ; Rabusetal. 2009 ; Raetz etal. 2009 ; Mislisetal. 2010 ; Scuderietal. 2010 ).Ourbest-ttingtransittimehas ameasurementuncertaintyof 11sandisoffsetfromthepredictedephemerisof Holmanetal. ( 2007 )by 4min,of Rabusetal. ( 2009 )by 142s,of Raetzetal. ( 2009 )by 16sandof Scuderietal. ( 2010 )by 64s.Giventhedifferencesinvarious ephemerides,observingbandpassesandLDmodels,wedonotconsiderthetransittime offsettobesignicantevidencefortransittimingvariations.Thisisconsistentwiththe recenttransittiminganalysisof Raetzetal. ( 2009 ).While Rabusetal. ( 2009 )raisedthe possibilityofsinusoidalvariationsinthetransitephemerisduetoanexomoon,theydo notndresultsofthemagnitudeproposedby Mislis&Schmitt ( 2009 ). DuetoTrES-2b'slargeimpactparameter,thetransitdurationisverysensitiveto changesinducedbyadditionalplanetsorotherbodiesinthesystem( Miralda-Escud e 2002 ). Mislis&Schmitt ( 2009 )and Mislisetal. ( 2010 )claimthatthetransitduration decreasedby3minbetween2006and2008,andtheyarguethatathirdbodyisa naturalexplanationforthischange.IfweincludeuncertaintyduetounknownLD models,wecannotdenitivelyruleouttheshorterdurationssuggestedby Mislis& Schmitt ( 2009 )and Mislisetal. ( 2010 ).Wheninterpretingputativedifferencesintransit duration/impactparameter,oneshouldbemindfulofdifferencesinthevariousltersand thestellarLDmodelused( Scuderietal. 2010 ).Indeed, Scuderietal. ( 2010 )report 5 Indeed,previousstudieshavefoundotherindicationsofpossiblestellaractivity, includingvariationsintheOOTux( O'Donovanetal. 2006b ; Raetzetal. 2009 ).Our observationsshownosignsofaputative"seconddip"( Raetzetal. 2009 )butonly extendfor 1haftertheendoftransit. 35

PAGE 36

nosignicantchangeintheorbitalinclination/transitdurationwhentheycomparetheir observationswiththoseof O'Donovanetal. ( 2006b )(takeninthesamelter).Since ourobservationsuseuniquebandpasses,itisnot(yet)possibletocomparethemto previousobservationswiththesamebandpass.SinceTrES-2isinthe Kepler eld, Kepler observationsshouldsoonshedlightonthismatter,astheobservationswillcover longertime-scalesandthereforemultipletransitsallwithinthesamepassband. Insummary,wendthatlargeground-basedobservatoriesarecapableofachieving high-precisiondifferentialphotometry.Systematiceffectsduetovariableatmospheric extinctioncanbeminimizedbytheuseofasmallfar-redorNIRbandpasschosen toavoidskyabsorptionlines.Thus,large,ground-basedobservatoriescanhelp characterizesuper-Earth-sizeplanetsdiscoveredbyongoingtransitsearches,including planetsinthehabitablezoneofmain-sequencestarsforwhichtransitslastseveral hours.Weanticipatethatthisinnovativetechniqueforhigh-precisionphotometrywill enhancetheabilityofcurrentandfuturelargeoptical/NIRobservatoriestostudythe propertiesofbothEarth-likeplanets(e.g.size,densityandorbit)andgiantplanets.For example,thehighphotometricprecisioncouldallowforthedetectionoftheoccultation ofhottransitinggiantplanetslikeTrES-3batopticalwavelengths.Whiletheoccultation depthprobesthetemperatureanddynamicsoftheplanet'satmosphere,thetimeand durationofoccultationprovideapowerfulprobeoftheorbit'seccentricity. Thetechniqueisevenmorepowerfulwhenobservationsinmultiplebandpasses canbeobtainednearlysimultaneouslyandthusduringthesametransit.Such observationscouldhelpverifythatthetransitofacandidateplanetisnotdueto stellarvariabilityplusrotationbyobservinginmultiplewavelengthsthatreducethe effectsofstellarvariabilityand/orenhancethecontrastofspotsorplage.Similarly, nearlysimultaneousobservationsinmultiplenarrowbandpassescouldbeusedto characterizetheatmospheresofgiantplanets.Forexample,the HubbleSpaceTelescope andtheHobby-EberlyTelescope(HET)haveusedtransmissionspectroscopyof 36

PAGE 37

HD209458bandHD189733btomeasureNa I absorptionintheplanetaryatmospheres ( Charbonneauetal. 2002 ; Redeldetal. 2008 ).FutureGTCobservationscouldbe usedtoperformsimilarmeasurementsofatmosphericfeaturessuchasNa I andK I absorptioninplanetstransitingbrightstars,especiallyonceplannedimprovementsin theOSIRISCCDcontrollersoftwarereducethedeadtimebetweenexposures. Consideringtheincreasingnumbersofknowntransitingexoplanets,wehopethat thisnoveltechniquewillbeanadditionalusefultooltohelpcharacterizeavarietyof planetaryproperties. 37

PAGE 38

Table2-1. RelativephotometryofTrES-2andTrES-3.Thisisasampleofthefulltable, whichisavailablewiththeonlineversionofthepaper(Supporting Information). Observed (nm)HJDRelativeFluxUncertainty TrES-2 790.22455007.58450.999630.00043 790.22455007.58710.999690.00043 790.22455007.59231.000060.00043 ... 794.42455007.58580.999560.00044 794.42455007.59100.999960.00044 794.42455007.59360.999840.00044 ... TrES-3 790.22455054.49760.998550.00048 790.22455054.50110.994720.00048 790.22455054.50470.989690.00049 ... 794.42455054.49930.998000.00049 794.42455054.50290.992660.00049 794.42455054.50640.989360.00049 ... Note .Thetimestampsincludedhereareforthetimesatmid-exposure,andtherelative uxhasbeencorrectedforslopes/trendsintheLCs(Section 2.4 ). 38

PAGE 39

Table2-2. SystemparametersofTrES-2andTrES-3. ParameterValue 1 a 2 b ( 1 + 2 ) c ( 1 + 2 ) d TrES-2 t 0 $ 2455000(HJD )7.64601 +0.00018 0.00021 7.64610 +0.00021 0.00019 7.64605 +0.00013 0.00013 7.64605 +0.00013 0.00013 b 0.8647 +0.0051 0.0413 0.8638 +0.0055 0.0335 0.8639 +0.0036 0.0349 0.8642 +0.0036 0.0349 D (d) 0.07369 +0.00203 0.00073 0.07478 +0.00224 0.00078 0.07415 +0.00188 0.00051 1 :0.07365 +0.00214 0.00061 2 :0.07481 +0.00176 0.00064 p 0.1254 +0.0050 0.0009 0.1245 +0.0054 0.0010 0.1249 +0.0048 0.0007 1 :0.1253 +0.0052 0.0008 2 :0.1245 +0.0046 0.0008 c 1 (xed) 0.2200.232 1 :0.220 1 :0.220 2 :0.232 2 :0.232 c 2 (xed) $ 0.084 $ 0.086 1 : $ 0.084 1 : $ 0.085 2 : $ 0.086 2 : $ 0.098 TrES-3 t 0 $ 2455000(HJD )54.52450 +0.00014 0.00015 54.52445 +0.00015 0.00015 54.52449 +0.00012 0.00010 54.52447 +0.00011 0.00011 b 0.8276 +0.0575 0.0082 0.8737 +0.0178 0.0453 0.8716 +0.0123 0.0455 0.8356 +0.0482 0.0073 D (d) 0.05779 +0.00080 0.00194 0.05554 +0.00202 0.00086 0.05608 +0.00160 0.00061 1 :0.05817 +0.00072 0.00203 2 :0.05652 +0.00069 0.00174 p 0.1672 +0.0071 0.0052 0.1641 +0.0067 0.0067 0.1662 +0.0046 0.0048 1 :0.1695 +0.0046 0.0049 2 :0.1631 +0.0048 0.0044 c 1 (xed) 0.6530.222 1 :0.220 1 :0.653 2 :0.222 2 :0.625 c 2 (xed) 0.263 -0.086 1 : $ 0.081 1 :0.263 2 : $ 0.086 2 :0.233 a Welabelthebandpasscentredon790.2nmas 1 forsimplicity. b Welabelthebandpasscentredon794.4nmas 2 forsimplicity. c Thebest-ttingparametersweredeterminedfromthejointanalysisofthetwoLCsforeachtarget,withonlytheLD coefcientsdeterminedseparatelyforthetwowavelengths.Seethetextforadditionaldetails. d SameasinColumn3,butalsowith D and p determinedseparatelyforeachwavelength. 39

PAGE 40

Figure2-1. NormalizedLCs(a),residuals(b),(c),andcolour(d)fornearlysimultaneous observationsat790.2and794.4 2.0nmofTrES-2asobservedonUT 2009June25.Thelledcirclesareobservationsandthelinesinpanel(a) showthebest-ttingmodels.Inpanel(a),the794.4-nmLChasbeen arbitrarilyoffsetby0.015.Panels(b)and(c)showresidualsfromthetsfor the790.2and794.4nmLC.Thecolouroftheresidualsisshowninpanel (d). 40

PAGE 41

Figure2-2. SameasinFigure 2-1 ,butforTrES-3asobservedonUT2009August10.In panel(a),the794.4-nmLChasbeenarbitrarilyoffsetby0.025. 41

PAGE 42

Figure2-3. Standarddeviationoftime-binnedresidualsforTrES-3asafunctionofthe numberofdatapointsperbin( n ).Thebluesquare(redtriangle)symbolsare thestandarddeviationsofthe790.2(794.4)nmbinnedLCresiduals.The solidcurvesshowthetrendexpectedforwhiteGaussiannoise( n 1 / 2 ).Each exposurewasseparatedby5.11min;thus,thepresentobservationsonly allowustotesttime-scalesofupto 40minduetothelimiteddurationof theobservations. 42

PAGE 43

CHAPTER3 PROBINGPOTASSIUMINTHEATMOSPHEREOFHD80606BWITH TUNABLEFILTERTRANSITSPECTROPHOTOMETRYFROMTHEGRAN TELESCOPIOCANARIAS Discoveriesofextrasolarplanetswhichtransittheirhoststarprovidevaluable opportunitiestomeasurethephysicalpropertiesofexoplanetaryatmospheres.The physicalcharacteristicsofanexoplanetaryatmospherecanbeprobedbytransmission spectroscopyobservedagainstthespectrumofthehoststar. Seager&Sasselov ( 2000 ), Brown ( 2001 )and Hubbardetal. ( 2001 )developedmodelsthatpredictedsuch absorption,particularlyfromNa I ,K I andotheralkalimetals.Subsequentrenements ofsuchmodelshaveconrmedthatintheopticalwavelengthregimethestrongest linesareexpectedfromtheNa I resonancelines( !! 589.6,589.0nm)andtheK I resonancelines( !! 769.9,766.5nm)(e.g. Barman 2007 ; Fortneyetal. 2010 ). 1 Inthe optical,thecoresoftheatomicfeaturesofNa I andK I arerelativelynarrow.Forthis reason,mediumtohighresolutionspectrographscanbeusedtocomparethein-transit stellarspectrumtotheout-of-transit(OOT)stellarspectrum.Theabsorptionofstellar photonsintheexoplanetaryatmosphereleadstoexcessabsorptioninthein-transit stellarspectrumwhencomparedtotheOOTspectrum.Inphotometricobservations,this leadsthentodeepertransitsandalargerapparentsizeoftheplanetattheabsorbing wavelengths( Brown 2001 ),withvariationsofordertheatmosphericscaleheight ( Fortney 2005 ).Suchmeasurementsinstrongopticaltransitionscanalsoconstrainthe atmosphericmetallicity,rainoutofcondensates,distributionofabsorbedstellaruxand photoionizationofatmosphericconstituents. 1 WecautionthattheselinesaremostprominentforhotJupiterlikeplanetswith acertainrangeofatmospherictemperatures.AtmospheremodelsgeneratedforHD 80606batthetimeoftransit[basedon Fortneyetal. ( 2010 )]donotpredictasignicant K I absorptionfeature,duetothelowequilibriumtemperatureof500K.Wereferthe readertoSection 3.3.4 forfurtherdiscussion. 43

PAGE 44

TherstdetectionofabsorptionduetoanexoplanetaryatmospherecamefromNa I observationsofHD209458busingtheSpaceTelescopeImagingSpectrograph(STIS) onboardthe HubbleSpaceTelescope ( HST )( Charbonneauetal. 2002 ).Unfortunately, thesubsequentfailureoftheSTISinstrumentpreventedsimilarobservationsformore than5years.Thus,attentionwasdirectedtowardsmakingsuchobservationsfromthe ground(e.g. Moutouetal. 2001 ; Winnetal. 2004 ; Naritaetal. 2005 ).Thesecond detectionofabsorptionduetoanexoplanetaryatmosphere,thistimefromtheground, wasalsomadeofNa I inobservationsofHD189733busingthe9.2-mHobby-Eberly Telescope(HET)( Redeldetal. 2008 ).FurtherdetectionsofNa I intheatmosphereof HD209458bweremadeusingarchivaldatafromthe8.2-mSubaruTelescope( Snellen etal. 2008 ),from HST by Singetal. ( 2008a )andfromKeckby Langland-Shulaetal. ( 2009 ).TherecentrepairofSTISandinstallationoftheCosmicOriginsSpectrograph (COS)onboard HST hasenablednewopticalandultraviolettransmissionspectrum observationsofexoplanetaryatmospheres,extendedexospheresandauroralemission (e.g. Linskyetal. 2010 ; Fossatietal. 2010 ; Franceetal. 2010 ). ComparingthesurprisinglyweakNa I absorptioninHD209458b( Charbonneau etal. 2002 ; Knutsonetal. 2007 )tothethreetimesstrongerNaIabsorptionofHD 189733b( Redeldetal. 2008 )suggeststhatthetwoplanetshavedifferentatmospheric structures.Theoristshavesuggestednumerousmechanismssuchasadjustments tothemetallicity,rainoutofcondensates,distributionofabsorbedstellaruxor photoionizationofsodium( Fortneyetal. 2003 ; Barman 2007 ).Inparticular, Barman etal. ( 2002 )suggestedthatnon-localthermodynamicequilibriumNalevelpopulations werethecauseoftheweakNafeatureobservedinHD209458b,andareanalysisofthe Knutsonetal. ( 2007 )databy Singetal. ( 2008a b )suggestedthatNacondensationor NaphotoionizationinHD209458b'satmospherewasthebestexplanationformatching thedata,giventheNalineshapestheyderived.Itisclearthatcomparisonsofthe atmosphericpropertiesofdifferenttransitingplanetswillbecriticaltounderstandingthe 44

PAGE 45

atmosphericpropertiesofexoplanetsasawhole.Althoughstillsmall,thelistofdetected atomsandmoleculesisgrowing.InadditiontoNa I ,severalmoleculeshavebeen detected,primarilyintheinfrared,withbothspace-basedandground-basedplatforms, includingCO,CO 2 ,H 2 OandCH 4 ( Swainetal. 2008 ; Swainetal. 2009 ; Snellenet al. 2010 ).Other HST observationsusingtheAdvancedCameraforSurveys(ACS) didnotdetectK I inHD189733b( Pontetal. 2008 ).Ifdetectionsofconstituentsinthe extendedexosphereareincluded,thenH I ,C II ,O I ,Mg II andothermetalshavealso beendetected( Vidal-Madjaretal. 2003 2004 ; Fossatietal. 2010 ; Linskyetal. 2010 ). Eachnewdetectionprovidesnotonlycompositionalinformation,butalsoanother windowintothephysicalpropertiesoftheexoplanetaryatmosphere(e.g.condensation, windspeedandphotoionization).Eventhoughatmospheremodelsdonotpredicta signicantK I featureinHD80606b,itremainsofgreatinteresttoobservationally determinethelevelofK I absorptioninitsatmosphere,sinceK I isgenerallypredicted tobethesecondstrongesttransmissionspectrumsignatureintheopticalwavelength range.Further,Na I andK I probedifferentlayersoftheatmosphere.Measurements ofK I cantestthehypothesisthatthelowabundanceofNa I onHD209458bmaybe duetoahigh-altitudelayerofcloudsorhaze.FindinglowabundanceforbothNa I and K I wouldbeconsistentwitheitherthecloudhypothesisorwiththephotoionization hypothesis,asbothareveryeasytoionize.FindingthatonlyNa I issignicantly depletedwouldpointtoalternativemodelswithcomplexatmosphericchemistry(e.g. incorporationintograins,oddtemperaturestructure,unexpectedmixingpatterns). Finally,inprinciple,futureobservationscouldprobetemporalvariabilityofNa I andK I duetohigh-speed,high-altitudewindsand/ordifferencesintheleadingandtrailinglimb ( Fortneyetal. 2010 ). Alloftheaboveatmosphericstudieswerebasedonobservationsusinghigh-resolution spectrographs.Here,wedescribeanewtechniquethatutilizesfast,narrow-band spectrophotometrywiththeOpticalSystemforImagingandlowResolutionIntegrated 45

PAGE 46

Spectroscopy(OSIRIS)installedonthe10.4-mGranTelescopioCanarias(GTC)to probethecompositionandotherpropertiesoftheatmospheresofexoplanetsthat transitbrightstars(Section 3.1 ).Fastlinespectrophotometrycanbemuchmore efcient(e.g. 34% withGTC/OSIRIS)thantypicalhigh-resolutionspectrographs ( 1-2percent)thankstotheuseofatunablelter(TF)ratherthandiffractiongratings. Further,thistechniquehasthepotentialtobelesssensitivetoseveralsystematicnoise sources,suchasseeingvariationsthatcauselinevariationsinwidespectrographslits (specicallyinnon-berfedspectrographs),atmosphericvariations(sincereference starswillbeobservedsimultaneously)and/orat-eldingerrors(sinceon-andoff-line dataareobtainedatthesamedetectorlocation).Thus,spectrophotometrywithaTF techniqueisparticularlywellsuitedforobservinganarrowspectralrangeofatomic absorptionfeatures,withoutsufferingfromtheinefcienciesorpotentialsystematic uncertaintiesofhigh-resolutionspectrographs. Herewepresentresultsofsuchobservationsofthe2010Januarytransitof HD80606busingtheGTCandtheOSIRISTFimager.HD80606bwasoriginally discoveredbyradialvelocityobservations( Naefetal. 2001 )andwasremarkable duetoitsveryhigheccentricity( e =0.93 ).Onlyseveralyearslaterdid Spitzer and ground-basedobservationsrevealthattheplanetpassesbothbehindandinfrontof itshoststar( Laughlinetal. 2009 ; Fosseyetal. 2009 ; Garcia-Melendo&McCullough 2009 ; Moutouetal. 2009 ).Spectroscopicobservationsrevealedthattheangular momentumaxisofthestellarrotationandthatoftheorbitalplanearemisaligned ( Moutouetal. 2009 ; Pontetal. 2009 ; Winnetal. 2009 ).Giventheinfrequenttransits andlongtransitduration( 12 h),follow-upobservationsarequitechallenging. Winn etal. ( 2009 ), Hidasetal. ( 2010 )and Shporeretal. ( 2010 )wereabletocharacterize transitsofHD80606bwithlongitudinallydistributednetworksofground-based observatories,and H ebrardetal. ( 2010 )observedthe2010Januarytransitusing the Spitzer spacecraft. 46

PAGE 47

The Spitzer observationsconstrainthethermalpropertiesoftheplanet'satmosphere ( Laughlinetal. 2009 ; H ebrardetal. 2010 ).Tothebestofourknowledge,the observationspresentedherearethersttoattempttodetectatmosphericabsorptionby HD80606b.WhileexistingatmospheremodelspredictthatHD80606bwould not have anysignicantK I featureduetoitshighsurfacegravityandcoldatmosphereatthetime oftransit(e.g.Section 3.3.4 ),ourobservationstestthisprediction.Eventhoughmodels donotpredictaK I feature,exoplanetobservationshaveatrackrecordofunexpected discoveries.Furthermore,inprinciple,dependingontheatoms/moleculesfoundinthe atmosphere,theseobservationscouldyieldinformationabouthowtheplanetcools, independentofanyobservationsofthethermalphasecurveofthissystem.Inprinciple, transmissionspectroscopyalsoprovidesawaytocharacterizetransitingplanetsin eccentricorbits,whicheitherdonotpassbehindtheirhoststarorwhicharetoocoolto detectviaoccultationwhentheydopassbehindthestar. Finally,wenotethatHD80606isoneofthebestsystemsformakingveryprecise spectrophotometricmeasurements.HD80606isthebrightestofthetransitingplanet hoststarswhichhaveacomparablybrightreferencestarverynearby( 20arcsec). Also,thelongdurationbetweenthesecondandthirdpointsofcontact( 6h)ofHD 80606bprovidestimetocollectalargeamountofin-transitdatainasingletransit.Thus, weexpectthatallelse(e.g.observingconditions)beingequal,HD80606bpermitsthe mostprecisespectrophotometricmeasurementsofanyknownsystem(atleastwith observationsofasingletransit). ThispaperpresentsextremelyprecisemeasurementsofthevariationinHD 80606b'sapparentradiuswithwavelengthneartheK I feature,whichinturncanhelpus testthepredictionsofatmospheremodels.Section 3.1 describesourobservationsand dataanalysisprocedures.WedescribetheresultsofourobservationsinSection 3.2 .In Section 3.3 weinterprettheresults,andwesummarizeourconclusionsanddiscussthe futureprospectsforthemethodinSections 3.4 and 3.5 47

PAGE 48

3.1Observations HD80606anditsnearbycompanion(HD80607)arebothbrightG5dwarvesof asimilarmagnitude(V 9)andcolour.Onthreenights,wemeasuredtheuxofboth HD80606(target)andHD80607(reference)simultaneously.Wecycledthrougha setoffourwavelengthsthroughouttheobservations.Onthenightof2010January 13-14,theplanetwasintransitforthedurationofourobservations,andwemeasurean "in-transit"uxratioofHD80606toHD80607foreachwavelength.Werepeatedthe observationson2010January15and2010April4,whentheplanetwasnottransiting HD80606,allowingustomeasuretheOOTuxratioofHD80606toHD80607for eachwavelength.Ourresults(Section 3.2 )arebasedontheratioofin-transitux ratio(targetoverreference)toOOTuxratio(targetoverreference).Anychangesin theEarth'satmospherefromonenighttothenextshouldaffectboththetargetand referencestarsimilarly.Bymakingdifferentialmeasurementsofthecolourduringthe sametransitandatsimilaratmosphericconditions,thismethodallowsforextremely precisemeasurementsofthetransitdepthatdifferentwavelengths.Whilenight-to-night variabilityintheatmosphericconditionsoreitherofthestarscouldcauseasystematic scalingofthetransitdepthmeasurements,therelativewavelengthdependenceofthe apparentplanetradiusislargelyinsensitivetoeitherofthesepotentialsystematics.We referthereadertoSections 3.3.9.1 and 3.3.9.3 forfurtherdiscussion. 3.1.1In-TransitandOut-of-TransitObservations WeobservedapartialtransitofHD80606bon2010January13-14andacquired baselinedataon2010January15and2010April4toestablishtheOOTuxratios. Forourobservations,weusedtheTFimagingmodeoftheOSIRISinstrumentinstalled onthe10.4-mGTC,whichislocatedattheObservatoriodelRoquedelosMuchachos ontheislandofLaPalma( Cepaetal. 2000 2003 ).IntheTFmode,theusercan specifycustombandpasseswithacentralwavelengthof651-934.5nmandafullwidth athalf-maximum(FWHM)of1.2-2.0nm.Theeffectivewavelengthdecreasesradially 48

PAGE 49

outwardfromtheopticalcentre;becauseofthiseffect,wepositionedthetargetand itsreferencestaratthesamedistancefromtheopticalcentreandonthesameCCD chip.Theobservedwavelengthsdescribedbelowrefertothelocationofthetarget(and reference)ontheCCDchip. Duringthetransitobservationsandbaselineobservationson2010January15, exposuresofthetargetanditsreferencestarcycledthroughfourdifferentwavelengths (allwithaFWHMof1.2nm):oneonthepredictedcoreoftheK I line(769.75nm);one totheblueside(768.60nm)andtworedwardsoftheK I feature(773.50and777.20 nm).AsthetuningsfortheTFaresetbytheordersorter(OS)lterused,ourbluest wavelengthisthebluestwavelengthwecouldobserveatinthewingoftheK I lineand stillobservewithinthesameOSlterasthe"on-line"wavelength(i.e.atthelocationof thecoreoftheK I line).WethenchosetwowavelengthsredwardsoftheK I lineinorder tosamplemoreofthestructure/wingsaroundtheK I line.Thereddestbandpasswas chosensinceweexpecttosee(foratypicalhotJupiter)amaximumdifferencebetween theuxratiointheon-linebandpassandaroundthatreddestbandpass.Inorderto maximizethesignal-to-noiseratiosintheon-linewavelengthandinthereddestoff-line wavelength,ineachsequenceweobservedon-linethreetimes,atthereddestoff-line wavelengthtwotimesandattheotheroff-linewavelengthsonetimeeach.Duringthe transit,theobservingsequencefromtheGTCwasasfollows:769.75,768.60,769.75, 773.50,769.75,777.20and777.20nm(repeat). Weemphasizethatthesewavelengthswerechosentobearoundthelocationofthe K I featureinHD80606b'satmosphere.InordertoobserveontheK I feature(which hasarestwavelengthof 769.9nm)intheframeoftheplanet,weaccountedforthe DopplershiftsduetotheEarth'smotionaroundtheSun,thesystem'sradialvelocity andtheplanet'snon-zeroradialvelocityduringtransit[ $ 59.6 kms 1 basedonvelocities from Winnetal. ( 2009 )].Afteraccountingfortheseeffects,theobservedwavelengths intheframeoftheplanetareredshiftedby0.16nmto769.91nm(on-line)and768.76, 49

PAGE 50

773.66and777.36nm(off-line).Theobservedwavelengthsintheframeofthestarare essentiallythesameasobservedonEarthduetothesmallsystemicvelocityoftheHD 80606planetarysystemandtheEarth'ssmallbarycentricvelocityonthenightofthe transit.Fortheremainderofthepaper,wereportthewavelengthsasobservedinthe frameoftheplanetwhendiscussingresultsfromthetransitobservations. Asimilarsequenceasdescribedabovewasusedforthebaselineobservations takenon2010April4,buttheobservedwavelengthswerecorrectedfortheDoppler shiftduetotheplanet'sorbitalvelocityonthatspecicdate( 23.9kms 1 )inorderto matchthewavelengthsobservedduringthetransit.Thus,thewavelengthsobservedon 2010April4(fromtheGTC)are770.00nm(on-line)and768.86,773.76and777.45nm (off-line). TransitobservationsofHD80606bbeganat22:28UTon2010January13(during ingress)andendedat7:15UTon2010January14(aroundthebeginningofegressand includingastronomicaltwilight),duringwhichtheairmassrangedfrom 1.08to1.72. Theobservingconditionswerephotometric,withaclearskyandadarkmoon.Nodata weretakenbetween5:20and5:50UTon2010January14duetorecalibrationofthe TFduringthattime.Theactualseeingvariedbetween0.7and0.9arcsecduringthe transitobservations,butweusedaslightdefocustoincreaseefciencyandreducethe impactofpixel-to-pixelsensitivityvariations.Therefore,thedefocusedFWHMofthe targetvariedfrom 0.9to2.3arcsec(7-18pixels)duringthetransit.Fortheportionof thelightcurveusedinouranalysis(Section 3.1.2 ),theFWHMwasmuchmorestable thanisindicatedbytherangegivenabove,withatypicalvaluebetween10and14pixels andameanvalueof12pixels.Evenwithanautoguidingsystem,thetarget'scentroid coordinatesshiftedby 9-10pixelsoverthecourseofthenight.Weused1 # 1binning andafastreadoutmode(500kHz)toreadoutasinglewindowof300 # 600pixels (locatedononeCCDchip)inordertoreducethedeadtimebetweenexposures.This windowisequivalenttoaeldofviewof 38 # 76arcsec 2 ,sotheonlystarsinoureld 50

PAGE 51

wereHD80606andasinglereferencestar,HD80607.Eachindividualobservationwas followedbyanaveragedeadtimeoflessthan4sforreadoutandtoswitchbetweenTF tunings.Weused10sexposures,resultinginanoverallcadenceofabout14sforeach observation.Duetotheshortexposuretimeused,theskybackgroundlevelwaslow enoughthatwedidnotneedtodiscardanyimagestakenduringastronomicaltwilight. Baselineobservationsweretakenfrom5:50to7:10UT(i.e.alsothroughthe beginningofastronomicaltwilight)on2010January15,butthedatawerehighly scattered,sowedonotincludeitinourprimaryanalysis. 2 Additionalbaseline observationstookplaceon2010April4from21:30(includingtheendofastronomical twilight)to0:00UT.Theobservingconditionswerephotometricandtakenduring greytime,usingthesameset-upasthein-transitobservationsdescribedabove. Duringtheobservations,theairmassrangedfrom 1.08to1.20,andtheactual seeingvariedbetween1.4and1.6arcsec(11-12.5pixels),sothetelescopewasnot intentionallydefocused.Thetarget'scentroidcoordinatesshiftedby 5-8pixelsduring theobservations.Theexposuretimewaschangedfromtheinitialexposuretimeof10s to8sandthenagainto11stocounteractvariationsintheseeingaswellasincreasing airmasswhileavoidingsaturationandmaintainingahighnumberofcounts.Inour analysis,wediscardthe10-sdatabecauseamajorityoftheimagesweresaturated. WetestedusingtheOOTuxratiosfromthe8and11sdataindividuallyinouranalysis andfoundthattheyproducedverysimilarresults.Thus,wecombinethe8and11sdata toestablishthenalOOTuxratios(Section 3.1.2 )andtoachievethelongestusable baselinepossible. 3.1.2DataReductionandAnalysis ObservationstakenwithOSIRISpriorto2010mid-Marchsufferedfromahigher thanexpectedlevelofdarkcurrentdespitetheshortexposuretimesused.Therefore, 2 Section 3.3.2 51

PAGE 52

weusedstandardIRAFproceduresforbiasanddarksubtractionandat-eldcorrection forthe2010JanuarytransitobservationsofHD80606.Wenotethattheat-eldsfor theseobservationsdidnotproducethepatternofhavingthetotalnumberofcounts inthedomeat-eldsdecreasingwithtimeasseenby Col onetal. ( 2010 ),soweuse almostall(65outof75)domeatsforeachobservedwavelengthinouranalysis(the 10domeatsnotincludedintheanalysiswereoverexposed).Anewdewarxedthe problemswiththedarkcurrentbeforethe2010Aprilobservationstookplace,sofor thebaselinedataweperformedstandardbiassubtractionandat-eldcorrection (combiningall133atstakenforeachobservedwavelength)anddidnotneedto subtractdarkframes. Becauseoftheverysmallreadoutwindowusedforourobservations,ourimagesdo notcontainthesky(OH)emissionringsthatoccurduetotheTF'ssmallbandpassand position-dependentwavelength.Therefore,weperformedsimpleaperturephotometry onthetargetandreferencestarusingthestandardIDLroutineAPER 3 forarangeof apertureradii.Wemeasuredthermsscatteroftheuxratio(equaltothetargetux dividedbythereferenceux)atthebottomofthetransit(forthe2010Januarydata) andfortheindividual8and11sdatatakenOOT(in2010April)ineachbandpass. Weconsideredtheresultsforeachbandpassandadoptedanapertureradiusof28 pixels(3.6arcsec)forthein-transitdataand32pixels(4.1arcsec)fortheOOTdata,as theseweretheapertureradiithattypicallyyieldedthelowestrmsscatter.Theradiiof theskyannulususedforthereductionofbothdatasetswere68-74pixelsinorderto completelyavoidanyuxfromthetargetorreferencestar.Wehaveincludedtheresults ofouraperturephotometryinTables 3-1 and 3-2 andillustratetheresultsinFigures 3-1 and 3-2 .Asillustrated,theuxineachbandpassdisplayedlargevariationsduring 3 Landsman1993;http://idlastro.gsfc.nasa.gov/ 52

PAGE 53

partsoftheobservations(particularlyduringpartsofthetransit),andwetakethisinto considerationinouranalysis(Section 3.2.1 ). Wepresenttherawin-transitlightcurvesinFigure 3-3 ,whichwerecomputedby dividingtheuxinthetargetaperturebytheuxinthereferencestarapertureand thennormalizingbytheweightedmeanOOTuxratioforeachbandpass(Section 3.2 fordetailsonthecomputationofthemeanuxratios).Inanattempttoreduce systematictrendsseeninourtransitlightcurves,weappliedtheexternalparameter decorrelation(EPD)technique(e.g. Bakosetal. 2007 2010 )toeachtransitand baselinelightcurve.Notethatforthetransitlightcurve,weonlyappliedEPDtothe 4hcenteredaroundmid-transit,or3:36UTon2010January14,asestimatedby Shporeretal. ( 2010 ). 4 Specically,wedecorrelatedeachindividuallightcurveagainst thefollowingparameters:thecentroidcoordinatesofboththetargetandreference, thesharpnessofthetargetandreferenceproles[equivalentto(2.35/FWHM) 2 ]and theairmass.AsillustratedinFigure 3-4 ,EPDremovedmostofthecorrelationsinthe in-transitdata.Forreference,weshowthecorrelationsbetweenthein-transitdata andthetarget'sFWHMandcentroidcoordinatesbothbeforeandafterEPDhasbeen appliedinFigure 3-5 .Forthebaselinedata,weperformedEPDforthe8-and11-sdata seriesseparately,butwethencombinedthetwodatasetstocomputetheweighted meanuxratioanditsuncertaintyforeachbandpassasdescribedinSection 3.2 .The resultsofthedecorrelationfortheOOTdataareillustratedinFigure 3-6 .Asaresult ofapplyingEPD,thermsscatterineachofthebandpassesimprovedbyasmuchas 25percent,butdecorrelatingthelightcurvesagainsttheaboveparametersdidnot completelyremovethesystematicsthatareseeninourdata.Inafurtherattemptto removesystematics,wealsotriedaquadraticdecorrelationagainstthesharpnessof 4 Thisephemerisisinbetweenthatgivenby Winnetal. ( 2009 )and H ebrardetal. ( 2010 ).Thechoiceofephemerisuseddoesnotsignicantlyaffectourresults. 53

