ebook img

Circumbinary planets - why they are so likely to transit PDF

1.5 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Circumbinary planets - why they are so likely to transit

Mon.Not.R.Astron.Soc.000,1–??(2014) Printed10June2015 (MNLATEXstylefilev2.2) Circumbinary planets - why they are so likely to transit David. V. Martin1(cid:63), Amaury. H.M.J. Triaud2,3,4 1ObservatoiredeGene`ve,Universite´deGene`ve,51chemindesMaillettes,Sauverny1290,Switzerland 2DepartmentofPhysics,andKavliInstituteforAstrophysicsandSpaceResearch,MassachusettsInstituteofTechnology,Cambridge,MA02139,USA 5 3CentreforPlanetarySciences,UniversityofToronto,DepartmentofEnvironmentalandPhysicalSciences,1265MilitaryTrail,Toronto,OntarioM1C1A4, 1 Canada 0 4FellowoftheSwissNationalScienceFoundation 2 n u Accepted.Received J 9 ABSTRACT ] P Transitsonsinglestarsarerare.Theprobabilityrarelyexceedsafewpercent.Further- E more, thisprobability rapidly approaches zeroat increasing orbital period.Therefore transit . surveyshavebeenpredominantlylimitedtotheinnerpartsofexoplanetarysystems.Herewe h demonstratehowcircumbinaryplanetsallowustobeattheseunfavourableodds.Byincorpo- p rating the geometry and the three-body dynamics of circumbinary systems, we analytically - o derive the probability of transitability, a configuration where the binary and planet orbits r overlap on the sky. We later show that this is equivalent to the transit probability, but at an t s unspecified point in time. This probability, at its minimum, is always higher than for single a star cases. In addition, it is an increasing function with mutual inclination. By applying our [ analyticaldevelopmenttoeclipsingbinaries,wededucethattransitsarehighlyprobable,and 2 in some case guaranteed. For example, a circumbinary planet revolving at 1 AU around a v 0.3 AU eclipsing binary is certain to eventually transit - a 100% probability - if its mutual 1 inclinationisgreaterthan0.6◦.Weshowthatthetransitprobabilityisgenerallyonlyaweak 3 function of the planet’s orbital period; circumbinary planets may be used as practical tools 6 forprobingtheouterregionsofexoplanetarysystemstosearchforanddetectwarmtocold 3 transitingplanets. 0 1. Key words: binaries: close, eclipsing, spectroscopic – astrometry and celestial mechanics: 0 celestialmechanics,eclipses–planetsandsatellites:detection,dynamicalevolutionandsta- 5 bility,fundamentalparameters–methods:analytical,numerical,statistical 1 : v i X 1 INTRODUCTION al.2011withtransitphotometry).Therearepresentlytentransit- r a ingcircumbinaryplanetsknown,allfoundbytheKeplertelescope Intheburgeoningsearchforextra-solarplanets,circumbinaryplan- (Welshetal.2014). etsrepresentsomeofthemostexoticsystemsfoundtodate.They Theadvantageoffindingcircumbinaryplanetsintransitisthat pose astronomers with interesting questions regarding their de- theycanyieldanunambiguousdetection,thankstoauniquesig- tectability (Schneider 1994), abundance (Armstrong et al. 2014; naturethatishardtomimicwithfalsepositives.Thephotometric Martin&Triaud2014),formation(Pierens&Nelson2013;Kley measurementoftheradiuscanbecomplementedwithtransittim- &Haghighipour2014),habitability(Haghighipour&Kaltenegger ing variations (TTVs), eclipse timing variations (ETVs) or spec- 2013; Mason et al. 2014), orbital dynamics (Leung & Hoi Lee troscopytoobtainthemassandbulkdensity,whichareimportant 2013)andstability(Dvorak1986;Dvoraketal.1989;Holman& fromaformationperspective.Transitsalsoopenthedoortoatmo- Wiegert1999). sphericcharacterisation(Seager&Deming2010),themeasureof Answerstothesequestionsarereliantonplanetdetections.So