PAGE 54

thetargetandreferenceproles,asthatwastheonlyparameterthatshowedapossible residualsystematicpatternafterEPDwasapplied.However,thequadraticdecorrelation didnotreducesystematicsinourlightcurvesanyfurther.Wediscussotherpotential sourcesofsystematicsindetailinSection 3.3.9 Becauseourgoalistocomparethedepthsofthetransitineachbandpass,therest ofouranalysisfocusesonthedatafromthebottomofthetransitaspresentedinFigure 3-3 andhighlightedinFigure 3-7 i.e.the 4hcenteredaroundmid-transit.Notethat thelightcurvesshowninFigure 3-7 havebeencorrectedusingEPD.Wealsodiscarded pointsthathadauxratiogreaterthan3 fromthemeanofthebottomofthetransit lightcurve.Thisresultedindiscardingfourpointsfromthereddestlightcurve(777.36 nm).Wealsodiscardedseveralexposuresfromeachwavelengththatwereunusable duetosaturation.ThedifferentpanelsinFigure 3-7 illustratethedeviationbetween themagnitudeoftheon-lineuxratiosandeachoftheoff-lineuxratios,whichwillbe discussedindetailinSections 3.2 and 3.3 Weestimatedtheuncertaintiesintheuxratiosbycomputingthequadraturesum ofthephotonnoiseforHD80606andHD80607,theuncertaintyinthesumofthe skybackground(anddarkcurrent,forthein-transitobservations)andthescintillation noiseforthetwostars.WeassumePoissonstatisticstocomputetheuncertaintyin theskybackground,andthenoiseduetoscintillationwasestimatedfromtherelation givenby Dravinsetal. ( 1998 ),basedon Young ( 1967 ).Wecautionthatthisempirical relationmightoverestimatescintillationforlargetelescopeslocatedatexcellentsites suchasLaPalma.Regardless,therelationdemonstratesthatscintillationisstillasmall contributiontothetotalerrorbudgetfortheseobservations.Theat-eldnoiseisalso negligiblecomparedtothephotonnoise,sowedonotincludeitinourdeterminationof themeasurementuncertainties.Basedontherelationgivenby Howell ( 2006 ),which computesthestandarddeviationofasinglemeasurementinmagnitudesandincludes acorrectiontermbetweentheerrorinuxunitsandtheerrorinmagnitudes,wend 54

PAGE 55

themediantotaluncertaintiesintheuxratioforeachexposuretobe0.538,0.532, 0.514and0.486mmagat768.76,769.91,773.66and777.36nm(overthebottomof thetransit),respectively.Thermsofthetransitlightcurveiscomparable,butslightly larger,withvaluesof0.585,0.667,0.631and0.662mmagforthosewavelengths.The mediantotaluncertaintiesfortheOOTobservationsarecalculatedinasimilarway, buttheuncertaintiesforthe8-and11-sdatasetswerescaledbytheuxratiosfor eachrespectivesetinordertocomputeaweighteduncertainty.Thus,themediantotal (weighted)uncertaintiesintheuxratioare0.657,0.650,0.627and0.592mmag,while theestimatedrmsisquitecomparable,withvaluesof0.562,0.605,0.554and0.586 mmagfor768.73,769.87,773.63and777.32nm,respectively. Thecompletephotometrictimeseriesforeachbandpassofthein-transitdata (uncorrectedandunnormalized)isreportedinTable 3-3 ,whilethephotometrictime series(bothbeforeandafterEPDwasapplied)forthetransitbottomandtheApril observationsarereportedinTables 3-4 and 3-5 .Theweightedmeanuxratiosfor boththein-transitandOOTdata(Section 3.2 )aregiveninTable 3-6 ,alongwiththeir uncertainties. 3.2Results AsillustratedinFigure 3-7 ,wecanseebyeyeahintofadeviationbetweenthe in-transituxratiosobservedattheon-linewavelengthandtheredoff-linewavelengths, butnocleardeviationisseenwhencomparedtothebluestoff-linewavelength.Despite evidenceoftime-correlatedsystematicsinourdata,weemphasizethattheerrorbars showninFigure 3-7 arebinnederrorbars,whichillustratethatourmeasurement uncertaintiesarelargerthananyresidualsystematicspresentinthelightcurvesand thatthedeviationsintheuxratiosbetweenthedifferentbandpassesarereal.Werefer thereadertoourdiscussionofpossiblesystematicsourcesinSection 3.3.9 InFigure 3-8 ,weplothistogramsofthe(unbinned)uxratiosatthebottomofeach ofthetransitlightcurves,wheretheuxratioshavebeennormalizedagainstthemean 55

PAGE 56

OOTuxratioforeachrespectivewavelength.Thesehistogramsfurtherillustratethat theuxratiosfortheon-lineandbluestoff-linelightcurvesarecomparable,butthered off-lineuxratios(particularlyforthereddestlightcurve)clearlylieatslightlyhigher valuescomparedtotheon-lineuxratios,indicatingasmallerapparentplanetaryradius atthosewavelengths. Ideally,whenonehasaccesstoeitheracompleteorpartialtransitlightcurve andbaselinedataacquiredimmediatelybeforeorafterthetransitevent,onecanta modeltothedataandestimatethetransitdepthfromthemodelresults.Duetothevery longdurationofHD80606b'stransit,wewerenotabletoacquirebaselinedataonthe nightofthetransit,therebymakingthistypeofanalysisimpractical.However,thanks toseveralrecentcampaignstoobserveacompletetransitofHD80606bandestablish accurateorbitalandphysicalparametersforthissystemvialight-curvemodeling( Winn etal. 2009 ; H ebrardetal. 2010 ; Hidasetal. 2010 ; Shporeretal. 2010 ),wedonot needtotamodeltoourpartiallightcurvetoachievethegoalsofthispaper.Instead, weconsideronlythemiddle 4hofthetransitlightcurveinouranalysis(compared tothefulldurationofthebottomofthetransit,whichis 6h),therebyminimizing systematiceffectsofstellarlimbdarkening(LD)asthestrongestLDoccursduring ingress,egressandrightafter/beforeingress/egress.Further,sincewedonotknow theLDmodelforthisstartotheprecisionofourobservations,addingsuchamodel wouldnotbeusefulforthisstudy.Thus,weassumethatLDisthesameoverallour bandpassesandthatthetransitephemeris,impactparameterandtransitduration donotvarywithwavelength.Theonlyparameterofwhichweassumechangeswith wavelengthistheapparentplanetradius( R p ). Toinvestigatehowtheapparentplanetradiuschangeswithwavelength,wesimply computetheweightedmeanin-transituxratio[ < % F / F > ,whichisproportional totheplanet-to-starradiusratio, ( R p / R ) 2 ],anditsuncertaintyforeachwavelength. 56

PAGE 57

Specically,wecomputetheweightedmeanas < % F / F > = n i =1 w i F i n i =1 w i (31) wheretheweights, w i ,areequalto1/ ( &" i ) 2 .Here, i istheestimatedphotometric uncertaintyweightedbysomewavelength-specicfactor( & )inordertoaccountforthe presenceofanyrednoiseineachindividualbandpass. Toillustratetheeffectofrednoiseonourmeasurementsandtheneedfora re-weightingfactor,thestandarddeviations( N )ofthein-transitandOOTtime-binned uxratiosareshowninFigures 3-9 and 3-10 asafunctionofbinningfactor( N )foreach bandpass.ThetheoreticaltrendexpectedforwhiteGaussiannoise( N 1 / 2 )isplotted asasolidcurve,andwecanseethatforthein-transitdatathermsdeviatesfromthe theoreticalcurveatlargebinningfactors,indicatingthatrednoiseispresentinmost bandpasses(beingtheleastsignicantinthebluestbandpass).However,fortheOOT data,ourphotometryappearstobegenerallyconsistentwiththephotonlimit(although thebluestlightcurvesuffersfromsmallnumberstatistics). Followingmethodsusedbye.g. Pontetal. ( 2006 )and Winnetal. ( 2007 ),we calculatedexplicitestimatesforboththewhite( w )andred( r )noiseineachbandpass bysolvingthefollowingsystemofequations: 2 1 = 2 w + 2 r (32) 2 N = 2 w N + 2 r (33) There-weightingfactor, & ,isthencomputedas r / ( w / ( N ) .Basedonourtstothe redandwhitenoise,wecomputedare-weightingfactorforeachbandpassandapplied itasstatedabove.Weimposedaminimumvaluefor & of1,particularlyforcaseswhere rednoisewasnegligible. 57

PAGE 58

TheuncertaintiesfortheOOTuxratioarealsoweightedbytheuxratio, F i sincetwodifferentexposuretimeswereusedduringtheOOTobservations.Finally,the uncertaintyontheweightedmeaniscomputedas < F / F > = # $ $ $ $ % 1 n i =1 w i (34) WeincludetheuncertaintyontheweightedmeanOOTuxratioinourcalculationofthe meannormalizedin-transituxratioanditsuncertainty.TheresultingspectrumofHD 80606b(thenormalizedweightedmeanin-transituxratiosasafunctionofwavelength) isshowninFigure 3-11 ,anditclearlyillustratesadifferencebetweentheuxratios forthebluestbandpassesandthoseforthereddestbandpasses.Whilewendno signicantdifferencebetweentheuxratiosmeasuredat768.76and769.91nm,we measuredifferencesof 3.02 3.02 # 10 4 and 8.09 2.88 # 10 4 betweenobservations at769.91and773.66and777.36nm. Welisttheweightedmeanin-transituxratios(normalizedbytheweightedmean OOTuxratios)aswellastheweightedmeanOOTuxratiosandtheiruncertainties inTable 3-6 .Inthistable,wealsoincludeourtstothewhiteandrednoise,aswell asourestimatesfor & .Whencalculatingthenormalizedin-transituxratioandits uncertainty,wealsoincludedthere-weighteduncertaintyforthemeanOOTuxratioin ourcalculation.TheerrorbarsfortheuxratiosgiveninTable 3-6 andshowninFigure 3-11 alsotakerednoiseintoaccount. 3.2.1EffectsofEarth'sAtmosphere Weconsidertheeffectofrandomatmosphericvariations(e.g.clouds)duringthe nightofthetransitaswellasduringtheAprilbaselineobservations.Asmentionedin Section 3.1.2 ,largevariationsintheabsoluteuxofboththetargetandreferencewere observedtowardsthebeginningandtheendofthetransitobservations,withafew largeuctuationsaroundthemiddleoftheobservationsaswell.Thus,tocheckifour 58

PAGE 59

measuredin-transituxratioswereaffectedbytheseuctuations,wecomputedthe weightedmeanin-transituxratioforeachbandpassafterexcludingoutlyingabsolute uxmeasurementsfromouranalysis.Wespecicallyexcludedanypointsthatwere greaterthan3 awayfromthemeanoftheattestpartofthespectrummeasuredfor eachbandpassandeachstar.Afterexcludingoutlyingpointsfromboththein-transit andAprilbaselinedata,wefoundthatthenewspectrumforHD80606bshowsavery similarshapeastheoriginalspectrum,albeitwiththeuxratiointhereddestbandpass differingthemostfromtheoriginalspectrum.However,westillmeasureasignicant differencebetweentheuxratiosintheon-lineandreddestbandpasses.Theseresults areincludedinTable 3-7 andshowninFigure 3-11 asthesolidcircles. 3.2.2Limb-DarkeningEffects SofarouranalysishasassumedthatLDisthesamebetweenourdifferent bandpasses,soLDshouldnotaffectthemeanuxratiosforeachbandpassdifferently. However,inprinciple,thereisalsothepossibilitythatLDcoefcientsvarysignicantly inandoutofnarrowspectrallines.Toinvestigatethepossibilitythatourspectrum's signatureisaresultofourprobinginandoutofHD80606'sstellarspectrallines,we havecomputedquadraticLDcoefcientsforeachofourbandpassesforagridof stellarmodels[usingPHOEBE; Pr sa&Zwitter ( 2005 )].Wethengeneratedtheoretical limb-darkenedlightcurvesforeachbandpassusingthestandardplanettransitmodelof Mandel&Agol ( 2002 ).WeusedstellarparametersanduncertaintiesforHD80606as givenby Winnetal. ( 2009 )toestimatearangeofLDcoefcientstouseinourmodels. WealsoinputplanetaryparametersanduncertaintiesforHD80606basgivenby H ebrardetal. ( 2010 ).Aftercomputinglight-curvemodelsfordifferentcombinationsof LDcoefcientsandplanetaryparameters,wecomputedthemeanmodeluxratioover thebottomofeachtransitlightcurve(the4hcenteredaroundmid-transit).Weinclude theresultingmodelspectruminFigure 3-11 assolidsquares.Thisparticularspectrum wascomputedbasedonusingamediansetofLDcoefcients,butallthemodelresults 59

PAGE 60

weresimilarovertherangeofLDcoefcientsused.ThemedianlinearandquadraticLD coefcients( u 1 u 2 ) are(0.392,0.229),(0.388,0.233),(0.391,0.230)and(0.376,0236)for the768.76-,769.91-,773.66-and777.36-nmbandpasses. WhilesmalldifferencesinLDexistbetweenthedifferentbandpasses,themean modeluxratiosdifferedbyonlyaverysmallamount( < 2 # 10 5 )betweenthedifferent bandpasses.Fromthis,weconcludethatLDismostlikelynotthecauseofthelarge variationsinourobservedspectrum.However,wenotethatPHOEBE(aswellasother LDcodes)hasnotbeencalibratedinandoutofnarrowspectrallines.Wealsonotethat themodelsshowthatthebottomofthelightcurveisinfactnotatduetoLD.However, basedonourcalculationofthemeanmodeluxratiooverthelimb-darkenedtransit bottomforeachbandpass,thisshouldnotaffectthemagnitudeofthevariationswe measureinourobservedspectrum.DuetoLDeffects,theoverallnormalizationofthe spectrummaybeaffected. 3.2.3TransitColour InFigure 3-12 wepresentthecolourofthenormalizedin-transituxratios, computedbydividingeachpointintheoff-linebandpassesbytheaverageofeach pairofon-linepointsaroundthoseoff-linepoints.Wendthatthecolourbetweenthe bluestbandpassandtheon-linebandpassisconsistentwithzero,withameanvalueof 6.30 6.04 # 10 5 (computedfollowingthemethoddescribedinSection 3.2 ).Themean colourofthe773.66nmandon-linebandpassesis $ 3.57 0.63 # 10 4 ,andthemean colourbetweenthereddestandon-linebandpassesis $ 8.99 0.62 # 10 4 Wealsopresentthestandarddeviationofeachcolourforanumberofbinning factorsinFigure 3-13 .Wendthatthetrendforeachcolourisconsistentwithhaving onlywhitenoiseineachofourcolours.Thisisalsoconrmedbyttingthewhiteand rednoiseexplicitlyforeachcolour.Consideringthattherednoiseisestimatedtobeless than 1 # 10 8 foreachcolour,whitenoiseclearlydominatestheuncertaintiesinthe transitcolour. 60

PAGE 61

AsexploredinSection 3.2.1 ,wealsocomputemeancoloursafterexcluding outlyingabsoluteuxmeasurementsfromouranalysis.Afterexcludingthosedata points,weestimatethemeancoloursbetweeneachoff-lineandtheon-linebandpasses tobe 1.79 6.60 # 10 5 $ 3.54 0.62 # 10 4 ,and $ 6.92 0.54 # 10 4 (frombluestto reddest).BoththesemeancoloursandthosediscussedaboveareplottedinFigure 3-14 .Thecoloursarecomparablebetweenthetwodatasets,withthecolourof thereddestbandpasshavingtheonlymeasurabledifferencebetweenthetwosets. Furthermore,Figure 3-14 illustratesthatnotonlyisthereasignicantchangeinthe colourduringtransit,butalsothatthemagnitudeofthechangeisequivalenttoalarge changeintheapparentplanetradius.Atthereddestwavelengths,weclearlymeasure achangeofover3percent(andasmuchas4.2percent,basedontheuxratiosthat donotexcludeoutlyingabsoluteuxmeasurements)intheapparentradiusoftheplanet comparedtotheplanet'sapparentradiusintheon-linebandpass. Overall,thesecoloursagreewiththemagnitudeanddirectionofthedifferences measuredbetweentheweightedmeanin-transituxratiosforthedifferentbandpasses (Section 3.2 ).Furthermore,thedifferencesbetweenthecolourofthebluesttoon-line bandpassesandthereddesttoon-linebandpasseshasgreaterthan 5 signicance. Sinceourcolourmeasurementsmatchthemagnitudeanddirectionofthedifferences intheuxratiosasmeasuredfromourspectrum,weconcludethatourmeasured spectrumofHD80606b'satmosphereisrealandthatthedifferencesintheuxratioare signicant. 3.3Discussion 3.3.1InterpretationofLight-CurveShape First,wecompareourlightcurve(integratedoverallbandpasses)tosimultaneous observationsfrom Spitzer ( H ebrardetal. 2010 )andotherground-basedobservatories ( Shporeretal. 2010 ).Inparticular, H ebrardetal. ( 2010 )identiedabumpinthe in-transitlightcurvethatoccurredwithinthehour before theirestimatedtimeof 61

PAGE 62

mid-transitandponderedwhetheritcouldbeduetoanexomoonorspotcrossing. Undertheexomoonhypothesis,themagnitudeofthebumpshouldbewavelength-independent. Ifthebumpwereduetoaspot,thenonewouldexpectanevengreaterfeatureinthe optical.Wedonotndanyevidenceforacoincidentbump[regardlessofwhether weadopttheephemerisof H ebrardetal. ( 2010 )or Shporeretal. ( 2010 )].Thus, thebumpisunlikelytobeduetoeitheranexomoonorstar-spot.Ifanything,we ndpossibleevidenceofabumpoccurring after mid-transit,butthisfeaturewasnot detectedby H ebrardetal. ( 2010 ).Ifweassumeourcandidatebumpisnotaresult ofinstrumentalsystematics,andwecompareourcandidatebumpasobservedin thedifferentwavebands,wendthatthesizeoftheputativebumpissmallestinthe bluestbandpass,providingfurtherevidenceagainstastar-spot.Furthermore,since themagnitudeofthebumpvariesslightlyforeachbandpass,thisprovidesadditional evidenceagainsttheexistenceofanexomoon.Futurehigh-precision,multi-wavelength observationscouldhelpprovideadditionalconstraintsonthelight-curveshape. 3.3.2ComparisontoPreviousObservations Next,wenotethatourmeasuredin-transituxratiosdifferslightlyfromtheuxratio givenby H ebrardetal. ( 2010 ).Thisisatleastpartlyduetothedifferentbandpasses used.Therecouldalsobeasystematicuncertaintyintheoverallnormalizationof ourtransitdepths.Ifourgoalhadbeentomeasurethetransitdepthprecisely,we wouldhaverequiredobservationstakenjustbeforeandafterthetransitevent.Inthis case,ground-basedobservationsspanningthefulltransitwerenotfeasibleduetothe extremelylongtransitduration.Thus,wenormalizedourin-transitlightcurvesbyOOT observationstakenonadifferentnight.Whileourobservationsresultedinaveryhigh precisionfordifferentialmeasurementsofthetransitdepthineachbandpass,achange intheobservingconditionsbetweennightscouldresultinthetransitdepthsallbeing affectedbyacommonscalingfactor. 62

PAGE 63

Toconrmthatthechangeintheapparentplanetaryradiuswithwavelengthis basedonarobustestimateoftheOOTuxratiodespiteusingbaselineobservations separatedbyfourmonthsfromthetransitobservations,weestimatedtheweighted meanin-transituxratiosasbefore,butnormalizedthemagainstthelowerqualityOOT datatakenon2010January15.Forreference,weincludetheresultsoftheaperture photometryforthisdatasetinTable 3-8 andtheuxratiosbeforeandafterEPDinTable 3-9 .WefoundthatdespitethelargescatterinthatOOTdata,thenormalizedin-transit uxratio(andtherefore,theapparentradiusoftheplanet)stillchangessignicantly withwavelengthandmaintainsthesameshapeasshowninFigure 3-11 .Weconclude thatthelargechangeintransitdepthfromthe768.76-and769.91-nmbandpassesto the773.66-and777.36-nmbandpassesisarobustresult.Wealsoemphasizethat theatmospherewasmuchmorestableduringtheAprilobservationsthantheJanuary baselineobservations,sowestillrelyontheAprilbaselineobservationsforourprimary analysis. Wealsotestedwhetherourresultsweresensitivetotheapertureradiususedfor photometry.Wetriedavarietyofannulifortheaperturesforboththein-transitandOOT datasets.Inallcases,weseethetrendofincreasinguxratiowithwavelengthand foundverysimilarresultstothosepresentedhere.Theonlydifferenceoccurredforthe largestapertures,inwhichcasetheweightedmeanin-transituxratioontheK I feature isslightlysmallerthantheuxratioatthebluestwavelength.Evenforthischoiceof apertures,theuxesinthe773.66-and777.36-nmbandpassesarenotsignicantly different(thoughtheerrorbarsareslightlylarger). 3.3.3LackofaK I LineCore AsillustratedinFigure 3-11 ,thereisnosignicantdifferencebetweenthe observationsacquiredinthecoreoftheK I lineandslightlytotheblue.Giventhe 1.2-nmFWHM,aDopplershiftof 200kms 1 wouldbeneededtoshiftthelinecore outoftheon-linebandpass.ThisisgreaterthantheescapespeedfromHD80606b 63

PAGE 64

( 121kms 1 ).Thus,weplacea3 limitonthestrengthoftheK I linecoreof 3 # 10 4 (forour1.2-nmFWHMbandpass). Byitself,thelackofalinecoreismostnaturallyexplainedbyalackofK I atthe altitudesprobedbytransmissionspectrophotometry.Thiscouldoccurif(1)thereisa signicantbulkunderabundanceofpotassium,(2)thepotassiumhascondensedinto cloudsand/ormolecules,(3)thereisacloudorhazelayerabovetheregioncapable ofcausingsignicantpotassiumabsorption,and/or(4)thepotassiumhasbeen photoionized( Fortneyetal. 2003 ).InthepreviouscaseofHD209568b,theoretical investigationsoftheunexpectedlyweakNa I absorptionshowedthattheobserved featuredepthisparticularlysensitivetotheextentofcloudformation( Fortneyetal. 2003 ).InthecaseofHD80606b,thehighlyeccentricorbitresultsinashheatingnear pericenterandextremetemperaturevariationsovertheorbitalperiod.Atthetimeof transit,thestar-planetseparationis 0.3AU,sotheequilibriumtemperatureis 500 K. Basedon Spitzer observations,coolingissufcientlyrapidthattheplanetisexpectedto havecooledbetweenpericenterandtransit( Laughlinetal. 2009 ).Thus,bothsodium andpotassiumarepredictedtohavecondensedintoclouds.Thus,weconcludethatthe lackofaK I corecouldeasilybeduetopotassiumhavingcondensedintocloudsbefore thetimeoftransit. 3.3.4PlanetaryAtmosphereModels Inanattempttomodelourobservations,weconsideredbothaconventional1D "cold"atmospheremodel( Fortneyetal. 2010 )(solidlineinFigure 3-11 )andasimilar model,butwitharbitraryadditionalheatingtoraisetheeffectivetemperatureby500 K(dottedlineinFigure 3-11 ).Bothmodelshavebeennormalizedtothestellarradius estimatedby H ebrardetal. ( 2010 ),andassumeastar-planetseparationof0.3AU (i.e.thedistancebetweenthestarandplanetwhentheplanettransits).Chemical equilibriumandastandardpressure-temperatureproleforHD80606bareassumed. Inthe"cold"atmospheremodel,theplanet's(apparent)radiusat10barwasadjustedto 64

PAGE 65

matchtheradiusmeasuredby H ebrardetal. ( 2010 )at4.5 m.Inthe"hot"atmosphere model,thetemperaturesintheupperatmosphererangefrom 300to500K,even withtheadditionalheating.Thehighertemperatureincreasestheobservedplanetary radiusatallwavelengths,andslightlyincreasesthepeaktotroughdistanceofthe features,buttheplanet'sradiuswasnotadjustedtomatchtheradiusfrom H ebrardet al. ( 2010 ).Atthesetemperatures,mostofthepotassiumisexpectedtohaveformed condensates,signicantlyreducingtheK I absorptionfeature.AstheinsetinFigure 3-11 illustrates,neitherthe"hot"atmospheremodelnorthe"cold"atmospheremodel predictsasignicantfeatureduetoK I absorption. 3.3.5ChangeinApparentRadiuswithWavelength WhilewedonotdetecttheK I core,wendrelativelylargedifferences( 3.57 0.63 # 10 4 and 8.99 0.62 # 10 4 )betweenthecoloursoftheon-linebandpass andthebandpassestothered(773.66and777.36nm).Cloudsandhazeswould suppressboththecoreandwingsoftheabsorptionfeature.Asimilarobservationfora typicalhotJupitercouldbereadilyinterpretedasstrongabsorptioninthewingsofthe potassiumlineduetoabsorptionbypressurebroadenedpotassiumatloweraltitudes, whilepotassiumathigheraltitudeshasbeenphotoionized( Fortneyetal. 2003 ). However,inourobservations,themagnitudeofthedifferenceinabsorptionat thetwoblueandtworedwavelengthsappearstoolargeforsuchamodel.Onecould expectsuchobservationstoprobetheloweratmosphereover 10scaleheights( H ), fromapressureof 100mbarto 1microbar.Assumingtheplanethasreacheda thermalequilibriumforthestar-planetdistanceatthetimeoftransitanda500Kupper atmospheretemperature,thescaleheightwouldbe H 20km.Thus,onemightexpect toseechangesintheapparentradiusoftheplanetontheorderof 200 km.Our observationssuggestamuchlargerchangeintheapparentradii(upto 4.2 percentor 2900 km)whencomparingobservationsintheK I linecoreandthereddestbandpass. Thescenariodescribedabovewouldsuggestthattheseobservationsprobed 145 65

PAGE 66

scaleheightsintheatmosphereofHD80606b,orpressuresoflessthan 10 55 bar, whichiswellintotheexosphere.Suchalargenumberofscaleheightsisnotrealistic, implyingthattheabsorptionisoriginatingfromapartoftheatmospheremuchhotter than500K.Fortunately,thetemperatureisexpectedtoriserapidlytothousandsof Kelvinaboveoneplanetaryradius( Yelle 2004 ). 3.3.6AbsorptionbyanExosphere BasedonthemodelofSection 3.3.4 ,wewouldestimatethatourobservationshave probed 145scaleheightsintheatmosphereofHD80606b,orapressureoflessthan 10 55 bar.However,theseestimatesassumeanatmospherictemperatureof500K. Yelle ( 2004 )ndsasteepriseinthetemperaturefrom 350to10000Kfrom1 R p to1.1 R p foraplanetat0.1AUfromtheSun.Ifweusetheirmodelasaroughguideline,andif weassumeatemperatureof2000Kbetween1and 1.04 R p forHD80606b,the2900 kmmeasuredchangeintheapparentradiuswouldimplythattheobservationsprobed 36scaleheights,ortoapressureoflessthan 10 14 bar.Regardlessofwhetherwe assumeatemperatureof500or2000K,theimpliedpressuresareindicativeofthose thatwouldexistinanexosphere. ThemodelsandopacitydatabaseofSection 3.3.4 arenotcompleteforthe temperatureandpressuresoftheexosphere.Theopacitydatabaseusedextendsto temperaturesof 2600Kand 1microbarandisnotintendedtodescribeopacity sourcesinanexosphereorwind(e.g. Vidal-Madjaretal. 2003 2008 ; Ballesteretal. 2007 ; Ehrenreichetal. 2008 ; LecavelierDesEtangsetal. 2008 2010 ; Ben-Jaffel& SonaHosseini 2010 ).Tothebestofourknowledge,anexosphericmodelthatpredicts thelocationandstrengthofabsorptionfeaturesarisingfromtheexospheredoesnot exist.Wehopethatourobservationswillstimulatetheoreticalmodelsfortheobservable effectsofexoplanetexospheresontransmissionspectroscopyandspectrophotometry. Giventheplanet'shighsurfacegravityandanyreasonablechoiceofplanetary parameters,a 4.2 percentchangeintheplanet'sapparentradiusrequiresavery 66

PAGE 67

dramaticchangeinthepressureatwhichtheslantopticaldepthreachesunity,between 770and777nm.Thus,weconcludethatabsorptionathighaltitudeandtemperatureis themostlikelyexplanationforthelargechangeintheapparentplanetradius. 3.3.7PossibilityofOtherAbsorbers Next,weconsiderwhetheranotherabsorbermightberesponsiblefortheobserved changeinapparentplanetradius.Methanecanbeactiveinthisregionofthespectrum. However,methanewouldbeunstableatthehightemperaturesofanexosphereor wind.BothofthemodelsinSection 3.3.4 includemethaneatalltemperaturesatwhich itwouldbestable,around < 1000K.Inthewavelengthregimethatweobserved,the opacityofmethaneislargestat778nmandsmallestat769nm,soitspresencewould producetheoppositetrendfromwhatisshowninthedata. Theobservedwavelengthswerealsochosentoavoidwatervapor(whichisalso unstableathightemperatures).Wearenotawareofanyotherabsorberwhichcould explainthelargechangeinapparentplanetradius,andconsiderK I themostlikely absorber.Nevertheless,wecannotruleoutthepossibilitythatHD80606b'sexosphere possessesanabsorberthatissomethingotherthanK I onaccountoftheincomplete opacitydatabase. 3.3.8AbsorptionbyaWind Ifthe 4.2 percentchangeintheapparentradiusisduetoabsorptionbyK I at highaltitude,thenitisnotobviouswhytheobservationsontheK I core(769.91nm) arenotsignicantlydifferentfromtheobservationsslightlytotheblue(768.76nm). Onepossibilityisthatthelinecorewasshiftedoutoftheon-linebandpass.Giventhe 1.2 -nmFWHM,thiswouldrequireaDopplershiftof 200 kms 1 .Ablueshiftof 225 kmwouldplacethecorehalfwaybetweenthe768.76-and769.91-nmbandpasses.A somewhatsmallerDopplershiftplusDopplerbroadeningmightalsoreducethesignal strength.Inanycase,thevelocitiesrequiredwouldbegreaterthantheescapespeed fromHD80606b( 121kms 1 ). 67

PAGE 68

Whileavelocityexceedingtheescapespeedissomewhatconcerning,itisnotout ofthequestionforawindbeingdrivenfromtheexosphere.Infact,similarobservations ofotherplanetsalsoappeartondanunexpectedlylargeDopplershift.Specically, alargeblueshifthasbeenfoundinallcases;e.g. Redeldetal. ( 2008 )foundan unexpectedblueshiftofthecoreoftheNa I absorptionforHD189733b. Snellenet al. ( 2010 )detecteda2kms 1 blueshiftintheupperatmosphereofHD209458bwith observationsofCO.Additionally, Holmstr ometal. ( 2008 )reportedLyman# absorption aroundHD209458batwavelengthoffsetscorrespondingtovelocitiesofseveral100km s 1 ,buttherewasnoinformationabouttheLyman# coreasitisnotobservabledue toEarth'sgeo-corona.MuchlikeourobservationsofHD80606b,thereisconsiderable uncertaintyregardingtheoriginoftheabsorptionandDopplershiftforHD209458b ( LecavelierDesEtangsetal. 2008 2010 ; Ben-Jaffel&SonaHosseini 2010 ).Proposed mechanismsincluderadiationpressureandinteractionwithastellarwind(e.g. Tianet al. 2005 ; Garc aMu noz 2007 ; Murray-Clayetal. 2009 ; Ekenb acketal. 2010 ),andin particularwenotethatmodelsofHD209458b'satmospherematchobservationsbetter ifitisassumedthatthelinecoreisobscured.Ourobservationscouldbeexplainedifa similarmechanismoperatesonHD80606bandheavyelements(i.e.potassium)are mixedintothewind. InthecaseofHD80606b,thedynamicsoftheexosphereandanyplanetary windisalmostcertainlyquitecomplex.Theplanethasthelargestsemimajoraxisof anyconrmedtransitingplanet( 0.455 AU),butitfollowssuchahighlyeccentricorbit ( e =0.93 )thatthestar-planetseparationofHD80606batperiastronis 2/3thatofHD 209458b.Thus,HD80606bexperiencesstrongandrapidheatingoftheatmosphere nearpericenter.Thelargeandrapidchangesintheincidentstellaruxandtemperature aswellasthestellarwinduxcouldleadtoepisodicmass-lossfollowingeachpericenter passage( Laughlinetal. 2009 ).BasedontheobservedX-rayux( Kashyapetal. 2008 ) andmass-losscorrelation( Woodetal. 2005 ),HD80606couldhaveamass-lossrate 68

PAGE 69

asmuchas 100timesstrongerthanHD209458,providingamuchstrongerstellar windtodriveawindfromHD80606b.TherapidcontractionandexpansionoftheRoche lobearoundeachpericentercouldfurthercomplicatethedynamicsoftheexosphere andplanetarywind. 3.3.9PotentialSystematics 3.3.9.1ExcludingTelluricAbsorption Theusualsuspectinground-basedobservationsisvariabilityinthetelluric absorption.Atourobservedwavelengthsthereisverylittleabsorption.Theonly twospeciesthatcontributeanyappreciableabsorptionarewaterandoxygen.In particular,thereisalackofabsorptionfromcarbondioxideormethaneinourobserved bandpasses.Oxygenisgenerallywellmixedintheatmosphere.Thus,weexpectany variabilityduetooxygenhasbeenremovedinourdatareductionprocedure,which normalizeseachobservationofHD80606bytheuxofHD80607takenatthesame timeandusingthesamebandpass.Thus,weruleoxygenabsorptionoutasapotential systematic. Sincewatercanbeveryanisotropicallydistributedintheatmosphere,onecould worrythatthe20-arcsecseparationbetweenHD80606andHD80607mightallowfor variationsinthewaterabsorptionthatarenotremovedbycalibration.However,thetwo bandpassestotheredofK I werespecicallychosentobeatwavelengthsthatavoid waterabsorption.Thus,eveninthescenariothattheon-lineandbluebandpasseswere contaminatedbywaterabsorption,westillmeasurea 2.7 percentchangeinthe apparentplanetradiusbetweenthetworeddestbandpasses(bothofwhichshouldbe substantiallyfreeoftelluricabsorption).Fromthis,weconcludethatourprimaryresult ofmeasuringalargechangeintheapparentradiuswithwavelengthisnottheresultof variabletelluricwaterabsorption. However,inanefforttocondentlyrulevariabletelluricabsorptionoutasapotential sourceforsystematics,weconstructanalternativemodelforthespectrumbasedon 69

PAGE 70

changingthelevelofwatervaporabsorption.Specically,weintegratedourbandpasses overhigh-resolutionmodeltransmissionspectrafortelluricwatervaporandoxygen. UsingtheTERRASPECcode(Benderetal.,inpreparation),wecomputedmodel transmissionspectrafortwodifferentairmasses(representingthemeanairmassover thetransitbottomasobservedinJanuaryandthemeanairmassduringthebaseline dataobservedinApril)andthreedifferentwatervaporlevels(1,5and10mm).Wethen integratedourbandpassesovereachspectraandcomputedtherelativetransmission forthedifferentbandpassesforeverypossiblecombinationofwatervaportowards HD80606andHD80607.Weintegratedovertheappropriatebandpassesforeach setofobservations,asthebandpassesusedinAprilwerecenteredatslightlydifferent wavelengthsthanfortheJanuaryobservations.Ourgoalwastodetermineifthe transmissionspectrumwouldhaveasimilarsignatureasourobservedspectrumifthere wasadifferenceinthewatervaportowardsHD80606andHD80607duringeither orbothoftheJanuaryandAprilobservations.Forexample,wetooktheintegrated transmissionforawatervaporlevelof10mm(towardsHD80606)dividedbythe integratedtransmissionforawatervaporlevelof1mm(towardsHD80607)basedon themeanairmassinJanuary.Then,wedividedthatresultbyasimilarratiobasedon thespectrafortheAprilobservations.Wecomputedthisratioforallcombinationsof watervaporandcomparedtheresults.Realistically,thewatervaporwasmostlikely below6mmforboththeJanuaryandAprilobservations[basedon Garc a-Lorenzoetal. ( 2010 )],butweapproachthisissuewithmuchcautionandthereforediscusstheresults forthe10-mmwatervaporlevelaswell. Fromourresults,wecanmakeseveralargumentsagainstvariablewatervapor absorptionand/orthedifferentwavelengthsobservedinJanuaryandAprilbeing thecauseofourspectrum'ssignature.First,sinceourmeasurementsaremultiply differential(comparingthetargettothereferencein-transittothetargettothereference OOT),weminimizeanysucheffectsfromourmeasurements.Secondly,evenifthe 70

PAGE 71

watervaporcolumntowardsHD80606andHD80607differedbyanaverageof10mm ononeofthenights,itwouldresultinadifferenceofonly0.0058percent(iftheJanuary watervapordifferedby10mm)or0.034percent(iftheAprilwatervapordifferedby 10mm)betweenthereddestandon-linebandpasses.Thisdifferenceintransmission basedonthewavelengthsobservedinAprilislessthanhalfoftheactualmeasured differencebetweentheuxratiosinthesetwobandpasses.Further,theseparation betweenHD80606andHD80607isonly20arcsec,soitisextremelyunlikelythatthe time-averagedwatercolumntowardsthetwostarswoulddifferby10mm.Further,an untenablylargewatercolumn,inconsistentwiththeobservationalconditions,would berequiredtoexplaintheobserveddifferenceof 8 # 10 4 betweentheon-lineand reddestbandpasses.Thirdly,eveniftheaveragewatercolumntowardsthetwostars diddifferbythatmuchononeofthenights,theresultingcoloursdifferfromwhatwe observe,i.e.thehypothesisthatourmeasurementsareprimarilyduetoatmospheric variabilitywouldpredictthetwobluestbandpassestobecomparableinsomecasesand differlargelyinothers,whilethetworeddestbandpassesarecomparableinallcases. Whilewedoobservetheuxratiosinthetwobluestbandpassestobecomparable, weseeasignicantdifferencebetweentheuxratiosforthetworeddestbandpasses. Further,themagnitudeofthedifferencesbetweenthebluestandreddestbandpasses isobservedtobemuchlargerthanwhatthedifferencewouldbeiftheywerecausedby variableatmosphericabsorption. Thus,weestimatethatforthefourdifferentbandpasses,theeffectofvariable atmosphericabsorptionwouldbelessthan(0.0024,0.0015,0.0075,0.0073percent) # [ < mmofH 2 OtowardsHD80606 > < mmofH 2 OtowardsHD80607 > ]/[10mmof H 2 O]atthewavelengthsandairmassobservedatinJanuary,orlessthan(0.004,0.027, 0.0041,0.0068percent)multipliedbythesameratiogivenaboveatthewavelengths andairmassobservedatinApril.However,assumingthatthedifferenceintransmission isnegliblebetweenthetargetandreferenceforboththeJanuaryandAprilobservations, 71

PAGE 72

thenthefactthatthestarswereobservedatdifferentwavelengthsandairmasseson thosenightsshouldbeirrelevant. Finally,wenotethatspectrophotometryusinganarrow-bandTFismuchlessprone tosystematicsthanspectroscopicobservations.Thelackofaslit,thesimultaneous useofaverygoodreferencestar,rapidswitchingbetweenbandpassesandthe multiplydifferentialnatureofourmeasurementshouldallminimizetheeffectsof telluricabsorption.WhiletheOHlinesarevariable,theskysubtractioninourdata reductionprocessremovestheemissiontoahighdegreeofprecision.Finally,wesee noevidence,inouratmospherictransmissionmodels,ofabsorbersthatcouldaccount forthesignaldetected. Torstorder,theeffectsofatmosphericextinctionarecorrectedbymeasuringux ratiosrelativetoHD80607.Weexpectnegligiblesecond-orderdifferentialextinction, sincethetargetandreferencestarsareofthesamespectraltypeandseparatedbyonly 20arcsec.Sincethemagnitudeofthiseffectscalesasthesquareofthelterbandpass, ouruseofsuchanarrowbandpassfurtherminimizessecond-orderdifferential extinction,allowingthistechniquetobeappliedtoothertargetswithreferencestars thatdifferintemperature. Wedonotconsiderdifferentialextinctiontobeaviableexplanationfortheeffect seeninFigure 3-11 .Nevertheless,weperformedanadditionalcheck,inwhichwedo notperformrelativephotometrybetweenthetargetandreference.Wecomparethe ratiooftheabsoluteuxofthereferencestarinthereddestbandpassandthebandpass centeredonK I asmeasuredonthenightofthetransittothesameratioasmeasured onthenighttheOOTobservationsweretaken(2010April4).Weestimatearatioof 0.98498 0.00093,equivalenttoacolourdeviationof 1.5percentbetweenthetwo nights.Thisprovidesanupperboundontheeffectsofatmosphericvariability,including differentialextinction.Theaccuracyofourprimaryanalysisshouldbeconsiderably higherthankstotheuseofrelativephotometrytocorrectforatmosphericvariability. 72

PAGE 73

3.3.9.2ExcludingInstrumentalEffects WithTFimaging,thephotonsforeachobservedwavelengthlandonthesame pixels,eliminatingconcernsaboutspatialvariationsintheat-elding.However,the normalizationoftheuxmeasurementsisaffectedbythewavelengthdependenceof thepixelsensitivity.Tominimizethiseffect,wetookdomeat-eldsforeachobserved wavelengthandcorrectedthescienceframestakenateachwavelengthwiththeir respectiveat-elds. Furthermore,weguardagainstpossiblenon-uniformityintheshuttermotion,which couldresultinthesystematiceffectofproducingslightlydifferentexposuretimesforthe targetandreferencestar,dependingonwheretheyarelocatedontheCCDchip.This systematiceffectismorenoticeableforshorterexposuretimes,soweguardagainst itbyfollowinganobservingsequencethatrepeatsaftersevenexposures,sothatthe subsequentsetofexposuresoccurswiththeshuttermotionintheoppositedirection. DependingontheorientationoftheTF,theobservedwavelengthcandriftdue totherotationoftheinstrumentduringtheobservations.Forourobservations,the TFwastunedbeforeobservations,inthemiddleofthetransitandattheendofthe observations.Nodriftslargerthan0.1nmoccurred. Finally,weconductedathoroughinvestigationintothepossibilityofsaturation and/ornon-linearityasasourceofsystematiceffects.Amajorityofthepeakcounts duringourobservationswerewellbelowthesaturationthreshold( 65000ADUs), andforstandardobservingmodeslinearityisguaranteedupto 65000ADUs,so non-linearityshouldnotbeanissue.However,asweuseanon-standardobserving modeonOSIRISinordertoreadouttheCCDsatthehighestratepossible(andthereby greatlyreducedeadtime),itisworthwhiletoinvestigatewhethernon-linearityisan issue.Therefore,wediscusshereseveralchecksfornon-linearity,wherewearbitrarily assumethat45000counts( 30821ADUs,basedonthegainof1.46 e perADU)isthe levelatwhichnon-linearitymightbegin. 73

PAGE 74

First,wecheckediftheaveragenumberofcountsfromtheat-eldstakenforeach bandpasshadalineardependencewithexposuretime,aswehadtakenat-eldsat severaldifferentexposuretimes.Wetalinetoallmeasurementsofthemeanat-eld counts(forvedifferentexposuretimes),andthenwetanotherlinetothedatabut excludedmeasurementsthatwerenearorabove45000counts.Toseeifincluding measurementsathighercountsresultedinnon-linearity,wecomparedtheslopes and y -interceptsofthetwobest-ttinglines.Aftercomparingthebest-ttingsolutions betweenthedifferentbandpassesandforthedifferentseriesofatstakeninJanuary andApril,wendthatitisnotobviousthatanyonesetofatsdisplayssignicant non-linearitycomparedtotheothers. Secondly,weinvestigatedttingaquadraticfunctiontotheat-eldcountsforboth theJanuaryandAprilat-elds.Aftercomparingthebest-ttingcoefcientsforthe differentbandpasses,wefoundthatatleastonesetofcoefcientsdeviatedsignicantly fromthecoefcientsfortheotherbandpasses.Whiletheremightbeanobviousoutlier intermsofonebandpassthatmightbeaffectedbynon-linearityforeachsetofats,the supposedoutlierisdifferentfortheJanuaryandAprilats.Weagainconcludethatitis notobviouswhich,ifany,ofourbandpassesisdisplayingsignicantnon-linearity. Thirdly,weinvestigatedthepossibilityofnon-linearitybycomputingthecolour (betweeneachoff-linebandpassandtheon-linebandpass)andseeinghowitvaried withrespecttotheaverageon-lineuxperpixel(estimatedbydividingthetotalabsolute on-linetargetuxbythetarget'sFWHMsquared).Wecomputedthemediancolour forexposureswheretheaverageuxperpixelwasbelow45000andforexposures wheretheaveragecountswereabove45000.Wethenestimatedthedifferenceinthe mediancolourforthosetwosetsofexposures.Wefoundthatinthenear-redbandpass (773.66nm),non-linearitymostlikelydoesnotplayarole,asthemediancoloursbelow andabovethe45000countleveldifferbyaninsignicantamount.However,inthe bluestandreddestcolours,wedoseeaslightcorrelation,withthemediancolours 74

PAGE 75

differingbycomparableamounts.Thisisnotwhatwemightexpecttoseeifnon-linearity werecausingsystematiceffectsinourobservedspectrum,sincewedonotobserve acomparablecolourdifferenceinthein-transituxratiobetweentheblue-on-lineand reddest-on-linecolours. WealsocomputedthecolourdeviationsfortheAprilbaselinedata,andwefound thatthecoloursbelowandabovethe45000countmarkareslightlylargerthanthe in-transitcolourdeviations(butthesewerecomputedusingacombineddatasetfor twodifferentexposuretimes,whichcouldaffecttheseestimates).Regardless,westill ndthatthesmallestdifferenceinthecoloursisinthenear-redbandpass,andthe differencesarecomparableforthebluestandreddestbandpasses,eventhoughinboth theobservedin-transitandOOTuxratiosweseethesmallestdifferenceintheux ratiobetweenthebluestbandpassesandthelargestbetweentheon-lineandreddest bandpass. Insummary,weconductedseveralchecksfornon-linearity.Weconcludethatany effectsofnon-linearityareeithertooinsignicanttoaffectourphotometryortheyarenot correlatedwiththedata. 3.3.9.3PossibleNon-PlanetaryAstrophysicalEffects Lastly,weconsiderpotentialastrophysicalsystematicssuchasstellarvariability. Observationsinallfourbandpasseswereobtainedduringthe same transit.Ifobservations usingdifferentbandpasseshadbeenmadeduringdifferenttransits,thentheinterpretation wouldbeambiguous,asstellarvariability(e.g.spotsthattheplanetdoesnotnecessarily passover)couldresultinapparentchangesinthein-transituxratio.Forthelarge changeinapparentradiustobeduetostellarvariability,therewouldneedtobea 4.2 percentchangeinthecolourofeitherthetargetorreferencestar.Suchlargevariability overasmallrangeofwavelengthsisaprioriunlikelyforsolar-likestars( H ebrardet al. 2010 ).However,assuggestedbythereferee,wehaveestimatedhowspottedHD 75

PAGE 76

80606wouldhavetobetocauseadifferenceof 8 # 10 4 intheuxratiosintheon-line andreddestbandpass. WecomputedtheblackbodyuxforHD80606( T e 5572K),thenintegratedthe uxovereachbandpasstoestimatethetotaluxobservedineachbandpass.Wethen completedsimilarcalculationsforaspotassumingatemperature1000Kcoolerthan HD80606andaspotradiusequaltotheplanet'sradius.Aftercomputingtheratioofthe integratedspotuxtotheintegratedstaruxforsome N spots,wefoundthatabout26 spotswiththeabovepropertieswouldhavetoexistonthesurfaceofHD80606during thetransitobservationsinordertoproduceadifferenceintheon-lineandreddestux ratiosofabout8 # 10 4 .Thatisequivalenttohaving 26percentofHD80606'ssurface coveredwithspots.Evenifthesystematictrendsweseeinourtransitlightcurvesare duetospotscominginandoutofviewonthesurfaceofthestar,thepercentofthe stellarsurfacecoveredbyspotsisunlikelytobeasmuchas26percent.Furthermore, ifthestarwasthisspotted,weshouldalsoseeadifferenceintheuxratiobetween thetwobluestbandpassesofover1 # 10 4 ,yetthedifferenceweobserveislessthan 6 # 10 5 Weconcludethatitispossibleforspotstoaccountforsomeofthevariations wemeasure,butthatHD80606is very unlikelytobespottedenoughtocausethe magnitude ofvariationswemeasure.Infact, Wrightetal. ( 2004 )measuredvalues of S HK =0.149 and logR $ HK = $ 5.09 forHD80606,whichindicatethatthestaris quiteinactive.Also, H ebrardetal. ( 2010 )monitoredHD80606anddetermineditisnot anactivestar.Specically,theyestimatethatthestarisphotometricallystableatthe levelofafewmmagintheopticalrangeonthetime-scaleofseveralweeks.Despite thesestatements,theystillattributethebumpintheirlightcurvetoaspotonthestellar surface.Consideringtheprecisionofourobservations(muchbetterthan1mmag),itis possiblethatweobserveduxvariationsthattheydidnothavetheprecisionto. 76

PAGE 77

Asanalcomment,wenotethatspotscouldaffectthenormalizationoftheoverall spectrum,asthespectrumcouldneedtobescaleddownward(i.e.decreasethe uxratiosorincreasethetransitdepths)toaccountfortheeffectofspots.However, theshapeofthespectrumwouldremainthesame,unlessover 26percentofthe star'ssurfacewascoveredwithspotsduringtransit.Asnotedbythereferee,alarge, long-livedpolarspotcouldexistonHD80606,whichwouldnotinducelargephotometric variationsbutcouldstillaffectourphotometry.Or,bothHD80606andHD80607orHD 80607alonecouldbespottedandcausetheobservedvariations.Duetothepossible variablenatureofHD80606(orHD80607),weencouragefutureOOTobservationsof HD80606andHD80607todetermineifsuchvariabilityiscommon. 3.4Conclusion Insummary,ourobservationsdonotmatchexistingmodels,duetotwobasic observations.Wendalargechangeinapparentplanetradiuswithwavelength,but donotobserveasignicantdifferencewheretheK I linecorewouldbeexpected. Ourobservationsplaceastronglimitonthestrengthofthelinecore(unlessithas beenDoppler-shiftedby 100 kms 1 ),yetimplylargevariationsinradiusover wavelengthsusuallydominatedbyK I absorption.Intheabsenceofotherviable absorbers,absorptionbyK I remainsthemostviableexplanation.Theatmospheric scaleheightofHD80606battransit( 20km)issignicantlysmallerthanthatofHD 209458bandHD189733b,yetthevariationinradiusislargerthanthatofHD209458b ( Singetal. 2008a ).Onepossiblemodelisabsorptionbypotassiumthatispartof ahigh-speedwindcomingofftheexosphere.Whilehigh-speedwindshavebeen observedforotherexoplanets,themechanismforpoweringsuchwindsisunclear. Weencouragefurthertheoreticalinvestigationstoimprovemodelsfortransmission spectroscopyofexoplanetexospheresingeneralandthespecicchallengeofHD 80606b. 77

PAGE 78

Finally,wehaveinvestigatedseveralpotentialsourcesofsystematiceffects. Thereisnosimpleorobvioussourcecausingthesystematicsinourdata.Further,any systematicsintroducedbythesourceswehaveinvestigatedhereproduceneitherthe samesignatureasourobservedspectrumnorthesamemagnitudeofdifferenceas thatofourmeasureduxratios.Whilewearecondentthatnoneofthesepossible sourcesofsystematiceffectscausestheshapeofourobservedspectrum,westill allowthepossibilitythatoneorsomecombinationofthesesystematicsmayaffectour measurementsand/ortheoverallnormalizationoftheobservedspectrum.Wealso acknowledgethatthetargetwasobservedataslightlydifferentsetofwavelengthsin JanuaryascomparedtoApril.Whilethedifferenceinwavelengthsissmall( 0.1nm), thereisstillthepossibilitythatthiscouldresultinsmalldifferencesineitherthetelluric absorptionorstellarspectra,whichinturncouldcausetheobservedspectrumthat wehaveattributedtoabsorptionfromtheatmosphereofHD80606b.Asanalnote, wehighlyencouragefollow-uptransitobservationsofHD80606btoconrmthesignal measuredhere.WenotethatthenextpartialtransitobservablefromLaPalmaoccurs on2012March3,duringwhichobservationspre-transitthroughthecompletersthalfof thetransitwillbepossible. 3.5FutureProspects FuturetransitobservationsatwavelengthsaroundK I inHD80606barepossible, butrequireconsiderablepatienceduetothelongorbitalperiod(111d).Observations ofthetransitdeptharoundotherabsorbingspeciescouldtesttheexosphereandwind models.SimilarobservationsofotherplanetswouldenableacomparisonofK I strength inboththewingandcoreasafunctionofstarandplanetproperties.Wenotethat shortlybeforesubmission,webecameawareofindependent,butsimilar,observations ofanotherexoplanet( Singetal. 2011 ).Boththeseandfutureobservationsofadditional exoplanetswillenablecomparisonsoftheatmosphericcompositionandstructure,as wellasstudiesofpotentialcorrelationswithotherplanetorhoststarproperties.Such 78

PAGE 79

observationswouldalsohelpimprovetheinterpretationoftheexistingHD80606b observations. Currently,onlytheOSIRISredTF(651-934.5nm)isavailableattheGTC.Oncethe blueTFisavailableforscienticobservations,itwillbepossibletoobserveadditional atmosphericfeatures,includingtheNa I featurepreviouslydetectedforHD209458b andHD189733b.ThelargeapertureoftheGTCmakesitpracticaltoperformsimilar observationsofseveralfainterhoststars.Thus,welookforwardtofutureobservations ofalargesampleoftransitingplanets.Thestrikingdiversityofexoplanetssuggeststhat itwillbefruitfultocompareNa I andK I observationstoidentifytrendswithstellarand planetproperties. Despitethecomplexinterpretationoftheseobservations,theveryhighprecision obtainedwiththeOSIRISnarrow-bandTFimageropensupnewavenuesofresearch forlargeground-basedobservatories.Indeed,themeasuredprecisionexceedsthatof Spitzer ( H ebrardetal. 2010 )andeventhe HST observationsforthegivenbandpass ( Pontetal. 2008 ).Thus,ground-basedobservationscannowcharacterizethe atmospheresofgiantplanetsusingspectrophotometry.Thephotometricprecision isalsosufcienttomeasureemittedand/orscatteredlightduringoccultationat multiplenear-infraredwavelengthsthatcouldimproveconstraintsonatmosphere modelsofshort-periodgiantplanets.Byprovidinghigh-precisionphotometryat multiplewavelengthsduringasingletransit,thetechniquecouldalsocontributeto theconrmationoftransitingplanetcandidates,suchasthoseidentiedby Kepler ( Boruckietal. 2010a ).Thetechniquecouldalsoimprovemeasurementsoftheimpact parameterandthusorbitalinclination( Col on&Ford 2009 ).Thiswouldbeparticularly valuableforsystemswithmultipletransitingplanets( Steffenetal. 2010 ),forwhichthe orbitalevolutiondependsontherelativeinclinationoftheorbits( Ragozzine&Holman 2010 ). 79

PAGE 80

SinceallNeptuneandsuper-Earth-sizedplanetswillhaverelativelylowsurface gravities,theycanmakegoodtargetsfortransmissionspectroscopy.Despiteasmaller transitdepththangiantplanets,thepotentiallylargeatmosphericscaleheightcan leadtoasubstantialsignalintransmission( Charbonneauetal. 2009 ),particularlyfor Neptune-sizedplanetsorbitingsub-solar-massstarsand/orsuper-Earth-sizedplanets orbitinglow-massstars.Previously,ithasgenerallybeenassumedthattheEarth's atmospherewillpreventground-basedfacilitiesfromachievingthehighprecisions necessarytomeasurebiomarkersonsuper-Earth-sizedplanetsandthatthe James WebbSpaceTelescope willprovidetherstopportunitytocharacterizeatmospheresof super-Earths( Demingetal. 2009 ).IfthechallengesofEarth'satmospherecouldbe overcome,thenground-basedobservatorieshaveseveraladvantages(e.g.muchlarger collectingarea,moremodernandsophisticatedinstrumentation,abilitytoadjustand upgradeinstruments).Theseobservationsdemonstratethatground-basednarrow-band photometryonlargetelescopescandelivertheprecisionnecessarytocharacterize super-Earth-sizeplanetsaroundbright,nearby,smallstars.Weencourageastronomers toconsiderafuturegenerationofinstrumentsspecicallydesignedforhigh-precision transitobservations,whichmayallowthecharacterizationofsuper-Earth-sizedplanets inupcominglargeground-basedobservatories[e.g.GiantMagellanTelescope(GMT), ThirtyMeterTelescope(TMT)andExtremelyLargeTelescope(ELT)]. 80

PAGE 81

Table3-1. Absolutetransitphotometryfrom2010January13. (nm)HJD F target F ref 768.762455210.442827898032501667 ... 769.912455210.441412079521081769 ... 773.662455210.441917899881603033 ... 777.362455210.442324418762188980 ... Note .Thewavelengthsincludedinthetablearetheobservedwavelengthsintheframe oftheplanet(seetextforadditionaldetails).Thetimestampsincludedhereareforthe timesatmid-exposure. F target and F ref aretheabsoluteuxmeasurementsofHD80606 andHD80607.Thefulltableisincludedonline(SupportingInformation),whileaportion isshownheresothereadercanseetheformattingofthetable. 81

PAGE 82

Table3-2. AbsoluteOOTphotometryfrom2010April4. (nm) t exp (s)HJD F target F ref 768.8682455291.424555505694951634 ... 770.0082455291.424356655545060730 ... 773.7682455291.424961251605466145 ... 777.4582455291.425370066256252784 ... Note .ColumnsaresimilartoTable 3-1 ,exceptthewavelengthsincludedinthetableare thewavelengthsasobservedfromtheGTC(seetextforadditionaldetails).Thesecond columncontainstheexposuretimefortheobservations,asobservationsbasedontwo differentexposuretimeswereincludedinouranalysis.Thefulltableisavailableonline (SupportingInformation),andaportionisshownheresothereadercanseethe formattingofthetable. 82

PAGE 83

Table3-3. Relativetransitphotometry. (nm)HJD F ratio Uncertainty 768.762455210.44281.115180.00086 ... 769.912455210.44141.116650.00124 ... 773.662455210.44191.116630.00104 ... 777.362455210.44231.115530.00091 ... Note .Thewavelengthsincludedinthetablearetheobservedwavelengthsintheframe oftheplanet(seetextforadditionaldetails).Thetimestampsincludedhereareforthe timesatmid-exposure. F ratio representstherelativeuxratiobetweenthetargetand referencestar(i.e. F target / F ref ).Thefulltableisavailableonline(SupportingInformation), andaportionisshownheresothereadercanseetheformattingofthetable. 83

PAGE 84

Table3-4. Normalizedphotometryfromaroundmid-transit. (nm)HJD F ratio F ratio Uncertainty ( raw )( corrected ) 768.762455210.56800.990330.989620.00054 ... 769.912455210.56700.990410.989120.00054 ... 773.662455210.56710.990360.989090.00052 ... 777.362455210.56750.992280.990700.00049 ... Note .Thewavelengthsincludedinthetablearetheobservedwavelengthsintheframe oftheplanet(seetextforadditionaldetails).Thetimestampsincludedhereareforthe timesatmid-exposure.Theuxratiosarepresentedbothbefore(raw)andafter (corrected)EPDwasapplied.Theuxratioshavealsobeennormalizedtotheweighted meanOOTuxratio(Table 3-6 ).Thefulltableisavailableonline(Supporting Information),andaportionisshownheresothereadercanseetheformattingofthe table. 84

PAGE 85

Table3-5. RelativeOOTphotometryfrom2010April4. (nm) t exp (s)HJD F ratio F ratio Uncertainty ( raw )( corrected ) 768.8682455291.42451.120961.120500.00058 ... 770.0082455291.42431.119511.118880.00058 ... 773.7682455291.42491.120561.119730.00056 ... 777.4582455291.42531.120561.119830.00053 ... Note .Thewavelengthsincludedinthetablearethewavelengthsas observedfromtheGTC(seetextforadditionaldetails).Thetimestamps includedhereareforthetimesatmid-exposure.Theuxratiosare presentedbothbefore(raw)andafter(corrected)EPDwasapplied.The fulltableisavailableonline(SupportingInformation),andaportionis shownheresothereadercanseetheformattingofthetable. 85

PAGE 86

Table3-6. Time-averageduxratiosandnoiseestimates. E (nm) P (nm) S (nm) < % F / F > < F / F > w r & In-transit 768.60768.76768.600.99014861.32 # 10 4 5.55 # 10 4 1.34 # 10 4 2.81 769.75769.91769.750.99019712.08 # 10 4 6.21 # 10 4 2.41 # 10 4 7.86 773.50773.66773.500.99049952.19 # 10 4 5.84 # 10 4 2.44 # 10 4 4.90 777.20777.36777.200.99100611.99 # 10 4 6.15 # 10 4 2.47 # 10 4 6.43 Out-of-transit 768.86768.791.12015318.85 # 10 5 5.62 # 10 4 1.24 # 10 9 1.00 770.00769.931.12001315.05 # 10 5 6.05 # 10 4 6.71 # 10 11 1.00 773.76773.691.12028078.46 # 10 5 5.54 # 10 4 4.65 # 10 9 1.00 777.45777.381.11962028.18 # 10 5 5.78 # 10 4 8.29 # 10 5 1.45 Note E istheobservedwavelengthfromtheGTC(i.e.fromtheEarth), P istheobservedwavelengthintheframeofthe planetand S istheobservedwavelengthintheframeofthestar.Valuesfor P arenotgivenfortheOOTobservations,as theplanetwasnottransitingandwasthereforenottechnicallyobserved.Thein-transitratiosrefertotherelativeuxratio betweenthetargetandreferencethathasbeennormalizedtotheweightedmeanOOTuxratios(givenatthebottomof thetable). 86

PAGE 87

Table3-7. Time-averageduxratiosandnoiseestimates(outlyingabsoluteuxesexcluded). E (nm) P (nm) S (nm) < % F / F > < F / F > w r & In-transit 768.60768.76768.600.99010214.83 # 10 5 4.88 # 10 4 3.90 # 10 5 1.00 769.75769.91769.750.99012181.76 # 10 4 6.29 # 10 4 2.03 # 10 4 6.44 773.50773.66773.500.99042922.15 # 10 4 5.82 # 10 4 2.40 # 10 4 4.80 777.20777.36777.200.99075482.42 # 10 4 5.89 # 10 4 2.90 # 10 4 7.88 Out-of-transit 768.86768.791.12025271.01 # 10 4 5.66 # 10 4 7.78 # 10 9 1.00 770.00769.931.12013271.20 # 10 4 5.89 # 10 4 1.13 # 10 4 2.19 773.76773.691.12037228.98 # 10 5 5.77 # 10 4 1.35 # 10 9 1.00 777.45777.381.11968495.99 # 10 5 5.82 # 10 4 3.36 # 10 9 1.00 Note .SameasinTable 3-6 ,buttheuxratioslistedherearethosecomputedafterexcludingoutlyingabsoluteux measurementsfromtheanalysis. 87

PAGE 88

Table3-8. AbsoluteOOTphotometryfrom2010January15. (nm)HJD F target F ref 768.602455211.74891077950962913 ... 769.752455211.748712107101082178 ... 773.502455211.74921020447911377 ... 777.202455211.749513013941161982 ... Note .ColumnsaresimilartoTable 3-1 ,exceptthewavelengthsincludedinthetableare thewavelengthsasobservedfromtheGTC(seetextforadditionaldetails).Thefulltable isavailableonline(SupportingInformation),andaportionisshownheresothereader canseetheformattingofthetable. 88

PAGE 89

Table3-9. RelativeOOTphotometryfrom2010January15. (nm)HJD F ratio F ratio Uncertainty ( raw )( corrected ) 768.602455211.74891.119471.120070.00128 ... 769.752455211.74871.118771.117750.00121 ... 773.502455211.74921.119681.119400.00131 ... 777.202455211.74951.119981.119100.00117 ... Note .ThecolumnsaresimilartoTable 3-5 .Thefulltableisavailableonline(Supporting Information),andaportionisshownheresothereadercanseetheformattingofthe table. 89

PAGE 90

Figure3-1. AbsoluteuxesofHD80606(a)andHD80607(b)asmeasuredon2010 January13-14.Thedifferentlightcurvesrepresenttheuxesasmeasured nearlysimultaneouslyinthedifferentbandpasses,withtheblack,blue, brownandredlightcurvesrepresentingthe769.91,768.76,773.66and 777.36nmdata.Thesedatahavenotbeencorrectedforairmassor decorrelatedinanyway.Notethebreakinthedataaround2hafter mid-transitduetorecalibrationoftheTF.Theverticalsolidlinesindicatethe expectedbeginningandendofthetransit,andtheverticaldottedlines indicatetheendofingressandthebeginningofegress[basedondurations estimatedby H ebrardetal. ( 2010 )andthetransitephemerisfrom Shporeret al. ( 2010 )].Theverticaldashedlinesindicatethe 4hintervalaround mid-transitthatouranalysisfocusedon(seetextforfurtherdetails). 90

PAGE 91

Figure3-2. SimilartoFigure 3-1 ,butfortheOOTdatatakenthenightof2010April4. Notethatthediscontinuityintheuxesaround2455291.48isduetoa changeintheexposuretime(from8to11s). 91

PAGE 92

Figure3-3. Transitlightcurvesasobservednearlysimultaneouslyindifferent bandpasseson2010January13-14.Theon-linelightcurve(769.91-nm)is showninblack,andtheoff-linelightcurves(768.76,773.66and777.36nm) areshowninblue,brownandred.Theuxratioforeachbandpasshas beennormalizedtotheweightedmeanOOTuxratioestimatedfromthe baselinedataacquiredin2010April,butthedatahavenotbeencorrected forairmassordecorrelatedinanyway.Theoff-linelightcurveshavebeen arbitrarilyoffsetby0.006,0.012and0.018,anderrorbarsarenotshownfor clarity.Theverticalsolid,dottedanddashedlinesarethesameasinFigure 3-1 92

PAGE 93

Figure3-4. Relativein-transituxrationormalizedtotherelativeOOTuxratioas measuredon2010April4.Therelativeuxbefore(a)andafter(b)EPDwas appliedisshown.Thedifferentcolorsrepresenttheuxratiosasmeasured inthedifferentbandpasses,withthecolorsthesameasinFigure 3-3 .Note thatEPDwasonlyappliedtothe 4hcenteredaroundmid-transit(i.e.the bottomofthetransitlightcurve).Thedatashownhavenotbeenbinned,but thedifferentlightcurveshavebeenoffsetarbitrarilyforclarity. 93

PAGE 94

Figure3-5. Correlationsbetweenthenormalizedin-transituxratioandthetarget FWHMand x and y centroidcoordinates,before(left-handcolumn)andafter (right-handcolumn)EPDhasbeenapplied.Allfourbandpassesareshown ineachpanel,withthecoloursthesameasinFigure 3-3 .Similarresults wereobtainedwhendecorrelatingthedataagainstthereferenceparameters butarenotshownhere. 94

PAGE 95

Figure3-6. RelativeOOTuxratioasmeasuredon2010April4.Therelativeuxbefore (a)andafter(b)EPDwasappliedisshown.Thedifferentcolorsrepresent theuxratiosasmeasuredinthedifferentbandpasses,withthecolorsthe sameasinFigure 3-3 .Notethesmallbreakinthedataaround2455291.48 wheretheexposuretimewaschanged.Thedatahavenotbeenbinned,but thedifferentlightcurveshavebeenoffsetarbitrarilyforclarity. 95

PAGE 96

Figure3-7. Correctedlightcurvesforobservationsofthebottomofthetransitas observednearlysimultaneouslyindifferentbandpasseson2010January 13-14.Ineachpanel,theblackpointsillustratethemeasurementstakenin theon-line(769.91-nm)bandpass.Wealsoshowmeasurementstakenin eachoftheoff-linebandpasses(768.76,773.66,777.36nm)ineachofthe respectivepanels(a,b,c)forcomparisontotheon-lineuxratios.Thedata shownherehavebeendecorrelated.Thecoloursandnormalizationsarethe sameasinFigure 3-3 ,butnooffsetshavebeenapplied.Here,wehave binnedthedataanderrorbarssimplyforclarity. 96

PAGE 97

Figure3-8. Histogramsofnormalizeduxratiosfromthebottomofthetransitlightcurve asshowninFigure 3-4 (b).Thehistogramsweregeneratedusingabinsize of0.5mmag.Eachpanelcomparestheon-lineuxratioswiththeoff-line uxratios.Ineachpanel,theblack(solid)histogramsrepresentthe769.91 nm(on-line)lightcurve.Theblue(dotted),brown(dashed)andred (dot-dashed)histogramsareforthe768.76,773.66and777.36nmlight curvesandareshowninpanels(a),(b)and(c),respectively.Panel(d) showsthehistogramsforallfourwavelengthsforfurthercomparison. 97

PAGE 98

Figure3-9. Standarddeviationofthetime-binneduxratiomeasurementsfromthe bottomofthetransit[e.g.asshowninFigure 3-4 (b)]asafunctionofthe numberofdatapointsperbin( N ).Panels(a),(b),(c)and(d)showthe standarddeviationsforthebinned769.91,768.76,773.66and777.36nm lightcurves.Theamountofbinningthatcouldbeperformedvariesforeach lightcurvesincethedifferentwavelengthswereobservedadifferentnumber oftimesinagivenobservingsequence(Section 3.1.1 ).Thesolidlinein eachpanelrepresentsthetrendexpectedforpurewhiteGaussiannoise ( N 1 / 2 ),normalizedtotheunbinnedstandarddevationmeasuredinour data.ThedottedlinesrepresentthetrendforGaussiannoisewhen normalizedtothetheoreticalnoiseforourobservations.Thedashedcurves aremodelsttedtothestandarddeviationthatincludebothwhiteandred noise.Theeffectofrednoiseisobviousinallbandpasses. 98

PAGE 99

Figure3-10. Standarddeviationofthetime-binnedOOTuxratiomeasurementsfrom 2010April[e.g.asshowninFigure 3-6 (b)]asafunctionofthenumberof datapointsperbin( N ).Panels(a),(b),(c)and(d)showthestandard deviationsforthebinned770.00,768.86,773.76and777.45nmlight curves.Thesolidlineineachpanelrepresentsthetrendexpectedforpure whiteGaussiannoise( N 1 / 2 ).Thedottedlinesrepresentthetrendfor Gaussiannoisewhennormalizedtothetheoreticalnoiseforour observations.Thedashedcurvesaremodelsttedtothestandard deviationthatincludebothwhiteandrednoise.Comparedtothein-transit observations,rednoisehasaveryminimaleffecthere.Deviationsbelow thecurvearelikelyduetosmallnumberstatistics.Theseresults demonstratethatnarrow-bandground-basedobservationscanprovidevery highprecisiondifferentialphotometry.Foragivenbandpass,thecombined precisionexceedsthatof Spitzer ( H ebrardetal. 2010 )or HST observations( Pontetal. 2008 ).Tothebestofourknowledge,these representthehighestprecisionphotometryfora1.2-nmbandpassfor groundorspaceobservations. 99

PAGE 100

Figure3-11. Normalizedweightedmeanin-transituxratioversusobservedwavelength (intheframeoftheplanet).Theopentrianglesrepresenttheuxratiosas computedforeachlightcurvedescribedinSections 3.1 and 3.2 .Thesolid circlesrepresenttheuxratioscomputedafterexcludingoutlyingabsolute uxvaluesforeachstarfromtheanalysis(Section 3.2.1 ).Notethatthe solidcircleshavebeenoffsetby0.25nmforclarity.Theverticalerrorbars includeafactortoaccountfortheeffectsofrednoiseinboththein-transit andOOTdata.The"errorbars"inthehorizontaldirectionindicatethe FWHMofeachbandpass.Thesolidsquaresrepresentthemeanin-transit uxratiosestimatedfromlimb-darkenedtransitlightcurvemodelsforHD 80606b.Thelinesshowthepredictionsofplanetaryatmospheremodels (Section 3.3.4 ).Theinsetgureshowstheatmospheremodelsonasmall verticalscale.WhileLDornight-to-nightvariability(ofEarth'satmosphere oreitherstar)couldaffecttheoverallnormalization,theobservedchangein theuxratiowithwavelengthisrobust. 100

PAGE 101

Figure3-12. Coloursofthenormalizedin-transituxratios.Thedifferentpanelsshow thecolourascomputedbetweeneachoff-linebandpassandtheon-line bandpass(afterbinningtheon-linedatatothenumberofpointsineachof theoff-linebandpasses).Thedashedlineineachpanelillustrateswhere thecolourequalszero.Thedatahasnotbeenexplicitlyoffset,andthere arenoobvioussystematicsseeninanyofthecolours. 101

PAGE 102

Figure3-13. Standarddeviationofthetime-binnedcolourmeasurementsfromthe bottomofthetransit(asshowninthedifferentpanelsinFigure 3-12 ).The differentpanelsshowthestandarddeviationsforthedifferentcoloursas presentedinthepanelsinFigure 3-12 ,withpanels(a),(b)and(c) respectivelyshowingthestandarddeviationsforthe768.76-769.91nm, 773.66-769.91nmand777.36-769.91nmcolours.Thesolidlineineach panelrepresentsthetrendexpectedforpurewhiteGaussiannoise ( N 1 / 2 ).ThedottedlinesrepresentthetrendexpectedforGaussian noisewhennormalizedtotheunbinnedtheoreticaluncertaintiesforthese observations.Thereisnoobviouspresenceofrednoiseatlargebinning factors. 102

PAGE 103

Figure3-14. Meancolourofthein-transituxratiosascomputedbetweeneachoff-line bandpassandtheon-linebandpass.Theopentrianglesrepresentthe coloursascomputedinSection 3.2.3 andillustratedinFigure 3-12 .The solidcirclesrepresentthecolourscomputedafterexcludingoutlying absoluteuxmeasurementsforeachstarfromtheanalysis(Section 3.2.1 ). Theerrorbarsrepresentthe1 uncertainties.Thedashedlineillustrates wherethecolourequalszero.Wearbitrarilysetthispointequivalenttoan apparentplanetradiusof1(i.e.weletthemeasuredradiusintheon-line bandpassbethebaselineradiusofHD80606b).Themeancoloursaround the773.66-nmbandpassareessentiallyequalforbothsetsofpoints,so thetwodatapointsappearasone. 103

PAGE 104

CHAPTER4 ASEARCHFORMETHANEINTHEATMOSPHEREOFGJ1214BVIA NARROW-BANDTRANSMISSIONSPECTROPHOTOMETRY Thediscoveryof"super-Earths",planetswithmassesbetween1.5and10M openedanentirelyneweldofexoplanetresearch.Previously,astronomershadbeen surprisedbythediscoveryof"hot-Jupiters",Jupiter-sizeplanetsinextremelyclose orbitsaroundtheirhoststars.Suchdiscoveriesinspirenewtheories,newmodels,and newobservations,sothatwemaytrytounderstandhowobjectslikesuper-Earthsand hot-Jupitersformandevolve.Furthermore,investigationsofsuchplanetshelpusbetter understandwhyourownSolarSystemdoesnotcontainanysuchplanets. Oneofthemorewell-studiedsuper-EarthsisGJ1214b,whichwasdiscovered inaground-basedtransitsurvey(MEarth)by Charbonneauetal. ( 2009 ).Because GJ1214borbitsanMdwarfstar,ithasalargerplanet-starradiusratiothanmost othersuper-Earthsdiscoveredtodate.Thisresultsinatransitdepththatisnearly 1.5%,whichimmediatelylendsitselftomakingGJ1214banexcellentcandidatefor atmosphericstudies.Furthermore,basedonitsobservedmassandradius,itwas believedthatGJ1214bwasrequiredtohaveasignicantatmosphere( Miller-Ricci& Fortney 2010 ).However,thereisadegeneracyinthemodelsduetoGJ1214b'sdensity andirradiation( Rogers&Seager 2010 ).Asaresult,itispredictedthatGJ1214bis eithercomposedofarocky/icecoresurroundedbyahydrogen-richatmosphere,a water/icecorewithawatervaporatmosphere,orarockycorewithathinatmosphere formedbyoutgassing. RecentstudieshaveattemptedtoconstrainGJ1214b'satmospherethrough transmissionspectroscopy(e.g., Beanetal. 2010 2011 ; Crolletal. 2011 ; Crosseld etal. 2011 ; D esertetal. 2011 ; Bertaetal. 2012 ; deMooijetal. 2012 ).Inamajority ofthesestudies,itwasfoundthatGJ1214bhasaat,featurelessspectrum.Withno evidenceofanysignicantfeatureseitherinoptical( 600-1000nm)ornear-infrared (1.1-1.7 m)wavelengths.Itisbelievedthatthelackofsignicantfeaturessupportsthe 104

PAGE 105

presenceofeitheraheavy,metal-richatmosphereorthickclouds/hazesthatproduce aconstantlevelofabsorptionacrossalargerangeofwavelengths.Oneexceptionisa studyconductedby Crolletal. ( 2011 ).Specically,theyreportedasignicantlydeeper transitdepthat 2.15 m,awavelengthwheremethanewouldbeaviablesourceof opacity. Toreconcilethesestudies,wepresentnarrow-bandphotometryofthreetransitsof GJ1214baroundapredictedmethaneabsorptionfeaturefoundintheoptical(predicted basedonthemodelsfrom Miller-Ricci&Fortney 2010 ).Thebroadmethanefeature thatwefocusonisfoundaround890nmandispredictedtocauseadditionalabsorption duringtransitatalevelof 0.1%(assumingahydrogen-richatmosphere).In ¤ 4.1 ,we describeourobservations(acquiredusingtheOSIRISinstrumentinstalledonthe10.4 mGTC).In ¤ 4.2 and ¤ 4.3 ,wedescribeourdatareductionandanalysisprocedures. Wepresentourresultsin ¤ 4.4 ,andwediscusstheimplicationsofourresultsonfuture atmosphericstudiesofplanetslikeGJ1214bin ¤ 4.5 .Finally,weconcludewitha summaryofourndingsin ¤ 4.6 4.1Observations WeobservedthreetransitsofGJ1214b,duringwhichweacquiredphotometry ofGJ1214andthreenearbyreferencestars(hereafterknownasRef1,Ref2,and Ref3).Wealsoacquiredadditionalbaseline(out-of-transit)photometryofthesame fourstarsonanightfollowingoneofthetransitevents.Forallobservations,weused thetunablelter(TF)imagingmodeonGTC/OSIRIStoacquirenear-simultaneous photometryintwonarrowbandpasses(FWHM=1.2nm)byalternatingbetweenthe twobandpassesduringtheobservations.TheTFimagerallowsforcustombandpasses withacentralwavelengthbetween651-934.5nmandaFWHMof1.2-2.0nmtobe specied.Therefore,wespecicallychoseourbandpassessothatonewaslocated inthecontinuum,totheleftofwhereweexpectthemethaneabsorptionfeatureto be.Theotherbandpasswaslocatedwithintheassumedmethaneabsorptionfeature, 105

PAGE 106

aroundtheexpectedpeakoftheabsorption.Thenalobservingsequencethatwe usedwastotaketwoexposuresinthecontinuumbandpass,twoexposuresinthe "methane"bandpass,andrepeat.Thespecicbandpassesusedaregiveninthe followingsections,astheyvariedslightlyforthedifferentobservations.Asdescribed in Col onetal. ( 2010 ),anotherfeatureoftheTFimagingmodeisthattheeffective wavelengthdecreasesradiallyoutwardfromtheopticalcenter,soweattemptedto positionthetargetandasingle"primary"referencestar(i.e.,onemostcomparablein brightnesstothetarget)atthesamedistancefromtheopticalcentersoastoobserve bothstarsatthesamewavelengths. 1 Theotherreferencestarswerethenobserved atslightlydifferentwavelengthsthanthetargetduetotheirbeingatdifferentdistances fromtheopticalcenter. Allobservationsused1 # 1binning,afastpixelreadoutrateof500kHz,aswellasa singlewindowlocatedononeCCDchip.Thesizeofthewindowvariedslightlyforeach observation,butwaschosentobelargeenoughsoastocontainthetargetandallthree referencestars.Furtherdetailsregardingeachspecictransitobservationaregivenin thefollowingsections. 4.1.12010July22Transit WeobservedatransitofGJ1214bon2010July22underphotometricconditions andduringbrighttime.Observationsbeganat00:26UTandendedat02:11UT. Theairmassrangedfrom 1.25to1.87.Theactualseeingvariedbetween0.9and 1.4arcsec,butaslightdefocuswasusedinordertoavoidsaturation.Theexposure time(forbothbandpasses)wassetto120s,withacorresponding 22sofdead timefollowingeachexposure.Thedeadtimeincludesthetimetoreadoutawindow 1 Duetotechnicalissues,thepositioningforsomeoftheobservationswasnotas expected,andthetargetandasinglereferencestarwerenotalwaysobservedatthe sameexactwavelengths.Seethefollowingsectionsforfurtherdetails. 106

PAGE 107

of849 # 3774pixelslocatedinCCD2ofOSIRIS,whichcontainedthetargetandthe threereferencestars.Thetargetandprimaryreferencestar(Ref1)werepositionedat distancesof2.86and3.32arcminfromtheopticalcenter,whichresultedinthestars beingobservedatdifferentwavelengths.Therefore,whilethetargetwasobservedat 878.45nmand884.96nm,theprimaryreferencestarwasobservedat876.45nmand 882.94nm.Likewise,theothertworeferencestarsintheeldwereobservedatslightly differentwavelengthsthanthetarget.Theseobservationswereconductedinqueue (service)mode. 4.1.22010August28and29TransitandBaselineObservations Observationsofthe2010August28transitofGJ1214btookplacefrom22:00UT (2010August28)to00:30UT(2010August29),duringbrighttimeandundermostly clearconditions(thoughsomedustwaspresent).Theairmassrangedfromabout1.26 to2.05.Theseeingwasvariablethroughouttheobservations,sothetelescopewas defocusedandtheexposuretimewaschangedtoavoidsaturation.Anexposuretime of100swasinitiallyused,butwhentheseeingbecameworse,theexposuretimewas increasedto150s(beforethetransitingress).Finally,theexposuretimewasincreased againto200stowardstheendoftheobservations(aftertransitegress).Thedead timefollowingeachexposurewastypicallyabout22s,whichincludedreadingouta singlewindowof850 # 4102pixelscontainingthetargetandreferencestars,locatedin CCD2ofOSIRIS.Asforthersttransitweobserved( ¤ 4.1.1 ),thetargetandaprimary referencestarwerepositionedatslightlydifferentdistancesfromtheopticalcenter(3.0 and3.3arcmin).Thisresultedinthetargetbeingobservedatwavelengthsof877.0nm and883.5nm,whiletheprimaryreferencewasobservedat875.7nmand882.2nm. Wenotethatbasedontheobservationspresentedin ¤ 4.1.1 ,inwhichwefoundthat theprimaryreference(Ref1)appearedtobevariable,wechoseadifferentstar(Ref3) asourprimaryreferencefortheseobservations.Finally,therearesomesmallgapsin thedatatowardsthebeginningoftheobservationsduetominortechnicalissues.Also, 107

PAGE 108

thereissomevignettinginthelastfewimagesduetothelowelevationofthetelescope, soweexcludethesefromouranalysis.Specically,alltheimageswithanexposuretime of200swerediscarded. Additionalbaseline(out-of-transit)photometrywasacquiredonthenightof2010 August29.Thesamesetupwasusedfortheseobservationsaswasusedforthe observationsconductedonthenightof2010August28.Observationsbeganat 21:17UT(2010August29)andendedat00:20UT(2010August30),duringwhich theairmassrangedfrom1.17to2.43.Anexposuretimeof150swasusedfor theseobservations,tomatchtheexposuretimeusedduringamajorityofthetransit observationsfromthepreviousnight.Asinthepreviousnight'sobservations,some vignettingoccurredtowardstheendoftheobservingrun,soweexcludetheselater imagesfromouranalysis.Wenotethatbothoftheseobservationswerealsoconducted inqueue(service)mode. 4.1.32011June11Transit Observationsofthe2011June11transitofGJ1214bwereconductedinvisitor mode.Theconditionswereclear,andobservationstookplaceinbrighttime.Observations beganat23:40UT(2011June10)andendedat02:50UT(2011June11),duringwhich timetheairmassrangedfromabout1.09to1.19andtheseeingvariedbetween1.0and 1.1arcsec.Asinglewindowof850 # 3250pixelscontainingthetargetandreference starswasreadout.Unlikethepreviousobservations,thestarswerelocatedinCCD1 ofOSIRIS.Anexposuretimeof100swasused,withacorresponding 19sofdead timefollowingeachexposure.Alsounlikeourpreviousobservations,wewereableto setthetargetandaprimaryreferencestar(Ref3,thesameprimaryreferenceasinthe August2010observations)atthesamedistancefromtheopticalcenter(3.2arcmin). Thus,boththetargetandprimaryreferencewereobservedatwavelengthsof877.0nm and883.5nm.Finally,wenotethatforthedurationoftheobservations,therewasa problemwiththeGTC'sprimarymirror.Onesegmentofthemirrorwouldnotstackwith 108

PAGE 109

theothersegments.However,asthisproblemhadthesameeffectonallthestars(i.e., eachstarhadaverysmall"copy"ofitselflocatedtowardsthebottomrightofthestar ontheCCD),weassumethatthephotometrywouldnotbeaffectedbythisproblemas aperturephotometrywouldstillincludethephotonsfromtheunstackedsegment. 4.2DataReduction Forallourdatasets,weusedstandardIRAFproceduresforbiassubtractionand at-eldcorrection.Fortheat-eldcorrection,weuseddomeatsthatweretaken aftereachobservationandforeachltersetting.However,sincethedomelightsdo notprovideauniformillumination,weaddedanilluminationcorrectiontothenal at-eldimagesforeachdataset.Asdiscussedin Col onetal. ( 2010 )and Col on& Ford ( 2011 ),theOSIRISTFproducesimageswithsky(OH)emissionrings.Therefore, weperformedskysubtractionusingtheIRAFpackageTFred 2 priortoperforming aperturephotometry.Thistaskspecicallymeasurestheskybackgroundwhileincluding theringsduetoskyemission.Afterrunningthistaskonallthescienceimages,we usedstandardIDLroutinestoperformaperturephotometryonthetargetandthethree referencestarsintheeld.Wetestedseveraldifferentsizeaperturesforeachdata set,andwesetournalaperturetobetheonethatyieldedthesmallestscatterinthe baseline(out-of-transit)uxratios(computedasthetargetuxdividedbythesumof thereferencestaruxes).Weconsideredtheindividualresultsforeachdatasetand appliedthesameaperturetoeachlter(foragiventransitobservation).Forthetransit observationsinJuly2010,August2010,andJune2011,weusedaperturesof42, 45,and52pixels(approximately5.3,5.7,and6.6arcsec).Theapertureusedforthe baselineobservationsfromAugust2010was56pixels,equivalentto7.1arcsec.We didnotneedtoincludeaskyannulusintheaperturephotometryprocessbecausethe 2 WrittenbyD.H.JonesfortheTaurusTunableFilter,previouslyinstalledonthe Anglo-AustralianTelescope;http://www.aao.gov.au/local/www/jbh/ttf/adv reduc.html 109

PAGE 110

TFredtaskhadalreadyremovedtheskybackground.Onceaperturephotometryhad beencompletedforeachdataset,wecontinuedwiththelightcurveanalysisdescribed inthefollowingsection. 4.3LightCurveAnalysis Wecomputedlightcurvesforeachbandpassandforeachobservationseparately bydividingthetotaluxmeasuredwithinthetargetaperturebythetotaluxofthe referencestar(s)foragivenbandpass/observation.Sincewehadonlythreereference starsinoureld,wecomputedlightcurvesforthetargetrelativetoeachofthereference starsaswellastodifferentcombinationsofthereferencestars.Whenusingmorethan onereferencestar,wecomputedthetotalweighteduxofthereferencestarensemble, whereweweightedtheuxfromeachreferencestarafterdiscardinganyoutlyingux measurements.Sinceallthereferencestarsweretypicallylocatedatdifferentdistances fromtheopticalcenterthanthetarget(exceptforthethirdtransit,observedinJune 2011,wheretheprimaryreferencestarwasindeedlocatedatthesamedistancefrom theopticalcenterasthetarget),weoptedtotestallpossiblelightcurvesinorderto comparetheresultsfromeachoneandtodetermineifanyspecicreferencestar(s) providedamorestablelightcurvethantheothers. 3 Aftercomputingtherelativeux ratios,wenormalizedeachlightcurvetothemeanbaseline(out-of-transit)uxratio measuredforeachbandpass.Wethencorrectedeachlightcurveagainstchanges inairmass.Also,fortheJune2011transit,weremovedanadditionalquadratictrend thatwaspresentoverthedurationoftheobservations.Finally,weperformedexternal parameterdecorrelation(e.g., Bakosetal. 2007 2010 )againstthe x and y centroid coordinatesofthetargetandthesharpness(i.e.theFWHM)ofthetarget'sprole.We notethatthecorrespondingobservationtimesforeachlightcurvewerecomputedfrom 3 WealsousedthebaselineobservationsfromAugust2010tocheckthestabilityof GJ1214relativetoeachreferencestar.Seethefollowingsectionforfurtherdetails. 110

PAGE 111

theUTCtimestampsgivenintheimageheaders.TheUTCtimeatmid-exposurewas thenconvertedtoBarycentricJulianDateinBarycentricDynamicalTime(BJD TDB)via anonlinecalculator. 4 Inordertocomputethephotometricuncertaintiesoneachdatapoint,weconsider thephotonnoiseofthetargetandthereferencestar(s)usedtocomputethelight curves,thenoiseintheskybackgroundaroundthetargetandthereference(s),and scintillationnoise.Wendthemedianphotometricuncertaintiesforallthelightcurves totypicallybebetween0.5and0.6mmag.Inallcases,thephotonnoiseofeitherthe targetorRef1dominatestheerrors,butingeneralthetargetandeachreferencestar haveacomparablelevelofphotonnoise.Finally,weinvestigatedwhetherrednoise hasasignicantpresenceinourdatabycomputingthestandarddeviationoftheux ratiosafterbinningthedata.Ingeneral,wefoundthatourdatatendtofollowthetrend expectedifonlywhiteGaussiannoisewerepresent.However,fortheobservationsfrom July2010,thereappearedtobesomeresidualsystematicsinthelightcurvesdueto theinclusionofRef1(theoneweusedasaprimaryreferenceforthatobservation).We attributethisrednoiseprimarilytovariabilityinthatstar,asthatspecicreferencestar consistentlyproducedlightcurvesthatwerehighlyvariablefromtransittotransit.Asa result,weinvestigatetheeffectofthesepotentialresidualsystematicsinourdataviaa "prayer-bead"analysis(e.g., D esertetal. 2009 ),andwediscussourresultsinfurther detailbelow. Forthethreetransitobservations,wetsyntheticmodelstoeachcorrectedlight curve.Specically,wefollowedanapproachsimilartothattakenby Col onetal. ( 2010 ), andweusedtheplanetarytransitlightcurvemodelsfrom Mandel&Agol ( 2002 ) totlimb-darkenedmodelstoourdata.Foreachlightcurve,wexedtheimpact 4 Asdescribedin Eastmanetal. ( 2010 ).Thecalculatorcanbefoundat http://astroutils.astronomy.ohio-state.edu/time/utc2bjd.html 111

PAGE 112

parameter( a cos i / R )andlimbdarkeningcoefcients( c 1 and c 2 ) 5 tovaluesgivenin theliterature.Thereasonforthisisthatduetotherelativelylongexposuretimesused (100-200s)incomparisontotheshorttransitduration( < 1h),weacquiredverylittle data,ifany,duringthetransitingressandegress,whichmakesitdifculttoconstrain parameterssuchastheimpactparameterandlimbdarkening.Wethentforthe followingparameters: timeofmid-transit( t 0 ) transitduration(fromrsttofourthcontact; ) planet-starradiusratio( R p / R ) baselineuxratio (linear)baselineslope ThelimbdarkeningcoefcientswerebasedonthecoefcientsfortheSloan z $ lteras reportedby Carteretal. ( 2011 ).Duringthettingprocess,wekeptthecoefcientsxed atself-consistentvalues,butwedidtestarangeofvaluesbasedontheuncertainties givenby Carteretal. ( 2011 ).Thevaluesthatweusedfortheimpactparameterwere takenfrom Bertaetal. ( 2011 ),asweretheinitialguessesforthemid-transittime,transit duration,andradiusratio.Wealsotestedarangeofinitialguessesfortheseparameters basedontheuncertaintiesgivenin Bertaetal. ( 2011 ).Weidentiedbest-ttingmodels viaaLevenberg-Marquardtminimizationscheme. 6 Wemodeledthelightcurvesforeachindividualtransitobservationfollowinga similarprocedureas Col onetal. ( 2010 ).Wethenappliedasimilaranalysistolight curvesthatweregeneratedbycombiningthedatafromallthreetransitobservations. 5 Here,wedenethelimbdarkeningcoefcientsas c 1 % u 1 + u 2 and c 2 % u 1 $ u 2 where u 1 and u 2 arethelinearandquadraticlimbdarkeningcoefcients. 6 Wespecicallyused mptfun ,whichispubliclyavailableat http://www.physics.wisc.edu/ craigm/idl/idl.html 112

PAGE 113

Tosummarizetheprocedurehere,wetmodelstoeachlightcurveforeachbandpass andeachtransitobservationindividually,thenusedthemodelstocorrectthedataand discardanypointslyinggreaterthan3 fromtheresiduals.Wethentmodelstothe correctedlightcurvesinajointanalysis,whereweforcedthemid-transittime,transit duration,baselineuxratioandbaselineslopetobethesameforbothlightcurves(for agiventransit).However,weallowedthedifferentlightcurvestohavedifferentvalues fortheradiusratioandlimbdarkeningcoefcients.Wethencorrectedtheindividual lightcurvesagain,thistimebasedontheresultsfromthejointanalysis.Analjoint analysiswasappliedtothe"nal"correctedlightcurves.Duringthisnalstage,wealso performeda"prayer-bead"analysis.Specically,weperformedacircularshiftonthe residualsforeachlightcurveandgeneratedsyntheticlightcurvesbyaddingtheshifted residualsbacktothebest-tmodel.Weappliedthejointanalysistoeachsynthetic lightcurve,andwecomparethedispersionofthebest-tparameterstotheformal1 errorsoneachttedparameter.Thisallowsustoinvestigatetheeffectofanyadditional systematicnoisesourcesinthedata,asdiscussedabove.Werepeatedthisprocess forthecombinedlightcurves,whereweconsideredallthreetransitstogether.Wealso performedthisanalysisforallthedifferentlightcurvesthatwehadcomputedforGJ 1214thatwerebasedondifferentcombinationsofreferencestars.Wepresentthe resultsfromouranalysisinthefollowingsection. 4.4Results Inthissection,werstdescribetheoverallresultsfromallourobservations, andthenwediscusstheresultsforeachindividualtransitobservationindetailin ¤ 4.4.1 $ 4.4.4 .Sinceweonlyhadthreereferencestarsinoureld,werstconsidered theresultsforourbaselineobservationsfromAugust2010todeterminewhetherany specicreferencestarorcombinationofreferencestarswasobviouslymorestable thanothers.InFigure 4-1 ,wepresentthelightcurvesforGJ1214relativetoeach combinationofreferencestarsfromtheAugust2010baselineobservations.Visual 113

PAGE 114

inspectionrevealsthatthelightcurvesareallfairlyscattered,butwendthatusing multiplereferencestarshelpstoreducethescatterandleveloutvariabilitydueto eitherstellaractivityand/ornon-idealweatherconditions(particularlyinthecaseof theseobservations,somedustwaspresentandcouldhaveaffectedourphotometry). However,includingRef1specicallytendstointroducethemostscatter,soweconclude thatweshouldgenerallydisregardresultsthatarebasedontheinclusionofRef1.This isfurthersupportedbyourotherobservations,asdiscussedbelow. Wepresentthelightcurvesandthecorrespondingbest-tmodelsforeachtransit observationinFigures 4-2 4-3 ,and 4-4 .InFigure 4-5 ,weshowthelightcurves generatedfromthecombinationofallthetransitobservations.Ineachgure,weagain showlightcurvesforGJ1214relativetoeachcombinationofreferencestars.Also, foreaseofcomparison,alloftheseguresareplottedwiththesamerangesonthe x and y -axes.AsalreadyillustratedinFigure 4-1 ,wendthatwhenweincludeRef1in theanalysis,thelightcurvestendtobemorescattered,particularlyinthecaseofthe August2010transit(Figure 4-3 ).Therefore,inthefollowingsections,wefocusonthe resultsfromthefollowingcases,whichwelaterrefertoascases2,3,and6:GJ1214 normalizedtoRef2,GJ1214normalizedtoRef3,andGJ1214normalizedtoRef2 andRef3. Beforewediscussthespecicresultsfromeachtransitobservation,inFigures 4-6 4-7 4-8 ,and 4-9 wepresentthebest-tradiusratiosforeachobservationand eachindividuallightcurve.Asbefore,foreaseofcomparison,allgureshavethesame rangesonthe x -and y -axes.Thebest-tvaluesandcorresponding3 errorbarsthat areshownarebasedonthemodelwiththesmallest $ 2 value.Whilenotillustratedhere, wealsocomparetheformalerrorswiththescatterinthettedparameters,computed fromthefullsetofmodelsgeneratedintheprayer-beadanalysis.Wendthatthe distributionofthebest-tradiusratiosoverallpermutationsisgenerallycomparable totheformal1 errorsfoundfromthemodelsttothersttwotransitobservations 114

PAGE 115

(July2010andAugust2010).However,forthethirdtransit(June2011),wendthat thedistributionisscattered,andissometimessmallerorlargerthantheformalerrors, dependingonwhichreferencestarisincludedintheanalysis.Thereisnotrendasto whichreferencestarcausesaskeweddistributionofbest-tradiusratios.Basedon visualinspectionofthelightcurves(Figure 4-4 ),wendthatthereisanunusualfeature presentineachofthelightcurvesat883.5nm,whichsuggeststhatactivityinGJ1214 maybetheculpritofsuchafeature.Despitethis,webelievethe3 errorbarsshownin Figures 4-6 $ 4-9 arelargelyrepresentativeofthetrueerrorsinthettedradiusratios. Whiletheerrorbarsmayinfactbelargerin some cases,wedonotexpectthemtobe signicantlylargerthanwhatisshowninthesegures. Whilenotshownhere,weconstructedsimilarguresfortheimpactparameter, thetransitduration,andthemid-transittime,andwefoundthat(a)asmallerimpact parameterwasgenerallypreferred,(b)thetransitdurationtendedtomatchvaluesfrom theliterature(withinafewminutes),and(c)themeasuredmid-transittimetypically deviatedfromtheexpectedtransitephemerisbylessthanoneminute. Inthefollowingsections,wediscusstheresultsfromeachspecictransitobservation, aswellasfromthecombinationofallthreetransits. 4.4.1Resultsfromthe2010July22Transit AsillustratedinFigures 4-2 and 4-6 ,fortheseobservationsweconsistently measuretransitdepths(andthereforeradiusratios)thatarelargerinthebluerbandpass (i.e.,theonelocatedinthecontinuum).Specically,wendthattheradiusratiosas measuredinourtwobandpassesconsistentlydifferbyslightlymorethan3 .Wend thisresultevenwhenwediscountmeasurementsthatincludeRef1(whichwefoundto bevariablefromobservationtoobservation).Also,lookingatjusttheresultsforcases 2,3,and6,wendthattheradiusratiosmeasuredineachbandpassareconsistent betweenthedifferentcases,andthattheradiusratiosmeasuredinthebluerbandpass areconsistentwiththatreportedby Bertaetal. ( 2011 ).Thisisinteresting,asnone 115

PAGE 116

ofthereferenceswereobservedatthesamewavelengthasthetargetinthistransit observation.Thissuggeststhatforthesereferencestars,observingthematdifferent wavelengthsthanGJ1214hasanegligibleeffectonourresults.Regardless,ournding thattheradiusratiosarelargerinthecontinuumbandpasscontradictswhatwewould expectifmethanewerepresentandabsorbinginGJ1214b'satmosphere. 4.4.2Resultsfromthe2010August28Transit InFigure 4-3 ,weclearlyseethatRef1introducedalargeamountofscatterduring theseobservations.Therefore,wedisregardanyresultsbasedonRef1.Lookingat justcases2,3,and6,wendthatincases2and6,thereappearstobenosignicant differenceinthetransitdepthsorradiusratiosbetweenthetwobandpasses.However, incase3,wendthattheradiusratiointheredbandpass(i.e.,themethaneband)is largerthanthatinthebluebandpassatalevelslightlygreaterthan3 .Correspondingly, theradiusratiomeasuredintheredbandpassisconsistentbetweenallthreecases, whiletheradiusratiofoundforthebluebandpassisconsistentonlyforcases2and6. However,despiteanyconsistencies,noneofthemeasuredradiusratiosforcases2,3, and6matchthevaluereportedby Bertaetal. ( 2011 ).Infact,allourradiusratiosare smallerthanthatfoundby Bertaetal. ( 2011 ).Theseresultssuggestthattheremay somestellaractivityinRef3thatisaffectingthelightcurveforGJ1214and/orthatin thiscase,thefactthatnoneofthereferenceswereobservedatthesamewavelength asthetargetisimportant.Also,becauseweconsistentlymeasureasmallerradiusratio than Bertaetal. ( 2011 ),itmaybethatGJ1214happenedtohavefewerstarspotson itssurfaceduringthisobservationthanitusuallydoes(assumingittypicallyhassome numberofspotsonitssurface).HavingfewerstarspotswouldimplythatGJ1214 appearedslightlybrighterthanittypicallydoes,sothatthefractionallossoflightdueto theplanetduringtransitwoulddecrease,resultinginashallowertransitandthereforea smallermeasuredplanet-starradiusratio(wealsoreferthereaderto Carteretal. 2011 forfurtherdiscussionontheeffectsofstarspotsonthetransitlightcurveforGJ1214b). 116

PAGE 117

4.4.3Resultsfromthe2011June11Transit Inthistransitobservation,weseeananomalousfeaturepresentineachofthelight curvesacquiredintheredderbandpass(Figure 4-4 ).Thefeatureisnotindicativeofa starspotcrossing,asthelightcurveappearstogetslightlydeeperaroundthebeginning ofthetransitegress.Also,thefeatureismostprominentwhenincludingRef1,sowe againconsideronlycases2,3,and6.Asthefeatureislessvisibleinthesecases,we donotdiscardanyofthesepointsinthettingprocess.Regardlessofthesourceof theanomalousfeature,wendthattheradiusratioisconsistentlylargerintheredder bandpassatalevelgreaterthan3 (Figure 4-8 ),whichsuggestspossibleadditional absorptionduetomethaneinGJ1214b'satmosphere.Furthermore,incases3and 6,wemeasurearadiusratiointhebluebandpassthatmatchesthatfrom Bertaetal. ( 2011 ).Thismatchesourresultsfromthersttransitobservation( ¤ 4.4.1 ).However,we ndadifferentradiusratiointhebluebandpassforcase2aswellasmuchlargerradius ratiosintheredbandpasscomparedtothosemeasuredinthersttransitobservation. Asinthesecondtransitobservation,itmightbethatwecanattributethesedifferences tovaryingstellaractivity(e.g.,differentnumbersofstarspotsonGJ1214'ssurface). Consideringthatwearendingdifferentresultsfromeachofourtransitobservations, webelievethatstellaractivitymaybeanextremelysignicantfactorinourobservations. Wediscussthisinfurtherdetailin ¤ 4.5 4.4.4ResultsfromAllTransits Despitethedifferentobservingconditionsforthethreetransitsthatweobserved, weinvestigatetheresultsfromcombiningallthreetransitstogether.Ourgoalwasto comparetheseresultswiththeresultsfoundfortheindividualtransitobservations. Inthiscase,wendthatthemeasuredradiusratiosarehighlycorrelatedwiththe individualreferencestars(i.e.,cases1,2,and3).However,whenincludingmorethan onereferencestarintheanalysis,wendthattheradiusratiosareconsistentlyslightly largerthanthatfoundby Bertaetal. ( 2011 ).Furthermore,theradiusratiosbetween 117

PAGE 118

thetwobandpassestendtobeconsistentforallcaseswheremultiplereferencesare included.Fromthiscombinedanalysis,wecannotconrmorrefutethepresenceof methaneatthesewavelengths.Asin ¤ 4.4.3 ,weconcludethatstellarvariability(likelyin thetarget and references)hasasignicanteffectonourmeasurements. 4.5Discussion WeobservedthreetransitsofGJ1214busingnarrow-bandphotometry,withthe goalofsearchingforextraabsorptionduetomethaneinGJ1214b'satmosphere. Fromouranalysis,wendthatallthreetransitsyieldinconsistentresultsinregards tooursearchformethaneabsorption.Specically,forthersttransit(July2010),we unexpectedlymeasureaslightlylargerplanet-starradiusratioinabandpassthatshould belocatedinthecontinuum(totheleftofthepredictedmethaneabsorptionfeature). Forthesecondtransit(August2010),wemeasureconsistentradiusratiosbetweenour twobandpasseswhenusingsomereferencestars,butwealsondacasewherethe radiusratioislargerinabandpassthatshouldbelocatedonamethaneabsorption feature.Regardless,theseradiusratiosdonotmatchvaluesfromtheliterature,asthey aresignicantlysmaller.Forthethirdtransit(June2011),weconsistentlyndalarger radiusratioinabandpassthatislocatedinthepredictedmethaneabsorptionband. Whilewemeasureasignicantdifferenceintheradiusratios,whichcouldindicateextra absorptionduetomethaneinGJ1214b'satmosphere,thevaluesfortheradiusratios foreachbandpassvaryquiteabitdependingonwhichreferencestarsareincludedin theanalysis.Finally,whenweconsiderallthreetransitstogether,weconsistentlynd nosignicantdifferencebetweentheradiusratiosmeasuredinourbandpasses(except forthecasewhereweonlyuseRef1,butthatstarhasappearedtohavesignicant transit-to-transitvariability). Ultimately,wendthattheresultsfromeachtransitobservationareinconsistent. However,wemustconsiderthefactthateachobservationwasconductedunderdifferent conditions(mostrelevantisthatintwoofthethreetransits,noreferencestarswere 118

PAGE 119

observedatthesamewavelengthsasGJ1214).Ifweconsiderthebest-casescenario, whichwouldbethecasewhereweobservedGJ1214andanotherreferencestarat thesamewavelength,thenthatleadsustothethirdtransitobservation(specicallythe thirdpanelinFigures 4-4 and 4-8 ),whereweobservedGJ1214andRef3atthesame wavelength.Forthisspeciccase,wemeasureradiusratiosthataredifferentbymore than3 ,withthelargerradiusratiobeingmeasuredinthebandpassthatislocated inthepredictedregionofmethaneabsorption.Furthermore,theradiusratiointhe continuumbandpassisanexcellentmatchwithvaluesfromtheliterature.Thissuggests thatwhenallourobservingcriteriaweremet,wewereindeedabletondpotential evidenceofmethaneinGJ1214b'satmosphere.However,whilesucharesultwould haveasignicantimpactonthecompositionofGJ1214b'satmosphere,wecannot dismissthefactthatourothertransitobservationsrevealthehighlyvariablenatureof suchstudies. Inthefollowingsections,wediscussthreepotentialcausesofvariabilityinour measuredradiusratios.WespecicallyfocusonvariabilitywithinGJ1214b(theplanet, specicallyitsatmosphere),GJ1214(i.e.,activityduetostarspots),andEarth(i.e.,our ownatmosphere).Whilewedonotdiscussinstrumentaleffectsasasourceofvariability here,wearecondentthatwehavealreadyaccountedforthemostsignicanteffects ( ¤ 4.3 )andthatanyresidualsystematicsduetotheinstrumentwouldhaveanegligible effectonourresults(e.g., Col onetal. 2012 ). 4.5.1VariabilityduetoGJ1214b'sAtmosphere BasedonearlymodelsforGJ1214b'satmosphere( Miller-Ricci&Fortney 2010 ), weestimatetheexpectedsignalduetomethaneabsorptioninourbandpasseswould be 0.1%(assumingahydrogen-richatmosphere).Lookingagainatourbest-case scenario(case3forthetransitobservationfromJune2011),therewemeasured planet-starradiusratiosof0.115656 0.000977and0.127522 0.000924inour blue(continuum)andred(methane)bandpasses.Thistranslatestoachangeinthe 119

PAGE 120

uxratiosbetweenourtwobandpassesatalevelof 0.29%,whichisnearlythree timeslargerthantheexpectedsignal.Weconsiderthatifmethanewerepresent andabsorbinginGJ1214b'satmosphere,thestrengthoftheabsorptionlinecould betime-variableifthereare,forexample,high-speedandhigh-altitudewindsin theatmosphere.However,tobeasvariableasrangingfrom0.1%tonearly0.3%in absorptionisunlikely,unlesstherewasmuchmoremethanepresentintheatmosphere thanpredicted.Weconcludethatwhilewemayhavemeasuredabsorptiondueto methane,thelevelofabsorptionwemeasuredislikelyfartoolargefortherenottobe othersourcesofvariabilityaffectingourresults.Furthermore,consideringthatrecent studiessupportaat,featurelessspectrumforGJ1214b(e.g., Beanetal. 2010 ; Berta etal. 2012 ),webelievethatasawholeGJ1214b'satmosphereislikelyquitestable. 4.5.2VariabilityduetoStellarActivity GJ1214hasbeenpreviouslyshowntohavesomelevelofactivityduetospots andares(e.g., Kundurthyetal. 2011 ),soitisnotsurprisingthatwendinconsistent resultsfromtransit-to-transit.However,ourobservationswereacquiredatfairlyred wavelengths,wherethecontrastbetweenstarspotsandthestellarphotosphereis minimized.Therefore,wemightexpecttondmoreconsistentresultsfromtransitto transitthanweactuallydid,unlessthenumberofspotspresentonGJ1214'ssurface variessignicantly.Together,thissuggeststhatperhapsthereferencestarsand/orGJ 1214aremorevariablethanpreviouslybelieved( ¤ 4.5.3 ).Wenotethatitisalsopossible thatourobservationsprobedapartoftheGJ1214'sspectrumthatishighlyvariable. Forinstance,TiOistypicallypresentinM4.5starsaround880nm( Kaler 2011 ),and,if variable,couldhaveaffectedourmeasurementsoftheplanet-starradiusratios.Inany case,stellaractivitysurelyplaguesstudiessuchastheonepresentedhere. 4.5.3VariabilityduetoEarth'sAtmosphere WeexpectthatanyvariabilityduetoEarth'satmosphereisremovedinourdata reductionprocedure,wherewenormalizeeachobservationofGJ1214bytheux 120

PAGE 121

ofareferencestartakenatthesametime.However,wehavetoconsiderthatinthe caseofourrsttwotransitobservations,GJ1214andeachofthereferencestarswere observedatdifferentenoughwavelengthsthatanyvariabilityduetospeciesinEarth's atmospheremaynothavebeenentirelyremoved.Toinvestigatethis,weconsiderthe lightcurveforRef2normalizedtoRef3(aftercorrectingforchangesinairmass),as theyarethetwomoststablereferencesinoureld.Eventhoughtheywereobserved atdifferentwavelengths,weusethislightcurvetodeterminehowstabletheEarth's atmospherewasduringeachobservation.Specically,wendthatfortheJuly2010 observations,theRef2/Ref3lightcurveisquitestable,andshowsnoanomalous featuresduetoeithervariabilityinthereferencesortheEarth'satmosphere.Thescatter inthelightcurveisalsosmallerthanthatmeasuredforGJ1214(outsideofthetransit event),soweconcludethatanyeffectsfromEarth'satmosphereweresufciently removedinthersttransitweobserved.IntheAugust2010observations,wend thattheRef2/Ref3lightcurveinthebluerbandpassincreasesslightlyaroundthe timeofmid-transit,thoughthelightcurveisfairlystableintheredderbandpassforthe durationoftheobservations.AssumingtheuxfromRef2andRef3wasstable,then thisfeaturecouldbearesultofvariabilityinEarth'satmosphereatthattime.Indeed, thiscouldbethesourceofthesmallerapparentplanet-starradiusratiomeasuredin thebluerbandpassforthecasewhereGJ1214wasnormalizedtoonlyRef3(Figures 4-3 and 4-7 ).But,consideringthatGJ1214andRef3wereobservedatmoresimilar wavelengthsthanGJ1214andRef2were,itwouldseemmorelikelythatthelightcurve forGJ1214normalizedtoRef2wouldhavehadananomalousradiusratio.Wesuggest thatinstead,thefeaturethatwendintheRef2/Ref3lightcurveisactuallydueto variabilityinoneofthereferencestarsandnottheEarth'satmosphere.Furthermore,we believethatthefeatureisduetoRef2,becausewendthattheRef2/Ref3lightcurve forthethirdobservation(June2011)alsoshowedananomalousfeatureinthebluer bandpassthatlookedverymuchlikeaare.Giventhattheuxratiosincreasedsharply 121

PAGE 122

(aroundtheendofthetransitegress),weconcludethattheuxofRef2increased relativetoRef3,andthereforeRef2istheactivereferencestarinboththeAugust2010 andJune2011observations.However,itisinterestingthatthefeaturesseenintheRef 2/Ref3lightcurvesoccuronlyinthebluerbandpass,andthelightcurvesintheredder bandpasswerestable.Evenifsuchvariabilityinthereferencestars'lightcurveisdue toEarth'satmosphereratherthanthereferencestarsthemselves,wedonotbelieve thatwecandisentanglewhethertheEarth'satmosphereoroneofthereferenceshasa greatereffectonourmeasurementsoftheplanet-starradiusratio. 4.6Conclusions Insummary,wendthatwecanreachvery-highprecisionswiththeGTC/OSIRIS tunablelter,evenforastarasfaintandactiveasGJ1214.Furthermore,theprecisions thatwereacharesuitableforatmosphericstudies.Inthiscase,weattemptedto searchformethaneinGJ1214b'satmospherebymeasuringtheradiusratiosintwo narrowbandpassesaroundthelocationofapredictedmethanefeature.Afteracquiring observationsofthreetransitsofGJ1214bandndingthattheobservationsdonot yieldconsistentresultsregardingthepresenceofabsorptionduetomethane,we concludethatvariabilityhasasignicanteffectonsuchstudies,regardlessofwhether thevariabilitycomesfromvariableabsorptioninanexoplanet'satmosphere,stellar activity,orvariableabsorptioninEarth'satmosphere. 122

PAGE 123

Figure4-1. LightcurvesforGJ1214frombaselineobservationsacquiredinAugust 2010.Ineachpanel,theblueandredpointsrepresentobservationsofGJ 1214at877.0nmand883.5nm.Recallthatfortheseobservations,no referencestarwasobservedatthesameexactwavelengthasGJ1214.The horizontaldottedlineineachpanelillustratesabaselineuxratioof1.0.The differentpanelsshowthelightcurvesforGJ1214whennormalizedto differentcombinationsofreferencestarsandarelabeledaccordingtowhich referenceswereusedtocomputethelightcurveshownineachspecic panel.Notethattheselightcurveswerecorrectedfortrendsinairmass,the shapeofthetarget'sprole,andthecentroidcoordinatesofthetarget,but noothercorrectionsweremadesothatwecouldascertaintheeffectsof differentreferencesonthetarget. 123

PAGE 124

Figure4-2. Lightcurvesandresidualsfromobservationsofthe2010July22transitof GJ1214b.Ineachpanel,theblueandredpointsrepresentobservationsof GJ1214at878.45nmand884.96nm.Again,recallthatforthese observations,noreferencestarwasobservedatthesameexactwavelength asGJ1214.Theblueandredsolidcurvesarethecorrespondingbest-t models.Thehorizontaldottedlineineachpanelindicatesthelevelatwhich theresidualswereoffset(forclarity).AsinFigure 4-1 ,thedifferentpanels showthelightcurvesforGJ1214whennormalizedtodifferentcombinations ofreferencestarsandarelabeledaccordingtowhichreferenceswereused tocomputethelightcurveshownineachspecicpanel. 124

PAGE 125

Figure4-3. SameasFigure 4-2 ,butforobservationsofthe2010August28transitofGJ 1214b.Forthistransit,theblueandredpointsrepresentobservationsofGJ 1214at877.0nmand883.5nm. 125

PAGE 126

Figure4-4. SameasFigure 4-2 ,butforobservationsofthe2011June11transitofGJ 1214b.AsfortheobservationspresentedinFigure 4-3 ,theblueandred pointsrepresentobservationsofGJ1214at877.0nmand883.5nm.Also, forthistransit,Ref3wastheonlyreferenceobservedatthesame wavelengthasGJ1214. 126

PAGE 127

Figure4-5. SimilartoFigures 4-2 4-3 ,and 4-4 ,buthereweshowthelightcurvesand thecorrespondingbest-tmodelsgeneratedfromcombiningallthreetransit observationsfromJuly2010,August2010,andJune2011. 127

PAGE 128

Figure4-6. Best-tplanet-starradiusratiosasdeterminedfrommodelsttothe2010 July22transitlightcurves(Figure 4-2 ).Ineachpanel,thehorizontalerror barsillustratethewidthoftheltersused,andtheverticalerrorbarsarethe formal3 errorsasdeterminedfromthemodels(seetextforfurtherdetails). Alsoshownineachpanelisahorizontaldashedline,whichindicatesthe radiusratioreportedby Bertaetal. ( 2011 ),andwhichweconsidertobea representativeradiusratioformostvaluesthathavebeenreportedinthe literature.Thecorrespondinghorizontaldottedlinesillustratetheuncertainty intheirradiusratio.Asinpreviousgures,eachpanelillustratesresultsfrom differentlightcurvescomputedbasedondifferentcombinationsofreference stars. 128

PAGE 129

Figure4-7. SameasFigure 4-6 ,butforradiusratiosasdeterminedfrommodelstto the2010August28transitlightcurves(Figure 4-3 ). 129

PAGE 130

Figure4-8. SameasFigure 4-6 ,butforradiusratiosasdeterminedfrommodelstto the2011June11transitlightcurves(Figure 4-4 ). 130

PAGE 131

Figure4-9. SimilartoFigures 4-6 4-7 ,and 4-8 ,butforradiusratiosasdeterminedfrom modelsttoallthreetransitobservations(Figure 4-5 ). 131

PAGE 132

CHAPTER5 VETTING KEPLER PLANETCANDIDATESWITHMULTICOLOR PHOTOMETRYFROMTHEGTC:IDENTIFICATIONOFANECLIPSING BINARYSTARNEARKOI565 Atpresent,thereareover180conrmedtransitingplanets,butonly 10%are estimatedtobeNeptune-sizeorsmaller. 1 The Kepler spacemission,whichlaunched in2009,isresponsibleforthediscoveryofamajorityoftheknowntransitingNeptuneandsuper-Earth-sizeplanets.Someofthesmallplanets Kepler hasdiscoveredto dateincludeKepler-4b(aNeptune-sizeplanet; Boruckietal. 2010b ),Kepler-9d(a super-Earth-sizeplanetinasystemwithtwoSaturn-sizeplanets; Holmanetal. 2010 ; Torresetal. 2011 ),Kepler-10b(arockyplanet; Batalhaetal. 2011 ),andKepler11-b, c,d,e,f,g(sixNeptune-tosuper-Earth-sizeplanets; Lissaueretal. 2011a ).Further, Kepler recentlydiscoveredover1000additionalsmallplanetcandidates( R p < 6 R ; Boruckietal. 2011b ).While80-95%ofthesecandidatesareexpectedtobetrue planets( Boruckietal. 2011b ; Morton&Johnson 2011 ),identifyingwhicharefalse positivesremainsachallenge. Themainsourcesoffalsepositivesarebackground(or,rarely,foreground)eclipsing binaries(EBs)orhierarchicalmultiplesystems.Dueto Kepler 'slargepoint-spread function( 6 $$ ),theuxfromastarthathasaneclipsingstellarorplanetarycompanion canbeblendedwiththeuxof Kepler 'stargetstarifthetwostarsarespatially co-alignedwitheachother.Inthesecases,itappearsthatthetargetstarhasatransiting companion.High-resolutionground-basedimaging(adaptiveopticsorspeckleimaging), spectroscopy,and Kepler 'scentroidanalysishelptoeliminatemanyblends,butcan struggleincaseswheretheblendedsystemisseparatedbylessthan 0.1 $$ from thetarget(e.g., Boruckietal. 2011b ).Further,radialvelocity(RV)follow-upisvery time-consumingforthetypical Kepler target,whichmaybefainterthan V 14( Batalha 1 TheExtrasolarPlanetsEncyclopedia;http://exoplanet.eu/ 132

PAGE 133

etal. 2010 ).IntheabsenceofRVmeasurements,thedetectionofsecondarytransits ordifferencesinthedepthsofindividualtransitscanalsobeusedtohelpruleouta blendedsystem.Here,weconsideranalternativetechniquethat(tothebestofour knowledge)wasrstdiscussedby Tingley ( 2004 )andrstdemonstratedobservationally by O'Donovanetal. ( 2006a ),whichistoruleoutblendsbymeasuringthetransitdepth indifferentbandpasses.Thisispossiblebecause,asshownby Tingley ( 2004 ),the colorchangeduringatransitevent(i.e.,thedifferenceinthetransitdepthmeasured atdifferentwavelengths)increasesasthecolorbetweenthedifferentcomponentsofa blendincreases.Therefore,observationsacquiredinmultiplebandpassescanbeused torejectaplanetcandidateifthemeasuredtransitdepthsindifferentbandpassesdiffer signicantly,whichcouldindicate,forexample,ablendwithastellarEBofadifferent spectraltypefromthetargetstar( Tingley 2004 ; O'Donovanetal. 2006a ; Torreset al. 2011 ).Wenotethatthe COROT spacetelescopeevenhasaprismbuiltinfor thepurposeofvetting COROT planetcandidateswithmulticolorphotometry,further demonstratingthevalueofsuchatechnique( Auvergneetal. 2009 ; Deegetal. 2009 ). Here,wedescribemultiwavelengthobservationsacquiredwiththeOpticalSystem forImagingandlowResolutionIntegratedSpectroscopy(OSIRIS)installedonthe10.4 mGranTelescopioCanarias(GTC)thatweusedtodeterminethetruenatureof( Kepler ObjectofInterest)KOI565.01. 2 KOI565.01( Kepler magnitudeof14.3)waspresented by Boruckietal. ( 2011a )asasuper-Earth-sizeplanetcandidate,withanestimated planetradiusof 1.6 R ,orbitinga1.068 R # starwithaperiodof2.34days.However, therecentarticleby Boruckietal. ( 2011b )listsKOI565.01asmostlikelybeingafalse positive,as Kepler measuredacentroidshiftof 8 $$ tothenorthofKOI565,indicating thatthestarthatisactuallydimmingislocatedonapixelthatisoffsetfromtheposition ofKOI565.Giventhatwedidnothavethisinformationatthetimeoftheobservations 2 AlsoknownasKIC7025846inthe Kepler InputCatalog. 133

PAGE 134

presentedhere,weoperateundertheassumptionthatwedidnotknowwhetherKOI 565.01wasatrueplanetorfalsepositive.Inthiscase,itwaspossibleforustoresolve thetruesourceofthetransitsignalthatcontaminated Kepler 'sphotometryofKOI565. WealsopresentmeasurementsofthecolorofKOI565andseveralnearbystarsduring thepredictedtransitevent,whichindependentlyconrmthatKOI565isinfactnotthe truehostofthetransitsignal.Moreimportantly,weshowthatthefalsepositivewould havebeenidentiedeveniftheseparationbetweenthestarswastoosmalltoeither spatiallyresolvethemorallowforthemeasurementofacentroidshift.Ourapproach offersanefcientfalse-positiveidenticationmethodthatishighlycomplementarytothe multi-colorfollow-upphotometrythatiscurrentlybeingconductedwith Spitzer atinfrared wavelengths(e.g., Fressinetal. 2011 )aswellastootherground-basedfollow-up techniques. Wedescribeourobservationsin ¤ 5.1 andthedatareductionandlightcurve(LC) analysisin ¤ 5.2 .In ¤ 5.3 wepresentourresultsanddemonstratethatcolorphotometry fromtheGTCcanbeusedtohelpidentifyfalsepositivesfromtransitsurveys.Finally,in ¤ 5.4 ,weconcludewithasummaryofourresultsandadiscussionofourplansforfuture observationsofadditional Kepler planetcandidateswiththeGTC. 5.1Observations Weobservedthetargetandseveralnearbystarsaroundthepredictedtimeof thetransiteventon2010September19usingtheOSIRIStunablelter(TF)imager installedonthe10.4mGTC.WiththeTFimager,custombandpasseswithacentral wavelengthbetween651-934.5nmandaFWHMof1.2-2.0nmcanbespecied.Inthis observingmode,theeffectivewavelengthdecreasesradiallyoutwardfromtheoptical center,sowepositionedthetargetanda"primary"referencestar(i.e.,mostcomparable inbrightnesswiththetarget)atthesamedistancefromtheopticalcentersothatboth starswouldbeobservedatthesamewavelengths.Several"secondary"reference starswerealsoobserved,buttheywereallatdifferentdistancesfromtheopticalcenter 134

PAGE 135

andthuswereobservedatslightlydifferentwavelengths.Duringtheobservations,we alternatedbetweentwobandpassescentredon790.2and794.3nm(atthelocation ofthetargetontheCCDchip)andwithFWHMof2.0-nm.Thesebandpasseswere specicallychosenastheyminimizeeffectsoftelluricabsorptionandemissionand yieldextremelyhighdifferentialprecisions,asdemonstratedby Col onetal. ( 2010 ).We used1 # 1binning,afastpixelreadoutrateof500kHz,andreadoutasinglewindow (containingthetargetandseveralreferencestars)locatedononeCCDchipof1415 # 2830pixels(equivalentto 3 $ # 6 $ or 56%oftheCCDchip)inordertodecreasethe deadtimebetweenexposures.Duetothefaintnessofthetarget( V 14.3)andthe narrowbandpassesused,theexposuretime(forbothltersettings)wassetto180s, witheachexposurefollowedbyapproximately21sofdeadtime. Theobservationsbeganat21:52UTon2010September19(duringbrighttime) andendedthefollowingmorningat01:55UT.Therewerethincirruscloudsaroundthe timeofobservations.Theairmassrangedfrom 1.07to2.27.Theactualseeingwas betterthan1.0 $$ ,butthetelescopewasintentionallydefocusedtoreducepixel-to-pixel sensitivityvariations,sothedefocusedFWHMofthetargetvariedbetween 1.3-2.0 $$ ( 10.2-15.5pixels).Thetelescope'sguidingsystemkepttheimagesalignedwithina fewpixelsduringtheobservations,withthetarget'scentroidcoordinatesshiftingbyless than2pixelsineitherdirection. Thepredictedmidtransittimebasedontheephemerisandorbitalperiodfrom Boruckietal. ( 2011a )was23:59UT(2455459.502BJD)on2010September19. However, Boruckietal. ( 2011b )presentedanupdatedephemerisandorbitalperiod,so thetransiteventweobservedoccurred 135minlaterthaninitiallypredicted;because ofthis,ourobservationsendedbeforemidtransit.Itshouldbenotedthattheupdated ephemerisfrom Boruckietal. ( 2011b )isstillconsistentwiththeuncertaintyinthe originalephemerisfrom Boruckietal. ( 2011a ).Furthermore,theephemeridesgiven in Boruckietal. ( 2011a )werebasedon 43daysofobservations,while Boruckietal. 135

PAGE 136

( 2011b )citedephemeridesbasedonamuchlongertimebaseline,thusallowingfor signicantlymorepreciseconstraintsontheephemerides. 5.2DataReductionandAnalysis StandardIRAFproceduresforbiassubtractionandat-eldcorrectionwereused. Intotal,95domeatsweretakenforeachltersetting.Wenotethatthedomelights donotproduceauniformillumination,soweaddedanilluminationcorrectiontothe nalat-eldimage.Duetothenarrowltersusedandposition-dependentwavelength, allimagescontainsky(OH)emissionrings.Therefore,weperformedskysubtraction onallimagesusingtheIRAFpackageTFred, 3 whichmeasurestheskybackground whileincludingtheringsduetoskyemission.Wethenperformedaperturephotometry onthetargetandseveralreferencestarsusingstandardIDLroutines.Wetested severaldifferentsizeaperturesandbasedournalchoiceofapertureonthatwhich yieldedthesmallestscatterintherelativeuxratios(i.e.,thetargetuxdividedbythe totalreferenceux)outsideoftransit.Thenalapertureradiususedinouranalysis is23pixels( 2.9 $$ ).Noskyannuluswasneeded,duetotheuseofTFred,which automaticallyremovestheskybackground.Theseprocedureswereperformedforeach lterseparately,butweconsideredtheresultsforeachbandpassandusedthesame apertureforeachlter.Wediscardedthree-fourimagestakenineachbandpass,dueto errorsinthereductionprocessthatprohibitedusfromperformingaperturephotometry ontheseimages. LCsforeachbandpasswerecomputedforKOI565bydividingtheuxmeasured withinthetargetaperturebythetotalweighteduxofanensembleofreferencestars. Weusedsixreferencestarsintotaltocomputethereferenceensembleux.Even thoughthesereferencestarswerelocatedatdifferentdistancesfromtheopticalcenter, 3 WrittenbyD.H.JonesfortheTaurusTunableFilter,previouslyinstalledonthe Anglo-AustralianTelescope;http://www.aao.gov.au/local/www/jbh/ttf/adv reduc.html 136

PAGE 137

wefoundthatusinganensembleofreferencestars,ratherthanjustasinglereference star,greatlyimprovedthesignal-to-noiseratio(S/N)ofourobservations.EachLCwas thennormalizedtothemeanbaseline(out-of-transit)uxratioasmeasuredineach bandpass.WecorrectedeachLCagainstchangesintheairmass,andinorderto accountforanyadditionalsystematicsintheLCs,weperformedexternalparameter decorrelation(e.g., Bakosetal. 2007 2010 )againsteachofthefollowingparameters: the X and Y centroidcoordinatesofthetargetontheimageframesandthesharpness ofthetarget'sprole[equivalentto(2.35/FWHM) 2 ]. Thephotometricuncertaintiesintherelativeuxratiosincludethephotonnoise ofthetargetandthereferenceensemble,thenoiseintheskybackgroundaroundthe targetandreferences,andscintillationnoise.Wecalculatedthemedianphotometric uncertaintiestobe0.987mmagforboththe790.2and794.3nmLCs,wherethephoton noiseofthetargetdominatestheerrors.Wealsoinvestigatedthepossibilityofrednoise withinourdatabycomputingthestandarddeviationoftheuxratiosafteraveraging thedataoverseveraldifferentbinsizes,andwefoundthatourdatafollowthetrend expectedforwhiteGaussiannoise. 5.3Results WepresenttheresultsofourphotometryinTable 5-1 andthecorresponding LCsforKOI565inFigure 5-1 4 Basedontheparametersofthecandidateplanet transitingKOI565givenby Boruckietal. ( 2011a ),weexpectedtomeasureatransit depthofapproximately182ppm(partspermillion).DuetothesomewhatlowS/Nofour observations(theresultofusingverynarrowbandpassestoobserveafainttargetwhile maintainingareasonableexposuretime),ourphotometricprecisions(0.978mmag) 4 WenotetheobservationtimesgivenaretheBarycentricJulianDatesinBarycentric DynamicalTime(BJD TDB),computedfromtheJulianDatesusingthecalculatorfound athttp://astroutils.astronomy.ohio-state.edu/time/utc2bjd.html( Eastmanetal. 2010 ). Theephemeridesgivenby Kepler areinthesametimecoordinatesystem. 137

PAGE 138

wereinsufcienttodetectsuchashallowtransit.However,wedidnotneedveryhigh precisionstodeterminethatanotherstarnearthetargetwasthetruesourceofthe dimmingthat Kepler observedinKOI565. Wefollowedsimilarproceduresasdescribedin ¤ 5.2 tocomputetheLCsofseveral starswithin 20 $$ ofthelocationofthetargetthatmighthavecontaminated Kepler photometryofKOI565,andwevisuallyinspectedtheirLCstoseeifanyshoweda transit-likesignalaroundthepredictedtimeofthetransit.Forreference,wepresentthe eldofviewaroundKOI565inFigure 5-2 .FromouranalysisofstarsnearKOI565,we determinedthatastar(KIC7025851)approximately15 $$ tothenorthofKOI565isthe truesourceofthetransitsignal,asweobservedasignicantdecrease( > 15%)inthe brightnessofthatstaratthetimeofthepredictedtransitevent.WepresenttheLCsfor KIC7025851inFigure 5-3 ,andthephotometryisalsogiveninTable 5-2 .Althoughwe werenotabletoobserveacompleteLC,basedonboththeminimumdepthandshape oftheLC,wededucethatKIC7025851isastellarEB.Thus,ouridenticationofKIC 7025851asastellarEBthatcontaminated Kepler 'sphotometryofKOI565isconsistent withthemagnitudeanddirectionofthecentroidshift,aswellastheeclipseephemeris fromthe Kepler data( Boruckietal. 2011b ).TheEBthereforehasaneclipseephemeris of2455459.5956BJDandorbitalperiodof2.340506days(asdeterminedby Borucki etal. 2011b ).Basedonthestarscolorsandrelativebrightnesses,weinferthatKIC 7025851isabackground(ratherthanforeground)EB. Next,weconsiderthecolors(790.2nm $ 794.3nm)ofthestarsduringthetransit event.InFigure 5-4 ,wepresentthecolorsofKOI565andKIC7025851.Wealso considerthecolorforan"unresolved"system,simulatingascenarioinwhichthetarget starandEBarephysicallyassociatedandthuscouldnotbespatiallyresolvedsothat alltheirlightiscombined.Forthiscase,wecombinetheuxfromKOI565withtheux fromKIC7025851ineachbandpassandthencomputethecolorfromtheLCsofthe 138

PAGE 139

unresolvedsystem. 5 Letting 1 betheapparentmagnitudeinthe790.2nmbandpass andletting 2 bethemagnitudeinthe794.3nmbandpass,wecalculatedthecolor indicesas 1 $ 2 = $ 2.5log F # 1 F # 2 (51) wherewehavetakentheaverageofeachpairofuxratiosinthe790.2nmLC( F # 1 )and dividedbythecorrespondingpointsinthe794.3nmLC( F # 2 ).AsillustratedinFigure 5-4 ,wedonotmeasureanappreciablechangeinthecolorofKOI565duringthetransit event,butthecolorofKIC7025851showsasignicantcolorchange.Tocomparethe colorchangesforeachstardirectly,wecalculatetheweightedmeancolorsandtheir uncertaintiesfortheintervalbeforethetransit(i.e.,theintervaltotheleftoftheleftmost dashedlineinFigure 5-4 )andcomparethosetothesamevaluesmeasuredduringthe partialtransitevent.ForKOI565,wecalculatethedifferenceinthemeancolorstobe 6.64 6.62 # 10 4 ,whichisconsistentwiththerebeingnodifferenceinthecolorsat alevelof1 .Onthecontrary,themeancolorsofKIC7025851differatacondence levelof 8.3 ,withadifferenceof 77.9 7.2 # 10 4 .Inthecaseofthehypothetical unresolvedsystem,whereweimaginethetargetandEBtobephysicallyboundsothe projectedseparationbetweenthetwostarsisundetectableandalltheirlightisadded together,wewouldstillmeasureanappreciablechangeinthecolor,withadifference of 47.9 9.6 # 10 4 atasignicanceof 3.8 .Thefactthatwemeasurethislargeof acolorchangeoversuchanarrowwavelengthregime( 4nm)duringthetransitevent clearlyindicatesanonplanetarysourceofthecolorchange,i.e.,astellarEBcomposed oftwostarswithverydifferenttemperatures.Forcomparison, Col onetal. ( 2010 )found noappreciabledifferencebetweenthein-transitandout-of-transitcolorsasmeasured 5 BasedontheseparationofthetargetandtheEB,itismostlikelythatonlyaportion ofthelightfromtheEBwasblendedwiththelightfromKOI565(Figure 5-2 ). 139

PAGE 140

fromthesamebandpassesusedhereforeitherTrES-2orTrES-3,bothofwhichhost Jupiter-sizeplanets. 5.4Discussion Inthisarticlewehavepresentedobservationsofa Kepler planetcandidateacquired intwoverynarrowbandpasses.Fromourobservations,weidentiedanearbystellar EBineclipseatthepredictedtimeofthetransitofthe Kepler candidate.Wealsoused ourobservationstomeasurethechangeinthecolorofahypotheticalunresolved source(composedofthe Kepler targetandthestellarEB)duringthetransitevent.The identicationofanearbystellarEBandthemeasuredcolorchangeduringthetransit eventseparatelyidentifythe Kepler candidateasafalsepositive,thusconrmingthe ndingsof Boruckietal. ( 2011b ).BasedontheLCsoftheresolvedstars,wededuce thatsomeofthelightfromabackgroundstellarEB(KIC7025851)contaminatedthe photometryofKOI565tomimicthetransitofasuper-Earth-sizeplanetaroundKOI565. Thetechniquewedescribeinthisarticleiscomplementarytootherfollow-up observationsoftransitingplanetcandidatesthatarecurrentlybeingconducted.For example, Spitzer isalsobeingusedforfollow-upof Kepler planetcandidateswitha wideinfraredbandpass(e.g., Fressinetal. 2011 ),andthe COROT spacetelescope hasaprismbuiltinforthepurposeofdetectingchangesinthecolorduringtransitsof COROT planetcandidatesviathreewideopticalchannels( Auvergneetal. 2009 ; Deeg etal. 2009 ).However,while Spitzer and COROT willgoofineinthenearfuture,our approachofacquiringground-basedtransitphotometrynearlysimultaneouslyinnarrow opticalbandpassescanbeusedindenitely.Whencomparedwithmulticolorphotometry acquiredwithotherground-basedtelescopes,ourtechniquehastheadvantageof beingabletoacquiremulticolorphotometryinasingletransitobservation.Wenote thatsomeotherground-basedinstrumentsarecapableofsimilarobservations,e.g., theSimultaneousQuadIRImager(SQIID; Ellisetal. 1993 )atKittPeakNational ObservatoryandULTRACAM( Dhillonetal. 2007 )attheWilliamHerschelTelescope. 140

PAGE 141

However,theGTC/OSIRIShasauniquecombinationofalargeeldofview,asuperior collectingarea,andawideselectionoflters;combined,theseallowveryefcient high-precisionmulticolorphotometryoffaint Kepler targets. MulticolorphotometrywiththeGTCisthusausefultoolforidentifyingfalse positivesintransitsurveys,sincethemagnitudeofthecolorchangeduringtransit canbeusedtoidentifynotonlybackground(orforeground)EBstars,butalsophysical triplestarsystems,whicharedifculttoreject,astheydilutethetransitdepthandresult inanegligiblecentroidshift.Inthecaseofphysicaltriplesthataredifculttoresolve spatially(e.g.,withseparationsoflessthan0.1 $$ ),high-precisionmulticolorphotometry canbeuseful,asameasurablecolorchangeduringtransitcouldindicateablend withastellarEBofadifferentspectraltypefromthatofthetargetstar( Tingley 2004 ; O'Donovanetal. 2006a ; Torresetal. 2011 ). Morton&Johnson ( 2011 )predictaslightly higherfalse-positiveratefor Kepler planetcandidates,duetophysicaltriplesversus backgroundEBs.Thisisinpartbecausephysicaltriplescanoftenmimicthetransit depthofNeptune-sizeplanetcandidates( Morton&Johnson 2011 ),andNeptune-size planetcandidatesdominatethecandidatesdiscoveredby Kepler ( Boruckietal. 2011b ). Thus,high-precisionmulticolorphotometrymaybeparticularlyusefulforrejectingfalse positiveswithintheclassofNeptune-sizeplanetcandidates. AblendwithastarthatishostingatransitingJupiter-orNeptune-sizeplanetwillbe moredifculttorejectwithmulticolorphotometry,asthemagnitudeofthecolorchange duringtransitwillbemuchsmallerthanforablendwithastellarEB.Themostdifcult scenariotorejectwithmulticolorphotometryisahierarchicaltriplesystem,wherea physicallyassociatedcompanionstarhasaplanetarycompanion.Ourmeasurements presentedinthisarticlewouldnothavebeensensitivetoablendwithastarhostinga Jupiter-sizeplanet.However,weemphasizethatourresultswerebasedonobservations intwonarrowbandpasseswithcentralwavelengthsthatdifferedbyonly 4nm.Future observationssimilartothoseherecouldbeconductedusingOSIRIS'sbroadbandlters. 141

PAGE 142

ThesebroadbandltersaremuchwiderthanthoseallowedbytheTFimagingmode, butarestillnarrowerthantypicalSloanltersbyafactorof 2-4,sotheystillreduce effectsofdifferentialextinctionandvariableatmosphericabsorption.Theadvantageof usingslightlywiderltersistoensurethathighS/Nareachievedevenforfaintertargets ( V 14-15)whilemaintainingareasonableexposuretime(lessthanafewminutes). Further,alargerwavelengthregimecanbecoveredbyobservingin,forexample,a bluerlter(e.g., 666nm)andaredderlter(e.g., 858nm),whichcanenhancethe changeinthecolorduringatransitevent.Observinginaredlter,inparticular,willalso helpreducestellarlimb-darkening(comparedwiththebroadopticalbandpassused by Kepler ; Col on&Ford 2009 ),sothatforcandidatesdeterminedtobetrueplanets, measuringthetransitdepthinaredbandpasswillimproveestimatesoftheplanet radiusanddensity(andthusbulkcomposition,assumingacertainmassrangeforthe candidateplanet; Valenciaetal. 2007 ).WhileTFimagingisparticularlywellsuitedfor high-precisiontransitphotometryofbrightertargets(e.g., Col onetal. 2012 2010 ),it isnotidealforfainterstars,duetothelongerexposuretimesrequiredtogetahighS/N. TheuseofOSIRIS'sbroaderlters,whichcoveralargerwavelengthregime,isthusone possiblewaytoboosttheS/Ntorejectblendswithbackground,foreground,orphysically associatedstarshostingtransitingplanets. Withtransitsurveyslike Kepler and COROT activelysearchingforandndingnew planetcandidates,itwillbevitaltouseallthetoolsathandtorejectfalsepositivesand determinethetruenatureofthecandidateplanets.Theseobservationsdemonstrate thatmulticolorphotometryfromtheGTCisoneadditionaltoolthatcanhelpwiththe identicationoffalsepositivesinthecomingyears. 142

PAGE 143

Table5-1. NormalizedphotometryofKOI565. BJD-2455000FluxratioUncertainty =790.2nm 459.414941.001120.00094 459.419590.999880.00093 459.424241.001470.00094 459.428900.999010.00096 459.433550.998650.00095 459.438211.000510.00096 459.442860.998840.00099 459.447510.999940.00098 459.452161.000650.00097 459.456821.000720.00098 459.461471.001220.00097 459.466121.001610.00097 459.470771.000090.00097 459.475431.000430.00097 459.480080.998570.00099 459.484730.999850.00097 459.489390.999410.00098 459.494040.999190.00099 459.498691.000700.00100 459.503350.999520.00102 459.508000.998490.00099 459.512651.000500.00098 459.517300.998980.00102 459.521961.000440.00102 459.526611.000170.00101 459.531260.998550.00102 459.535911.002570.00104 459.540570.999450.00105 459.545221.000190.00104 459.549870.999630.00112 459.554530.999800.00104 459.559181.000760.00104 143

PAGE 144

Table 5-1 .Continued BJD-2455000FluxratioUncertainty =794.3nm 459.417260.999360.00093 459.421920.999460.00094 459.426570.998880.00094 459.431230.999440.00095 459.435880.998020.00094 459.440531.000950.00096 459.445181.001290.00100 459.449841.001520.00097 459.454491.000440.00096 459.459141.001390.00100 459.463801.000780.00097 459.468451.000380.00096 459.473101.001390.00096 459.477750.999620.00097 459.482410.999430.00100 459.487060.998420.00096 459.491710.997620.00096 459.496371.000630.00097 459.501021.000850.00099 459.505671.000620.00099 459.510321.001190.00097 459.514981.000910.00102 459.519631.001160.00099 459.524281.001160.00100 459.528931.000040.00100 459.533591.000640.00101 459.538240.999290.00103 459.542891.000830.00111 459.547550.999090.00101 459.552201.001030.00102 459.556850.998290.00101 459.561501.000300.00103 a ThetimestampsincludedherearetheBarycentricJulianDatesinBarycentric DynamicalTime(BJD TDB)atmid-exposure.Theuxratiosincludedherearethose thathavebeencorrectedusingexternalparameterdecorrelationandnormalizedtothe baseline(out-of-transit)uxratios( ¤ 5.2 ). 144

PAGE 145

Table5-2. NormalizedphotometryofKIC7025851. BJD-2455000FluxratioUncertainty =790.2nm 459.414940.999230.00097 459.419591.000360.00096 459.424241.000930.00097 459.428901.000280.00099 459.433550.999610.00098 459.438211.000160.00098 459.442860.998320.00102 459.447510.998260.00101 459.452161.000620.00100 459.456821.000730.00100 459.461471.000410.00099 459.466121.000800.00100 459.470771.001210.00100 459.475430.998730.00100 459.480081.001700.00102 459.484731.000280.00100 459.489390.999670.00100 459.494040.999250.00101 459.498690.999480.00103 459.503351.004730.00104 459.508000.998240.00102 459.512650.994340.00101 459.517300.984030.00106 459.521960.974030.00106 459.526610.961130.00106 459.531260.949220.00107 459.535910.935340.00111 459.540570.916300.00113 459.545220.899720.00113 459.549870.882500.00124 459.554530.864310.00116 459.559180.844280.00118 145

PAGE 146

Table 5-2 .Continued BJD-2455000FluxratioUncertainty =794.3nm 459.417261.000500.00097 459.421920.998550.00097 459.426570.998620.00097 459.431231.000380.00099 459.435881.000110.00098 459.440531.000510.00099 459.445180.999810.00104 459.449840.998320.00100 459.454490.999830.00099 459.459141.001640.00103 459.463801.000590.00101 459.468451.001350.00099 459.473101.001550.00100 459.477751.000540.00100 459.482411.001080.00104 459.487060.998060.00099 459.491710.999260.00099 459.496371.000790.00101 459.501020.998550.00103 459.505670.996830.00102 459.510320.994150.00101 459.514980.988670.00106 459.519630.977850.00103 459.524280.964870.00105 459.528930.952940.00107 459.533590.936190.00108 459.538240.918840.00112 459.542890.902590.00122 459.547550.884290.00112 459.552200.865690.00115 459.556850.845420.00115 459.561500.827190.00118 a ThecolumnsarethesameasinTable 5-1 146

PAGE 147

Figure5-1. NormalizedLCsfornearlysimultaneousobservationsat790.2 2.0nm (circles)and794.3 2.0nm(squares)ofKOI565asobservedon2010 September19.The794.3nmLChasbeenoffsetforclarity.Thevertical dashedlinesindicate(fromlefttoright)thepredictedbeginningofingress andmid-transittime(basedon Boruckietal. 2011b ).Seetheelectronic editionofthe PASP foracolorversionofthisgure. 147

PAGE 148

Figure5-2. ImagefromGTC/OSIRISobservationsat790.2nmcontainingpartofthe eldofviewaroundKOI565.Thisisonlyasmallportionoftheobserved eldofview,sothesixreferencestarsusedinouranalysisarenotshown here.Thetargetislocatedat # =19 h 17 m 26.05 s % =42 % 31 $ 34.3 $$ ,andKIC 7025851islocatedat # =19 h 17 m 26.30 s % =42 % 31 $ 48.9 $$ .Thesmallcircles aroundKOI565andKIC7025851indicatethesizeoftheaperturesusedin ourphotometry( r 23pixels 2.9 $$ ).ThelargercirclearoundKOI565has aradiusof8 $$ andisincludedsimplytoillustratewhichstarsarelocated withinadistance8 $$ fromthetarget,as Boruckietal. ( 2011b )measureda centroidshiftof8 $$ tothenorthofthetarget.Thepositionofthestarcausing thecentroidshiftistypicallyataslightlyfurtherdistance,sothepositionof KIC7025851relativetoKOI565isconsistentwithmeasurementsfrom Kepler .Seetheelectroniceditionofthe PASP foracolorversionofthis gure. 148

PAGE 149

Figure5-3. SimilartoFigure 5-1 ,butforKIC7025851.Whileonlyapartialeclipsewas observed,theminimumdepthoftheLCiscomparablewithwhatisexpected forastellarEB.Seetheelectroniceditionofthe PASP foracolorversionof thisgure. 149

PAGE 150

Figure5-4. Colorsascomputedbetweenthe790.2and794.3nmobservationsof ( a ) KOI565, ( b ) astellarEB(KIC7025851),and ( c ) foran"unresolved"system (thetargetlightcombinedwiththelightfromtheEB).Ineachpanel,the verticaldashedlinesindicate(fromlefttoright)theapproximatebeginningof ingressandthemid-transittime(basedon Boruckietal. 2011b ).The verticalscaleisthesameforeachpanelforeaseofcomparison.Thereisno changeinthecolorseenforKOI565,butfortheEBaswellasthe hypotheticalunresolvedsystem,wemeasureanappreciabledifferencein thecolorduringthetransitevent. 150

PAGE 151

CHAPTER6 CONSTRAININGTHEFALSEPOSITIVERATEFOR KEPLER PLANET CANDIDATESWITHMULTI-COLORPHOTOMETRYFROMTHEGTC The Kepler spacemissionhasdiscovered,todate,61transitingplanetsaswellas over2,000planetcandidatesand2,000eclipsingbinarystars( Batalhaetal. 2012 ; Pr sa etal. 2011 ; Slawsonetal. 2011 ). 1 Withsuchavastnumberofplanetcandidates,it canbedifculttodecidewhichtofocusfollow-upeffortson.Recentstudieshavetried toaddressthisissuebyestimatingthefalsepositiverateforthe Kepler sample.Based onthelistof1,235 Kepler planetcandidatespublishedby Boruckietal. ( 2011b ),ithas beenpredictedthatasmanyas 95%ofthesecandidatesaretrueplanets( Morton& Johnson 2011 ).However,previousstudieshavenottakenintoaccounthowdifferent subsetsof Kepler targetsmayhavedifferentfalsepositiverates.Forexample,thereisa rapidriseinthenumberofdetachedeclipsingbinarystarsthathavebeendiscoveredby Kepler atorbitalperiodsoflessthan 3days,andsuchsystemscanmimicplanetary transits( Pr saetal. 2011 ; Slawsonetal. 2011 ).Thissuggeststhattheremaybe correspondingchangesinthefalsepositiveratewithorbitalperiod.Becausethe probabilityofobservingatransiteventincreasesastheorbitalperiodoftheplanet decreases,manyof Kepler 'splanetcandidateshaveshortperiods.Thusitisnecessary tobecautiouswhenestimatingfalsepositiveratesoverthewhole Kepler sample. WhileobservationalstudieswithwarmSpitzer supportpredictionsoflowfalsepositive ratesovertheentire Kepler sample( D esertetal. 2012 ),biasesintargetselectioncan affectobservationally-constrainedfalsepositiverates( ¤ 6.6 ).Wealsonotethatarecent imagingstudyhasfoundthatnearly42%oftheirsampleof98 Kepler planetcandidate hostshasavisualorboundcompanionwithin6arcsecondsofthetargetstar(Lillo-Box 1 Uptodatecatalogscanbefoundathttp://kepler.nasa.gov 151

PAGE 152

etal.,inpreparation).Suchstudiesemphasizetheneedforfollow-upimagingtoexclude blendscenariosimitatingplanetcandidatesorcontaminatedtransitdepths. Thefalsepositivescenariosweconsiderinthispaperarethosethatresultfrom stellareclipsingbinariesthatareeitherinthebackground(or,inrarecases,foreground) orboundtothetargetstarandarenotalwayseasytoidentifywith Kepler duetotheux fromthedifferentstarsbeingblendedtogetherwithin Kepler 'sphotometricaperture( 6 arcsec).Asdiscussedby,e.g., Col on&Ford ( 2011 ),differenttechniquescanbeusedto eliminatemany,butnotall,blends.Multi-colortransitphotometryisanefcientmethod forrecognizingblendsthatcannotbespatiallyresolved,asmeasuringthetransitdepth indifferentbandpasses(i.e.thetransitcolor)allowsonetotesttheplanethypothesis. Thisispossiblesincethemagnitudeofthetransitcolorchangesaslongastheblended starshavesignicantlydifferentcolors. Col on&Ford ( 2011 )presentedmulti-colortransitphotometryofa Kepler target, KOI565.01,thatwasrstannouncedby Boruckietal. ( 2011a )tobeasuper-Earth-size planetcandidatebutlaterrecognizedasalikelyfalsepositiveduetomeasurementsofa centroidshiftawayfromthelocationofthetargetontheCCDduringtransit( Boruckiet al. 2011b ).In Col on&Ford ( 2011 ),weusednear-simultaneousmulti-colorobservations acquiredusingthenarrow-bandtunablelterimagingmodeontheOpticalSystemfor ImagingandlowResolutionIntegratedSpectroscopy(OSIRIS)installedonthe10.4-m GranTelescopioCanarias(GTC)toconrmthatKOI565.01isindeedafalsepositive, aswebothresolvedastellareclipsingbinary 15arcsecfromthetargetandmeasured acolorchangeinthe"unresolved"target+eclipsingbinarysystem.Thus, Col on&Ford ( 2011 )demonstratedthecapabilityoftheGTC/OSIRISforefcientvettingofplanet candidatesviaitscapabilitiesfornear-simultaneousmulti-colorphotometrywithina singletransitevent. Inthispaper,wepresentobservationsoffour Kepler planetcandidatesspecically selectedtohavesmallradiiandshortorbitalperiods( R p < 5 R and P < 6 days). 152

PAGE 153

Measuringthefalsepositiverateforthisextremesubsetofplanetcandidateswillallow ustotestifthereisacorrelationbetweenthefalsepositiverateanddifferentplanetary andstellarproperties.Asin Col on&Ford ( 2011 ),theobservationspresentedherewere acquiredwiththeGTC/OSIRIS.However,theseobservationsusedbroadbandlters inlieuofthenarrow-bandtunableltersinordertocollectmorephotonsandtoobtain greaterwavelengthcoverage(andtherebyprobegreatercolordifferences).In ¤ 6.1 we discussourtargetselectioncriteria,andwediscussthecorrespondingobservations forourfourtargetsin ¤ 6.2 .In ¤ 6.3 and ¤ 6.4 wedescribeourdatareductionandlight curveanalysisprocedures,andwepresentresultsforeachtargetin ¤ 6.5 .Weincludea discussionofourresultsandhowtheyrelatetothedistributionofeclipsingbinariesthat havebeendiscoveredby Kepler aswellaspreviousestimatesofthefalsepositiverate forthe Kepler samplein ¤ 6.6 .Inparticular,wediscusstheoreticalestimatesfrom Morton &Johnson ( 2011 )andobservationalconstraintsfromstudiesby D esertetal. ( 2012 )and Santerneetal. ( 2012 ).Finally,wesummarizeourresultsandconclusionsin ¤ 6.7 ,and wealsodiscussourplansforfutureobservationsofadditional Kepler targetswiththe GTC. 6.1TargetSelection Recentstudies(e.g., Lissaueretal. 2011b 2012 )havedemonstratedthata majorityoftheplanetcandidatesin Kepler 'smulti-planetcandidatesystemsshouldin factberealplanets.Consideringthesestudies,forourprogramwetargetonlythose candidatesfoundinsinglesystemsasdeterminedby Boruckietal. ( 2011b ).Fromthe sampleofcandidatesinsinglesystems,weselectedtargetsbasedonthefollowing criteria: orbitalperiod ( P ) < 6 days planetradius ( R p ) < 6 R transitdepth(atcenteroftransit; % ) > 500ppm transitduration(rsttofourthcontact; ) < 2.5h 153

PAGE 154

Kepler magnitude(Kp) < 15.5 vettingag > 1 Asdiscussedintheprevioussection,thereisasignicantriseinthenumberof detachedeclipsingbinarystarscomparedtoplanetcandidatesatshortorbitalperiods. ThisisalsoillustratedinFigure 6-1 ,whereweshowhistogramsofthenumberof planetcandidates,detachedeclipsingbinaries,andall"other"eclipsingbinaries (i.e.allbinariesthatarenotlistedasdetachedin Slawsonetal. 2011 )asafunction oforbitalperiod.Therefore,wefocusonconstrainingthefalsepositiverateonly forplanetcandidateswithshortorbitalperiods,wherethepresenceofeclipsing binariesisgreatest.Ourdecisiontofocusonsmallcandidatesisduetotherebeing adominantpopulationofNeptune-sizeorsmallercandidatesinthe Kepler sample. Wesetconstraintsonthetransitdepth,transitdurationand Kepler magnitudedueto limitationsofourobservingtechnique,soastomeasurethetransitdepthandcolor precisely.Furthermore,wewanttodothiswhilemaintainingareasonableobserving cadence,hencethelimitsonthe Kepler magnitudeofourtargets.Finally,weexclude candidatesthathavevettingagsof1,asthosehavebeenpreviouslyconrmedas planets. Alongwiththeaboveconstraintsontheplanetaryandstellarproperties,wehave severalobservationalconstraintsduetolimitationsoftheGTC/OSIRIS.First,werule outcandidatesthathavenoobservabletransitsfromthelocationoftheGTC(which islocatedonLaPalmaattheObservatoriodelRoquedelosMuchachos).Then,we excludethosethatdidnothavemultipletransitsobservableduringbrightorgreytime (duringthe2011observingseason).Wealsoexcludetransiteventsthatoccuroutside altitudesof 35-72degrees,soastoavoidhighairmassandvignettingthatoccursat highaltitudesduetolimitationsoftheGTCdome. 154

PAGE 155

Finally,wedid not checkthe Kepler lightcurves 2 foreachtargetthattallofthe abovecriteriapriortoacquiringobservations.However,aswediscussin ¤ 6.7 ,for futureobservationsweplantocheckthe Kepler lightcurvesforsecondaryeclipsesand V-shapedeclipses,eitherofwhichcouldindicate apriori thatthetargetisnotatransiting planetbutinsteadastellareclipsingbinary. Theseselectioncriterialedustoacquireobservationsoffourplanetcandidates betweenAprilandSeptember2011:KOI( Kepler ObjectofInterest)225.01,420.01, 526.01and1187.01.InTable 6-1 welistsomeofthepropertiesofeachofourtargets asdeterminedby Boruckietal. ( 2011b ). Boruckietal. ( 2011b )aggedKOI225.01as possiblyhavingellipsoidalvariations.KOI1187.01hastheshortestperiodofallthe planetcandidatesannouncedby Boruckietal. ( 2011b ).InFigures 6-2 6-3 and 6-4 we illustratedifferentpropertiesofourtargetsincomparisontothesampleof1,235KOIs andthecorresponding997hoststarsfromwhichourtargetswerechosen.Thesample ofKOIsobservedwithwarmSpitzer ( D esertetal. 2012 )isalsoillustratedinthese guresforcomparison(wediscussthe Spitzer sampleinmoredetailin ¤ 6.6.1 ).Our targetsandthecorrespondingobservationsaredescribedindetailin ¤ 6.2 below. 6.2Observations Weacquiredphotometryofeachtargetandseveralnearbyreferencestarsaround thepredictedtimeofatransitevent.Foreachobservation,weusedtheGTC/OSIRIS toacquirenear-simultaneousmulti-colorphotometrybyalternatingbetweentwo broadband"ordersorter"lters 3 thatwerecustommadeforOSIRIS:666 36nmand 858 58nm.Bothbandpasseswerespecicallychosensoastominimizeeffectsof telluricabsorptionandemissionwhilealsoallowingforamplewavelengthcoverage.In ordertodecreasedeadtime,allobservationsused1 # 1binning,afastpixelreadout 2 Availableathttp://archive.stsci.edu/kepler/ 3 http://www.gtc.iac.es/en/pages/instrumentation/osiris.php 155

PAGE 156

rateof500kHzandasinglewindowlocatedononeCCDchip.Thesizeofthewindow variedforeachobservation,buteachwindowwaslargeenoughtocontainthetarget andseveralreferencestars.WenotethatallobservationsexceptthoseforKOI1187 wereconductedinqueue(service)mode.Inthefollowingsections,wedescribespecic detailsregardingeachtargetanditsrespectiveobservations. 6.2.1KOI225.01 WeobservedthetransiteventofKOI225.01on2011April13underclear conditionsandduringdarktime,withobservationsbeginningat03:10UTandending at06:12UT,duringwhichtimetheairmassrangedfrom 1.66to1.07.Theexposure timewas50s(forbothbandpasses),with 40sofdeadtimefollowingeachexposure. Seeingvariedbetween1.2and2.0arcsecuntilabout05:00UT,atwhichpointthe seeingimprovedto < 1arcsecandaslightdefocuswasintroducedinordertoavoid saturation.Adiffuse,darkbandwaspresentinalltheimages(aresultofaverybright starlocatedjustoutsidetheCCD),sothetargetwaspositionedontheCCDsuchthat itwasoutsidethisregionandalsoavoidedbadpixelsinthecenteroftheCCDchip. Duringtheobservations,thetarget'scentroidcoordinatesshiftedby < 9pixelsinthe x -directionand 2pixelsinthe y -direction.Severalimageswerelostduetotechnical issues,andafewoftheimagestowardstheendoftheobservationswerediscardeddue tothebeginningoftwilight. 6.2.2KOI420.01 Observationsofthe2011September13transitofKOI420.01tookplacefrom21:48 UT(2011September13)to01:30UT(2011September14).Observationsoccurred duringbrighttimeandunderphotometricconditions,andtheairmassrangedfrom1.07 to1.84.Theexposuretimewassetto40s,withacorresponding20sofdeadtime followingeachexposure.Theseeingroughlyrangedfrom1.1-1.8arcsecthroughout theobservations,thoughatthebeginningoftheobservationstheseeingconditions changeddrasticallyenoughfromoneimagetoanotherthatsomeofthereferencestars 156

PAGE 157

weresaturated.Weexplicitlyexcludeanysaturatedreferencestarsfromouranalysis ( ¤ 6.4 ).Theseeingstabilizedandimprovedafterabout23:00UT,enoughsothataslight defocuswasimplementedtoavoidsaturation.Thetarget'scentroidcoordinatesshifted by < 4pixelsineitherdirectionduringtheobservations. 6.2.3KOI526.01 The2011September11transitofKOI526.01wasobservedunderphotometric conditionsandduringbrighttime.Observationsbeganat21:37UTon2011September 11andendedat00:00UTon2011September12.Theairmassrangedfrom1.05 to1.22andtheseeingwasstablebetween0.6and0.8arcsec.Aslightdefocuswas implemented,yieldingadefocusedFWHMof0.8-1.0arcsec.Anexposuretimeof10 swasused,with20sofdeadtimebetweenexposures.AsinobservationsofKOI 225.01( ¤ 6.2.1 ),adarkbandcausedbyabrightstaroutsidetheCCDwindowwas presentinallimages,soweplacedthetargetstarappropriatelyfarawayfromtheband sothatthephotometrywouldnotbeaffected.Thecentroidcoordinatesofthetarget shiftedbylessthan4and2pixelsinthe x -and y -directions,respectively.Twiceduring theobservationstheprimarymirrorsegmentslostalignmentandproduceddistorted images:rstaround22:30UTandagainat00:00UT.Wediscardedafewimagesfrom therstinstance,buttheissuewiththemirrorcouldnotbereadilyxedduringthe secondinstance,andtheobservationswereforcedtoendearly,aroundthetimeofthe transitegress. 6.2.4KOI1187.01 Weobservedthe2011June12transiteventofKOI1187.01,withobservations takingplacefrom22:31UT(2011June11)to02:01UT(2011June12).Observations tookplaceduringbrighttimeandunderclearconditions,andtheairmassrangedfrom 1.90to1.05.Seeingwasexcellentandrangedfrom0.6to0.9arcsecduringthenight. Duetotheexcellentseeingconditions,theexposuretimewassetto5s.Therewere24 157

PAGE 158

sofdeadtimefollowingeachexposure.Duringtheobservations,thetarget'scentroid coordinatesshiftedby < 4pixelsineitherdirection. 6.3DataReduction Weusedstandard IRAF proceduresforbiassubtractionandat-eldcorrection (usingdomeatstakenforeachbandpass)foreachtarget.Duetononuniform illuminationbythelamp,weaddedanilluminationcorrectiontothenalat-eldimage foreachtarget(exceptKOI225).Theilluminationcorrectionwasperformedusingthe IRAF task mkillumat withinthe noao imred ccdred package,whichremovesthelarge scaleilluminationpatternfromtheateldbysmoothingtheateldimage.ForKOI 225,thenalat-eldforthe666nmltershowedastronggradientafterperforming anilluminationcorrection,sowechosetousetheat-eldimageasitwas.Tobe consistent,wedidnotperformanilluminationcorrectiononthe858nmnalat-eld forKOI225either.WeusedtheIDLAstronomyUser'sLibrary 4 implementationof DAOPHOT ( Stetson 1987 )toperformaperturephotometryoneachtargetandseveral nearbyreferencestarsaswellasonotherpotentialsourcesofthetransitsignal(i.e. starswithin 20arcsecofeachtarget).Foreachtarget,wetesteddifferentapertures andchoseanalaperturebasedonthatwhichresultedinthesmallestscatterinthe baseline(out-of-transit)uxratios(i.e.,thetargetstaruxdividedbytheensemble referencestarux).ForKOI225,420,526and1187ournalaperturewas25,23,15 and14pixels(equivalenttoapproximately3.2,2.9,1.9and1.8arcsec).Skybackground subtractiontookplaceduringtheaperturephotometryprocess,withannulichosento befarenoughawayfromeachstarthattheuxfromagivenstarwouldnotbeincluded withintheskyannulus.Whiletheseprocedureswereperformedforeachbandpass separately,weconsideredtheresultsforeachbandpassandusedthesameaperture andskyannulusforeachdatasetforagiventarget.Notethatinmanycasesasmaller 4 http://idlastro.gsfc.nasa.gov/ 158

PAGE 159

apertureandskyannuluswereusedforthestarswithin 20arcsecofagiventarget duetothesmallseparationsbetweensomeofthestarsandthepotentialforblending. Forreference,wepresentareducedscienceimageshowingtheeldofviewaround KOI1187inFigure 6-5 .Onceaperturephotometryandskysubtractionwerecompleted, weproceededwithlightcurveanalysisasdiscussedin ¤ 6.4 below. 6.4LightCurveAnalysis Foreachtarget,wecomputedlightcurvesforeachbandpassbydividingthetotal uxmeasuredfromthetargetbythetotalweighteduxofanensembleofreference stars,wheretheuxfromeachreferencestarwasweightedaccordinglyafterdiscarding outlyinguxvalues(resultingfromeithertechnicalissuesorvariableskyconditions). Duringtheanalysis,severalreferencestarswerefoundtobeobviouslyvariableor saturated,soweexcludedthosestarsfromtheanalysis.Thisresultedinususing10, 4,2and9referencestarsintheeldnearKOI225,420,526and1187tocompute thereferenceensembleux.Next,eachlightcurveforeachtargetwasnormalizedto themeanbaselineuxratioasmeasuredineachbandpass.Thelightcurveswere thencorrectedforchangesinairmassaswellasfordriftsinthecentroidcoordinates ofthetargetontheCCDandthesharpnessofthetarget'sprole[(2.35/FWHM) 2 ].The lattercorrectionsweredoneviaexternalparameterdecorrelation(e.g., Bakosetal. 2007 2010 ).WenotethatduetothelackofdataforKOI526duringthetransitegress andpost-transit,ourattemptstocorrectthoselightcurvesproducedskewedresults. Therefore,allfurtheranalysisforKOI526isbasedontheuncorrectedlightcurves. Thecorrespondingobservationtimesforeachlightcurveforeachtargetwere computedfromtheUTCtimestampsgivenintheimageheaders.Weconvertedthe UTCtimeatmid-exposuretoBarycentricJulianDatesinBarycentricDynamicalTime (BJD TDB)viaanonlinecalculatordescribedin Eastmanetal. ( 2010 )inordertomatch 159

PAGE 160

thetimecoordinatesystemthattheephemeridesfrom Boruckietal. ( 2011b )aregiven in. 5 Thephotometricuncertaintieswerecomputedfromthephotonnoiseofthetarget starandthereferencestarensemble,thenoiseintheskybackgroundaroundthetarget andeachreferenceusedintheensemble,andscintillationnoise.Theresultingmedian photometricuncertaintiesforeachlightcurveforeachtargetaregiveninTable 6-2 Ineachcase,thephotonnoiseofthetargetistheprimarysourceoferror.Finally,in ordertoinvestigatethepresenceofrednoiseinourdata,wecomputedthestandard deviationoftheuxratiosfordifferentbinsizesandfoundthatallourdatafollowthe trendexpectedforwhite(Gaussian)noise.Despitethisresult,wealsoinvestigate potentialresidualsystematicsviaa"prayer-bead"analysis,whichwediscussbelow. Aftercorrectingthelightcurvesforeachtarget(exceptforKOI526asdiscussed above),weproceededtotsyntheticmodelstoeachlightcurve,followingtheapproach takenby Col onetal. ( 2010 ).Weassumedeachobservedtransiteventwasaplanetary transit,andweusedtheplanetarytransitlightcurvemodelsfrom Mandel&Agol ( 2002 ) totlimb-darkenedmodelstoourdata.Specically,foreachtarget,wetforthe followingparameters: timeofmid-transit( t 0 ) transitduration(rsttofourthcontact; ) impactparameter( a cos i / R ) planet-starradiusratio( R p / R ) twolimbdarkeningcoefcients( c 1 and c 2 ) baselineuxratio (linear)baselineslope 5 Thecalculatorisavailableathttp://astroutils.astronomy.ohio-state.edu/time/utc2bjd.html 160

PAGE 161

Thelimbdarkeningcoefcientsthatwetforaredenedas c 1 % u 1 + u 2 and c 2 % u 1 $ u 2 where u 1 and u 2 arelinearandquadraticlimbdarkeningcoefcients.ForKOI526,we notethatweheldthetransitdurationxedtothevaluefrom Boruckietal. ( 2011b )in ordertotamodeltoourpartialtransit. Initialguessesforthemid-transittime,transitduration,impactparameterand radiusratiowerebasedonthevaluesandtheircorrespondinguncertaintiesasgiven in Boruckietal. ( 2011b ).Valuesforthelimbdarkeningcoefcientswereinterpolated fromthe Claret&Bloemen ( 2011 )modelsfortheSloan r $ and z $ ltersandarebased onthestellarparametersgivenin Boruckietal. ( 2011b ).Duetothepotentiallylarger thanestimateduncertaintiesinthestellarparameters(e.g., Boruckietal. 2011b ),we didnotallowthelimbdarkeningcoefcientstobecompletelyfreeparametersinthe ttingprocess.Rather,wekeptthecoefcientsxedatself-consistentvaluesduring thettingprocess,butwetestedarangeofxedvaluesforthecoefcients.Similarly, wetestedarangeofinitialguessesfortheotherparametersbasedontheuncertainties fortheplanetaryandstellarparameters.Best-ttingmodelswereidentiedviaa Levenberg-Marquardtminimizationscheme. 6 Ourlightcurvettingprocedureforeachtargetisasfollows.First,wetmodelsto eachlightcurveindividually,correctedthedataagainstthebest-tbaselineuxratioand slope,subtractedthebest-tmodelsanddiscardedanydatapointslyinggreaterthan3 fromtheresiduals.Then,weusedthecorrectedlightcurvesandtmodelstothemina jointanalysis,whereweforcedtheimpactparameter,transitduration,mid-transittime, baselineuxratioandbaselineslopetobethesameforbothlightcurves.However,we alloweddifferentvaluesfortheradiusratioandlimbdarkeningcoefcientstobettedto thedifferentlightcurves.Theresultsfromthejointanalysiswerethenusedtocorrect 6 Wespecicallyused mptfun ,whichispubliclyavailableathttp://www.physics.wisc.edu/ craigm/idl/idl.html 161

PAGE 162

theindividuallightcurvesasintherststepdescribedabove.Analjointanalysis wasthenappliedtothe"nal"correctedlightcurves.Duringthisnalstage,wealso performeda"prayer-bead"analysisaswasdonein Col onetal. ( 2010 ).Specically,we performedacircularshiftontheresidualsforeachlightcurve(computedafterremoving thebest-tmodel)andconstructedsyntheticlightcurvesbyaddingtheshiftedresiduals backtothebest-tmodel.Thejointanalysisdescribedabovewasappliedtoeach syntheticlightcurve,andweusethedispersionofthebest-tparameterstocalculate uncertaintiesoneachparameter.Thisaccountsforanyadditionalsystematicnoise sourcesinthedata.Wepresentresultsfromourlightcurveanalysisin ¤ 6.5 6.5Results Wepresentthelightcurvesandthecorrespondingbest-tmodelsforeachtargetin Figures 6-6 6-7 6-8 and 6-9 .Whilenotshownhere,wehadalsogeneratedlightcurves forthepotentialsourcesofeachtransitsignal(i.e.starsthatcouldhavebeenblended withthetargetwithin Kepler 'saperture)followingasimilarprocedureasdescribed above.Wegeneratedlightcurvesfor10,3,3and6starswithin 20arcsecofKOI225, 420,526and1187.Uponvisualinspection,wefoundthatnoneoftheselightcurves showedatransitsignalduringthetimeofthetransitevent,indicatingthatthetransits thatweobservedeitheroccurduetoanobjecttransitingthetargetstaroranobject transitinganunresolvedstarthatisblendedwiththetargetstar.Wealsopresentthe transitcolor(666nm $ 858nm)ofeachtargetinthebottompanelofFigures 6-6 $ 6-9 whichwascomputedbytakingtheaverageofeachpairofuxratiosinthe666nmlight curveanddividingbythecorrespondingpointsinthe858nmlightcurve.Therefore, inFigures 6-6 $ 6-9 ,apositivecolorindicatesareddertransitandanegativecolor indicatesabluertransit. InTable 6-3 ,wereportbest-tmodelparametersforeachtargetbasedonthe modelwiththesmallest $ 2 value,butwenotethattheparameteruncertaintiesare basedonthefullsetofbest-tvaluesandtheircorrespondinguncertaintiesasfound 162

PAGE 163

duringthe"prayer-bead"analysis.Specically,foragivenparameter,theuppererrorbar isaresultofsubtractingthebest-tvalue(fromthemodelwiththesmallest $ 2 value) fromthemaximumsumofattedvalueanditsassociatedmeasurementuncertainty asdeterminedfromthefullsetofmodelscomputedduringthe"prayer-bead"analysis, andlikewiseforthelowererrorbar.Ingeneral,wendthattheformal1 errorsforthe best-tparametersarecomparabletothoseerrorscomputedbasedontheresultsfrom the"prayer-bead"analysis,whichindicatesthatanyresidualsystematicsinthedata(i.e. thosenotremovedbyexternalparameterdecorrelationorairmasscorrections)havea negligibleeffectonourresults.Furthermore,thedistributionofeachbest-tlightcurve parameteroverallpermutationsissmallerthantheuncertaintiesgiveninTable 6-3 whichfurtherindicatesthattheresultspresentedherearerobust. Finally,assuggestedbythereferee,weconsiderwhetherwecanuseastrometry fromourimagestoprovideadditionalconstraintsonthepropertiesofthetargetswe observed.ApreliminaryastrometricanalysisdoesnotsuggestthatanyoftheKOIswe observedareduetoablendwithanotherstarthatwasresolvedbyourobservations. Furthermore,despitethelargeapertureusedby Kepler ,theastrometricprecisions from Kepler areextremelyhigh( < 0.004arcsecondsforasingle30minexposure; Monetetal. 2010 ),andtothebestofourknowledge,nosignicantcentroidshiftswere measuredby Kepler foranyofourKOIs. Wediscussspecicresultsforeachtargetindividuallyinthefollowingsections. 6.5.1KOI225.01 ThelightcurvespresentedinFigure 6-6 areprominentlyV-shaped,whichalone suggestsapossiblenon-planetarytransitevent.Consideringtheshapeofthelight curvesandthesignicantdifferenceinthetransitdepthsasmeasuredintheGTC bandpasses,wededucethatKOI225.01ismostlikelyastellareclipsingbinarysystem. Specically,wendthatthebest-tplanet-starradiusratiosdifferatalevelof > 11 whichisaclearindicationthatthisobjectisnotaplanet.Thetransitdepthinthe Kepler 163

PAGE 164

bandpassisalsosignicantlydifferentfromthosemeasuredintheGTCbandpasses. Inparticular,thetransitdepthisnotablydifferentbetweenthe Kepler bandpassandthe blueGTCbandpass(around666nm),eventhoughtheyprobefairlysimilarwavelength regimes.Weprimarilyattributethisdifferencetodilutioninthe Kepler photometry,since thereisarelativelybrightstarnearKOI225thatappearstohaveslightlycontaminated oneofthepixelsusedinthe Kepler photometry. 7 Asnotedin Boruckietal. ( 2011b ),this target'slightcurvehadpossibleellipsoidalvariations,alsosuggestingthatthesystem containsaneclipsingbinary.Furthermore,afterourobservationshadbeenconducted, Slawsonetal. ( 2011 )listedthistargetintheireclipsingbinarycatalog. 8 Finally, Or& Dreizler ( 2012 )recentlyconductedanindependentanalysisofthe Kepler datasetand identiedKOI225.01asafalsepositiveduetothedetectionofsignicantdifferences betweentheoddandeveneclipseevents. SincewemeasureadeepereclipseinthebluerGTCbandpass,thisindicatesthat duringtheeclipsemorebluelightisblocked,andthereforethesecondary(eclipsing) componentisredderthantheprimarystar.InTable 6-3 ,weprovideanupdated ephemerisandeclipseduration,thoughwenotethatlikelyduetothenon-planetary transitlightcurveshape,thebest-tmodelyieldedanimpactparameterthatreached theupperboundarylimitsduringthettingprocess.Therefore,theresultingbest-tlight curveparametersarelikelysomewhatskewed,butthisdoesnotchangethesignicance ofourresults. Slawsonetal. ( 2011 )reportaperiodthatistwiceaslongasthatfoundby Borucki etal. ( 2011b ).Ifcorrect,theneitherwhat Boruckietal. ( 2011b )believedweretransit 7 Asdeterminedfromtargetpixellesdownloadedfromhttp://archive.stsci.edu/kepler/ 8 Wereferthereaderto Slawsonetal. ( 2011 )forfurtherdetailsonhowsomeKOIs wererejectedasplanetcandidatesandsubsequentlyaddedtotheeclipsingbinary catalog. 164

PAGE 165

eventswereactuallyacombinationofprimaryandsecondaryeclipses or theeclipsing binaryhasalargeenoughinclinationthatalleclipseeventsareprimaryeclipsesandthe secondariesareunobservablefromourline-of-sight.Ifweassumethattheperiodfound by Slawsonetal. ( 2011 )iscorrect(andwasdetermined,forinstance,bymeasuring differentdepthsforsuccessivetransits,whichledtothisKOIbeingrejected),thenitis necessarytoconsiderhowthisaffectsanycorrelationbetweentheorbitalperiodandthe falsepositiveratefor Kepler targets.Wereferthereaderto ¤ 6.6 forfurtherdiscussion. 6.5.2KOI420.01 AsillustratedinFigure 6-7 ,thetransitdepthsasobservedintheGTCbandpasses arecomparable,andthereisnotasignicantchangeinthemeasuredcolorduring transit.Theplanet-starradiusratiosmeasuredfromtheGTClightcurvesareconsistent within 2.8 ,andwedidnotresolveanypotentialsourcesofthetransitsignalwithin 20arcsecofthetarget,sowededucethatthistargetis not afalsepositiveandis insteada validated planet. 9 Thebest-tparametersfromourlightcurvemodelsare giveninTable 6-3 andincludeanupdatedephemeris,transitduration,andimpact parameter.Wenotethattheimpactparametergivenin Boruckietal. ( 2011b )hasan associateduncertaintyof1,and Batalhaetal. ( 2012 )reportanimpactparameterof0.57 witharelativelylargeuncertaintyof0.52,soweprovideamuchstrongerconstrainton thisparameterhere.Finally,basedonourmeasuredplanet-starradiusratiosandthe revisedstellarradius(0.69 R # comparedtothevalueof0.83 R # reportedby Borucki etal. 2011b )givenin Batalhaetal. ( 2012 ),wendKOI420.01tohavearadiusthatis between 4.15and4.87 R (slightlylargerthanthevalueof3.65 R foundby Batalha etal. 2012 ). 9 Weusetheterm validated tomeanthattheplanetcandidateismostlikelya planetbasedonourobservationsbutitisnota conrmed planetbecausethereisnot independentevidence(e.g.,Dopplerortransittimingvariations)fortheplanetmodel. 165

PAGE 166

6.5.3KOI526.01 DespiteonlyobservingapartialtransitwiththeGTC(asillustratedinFigure 6-8 ), wendthatthetransitdepthsclearlymatcheachother,withplanet-starradiusratios thatareconsistentwithin 1.2 .Whilethe Kepler depthismuchshallower,there appearstobenosignicantcontaminationfromnearbystarsinthe Kepler photometry (seefootnotein ¤ 6.5.1 ).Giventhateventhetransitdepthsinthe Kepler andtheblue GTCbandpassaresignicantlydifferent,andthatdilutionappearstonotbethesource, webelievethisobjectwarrantsfurtherinvestigationbeyondthescopeofthispaper. Forinstance,itispossiblethatstellarvariabilityinthetargetstarorablendedstar impactedboththeGTCand Kepler 'smeasuredtransitdepths. 10 Regardless,since thedepthsbetweentheGTCbandpassesareconsistent,thistargetstillpassesour validationtest.AsinthecaseforKOI420.01,wefoundnostarswithin 20arcsec ofthetargetthatshowedatransitsignalattheexpectedtimeofthetransit,sowe also validate KOI526.01asaplanet.Table 6-3 includesanupdatedephemerisand impactparameter(recallthatthetransitdurationwasheldxedduetottingonlya partialtransit).Again,asforKOI420.01, Boruckietal. ( 2011b )foundanuncertainty of1ontheimpactparameter,and Batalhaetal. ( 2012 )reportanimpactparameterof 0.80 0.34,soourobservationsandmodelsprovideamuchstrongerconstraintonthe impactparameter,whichwendhasanassociateduncertaintyof0.0771.Basedon ourmeasuredplanet-starradiusratiosandassumingastellarradiusof0.92 R # (from Batalhaetal. 2012 ,andslightlylargerthanthevalueof0.80 R # reportedby Boruckiet al. 2011b ),wecalculatethatKOI526.01hasaradiusbetween 6.37and7.16 R .We notethatthisisovertwicetheradiusof3.11 R measuredby Batalhaetal. ( 2012 ). 10 Kepler lightcurvesforthistargetshowbaselinevariabilityatalevelof 1%. 166

PAGE 167

6.5.4KOI1187.01 AsinthecaseofKOI225.01,thelightcurvesforKOI1187.01(showninFigure 6-9 )appeartobefairlyV-shaped.Furthermore,visualinspectionoftheGTClight curvesshowsthatthereisclearlyasignicantdifferenceintheGTCtransitdepths. However,wendthatthebest-tplanet-starradiusratiosonlydifferatalevelof 1.8 ,likelyduetothelargeuncertaintyintheradiusratiomeasuredforthe858 nmlightcurve.Weattributethislargeuncertaintytoacombinationoftherelatively poorphotometricprecisionsachievedforthistargetcombinedwiththedegeneracy betweentheimpactparameter(measuredtobenearlyequalto1)andtheplanet-star radiusratio.Toreconcilethesemeasurementswithwhatwendvisually,wecompute weightedmeancolorsandtheiruncertaintiesforthein-transitandout-of-transitdata. Wendameanin-transitcolorof-0.00425 0.00050andameanout-of-transitcolorof -0.000160 0.000212,whichdiffersignicantlyatalevelof 5.8 .Therefore,despite theconsistentmeasuredradiusratios,thereisanobviouslysignicantcolorchange duringthetransitevent,sowearguethatKOI1187.01isinfactafalsepositiveandnota planet. ContrarytoKOI225.01,thetransitofKOI1187.01isdeeperintheredderGTC bandpassthaninthebluerGTCbandpass.Thisimpliesthatduringtheeclipse,more redlightisblockedthanbluelight,andthesmaller(eclipsing)componentisbluerthan theprimarystar.Inthiscase,KOI1187mayconsistofanevolvedgiantstarthatis redderandseveralmagnitudesbrighterthantheeclipsingstar. Slawsonetal. ( 2011 ) alsolistthistargetintheeclipsingbinarycatalogasarejectedKOI(whichwewere alsonotawareofpriortoobservingthistarget).Also,justasforKOI225.01, Slawson etal. ( 2011 )foundanorbitalperiodforKOI1187.01thatistwiceaslongasthatfound by Boruckietal. ( 2011b ).Asdiscussedin ¤ 6.5.1 aboveandbelowin ¤ 6.6 ,suchresults haveramicationsonanycorrelationbetweentheorbitalperiodandfalsepositiverate. 167

PAGE 168

6.6Discussion Ofthefour Kepler planetcandidatespresentedinthispaper,weidentiedtwo asfalsepositivesandprovidefurtherevidencesupportingtheplanethypothesisfor twocandidates.Thissuggestsafalsepositiveratethatismuchhigherthanhasbeen previouslypredictedfortheBoruckietal.(2011b)KOIcatalog(Morton&Johnson2011; ¤ 6.6.1 ),soweconsiderwhatcouldcausesuchalargefalsepositiverateforoursample. ReferringbacktoFigure 6-1 ,weseethatourtwofalsepositives,KOI225.01and KOI1187.01,haveshorterorbitalperiodsthanthetwoKOIswevalidatedasplanets. Therefore,ourndingssuggestthatthefalsepositiverateforthe Kepler samplevaries signicantlywithorbitalperiodandislargestattheshortestperiods( P < 3days),which iswhatonewouldexpect apriori duetotheriseinthenumberofdetachedeclipsing binariesattheseshortperiods(Figure 6-1 ). Wendnosignicantcorrelationbetweenthefalsepositiverateandplanetradius orotherpropertiesofthehoststar.Figure 6-2 illustratesthatthereseemstobenotrend inthefalsepositiveratewithplanetradius,astheKOIsweidentiedasfalsepositives havethesmallestandlargestapparentradii(asmeasuredby Kepler )inoursample. Similarly,thereappearstobenotrendwiththeeffectivetemperatureofthehoststar, althoughwenotethatbothfalsepositiveswerealsothefaintesttargetsinoursample, asillustratedinFigure 6-4 ( ¤ 6.6.1 ).Finally,weemphasizethatbasedonthestellar parametersfromtheKIC,allourtargetsarelikelyFGKdwarfs,whicharetheprimary targetsofthe Kepler mission. Onefactorthatmustbeconsideredwhenmakingsuchconclusionsisthatwe donottakeintoaccountthelevelofuncertaintyinthestellarparametersthatare drawnfromtheKIC.Anyuncertaintyinthestellarparameterswouldobviously affectthedistributionoftheplanetaryradiiandstellarmagnitudesandtemperatures discussedhere.However,ifthestellarradiiintheKIChavebeensystematicallyoverorunder-estimated,thenallplanetaryradiiwouldsimplyscaleupordown,andour 168

PAGE 169

conclusionsaboutthelackofacorrelationbetweenthefalsepositiverateandplanet radiuswouldremainthesame. 11 Additionally,thereistheissuethatifagivenplanetcandidateisactuallyaneclipsing binarywithanorbitalperiodthatistwiceaslongaswasinitiallyexpected,thenthis wouldimplythatwearenotprobingplanetcandidateswithorbitalperiodsofless than6days,butinsteadweareactuallyprobingasampleofeclipsingbinarieswith periodsaslongas12days.However,sincethereisstillacomparablepopulationof eclipsingbinariesandplanetcandidateswithlongerperiods(e.g.outtoatleast10 days,asillustratedinFigure 6-1 ),webelievethisperioddiscrepancywouldnotaffect ourconclusionthatapopulationofshort-periodeclipsingbinariescansignicantly contaminateshort-periodplanetcandidates.Candidatesthatareinitiallyidentied asplanetsbutareactuallyeclipsingbinarieswilltendtohaveshorterapparent periods(theirperiodswilldoublewhentheyareidentiedaseclipsingbinaries),which strengthensourargumentforcontaminationoftheplanetsampleatshortperiods,since thiseffectwouldfurtherreducetheratioofplanetstoeclipsingbinariesatperiodsofless thanapproximately3days. Finally,wenotethatwiththetechniquepresentedhere,itisnotpossibleforusto identifycaseswhereaplanetcandidateisactuallyabinarystarcomposedoftwostars thathavethesametemperature,astherewouldbenomeasurablecolorchangeduring thetransitevent.Thissuggeststhatsomeofourpotentiallyvalidatedplanetscouldstill befalsepositives.However, Kepler 'sphotometryshouldbeabletodistinguishifthe transitdepthsdifferbetweeneveryothertransit(i.e.odd-eventransitdepths),which wouldidentifyacandidateasafalsepositive.Incaseswhereasignicantodd-even 11 Somehighsignal-to-noiseratioeclipsingbinariesaremorelikelytohavecorrectly determinedradii,sosomeshort-periodplanetcandidatesthatareactuallyeclipsing binariesarecorrespondinglymorelikelytohaveaccurateradii. 169

PAGE 170

ratioismeasured,thiscanalsobeusedtoreconciletheissuewiththeorbitalperiods describedabove.Tothebestofourknowledge,nosignicantodd-evenratiowasfound forourvalidatedplanets,KOI420.01andKOI526.01. Recently,anewlistof Kepler planetcandidateswasannounced,bringingthe totalnumberofcandidatesto2,321( Batalhaetal. 2012 ).AsillustratedinFigures 6-1 $ 6-4 and 6-10 ,thenewcatalogfollowsthesamegeneraldistributionoforbital periods,planetradii,stellarmagnitudes,temperatures,andGalacticlatitudesasthe Boruckietal. ( 2011b )catalog,butthereisnotablyagreaternumberofsmallerand shorter-periodplanetcandidatesinthenewcatalog.Thisservestoemphasizethe needtoobservationallyconstrainthefalsepositiverateforsuchsmall,short-period planets.Wenotethatthisnewcatalogwasgeneratedusingimprovedvettingmetrics,so itshouldhaveahigherdelitythanpreviousones(i.e.,theBoruckietal.2011bcatalog). However,theKOIsfromthe Boruckietal. ( 2011b )cataloghavenotbeenvettedagainst theimprovedmetricsdescribedin Batalhaetal. ( 2012 ),sobothKOI225.01andKOI 1187.01arestillincludedintheupdatedcatalogaspotentialplanetcandidates.Also, Batalhaetal. ( 2012 )remarkthatpotentialKOIsarenotvettedagainsttheshapeoftheir lightcurves,soplanetcandidateswithV-shapedtransitlightcurves(suchasKOI225.01 andKOI1187.01)arenotimmediatelyrejected.Givenongoingparalleleffortstoidentify planetcandidatesandKOIsandthat Slawsonetal. ( 2011 )previouslyidentiedbothKOI 225.01andKOI1187.01aseclipsingbinaries,werecommendthattheeclipsingbinary catalogbeconsultedbeforefollowingupanygivenplanetcandidate. 6.6.1ComparisontoTheoreticalStudies Arecentstudyby Morton&Johnson ( 2011 )providedtheoreticalestimatesofthe falsepositiveratefor Kepler planetcandidates. Morton&Johnson ( 2011 )specically estimatedthatnearly90%ofthe1,235candidatespresentedby Boruckietal. ( 2011b ) hadafalsepositiveprobabilityoflessthan10%.Forourtargetsinparticular, Morton& Johnson ( 2011 )computedthefollowingfalsepositiveprobabilities:0.01(KOI225.01), 170

PAGE 171

0.04(KOI420.01),0.03(KOI526.01)and0.03(KOI1187.01).Itisinterestingtonote thatthetargetwiththelowestfalsepositiveprobabilityisonethatendedupbeingafalse positive.However,wenotethatthefalsepositiveprobabilitiesfrom Morton&Johnson ( 2011 )areonlyvalidfornonV-shapedtransitsignals.Therefore,consideringthatboth KOI225.01andKOI1187.01appeartohavesomewhatobviouslyV-shapedtransits, thelowfalsepositiveprobabilitiescomputedforthesetargetsby Morton&Johnson ( 2011 )arenotappropriate.Arecentstudyby Morton ( 2012 )improvesupontheanalysis from Morton&Johnson ( 2011 )bytakingthetransitshapeintoaccountwhencomputing thefalsepositiveprobabilityforagiventarget.Fromtheirnewanalysis, Morton ( 2012 ) calculateafalsepositiveprobabilityof > 0.99forKOI225.01and0.76forKOI1187.01, whichisconsistentwithourobservations.Thecaveataboutthetransitshapealso affectstheoverallassumedfalsepositiverateforthe Kepler sample,as Morton& Johnson ( 2011 )didnotseparateplanetcandidateswithV-shapedtransitsignalsfrom thosewithnon-V-shapedtransitsignals.Thismeansthatweshouldnotinterprettheir ndingsas90-95%of Kepler planetcandidatesareplanets.However,wecanusetheir averagedfalsepositiveprobabilitiesasastartingpointfortheoverallfalsepositiverate forthe Kepler sample.Despiteanycaveats,weestimatetheprobabilityofdetecting2 or3falsepositivesoutof4or5targetsobserved.Assumingabinomialdistributionand afalsepositiverateof10%for Kepler planetcandidates( Morton&Johnson 2011 ),we estimatethattheprobabilityofdetecting2falsepositivesfromasampleof4targetsis lessthan1%(andlessthan5%fordetecting3falsepositivesfrom5targets).Thus,we usethistoemphasizethatthefalsepositiverateforcurrent Kepler planetcandidates withradiilessthan 5 R andorbitalperiodslessthan 6.0daysislikelymuchhigher than10%. Inregardstothefull Kepler sample, Morton&Johnson ( 2011 )foundthatthe falsepositiveprobabilityvariedwiththedepthofthetransitaswellasthemagnitude andGalacticlatitudeofthetargetstar,buttheydidnotinvestigatehowtheprobability 171

PAGE 172

mightvarywithorbitalperiodortransitduration(thoughtheyclaimthatsuchproperties wouldservetodecreasetheirprobabilityestimatesandthustheirestimatesareupper limits).Specically,ourresultssupporttheirndingthatthefalsepositiverateincreases slightlywithmagnitude,asourtwofaintesttargetsendedupbeingfalsepositives. Theyalsondageneralincreaseinthefalsepositiveprobabilitywithincreasingtransit depth(thoughtherearesomelocalminimaandmaxima;e.g.,Figure7inMorton& Johnson2011)andwithdecreasingGalacticlatitude.Inourcase,thetargetwiththe largesttransitdepth(asmeasuredinthe Kepler bandpass)wasKOI420.01,which wevalidatedasaplanet.Whilewendnoobviouscorrelationinoursamplebetween Kepler 'smeasuredtransitdepthsandthefalsepositiverate,dilutionisanimportant factorthathastobeconsidered(e.g.,asillustratedbyourlightcurvesforKOI225.01 inFigure 6-6 ),assomeofthedepthsmeasuredby Kepler maybeunderestimated.In regardstoGalacticlatitude,inFigure 6-10 weshowthecumulativedistributionofthe Galacticlatitudesforthe Boruckietal. ( 2011b )KOIlist,the Batalhaetal. ( 2012 )KOI listandthe Slawsonetal. ( 2011 )catalogofeclipsingbinaries,whichillustratesthat thesampleshavethesamegeneraldistribution.Furthermore,ourfalsepositivesare indeedlocatedatlowerlatitudes,whichsupportstheargumentby Morton&Johnson ( 2011 )thatthereisaslightincreaseinthefalsepositiveratewithdecreasingGalactic latitude.WepresentasimilargureinFigure 6-11 ,whereweshowthecumulative distributionoftheGalacticlatitudesforthe Kepler eclipsingbinariesalongwithseparate distributionsforstarsthathostshort-period( P ) 2days)andlong-period( P > 2days) planetcandidates.Weseethatthedistributionforthestarsthathostshort-period candidatesmatchestheeclipsingbinarydistributionsmoresothanthesampleof long-periodcandidates.Therefore,Figure 6-11 furtherillustratesthelikelycontamination ofshort-periodplanetcandidatesbyeclipsingbinaries.InFigure 6-12 ,weillustratehow thenumberofeclipsingbinariesvarieswithGalacticlatitude,andwendthatoverallthe fractionof Kepler targetsthatareeclipsingbinariesisaboutthesameasafunctionof 172

PAGE 173

Galacticlatitude.Thisfurtherimpliesanearlyconsistentpresenceofeclipsingbinaries thatcouldplaguethe Kepler planetcandidatesample. 6.6.2ComparisontoObservationalStudies D esertetal. ( 2012 )haveusedwarmSpitzer follow-upofKOIstondalowfalse positiverateconsistentwiththeestimatesfrom Morton&Johnson ( 2011 ).Overall,our sampleincludesshorterperiodplanetsthantheirsample,aswellasfaintertargets (excludingtheirsampleofMdwarfstars).Giventhatonlyplanetcandidateswiththe shortestorbitalperiods( << 2days)arelikelytobesignicantlycontaminatedby non-detachedeclipsingbinaries,itisnotentirelysurprisingthat D esertetal. ( 2012 ) foundalowfalsepositiveratefortheirsampleofplanetcandidateswithperiodsof 1.8 $ 100days. Recently, Santerneetal. ( 2012 )presentedaDopplerstudyofshort-periodgiant Kepler planetcandidates.Theytargetedasampleof46 Kepler planetcandidateswith transitdepthsgreaterthan0.4%,orbitalperiodslessthan25days,andhoststars brighterthan14.7magnitude(Kp).Basedontheirradialvelocityfollow-upobservations, Santerneetal. ( 2012 )estimateafalsepositiverateof34.8 6.5%forthissubsetof Kepler planetcandidates.Whileweprobedadifferent(andsmaller)populationof Kepler planetcandidates,ourobservedfalsepositiverateof50%(withlargeuncertaintydueto oursmallsamplesize)forsmall,short-period Kepler planetcandidatesissupportedby theirobservations.Thestudyby Santerneetal. ( 2012 )alsosupportsourargumentthat differentpopulationsof Kepler targetslikelyhavedifferentfalsepositiveratesassociated withthem. 6.7Conclusion Weacquiredmulti-colortransitphotometryoffoursmall( 2.5 $ 4.9R ),short-period ( 0.37 $ 6.0days) Kepler planetcandidateswiththeGTC.Basedonthetransitcolor, weidentiedtwocandidatesasfalsepositives.Fortwo,wendfurtherevidence supportingtheplanethypothesis,consistentwith validated planets.Wealsoremind 173

PAGE 174

thereaderofKOI565.01,whichwasaplanetcandidateobservedinasimilarfashion andthatwasalsofoundtobeafalsepositive(albeititwasselectedfromtherstKOI catalogpublishedbyBoruckietal.2011a;Col on&Ford2011).Whilewendahigh falsepositiverate(2/4or3/5,ifweincludeKOI565)inoursmallsample,wecaution thatthisislikelynotrepresentativeoftheentiresampleof Kepler planetcandidates, duetothesmallnumberoftargetsweobservedandthespecicpropertiesofthese candidates(e.g.theorbitalperiodandsize).Nevertheless,ourresultsdemonstrate theimportanceofconsideringthesepropertieswhenevaluatingthefalsepositive probabilityofspecicsystems.Whileourndingsseemtocontradictthetheoretical estimatesfrom Morton&Johnson ( 2011 ),thelowfalsepositiveratethattheyestimate isbasedontheassumptionthatallcandidateshadpassedpreliminaryfalse-positive vettingmetricsbasedon Kepler photometryandastrometry.Thus,ifweconsiderthe obviouslyV-shapedtransitsforKOI225.01andKOI1187.01toimplyafalsepositive naturefortheseKOIs,thenaccordingto Morton&Johnson ( 2011 )thelowfalsepositive probabilitiesforthesetargetsarenotaccurate.Thefalsepositiverateforoursampleis alsomuchlargerthantheobservationalconstraintsfrom D esertetal. ( 2012 )thatpredict thatthefalsepositiverateismuchlessthan10%.Thisislikelypartlyaresultofdifferent targetsbeingprobedbythedifferentstudies.Therecentstudyby Santerneetal. ( 2012 ) furthersupportsthisidea,astheyfounda 35%falsepositiverateforshort-periodgiant Kepler planetcandidates.Weplantocontinueobservingsmall,short-periodKOIswith theGTCinordertoimprovethesamplesizeofourstudy.Theobservationspresented here,aswellasfutureobservationswiththeGTC,greatlycomplementfollow-upof KOIsdonewithwarmSpitzer aswellasotherobservatories.Aswecanexpectthe numberof Kepler planetcandidatestocontinuetoincrease,wecanuseresultsfromall suchstudiestopinpointwhichtargetsarethebesttofollow-upinordertomaximizethe scienceoutputfromthe Kepler mission. 174

PAGE 175

Table6-1. KOIProperties. StarPlanetCandidate KOIKICKpT e (K) b (deg) P (days) (hr) % (ppm) R p ( R ) R p / R 225580157114.78460379.1770.8385981.245225714.90.04932 420835253714.247468716.4126.0104012.258227004.30.0474 526915763414.427546712.4262.1047191.76399262.60.0305 1187384897214.489528610.8460.37052850.777818352.50.03961 Allvaluesarefrom Boruckietal. ( 2011b ).TheKICnumberreferstothe Kepler InputCatalognumberforeachtarget.Note that b istheGalacticlatitudeoftheKOIhoststar, isthetransitdurationand % isthetransitdepthasmeasuredinthe Kepler bandpass. 175

PAGE 176

Table6-2. PhotometricPrecisions. KOI 666 nm (ppm) 858 nm (ppm) 225476540 420501428 52612231100 118714471295 176

PAGE 177

Table6-3. Best-FitModelParameters. KOI t 0 a cos i / R R p / R R p / R c 1 c 2 c 1 c 2 (BJD $ 2454900)(hr)(666nm)(858nm)(666nm)(666nm)(858nm)(858nm) 225.01764.70659 +0.00096 0.00026 1.7466 +0.0214 0.0871 1.00.1739 +0.0034 0.0012 0.1318 +0.0026 0.0015 0.65480.01520.5034-0.0398 420.01918.49094 +0.00039 0.00102 2.1188 +0.0560 0.0379 0.6270 +0.0898 0.1498 0.0552 +0.0021 0.0021 0.0647 +0.0014 0.0013 0.75810.47290.5837-0.2011 526.01916.46330 +0.00141 0.00170 1.7639(xed)0.7456 +0.0771 0.0771 0.0714 +0.0037 0.0029 0.0635 +0.0035 0.0025 0.69200.24660.53390.0869 1187.01824.52863 +0.00049 0.00045 0.5066 +0.0390 0.0285 0.9345 +0.0427 0.0423 0.0469 +0.0074 0.0070 0.0924 +0.0174 0.0173 0.68950.23870.53700.0886 Notethatthetimeofmid-transit, t 0 ,istechnicallygiveninBJD TDB(BarycentricJulianDateinBarycentricDynamical Time).Seetextforfurtherdetails. 177

PAGE 178

Figure6-1. Histogramsillustratingthenumberof Kepler planetcandidates(solidblue line;basedon Boruckietal. 2011b )anddetachedandall"other"eclipsing binaries(dottedredanddashedblacklines,respectively;basedon Slawson etal. 2011 )asafunctionoforbitalperiod.Wealsoshowahistogram (dash-dotgreenline)basedontherecentlyreleasedlistof2,321KOIsfrom Batalhaetal. ( 2012 ).Therighthandy-axisshowsthepercentageofsystems relativetothetotalnumberoftargetstarsobservedby Kepler inQ1.Note thatonlysystemswithorbitalperiodsofupto10daysareshown.While thereisasubstantialnumberof"other"(primarilysemi-detachedand overcontact)eclipsingbinariesattheshortestorbitalperiods(whichmaybe lesslikelytobeaggedaspotentialplanets),thereisstilla greater-to-comparablenumberofshort-perioddetachedeclipsingbinaries comparedtoshort-periodplanetcandidates,whichindicatesthepotentialfor eclipsingbinariestoinltratetheshort-periodplanetcandidatesample.The orbitalperiodsoftheKOIsweobservedareindicatedwithvertical long-dashedgraylinesandarelabeledaccordingly. 178

PAGE 179

Figure6-2. Radiusversusorbitalperiodof Kepler planetcandidates.Thelocationsof thefourKOIsweobservedareindicatedwithlledredsquaresandare labeledaccordingly.ThelocationsofthetargetsobservedwithwarmSpitzer by D esertetal. ( 2012 )areindicatedwithlledbluesquares.Thelledgray circlesandtheopengraycirclesrespectivelyrepresentallKOIspresented by Boruckietal. ( 2011b )and Batalhaetal. ( 2012 )outtoorbitalperiodsof 200days.NotetheprominenceofKOIssmallerthan 6 R 179

PAGE 180

Figure6-3. Signal-to-noiseratio(SNR)pertransitasafunctionoforbitalperiodfor Kepler planetcandidates.TheSNRpertransitforeachKOIwascomputed fromthetransitdepth,thetransitdurationandeitherthe6-hourcombined differentialphotometricprecision(CDPP)fromQ3data(forthe Boruckietal. 2011b list)orthe6-hourCDPPfromQ1-Q6data(forthe Batalhaetal. 2012 list).ThecolorsandsymbolsarethesameasinFigure 6-2 180

PAGE 181

Figure6-4. Kepler magnitude(Kp)versuseffectivetemperatureforthe997KOIhost starspublishedin Boruckietal. ( 2011b )andtheadditional926KOIhost starspublishedin Batalhaetal. ( 2012 ).Thecolorsandsymbolsarethe sameasinFigures 6-2 and 6-3 181

PAGE 182

Figure6-5. ImageoftheeldofviewaroundKOI1187asacquiredusingthe GTC/OSIRISandthe666/36nmordersorterlter.Thetargetislabeled accordinglyandiscontainedwithinablackcircleequivalenttotwicethe apertureusedinthereductionprocess.Theunlabeledstarscontainedin blackcirclesaretheninereferencestarsusedinouranalysis.Thelargerred circlearoundKOI1187hasaradiusof 20arcsec,andthesixbrightest starslocatedwithinthiscircle(excludingthetarget)arethestarsthatwe investigatedaspotentialsourcesofthetransitsignal. 182

PAGE 183

Figure6-6. Lightcurves,residuals,andcolorbasedonobservationsofthe2011April13 transitofKOI225.01acquiredat666nm(inblue)and858nm(inred).The circlesareobservationsandthesolidcurvesarethecorrespondingbest-t models(seetextforfurtherdetails).Representativephotometricerrorbars foreachlightcurveareshownontheleft-handsideofthetoppanel.The blackhorizontaldashedlineillustratesthedepthofthetransitasmeasured inthe Kepler bandpass(Table 6-1 ).Residualsfromthetsforeachlight curvearealsoshowninthetoppanel,withahorizontaldottedlineindicating thelevelatwhichtheresidualswereoffset(forclarity).Thebottompanel showsthecolorascomputedfromthetwolightcurvesaswellasa representativeerrorbarontheleft-handsideofthepanelandahorizontal dottedlinethatillustratesacolorofzero(forreference). 183

PAGE 184

Figure6-7. SameasFigure 6-6 ,butforthetransitofKOI420.01asobservedon2011 September13. 184

PAGE 185

Figure6-8. SameasFigure 6-6 ,butforthetransitofKOI526.01asobservedon2011 September11. 185

PAGE 186

Figure6-9. SameasFigure 6-6 ,butforthetransitofKOI1187.01asobservedon2011 June12. 186

PAGE 187

Figure6-10. Cumulativefractionofplanet-hostingstarsfrom Boruckietal. ( 2011b ,blue) and Batalhaetal. ( 2012 ,gray)anddetached(red)andall"other"(primarily semi-detachedandovercontact;black)eclipsingbinariesfrom Slawsonet al. ( 2011 )asafunctionofGalacticlatitude.Thecumulativefunctionswere normalizedagainst997starsforthe Boruckietal. ( 2011b )sample,1,790 starsforthe Batalhaetal. ( 2012 )sample(whichincludesthe Boruckietal. 2011b sample),1,266detachedeclipsingbinaries,and901"other" eclipsingbinaries(from Slawsonetal. 2011 ).Thegrayverticaldashed linesindicatetheGalacticlatitudeofthetargetsweobservedwiththeGTC, witheachtargetmarkedaccordingly.Therateofplanet-hostingstarsis slightlylargertowardshigherlatitudes,whiletherateofeclipsingbinariesis slightlylargeratlowerlatitudes,butingeneralthetwosamplesarenot signicantlydifferent.BoththeKOIswefoundtobefalsepositivesare locatedatlowerlatitudes. 187

PAGE 188

Figure6-11. SimilartoFigure 6-10 .Here,weshowthecumulativefractionofstars hostingshort-period(P ) 2days;long-dashedbluecurve)andlong-period (P > 2days;dottedbluecurve)KOIs(basedonthefulllistof2,321KOIs from Batalhaetal. 2012 )alongwiththecumulativefractionoftwoeclipsing binarypopulationsfrom Slawsonetal. ( 2011 )showninFigure 6-10 asa functionofGalacticlatitude.Thecurvesfortheplanet-hostingstarswere normalizedagainststarshosting158short-periodand2,163long-period planetcandidates.Wendthattheshort-periodplanetpopulationismore consistentwiththeeclipsingbinarypopulationsthanthelong-periodplanet population,whichfurthersupportsourargumentforcontaminationofthe planetpopulationbyeclipsingbinariesatshortperiods.Conversely,the long-periodplanetpopulationislesslikelytobeinltratedbyeclipsing binaries.SeethecaptiontoFigure 6-10 foradditionaldetails. 188

PAGE 189

Figure6-12. Histogramsofthenumberofdetached(red)andall"other"(gray)eclipsing binariesfrom Slawsonetal. ( 2011 )asafunctionoforbitalperiod.Each panelshowsadifferentbinofGalacticlatitude.AsinFigure 6-1 ,the righthandy-axisshowsthepercentageofsystemsrelativetothetotal numberoftargetstarsobservedby Kepler inQ1.Thereisaroughly consistentnumberofeclipsingbinariesineachlatitudebin(exceptforthe highestlatitudes),whichsuggestsaconsistentpresenceofeclipsing binariesthatcouldcontaminatepotentialplanetcandidates.Whilethere arefewereclipsingbinariesatthehighestlatitudes,therearealso correspondinglyfewerplanetcandidatesatthoselatitudes(Figure 6-10 ). 189

PAGE 190

CHAPTER7 SUMMARYANDCONCLUSIONS Withtheunexpecteddiscoveryofhundredsofextrasolarplanetsinthepasttwo decades,astronomershavebecomeoverwhelmedwiththenumberofexoplanetsto study.Inthisdissertation,Iaddressedtheissueofhowwecanefcientlycharacterize theorbital,physicalandatmosphericpropertiesoftransitingexoplanetsinparticular. Mystudiesfocusedonusingwhatiscurrentlytheworld'slargest,ground-based,fully steerable,single-aperture,opticaltelescope-the10.4mGTC-toacquiremulti-color photometryofexoplanettransits.InChapter2( Col onetal. 2010 ),Ipresentedanovel techniquetoacquirehigh-precision,multi-color,narrow-bandphotometryofthetransits ofgiantexoplanets(TrES-2bandTrES-3b)usingthetunablelterimagingmodeon theGTC/OSIRIS.Usingthistechnique,Iobtainednear-simultaneousobservationsin twonear-infraredbandpasses(790.2 2.0nmand794.4 2.0nm)specicallychosen toavoidwatervaporabsorptionandskyglowsoastominimizetheatmosphericeffects thatoftenlimittheprecisionofground-basedphotometry.Fromtheseobservations, IachievedphotometricprecisionsfortheTrES-2observationsof0.343and0.412 mmagforthe790.2and794.4nmlightcurves,whiletheprecisionsoftheTrES-3 observationswere0.470and0.424mmagforthe790.2and794.4nmlightcurves, respectively.Theseprecisionsaresomeofthehighestprecisionsachievedtodateusing aground-basedtelescopeandallowforthedetectionofsuper-Earth-sizetransitsas wellasthesmallsignalsduetoabsorptionintheatmospheresoftransitingexoplanets (comparedtotheprecisionsthataretypicallyreachedwithground-basedobservatories, 1mmag). InChapter3( Col onetal. 2012 ),Ipresentedobservationsthatdemonstrated thatthetechniquepresentedinChapter2( Col onetal. 2010 )canindeedbeused tocharacterizetheatmospheresoftransitingexoplanets.Specically,Iconducted asearchforpotassium(K I ,whichisexpectedtobethesecondstrongestoptical 190

PAGE 191

absorberintheatmospheresofgiantexoplanets)intheatmosphereofHD80606b, aJupiter-sizeplanetinaveryeccentric( e =0.93),long-period(111d)orbit.Using theGTC/OSIRIS,Iacquirednarrow-bandphotometryinfourbandpassesaround theK I absorptionfeatureduringatransitofHD80606b.Fromtheseobservations, Iobtaineddifferentialphotometricprecisionsof 2.08 # 10 4 forthein-transitux ratiomeasuredat769.91nm,whichprobestheK I linecore.Ifoundnosignicant differenceinthein-transituxratiobetweenobservationsat768.76nmand769.91 nm.Yet,Ifoundadifferenceof 8.09 2.88 # 10 4 betweentheseobservationsand observationsatalongerwavelengththatprobetheK I wing(777.36nm).Furthermore, the777.36 $ 769.91nmcolorduringtransitindicatesasignicantcolorchange,with ameanvalueof 8.99 0.62 # 10 4 .Thislargechangeinthecolorisequivalentto a 4.2%changeintheapparentradiusoftheplanetwithwavelength,whichismuch largerthantheatmosphericscaleheight.Thisimpliestheobservationsprobedthe atmosphereatverylowpressuresaswellasadramaticchangeinthepressureatwhich theslantopticaldepthreachesunitybetween 770and777nm.Ihypothesizedthatthe measuredexcessabsorptionmaybeduetoK I inahigh-speedwindbeingdrivenfrom HD80606b'sexosphere,whichinturnisaresultofthehighlyeccentricorbitcausing extremetemperaturechangesinHD80606b'satmosphere.Thisstudyprovidedoneof therstdetectionsofabsorptionduetoK I inanexoplanetatmosphere. InChapter4,IextendedthestudydoneinChapter3( Col onetal. 2012 )to super-Earth-sizeplanets.IpresentedobservationsofGJ1214b,a 2.7 R planet orbitinganMdwarfstar.Mygoalwastomeasureabsorptionduetomethane(CH 4 )in GJ1214b'satmosphereinordertoresolverecentstudiesthatdisagreewitheachother regardingthepresenceofmethaneinthisplanet'satmosphere.IusedtheGTC/OSIRIS toacquirenarrow-bandphotometryintwobandpasses,oneinthecontinuumandone withintheassumedmethaneabsorptionfeaturearound890nm(predictedbasedonthe modelsof Miller-Ricci&Fortney 2010 ).Usingobservationsfromthreedifferenttransits 191

PAGE 192

ofGJ1214b,Ifoundthatstellaractivityhadasignicanteffectonmyobservations.For eachtransit,Ifoundthatthemeasuredplanet-starradiusratioswerehighlydependent onwhichreferencestarswereincludedintheanalysis.Frompreviousstudies(e.g., Beanetal. 2010 ; Bertaetal. 2012 ),itisbelievedthatGJ1214b'satmosphereisquite stable.Furthermore,Iinvestigatedthelightcurvesforeachcombinationofreference starsandconcludedthatanyeffectsfromtheEarth'satmosphereweresufciently removedduringtheanalysis.Therefore,Iconcludedthatstellarvariabilitywithinthe targetstarand/orthereferencestar(s)istheprimarysourceofthevariabilitywithinthe measuredradiusratios.Asaresult,Iamnotabletoconrmorrefutethepresence ofmethaneintheatmosphereofGJ1214b.Thisstudyrevealsthesensitivenatureof atmosphericstudiesanddemonstratesthatallviablescenariosneedtobeconsidered whenconductingsuchstudies. Chapter5( Col on&Ford 2011 )focusesontheuseofnarrow-band,multi-color photometrytocharacterizetransitingplanetcandidates,i.e.thosediscoveredbygroundandspace-basedsurveys(like Kepler )butnotyetconrmedviameasurementsof theirmass.Sinceplanetarytransitsshouldbelargelyachromaticwhenobservedat differentwavelengths(excludingthesmallcolorchangesduetostellarlimbdarkening), itispossibletousetheobservedtransitcolortoidentifycandidatesaseitherfalse positives(e.g.,ablendwithastellareclipsingbinaryeitherinthebackground/foreground orboundtothetargetstar)orvalidatedplanets.InChapter5( Col on&Ford 2011 ),I reportedthediscoveryofaneclipsingbinarystarnearKOI565(previouslya Kepler planetcandidate)basedonobservationsofKOI565andseveralnearbystarsacquired intwonarrowbandpassesnearlysimultaneouslywiththeGTC/OSIRIS.Iusedthe individualphotometryineachbandpass,aswellasthecolorsofKOI565andother nearbystars,todeterminethatthesourceofthetransitsignalinitiallydetectedby Kepler isnotduetoasuper-Earth-sizeplanetaroundKOI565.Instead,Ifoundthesourcetobe abackgroundeclipsingbinarystarlocated 15 $$ tothenorthofKOI565. 192

PAGE 193

FollowingthestudypresentedinChapter5( Col on&Ford 2011 ),inChapter6 (Col onetal.,accepted)Ipresentedanextendedsampleofplanetcandidatesthat wasobservedusingthe"ordersorter"ltersontheGTC/OSIRIS(ratherthantheTF imagingmode,soastoprobeagreaterwavelengthrange).Iacquiredmulti-color transitphotometryoffoursmall( R p < 5 R )short-period( P < 6 days) Kepler planet candidates.MyresultsincludedtheidenticationofKOI225.01andKOI1187.01as falsepositivesandthetentativevalidationofKOI420.01andKOI526.01asplanets. Theseresultssuggestahigherfalsepositiverateforthesmall,short-period Kepler planetcandidatesthanhasbeentheoreticallypredictedorobservationallydetermined byotherstudieswhichconsiderthe Kepler planetcandidatesampleasawhole(e.g. Morton&Johnson 2011 ; D esertetal. 2012 ).Ialsoinvestigatedhowthefalse positiverateformysamplevarieswithdifferentplanetaryandstellarproperties.My resultssuggestthatthefalsepositiveratevariessignicantlywithorbitalperiodand islargestattheshortestorbitalperiods( P < 3 days),wherethereisacorresponding riseinthenumberofdetachedeclipsingbinarystars(i.e.,systemsthatcaneasily mimicplanetarytransits)thathavebeendiscoveredby Kepler .However,Ididnotnd signicantcorrelationsbetweenthefalsepositiverateandotherplanetaryorstellar properties.Whilemysamplesizeisnotyetlargeenoughtodetermineiforbitalperiod playsthelargestroleindeterminingthefalsepositiverate,Ihaveplanstoparticipatein futureobservationsofadditional Kepler candidates.Withlimitedtelescoperesources,it isimperativethatfalsepositivesareidentiedasefcientlyaspossible,andtheGTCis aprimetoolthatcanbeusedforthisduetoitslargeeldofview,asuperiorcollecting area,anditscapabilityforacquiringhigh-precision,multi-colorphotometryinasingle transitobservation.Withtheidenticationoffalsepositivescomesacorresponding conrmationofcandidatesastrueplanets,whichultimatelyallowsforconstraintsonthe knownplanetpopulationasawhole. 193

PAGE 194

Ultimately,Ihaveshownthatlargeground-basedtelescopesliketheGTCcan beveryefcienttoolsforcharacterizingextrasolarplanets.IhaveusedtheGTCto reachsomeofthehighestphotometricprecisionsachievablewithground-based telescopes(todate),toprobetheatmospheresofbothgiantandsuper-Earth-size planets,andtoidentifyfalsepositivesandvalidateplanetswithinthesampleof Kepler planetcandidates.Mostsignicantly,IachievedoneoftherstdetectionsofK I in theatmosphereofanextrasolarplanet,whichhasimplicationsparticularlyforthe structureandcompositionoftheatmospheresofexoplanetsoneccentricorbits.Ialso demonstratedthateclipsingbinarystarsmostcertainlycontaminatethepopulationof short-period Kepler planetcandidates,whichhasrepercussionsonthesciencebeing donewiththefullplanetcandidatepopulationdiscoveredby Kepler .Consideringthat theeldofexoplanetresearchwilllikelycontinuetogrowrapidly,thestudiesthatIhave presentedinmydissertationpromotethecontinueduseoftheGTCandotherlarge ground-basedobservatoriesforexoplanetstudies.Furthermore,investigationsofthe differentuniqueclassesofexoplanets(namelyhot-Jupitersandsuper-Earths)ultimately helpusbetterunderstandwhynosuchplanetsexistinourownSolarSystem. 194

PAGE 195

REFERENCES Ammler-vonEiff,M.,Santos,N.C.,Sousa,S.G.,Fernandes,J.,Guillot,T.,Israelian,G., Mayor,M.,&Melo,C.2009,A&A,507,523 Auvergne,M.,etal.2009,A&A,506,411 Bakos,G. A.,etal.2007,ApJ,670,826 Bakos,G. A.,etal.2010,ApJ,710,1724 Ballester,G.E.,Sing,D.K.,&Herbert,F.2007,Nature,445,511 Barman,T.S.,Hauschildt,P.H.,Schweitzer,A.,Stancil,P.C.,Baron,E.,&Allard,F. 2002,ApJL,569,L5 Barman,T.2007,ApJL,661,L191 Batalha,N.M.,etal.2010,ApJL,713,L109 Batalha,N.M.,etal.2011,ApJ,729,27 Batalha,N.M.,Rowe,J.F.,Bryson,S.T.,etal.2012,arXiv:1202.5852 Bean,J.L.,Miller-RicciKempton,E.,&Homeier,D.2010,Nature,468,669 Bean,J.L.,D esert,J.-M.,Kabath,P.,etal.2011,ApJ,743,92 Ben-Jaffel,L.,&SonaHosseini,S.2010,ApJ,709,1284 Berta,Z.K.,Charbonneau,D.,Bean,J.,etal.2011,ApJ,736,12 Berta,Z.K.,Charbonneau,D.,D esert,J.-M.,etal.2012,ApJ,747,35 Borucki,W.J.,Scargle,J.D.,&Hudson,H.S.1985,ApJ,291,852 Borucki,W.J.,etal.2009,Science,325,709 Borucki,W.J.,etal.2010a,Science,327,977 Borucki,W.J.,etal.2010b,ApJL,713,L126 Borucki,W.J.,etal.2011a,ApJ,728,11 Borucki,W.J.,etal.2011b,ApJ,736,19 Brown,T.M.2001,ApJ,553,1006 Carter,J.A.,Winn,J.N.,Holman,M.J.,etal.2011,ApJ,730,82 Cepa,J.,etal.2000,Proc.SPIE,4008,623 Cepa,J.,etal.2003,Proc.SPIE,4841,1739 195

PAGE 196

Charbonneau,D.,Brown,T.M.,Noyes,R.W.,&Gilliland,R.L.2002,ApJ,568,377 Charbonneau,D.,etal.2009,Nature,462,891 Claret,A.,&Bloemen,S.2011,A&A,529,A75 Col on,K.D.,&Ford,E.B.2009,ApJ,703,1086 Col on,K.D.,Ford,E.B.,Lee,B.,Mahadevan,S.,&Blake,C.H.2010,MNRAS,408, 1494 Col on,K.D.,&Ford,E.B.2011,PASP,123,1391 Col on,K.D.,Ford,E.B.,Redeld,S.,etal.2012,MNRAS,419,2233 Croll,B.,Albert,L.,Jayawardhana,R.,etal.2011,ApJ,736,78 Crosseld,I.J.M.,Barman,T.,&Hansen,B.M.S.2011,ApJ,736,132 Daemgen,S.,Hormuth,F.,Brandner,W.,Bergfors,C.,Janson,M.,Hippler,S.,& Henning,T.2009,A&A,498,567 Deeg,H.J.,Gillon,M.,Shporer,A.,etal.2009,A&A,506,343 Deming,D.,etal.2009,PASP,121,952 deMooij,E.J.W.,Brogi,M.,deKok,R.J.,etal.2012,A&A,538,A46 D esert,J.-M.,LecavelierdesEtangs,A.,H ebrard,G.,Sing,D.K.,Ehrenreich,D.,Ferlet, R.,&Vidal-Madjar,A.2009,ApJ,699,478 D esert,J.-M.,Bean,J.,Miller-RicciKempton,E.,etal.2011,ApJL,731,L40 D esert,J.-M.,Charbonneau,D.,Fressin,F.,&Torres,G.2012,AmericanAstronomical SocietyMeetingAbstracts,219,#414.02 Dhillon,V.S.,Marsh,T.R.,Stevenson,M.J.,etal.2007,MNRAS,378,825 Dravins,D.,Lindegren,L.,Mezey,E.,&Young,A.T.1998,PASP,110,610 Eastman,J.,Siverd,R.,&Gaudi,B.S.2010,PASP,122,935 Ehrenreich,D.,etal.2008,A&A,483,933 Ekenb ack,A.,Holmstr om,M.,Wurz,P.,Grie§meier,J.-M.,Lammer,H.,Selsis,F.,& Penz,T.2010,ApJ,709,670 Ellis,T.,Drake,R.,Fowler,A.M.,etal.1993,Proc.SPIE,1765,94 Fortney,J.J.,Sudarsky,D.,Hubeny,I.,Cooper,C.S.,Hubbard,W.B.,Burrows,A.,& Lunine,J.I.2003,ApJ,589,615 196

PAGE 197

Fortney,J.J.2005,MNRAS,364,649 Fortney,J.J.,Shabram,M.,Showman,A.P.,Lian,Y.,Freedman,R.S.,Marley,M.S.,& Lewis,N.K.2010,ApJ,709,1396 Fossati,L.,etal.2010,ApJL,714,L222 Fossey,S.J.,Waldmann,I.P.,&Kipping,D.M.2009,MNRAS,396,L16 France,K.,Stocke,J.T.,Yang,H.,Linsky,J.L.,Wolven,B.C.,Froning,C.S.,Green, J.C.,&Osterman,S.N.2010,ApJ,712,1277 Fressin,F.,Torres,G.,Desert,J.-M.,etal.2011,ApJS,197,5 Fressin,F.,Torres,G.,Rowe,J.F.,etal.2012,Nature,482,195 Garc a-Lorenzo,B.,Eff-Darwich,A.,Castro-Almaz an,J.,Pinilla-Alonso,N., Mu noz-Tu n( on,C.,&Rodr guez-Espinosa,J.M.2010,MNRAS,405,2683 Garcia-Melendo,E.,&McCullough,P.R.2009,ApJ,698,558 Garc aMu noz,A.2007,P&SS,55,1426 Gautier,T.N.,III,Charbonneau,D.,Rowe,J.F.,etal.2012,ApJ,749,15 Gibson,N.P.,etal.2009,ApJ,700,1078 H ebrard,G.,etal.2010,A&A,516,A95 Hidas,M.G.,etal.2010,MNRAS,808 Holman,M.J.,etal.2007,ApJ,664,1185 Holman,M.J.,etal.2010,Science,330,51 Holmstr om,M.,Ekenb ack,A.,Selsis,F.,Penz,T.,Lammer,H.,&Wurz,P.2008,Nature, 451,970 Howell,S.B.2006,HandbookofCCDastronomy,2nded.,byS.B.Howell.Cambridge observinghandbooksforresearchastronomers,Vol.5Cambridge,UK:Cambridge UniversityPress,2006ISBN0521852153 Hubbard,W.B.,Fortney,J.J.,Lunine,J.I.,Burrows,A.,Sudarsky,D.,&Pinto,P.2001, ApJ,560,413 Johnson,J.A.,Winn,J.N.,Cabrera,N.E.,&Carter,J.A.2009,ApJL,692,L100 Kaler,J.B.2011,StarsandtheirSpectra,AnIntroductiontotheSpectralSequence, ISBN0521899540,CambridgeUniversityPress Kashyap,V.L.,Drake,J.J.,&Saar,S.H.2008,ApJ,687,1339 197

PAGE 198

Knutson,H.A.,Charbonneau,D.,Noyes,R.W.,Brown,T.M.,&Gilliland,R.L.2007, ApJ,655,564 Kundurthy,P.,Agol,E.,Becker,A.C.,etal.2011,ApJ,731,123 Langland-Shula,L.E.,Vogt,S.S.,Charbonneau,D.,Butler,P.,&Marcy,G.2009,ApJ, 696,1355 Laughlin,G.,Deming,D.,Langton,J.,Kasen,D.,Vogt,S.,Butler,P.,Rivera,E.,& Meschiari,S.2009,Nature,457,562 LecavelierDesEtangs,A.,Vidal-Madjar,A.,&Desert,J.-M.2008,Nature,456,E1 LecavelierDesEtangs,A.,etal.2010,A&A,514,A72 L eger,A.,etal.2009,A&A,506,287 Linsky,J.L.,Yang,H.,France,K.,Froning,C.S.,Green,J.C.,Stocke,J.T.,& Osterman,S.N.2010,ApJ,717,1291 Lissauer,J.J.,etal.2011a,Nature,470,53 Lissauer,J.J.,Ragozzine,D.,Fabrycky,D.C.,etal.2011b,ApJS,197,8 Lissauer,J.J.,Marcy,G.W.,Rowe,J.F.,etal.2012,ApJ,750,112 Mandel,K.,&Agol,E.2002,ApJL,580,L171 Miller-Ricci,E.,&Fortney,J.J.2010,ApJL,716,L74 Miralda-Escud e,J.2002,ApJ,564,1019 Mislis,D.,&Schmitt,J.H.M.M.2009,A&A,500,L45 Mislis,D.,Schr oter,S.,Schmitt,J.H.M.M.,Cordes,O.,&Reif,K.2010,A&A,510, A107 Monet,D.G.,Jenkins,J.M.,Dunham,E.W.,etal.2010,arXiv:1001.0305 Morton,T.D.2012,arXiv:1206.1568 Morton,T.D.,&Johnson,J.A.2011,ApJ,738,170 Moutou,C.,Coustenis,A.,Schneider,J.,StGilles,R.,Mayor,M.,Queloz,D.,&Kaufer, A.2001,A&A,371,260 Moutou,C.,etal.2009,A&A,498,L5 Muirhead,P.S.,Johnson,J.A.,Apps,K.,etal.2012,ApJ,747,144 Murray-Clay,R.A.,Chiang,E.I.,&Murray,N.2009,ApJ,693,23 198

PAGE 199

Naef,D.,etal.2001,A&A,375,L27 Narita,N.,Suto,Y.,Winn,J.N.,Turner,E.L.,Aoki,W.,Leigh,C.J.,Sato,B.,Tamura, M.,&Yamada,T.2005,PASJ,57,471 O'Donovan,F.T.,Charbonneau,D.,Torres,G.,etal.2006a,ApJ,644,1237 O'Donovan,F.T.,etal.2006b,ApJL,651,L61 Or,A.,&Dreizler,S.2012,arXiv:1206.5347 Pont,F.,Zucker,S.,&Queloz,D.2006,MNRAS,373,231 Pont,F.,Knutson,H.,Gilliland,R.L.,Moutou,C.,&Charbonneau,D.2008,MNRAS, 385,109 Pont,F.,etal.2009,A&A,502,695 Pr sa,A.,&Zwitter,T.2005,ApJ,628,426 Pr sa,A.,Batalha,N.,Slawson,R.W.,etal.2011,AJ,141,83 Rabus,M.,Deeg,H.J.,Alonso,R.,Belmonte,J.A.,&Almenara,J.M.2009,A&A,508, 1011 Raetz,S.,etal.2009,AstronomischeNachrichten,330,459 Ragozzine,D.,&Holman,M.J.2010,arXiv:1006.3727 Redeld,S.,Endl,M.,Cochran,W.D.,&Koesterke,L.2008,ApJL,673,L87 Rogers,L.A.,&Seager,S.2010,ApJ,716,1208 Santerne,A.,D az,R.F.,Moutou,C.,etal.2012,arXiv:1206.0601 Scuderi,L.J.,Dittmann,J.A.,Males,J.R.,Green,E.M.,&Close,L.M.2010,ApJ,714, 462 Seager,S.,&Sasselov,D.D.2000,ApJ,537,916 Shporer,A.,etal.2010,ApJ,722,880 Sing,D.K.,Vidal-Madjar,A.,D esert,J.,LecavelierdesEtangs,A.,&Ballester,G. 2008a,ApJ,686,658 Sing,D.K.,Vidal-Madjar,A.,LecavelierdesEtangs,A.,D esert,J.-M.,Ballester,G.,& Ehrenreich,D.2008b,ApJ,686,667 Sing,D.K.,etal.2011,A&A,527,A73 Slawson,R.W.,Pr sa,A.,Welsh,W.F.,etal.2011,AJ,142,160 199

PAGE 200

Snellen,I.A.G.,Albrecht,S.,deMooij,E.J.W.,&LePoole,R.S.2008,A&A,487,357 Snellen,I.A.G.,deKok,R.J.,deMooij,E.J.W.,&Albrecht,S.2010,Nature,465, 1049 Southworth,J.2008,MNRAS,386,1644 Southworth,J.,etal.2009,MNRAS,396,1023 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 Sozzetti,A.,etal.2009,ApJ,691,1145 Steffen,J.H.,etal.2010,ApJ,725,1226 Stetson,P.B.1987,PASP,99,191 Swain,M.R.,Vasisht,G.,&Tinetti,G.2008,Nature,452,329 Swain,M.R.,Vasisht,G.,Tinetti,G.,Bouwman,J.,Chen,P.,Yung,Y.,Deming,D.,& Deroo,P.2009,ApJL,690,L114 Tian,F.,Toon,O.B.,Pavlov,A.A.,&DeSterck,H.2005,ApJ,621,1049 Tingley,B.2004,A&A,425,1125 Torres,G.,Winn,J.N.,&Holman,M.J.2008,ApJ,677,1324 Torres,G.,etal.2011,ApJ,727,24 Valencia,D.,Sasselov,D.D.,&O'Connell,R.J.2007,ApJ,665,1413 Valenti,J.A.,&Fischer,D.A.2005,ApJS,159,141 Vidal-Madjar,A.,LecavelierdesEtangs,A.,D esert,J.-M.,Ballester,G.E.,Ferlet,R., H ebrard,G.,&Mayor,M.2003,Nature,422,143 Vidal-Madjar,A.,D esert,J.-M.,LecavelierdesEtangs,A.,H ebrard,G.,Ballester,G.E., Ehrenreich,D.,Ferlet,R.,McConnell,J.C.,Mayor,M.,&Parkinson,C.D.2004,ApJL, 604,L69 Vidal-Madjar,A.,LecavelierdesEtangs,A.,D esert,J.-M.,Ballester,G.E.,Ferlet,R., H ebrard,G.,&Mayor,M.2008,ApJL,676,L57 Winn,J.N.,Suto,Y.,Turner,E.L.,Narita,N.,Frye,B.L.,Aoki,W.,Sato,B.,&Yamada, T.2004,PASJ,56,655 Winn,J.N.,etal.2007,AJ,133,1828 Winn,J.N.,etal.2009,ApJ,703,2091 200

PAGE 201

Wood,B.E.,M uller,H.-R.,Zank,G.P.,Linsky,J.L.,&Redeld,S.2005,ApJL,628, L143 Wright,J.T.,Marcy,G.W.,Butler,R.P.,&Vogt,S.S.2004,ApJS,152,261 Yelle,R.V.2004,Icarus,170,167 Young,A.T.1967,AJ,72,747 201

PAGE 202

BIOGRAPHICALSKETCH KnicoleDawnCol onwasborninFaireld,Californiain1985.Beforeshewaseven veyearsold,herfamilyknewshehadaninterestinscience,asshewouldoften"listen" forearthquakesbylayingontheoorandputtinghereartotheground.Shemoved withherfamilytoFloridain1990,and,aftersurvivingHurricaneAndrew,movedto NewJerseyin1992.KnicolegraduatedfromRancocasValleyRegionalHighSchoolin 2003.Beforegraduating,shespentthesummerof2002participatingintheNewJersey Governor'sSchoolintheSciences,whichwasamonthlongscience-intensiveprogram heldatDrewUniversity.Thisprogramhelpedfuelherinterestinastronomy(alongwith variousbooksandmovies,like Contact byCarlSagan).ShethenattendedTheCollege ofNewJersey(TCNJ),whereshegraduated magnacumlaude withaB.S.inphysics andaminorinmathematicsin2007.DuringhertimeatTCNJ,sheparticipatedintwo NationalScienceFoundationResearchExperienceforUndergraduatesprograms. Therstprogramtookplaceinthesummerof2005atLehighUniversity,whereshe workedwithDr.MichaelStavolaoninvestigatingthepropertiesofsemi-conducting materials.ThesecondprogramtookplaceatAreciboObservatoryinthesummerof 2006,wheresheworkedwithDr.MayraLebronandstudiedthepropertiesofmasersin astar-formingregion.KnicolebegangraduateschoolattheUniversityofFloridain2007 andearnedaM.S.inastronomyin2009.ShewasalsoawardedaNationalScience FoundationGraduateResearchFellowshipin2009.KnicolewillreceiveaPh.D.in astronomyfromtheUniversityofFloridainAugust2012.Aftergraduating,shewillstart apostdoctoralpositionattheUniversityofHawaiiatManoa,whereshewillbeworking withDr.EricGaidostosearchforandcharacterizetransitingextrasolarplanetsaround Mdwarfstars. 202