the Rossiter-McLaughlin effect (Queloz et al. 2000; Fabrycky & far there have been reported discoveries from several techniques, Winn2014),andthedetectionofexomoons(Kippingetal.2012). includingPSRB1620-26(Thorsettetal.1999withpulsartiming), Itwillbeshowninthispaperthatcircumbinaryplanets,be- HD202206(Correiaetal.2005withradialvelocimetry),DPLeo- yond their exoticity, are useful astronomical tools. Their particu- nis(Qianetal.2010witheclipsetimingvariations),Ross458(Bur- largeometryandorbitaldynamicsleadtopotentiallymuchhigher gasser et al. 2010 with direct imaging), and Kepler-16 (Doyle et transitprobabilitiesincomparisonwithsinglestars.Thereisalsoa weakerdependenceonorbitalperiod,allowingustoextendtransit studiestotheouterregionsofstellarsystems. (cid:63) E-mail:[email protected] Thepaperisstructuredasfollows.InSect.2weintroducethe (cid:13)c 2014RAS 2 MartinandTriaud geometryofcircumbinaryplanets.NextinSect.3weanalysethe y orbitaldynamicsofcircumbinarysystemsandtheeffectsontheir observability.Wethendefinetheconceptoftransitabilityandana- lyticallyderiveacriterionforitsoccurrenceinSect.4.Following this,weconvertthiscriterionintotheprobabilityofacircumbinary f systemexhibitingtransitabilityinSect.5,similartotheworkdone 2 periapse fwoersainnagllyessetathrse(sBpoecruiaclkcia&seSoufmemcleiprssi1n9g8b4i;nBarairense.s2007).InSect.6 a ( 1 - e ) Asanobserver,theobservablequantityisatransit,nottran- ω sitability.ThisiswhyinSect.7weconnectthetwoconcepts,ver- x ifyingthatasystemexhibitingtransitabilityiseffectivelyguaran- a ( 1 + e ) 1 teed to transit, albeit at an unspecified point in time. Some illus- trativetransitwaittimesarecalculated,revealingthattheymaybe withinafewyearsformanysystems.InSect.8wediscusssomeap- apoapse plicationsandlimitationsofourwork,beforeconcludinginSect.9. 2 GEOMETRY WewilltreatacircumbinarysystemasapairofKeplerianorbits Figure1.Planarorbitalelementsofatwo-bodysystem. inJacobicoordinates,withtheadditionoffirst-orderdynamicalef- fects(Sect.3).Theinnerorbitisthestellarbinary(subscript“bin”). z Theouterorbitistheplanetaroundthebinarycentreofmass(sub- script“p”).EachKeplerianorbitisanellipsecharacterisedbyfour orbital elements: the semi-major axis a, eccentricity e, argument towards observer ofperiapsisωandtrueanomaly f.Thesequantitiesaredefinedin Fig.1.Thissetoffourisnotunique,andoftenwewillusethepe- ΔI riodT insteadofthesemi-majoraxis.Theorientationofeachorbit inthreedimensionsisdefinedusingtwoextraangles:theinclina- tionIandlongitudeoftheascendingnodeΩ.InFig.2wedepictthe I 3Dorientationofabinaryandplanetorbit.Wetaketheobserverto bin belookingdownthez-axis.Aneclipsingbinary,forexample,cor- y respondstoI ≈π/2.Throughoutthispaperweuseradiansunless bin otherwisespecifiedwitha◦symbol. Ω bin Theorientationoftheplanetaryorbitwithrespecttothebinary ΔΩ I p ischaracterisedbytwoquantities:themutualinclination Ω p cos∆I=sinIbinsinIpcos∆Ω+cosIbincosIp, (1) xx sky plane andthemutuallongitudeoftheascendingnode Figure2.Acircumbinaryplanetinamisalignedorbit(blue,outer)around abinarystarsystem(pink,inner).Themisalignmentischaracterisedbythe ∆Ω=Ωbin−Ωp, (2) mutualinclination,∆I,andthemutuallongitudeoftheascendingnode,∆Ω. Theobserverislookingdownthez-axisfromabove,andhencethegreyx-y whicharealsoshowninFig.2.Whenusingtransitphotometryor planedenotestheplaneofthesky. radial velocimetry, the observer is sensitive to ∆Ω but not to the individual quantities Ω and Ω . Throughout this paper can we bin p thereforetakeΩ =0andallowΩ tovary. bin p the binary plane (∆Ω = 0 to 2π), whilst maintaining ∆I = const (Schneider1994;Farago&Laskar2010;Doolin&Blundell2011). TheprecessionperiodT accordingtoSchneider(1994)is 3 DYNAMICORBITS prec A static Keplerian orbit is insufficient for accurately describing a 16(cid:32) a (cid:33)2 1 circumbinary planet. Owing to perturbations from the binary, the Tprec=Tp 3 ap cos∆I, (3) orbital elements defined in Sect. 2 vary on observationally rele- bin vanttimescales.Weincludeinourderivationthemostprominent where the stars are assumed to be of equal mass. An alternative, oftheseeffects:aprecessionintheplanet’sorbitalplane.Thisbe- morecomplexderivationcanbefoundinFarago&Laskar(2010). haviourisdescribedbyatime-variationin∆I,and∆Ω.Werestrict Theplanetorbitisstableaslongasitisnottooclosetothebinary. ourselvestocircularbinariesandplanets1.Inthiscase,theorbital AnapproximatecriterionfromtheworkofDvorak(1986);Dvorak planeoftheplanetrotatesataconstantratearoundthenormalto etal.(1989);Holman&Wiegert(1999)is 1 Itistechnicallyamisnomertospeakofcircularcircumbinaryorbits,since ever,thisonlyhasaverysmalleffectonthetransitgeometry.SeeSect.8.5.2 perturbationsfromthebinarycauseeptovaryevenifinitiallyzero.How- forfurtherdetail. (cid:13)c 2014RAS,MNRAS000,1–?? Analyticcircumbinarytransitprobability 3 130 1 125 0.8 120 ) 0.6 g 115 e π 0.4 (d 110 / ) 0.2 p105 Ω I ∆ ( 0 100 n si −0.2 95 I ∆ −0.4 90 −0.6 0 20 40 60 t (ydeaayrss) −0.8 Figure4.VariationofIpovertimefortwocircumbinarysystems(ap=0.3 −1 AUinblackandwhitedashes,ap = 0.6AUinblue).Thehorizontalred −1 −0.5 0 0.5 1 linedenotesthebinary’sorbitalplaneinclinationonthesky,Ibin.Thegrey ∆Icos(∆Ω)/π regioncorrespondstowhentheplanetisintransitability. Figure3.Surfacesofsectionofthemutualinclination,∆I,andmutuallon- gitudeoftheascendingnode,∆Ω,betweenthebinaryandplanetorbital planes. a (cid:38)3a . (4) p bin InFig.3wedemonstratehowprecessioneffects∆I and∆Ω usingnumericalN-bodyintegrations2.Weranasetofsimulations with∆Ωstartingat90◦and∆Ivariedbetween0◦and180◦insteps of10◦ and,eachcorrespondingtoadifferentcurveinFig.3.The starsareofmass1M and0.5M witha =0.07AU.Theplanet (cid:12) (cid:12) bin Figure5.Anexamplecircumbinarysystemexhibitingtransitability. isamasslesstestparticlewitha =0.3AU.Thegreen,innercurves p areforprogradeorbitswithclockwiseprecession.Theblue,outer curvesareforretrogradeorbitswithanti-clockwiseprecession.The gapbetweenthegreenandbluecurvescorrespondsto∆I = 90◦, expressionsinEq.3producesprecessionperiodsof6.0yrand67.1 yr,showingittobereasonablyaccurate. i.e.forapolarorbitwheretheprecessionperiodbecomesinfinitely long(Eq.3). Theseorbitaldynamicshaveobservationalconsequences.The inclinationplanetonthesky,I varieswithtimeaccordingto 4 CRITERIONFORTRANSITABILITY p (cid:32) (cid:33) 2π Transitability is an orbital configuration where the planet and bi- I =∆Icos t +I , (5) p T bin naryorbitsintersectonthesky,liketheexampleshowninFig.5. prec In this scenario transits are possible but not guaranteed on every wheretistime.AnexampleisshowninFig.4fortwocircumbi- passageoftheplanetpastthebinary,becauseoftherelativemotion narysystems.Thebinaryinbothsystemshasequalmassstarswith ofthethreebodies.ThisterminologywasfirstintroducedinMartin M = M = 1M , a = 0.1 AU, I = 110◦, and Ω = 0◦. A B (cid:12) bin bin bin &Triaud(2014),wewhereformallydefinedandelaboratedupon Theplanetisamasslessbodywithstartingvalues I = 130◦ and p a concept that had been already used in several studies (Schnei- Ω = 0◦.Themutualinclinationis20◦.Thetwoplanetsshownin p der 1994; Welsh et al. 2012; Kratter & Shannon 2013). In Fig. 4 thefigurehavedifferentvaluesfora :0.3AUfortheblack,dashed p thegreyregiondenotesthetimespentintransitability.Duringeach sinusoidand0.6AUfortheblue,solidsinusoid. precessionperiod,therewillbezero,oneortwointervalsoftran- ThemaximumandminimumvaluesofI areindependentof p sitability,orpermanenttransitabilityinonlytwoscenarios:1)I a . The planet semi-major axis does, however, strongly influence bin p andI arebothveryclosetoπ/2and2)polarorbitswhereI =0 theprecessionperiod:T = 5.6yrfora = 0.3AUand65.9yr p bin prec p andI =π/2. fora =0.6AU,accordingtotheN-bodysimulation.Theanalytic p p We work to derive a criterion that predicts whether or not a circumbinaryplanetwillentertransitabilityatanypointduringthe 2 All N-body simulations in this paper are done using a fourth-order precessionperiod.Asafirstapproximation,weknowthattheplanet Runge-Kuttaalgorithm,whereenergylossduetoitsnon-symplecticnature isintransitabilitywhentheplanetorbitisperpendiculartotheplane waskepttonegligiblelevels. ofthesky: (cid:13)c 2014RAS,MNRAS000,1–?? 4 MartinandTriaud π sky! I = . (6) p 2 plane Thisisthemostconservativecasepossible,sinceitignoresthefi- towards! nite extent of the binary. According to Eq. 5, I is guaranteed to p observer reachπ/2if |90°-I | p X ∆I>(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)π2 −Ibin(cid:12)(cid:12)(cid:12)(cid:12)(cid:12). (7) XB A ΔI |90°-Ibin| Xp I This is the first-order criterion for transitability. Three things are bin apparent:1)mutualinclinationsaidtransitability,whichiscontrary toconventionalviewsontransitgeometries,2)thiscriterionisin- dependentoftheplanetperiodand3)thiscriterioniseasiesttoful- Figure6.Aside-onviewofacircumbinarysysteminthelimitingcaseof filatI ≈ π/2,i.e.foreclipsingbinaries.Thiscriterionwasalso transitability,where∆Ω = 0.Thetwoextremeverticalpositionsofeach bin derived by Schneider (1994), who was the first author to analyse starontheskyaredrawnindifferentcolours.Inthisexample,theplanetis barelyintransitabilityonthesecondarystar,butnotontheprimary. circumbinarytransitprobabilitiesinthepresenceofprecession. Thesecondlevelofcomplexityistoincludethefullextentof the stellar orbits, meaning that a value of I offset from π/2 may p stillexhibittransitability.Considerthelimitingcaseoftransitabil- aretheindividualsemi-majoraxesforthetwostarsand ity.Thisiswhentheplanetandbinaryorbitsbarelyoverlapwhen M |Ip−π/2|isataminimum(dIp/dt=0).Wecalculatetheorientation µB,A= MA+B,AMB (14) ofthebinaryandplanetorbitsinthisconfiguration.TakeEq.1and rearrangeittoisolatethetermcontaining∆Ω: arethereducedmasses.Similarlyfortheplanet, (cid:12) (cid:12) (cid:12) π(cid:12) cos∆I−cosI cosI Xp=apsin(cid:12)(cid:12)(cid:12)Ip− 2(cid:12)(cid:12)(cid:12). (15) cos∆Ω= sinI sinbinI p. (8) ThereistransitabilityonstarsAand/orBwhen bin p InEq.8onlyI and∆Ωaretime-dependentquantities.Differenti- p atingbothsidesofEq.8withrespecttotimeleadsto X <X , (16) p A,B Accordingtotheorbitaldynamics(Eq.5),theplanetwillentertran- −sin∆Ωd∆Ω =(cid:34)cosI sinI dIp sinI sinI sitabilityatsomepoint(fulfillingEq.16)ifthefollowingcriterion dt bin p dt bin p ismet: (cid:35) /−(si(nc2osI∆Isi−n2coIs)I.bincosIp)sinIbincosIpddItp (9) ∆I>(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)π2 −Ibin(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)−βA,B(Ibin), (17) bin p wheretheangleβ isafunctionofthebinaryextentonthesky,as BysubstitutingdI /dt=0intoEq.9weget A,B p seenbytheplanet.Thisisthesecond-ordercriterionfortransitabil- ity.CombineEqs.11and17toget d∆Ω sin∆Ω =0. (10) dt (cid:12) (cid:12) (cid:12) (cid:12) (cid:12)π (cid:12) From Sect. 3 it is known that d∆Ω/dt = const (cid:44) 0, and hence (cid:12)(cid:12)Ibin−Ip(cid:12)(cid:12)>(cid:12)(cid:12)(cid:12)2 −Ibin(cid:12)(cid:12)(cid:12)−βA,B(Ibin). (18) sin∆Ω = 0, implying that ∆Ω = 0. This means that the limiting ByinsertingEqs.12and15intoEq.16andrearrangingtomatch caseoftransitabilityoccurswhentheascendingnodesofthebinary theformofEq.18,weobtain andplanetorbitsarealigned.Thissimplifiesthegeometryandcal- culations.AccordingtoSect.3,∆Ωisguaranteedtoequalzeroat (cid:34) (cid:12) (cid:12) (cid:35) sinocmlienaptoioinntcdaulrciunlgattihoenpinreEceqs.s1ioinspsiemripoldifi.Wedhteon∆Ω=0themutual βA,B(Ibin)=sin−1 aaAp,B sin(cid:12)(cid:12)(cid:12)(cid:12)π2 −Ibin(cid:12)(cid:12)(cid:12)(cid:12)+ RaAp,B , (19) wherethequantitiesinsidethesquarebracketsaresufficientlysmall ∆I=|I −I |. (11) thatwecanusethesmallangleapproximation. bin p Thesizeofβ determineshoweasyitisforasystemtoful- A,B InFig.6weshowacircumbinarysysteminthelimitingcase filthetransitabilitycriterioninEq.17.Dependingontheseparate oftransitability.Thisisa“side-on”viewoftheorbit,withtheob- valuesofβ andβ ,itispossibletofulfilthecriterionforjustone A B serverlocatedtotherightofthepage.Usingthisdiagram,wedefine ofthestars.Generallyβ >β exceptforeclipsingbinaries. B A halftheprojectedheightsontheskyofthetwostellarorbitstobe Totestthevalidityofthesecond-ordertransitabilitycriterion werannumericalN-bodysimulationsonthousandsofhypothetical (cid:12) (cid:12) (cid:12) π(cid:12) circumbinarysystems.ThedetailsareshowninAppendixA.The XA,B=aA,Bsin(cid:12)(cid:12)(cid:12)Ibin− 2(cid:12)(cid:12)(cid:12)+RA,B, (12) analyticcriterionisshowntobeveryaccurate,withanerrorless than0.1%.Allerrorcaseswerenearthelimitoftheinequalityin where Eq.17.Errorsariseduetosmallvariationsinthesemi-majoraxis a =a µ (13) and eccentricity, which in the limiting case of transitability may A,B bin B,A (cid:13)c 2014RAS,MNRAS000,1–?? Analyticcircumbinarytransitprobability 5 leadtoacontraryresulttothepredictionofEq.17.Thisiselabo- (a) rateduponinSect.8.5.2. celestial! sphere 5 PROBABILITYOFTRANSITABILITY planet at! Foragivensetoforbitalparameters,wecancalculatetheproba- inclination! extremum bilitythatagivenobserverwillobservetransitability,atsometime ϑ A duringtheprecessionperiod.Asaninitialapproximation,weuse ζ A the first-order transitability criterion (Eq. 7). The orientation of a A circumbinarysystemontheskyisuniformlyrandom.Ittherefore followsthatcosI hasauniformdistribution,andhencetheprob- bin abilitydensityfunctionis ϑAζAA 2ΔI p(Ibin)=sinIbin. (20) 2ΔI+2ζAA ByintegratingthisbetweentheboundsspecifiedbyEq.7,weob- tainanapproximateprobabilityoftransitability: planet at! inclination! (cid:90) π/2 extremum P = sinI dI A,B bin bin π/2−∆I =sin(∆I). (21) The probability is period-independent and non-zero, except for strictlycoplanarsystems. (b) Thenextstepistoincludethefiniteextentofthebinaryor- R bit.Theprobabilityoftransitabilityisafunctionofthesizeofthe A solid angle subtended on the celestial sphere such that the planet a a p ϑ and stellar orbits are seen overlapping, including the full orbital A A evolution.WedemonstratethisinFig.7a.Tosimplifythediagram, ΔI ζAA a weonlydrawtheanglesfortransitabilityontheprimarystar,but A R thecalculationproceedsidenticallyforthesecondary.Theplanet A orbitisshownintwodifferentpositions,correspondingtothetwo extrema of I separated by 2∆I (Eq. 5). The angles ζ and θ p A,B A,B Figure7.In(a)weshowaside-onviewofacircumbinarysystemwiththe aretheanglessubtendedbytheplanetintransitabilityoneachstar. They are functions of how big the stellar orbit is, as seen by the planetshownatitstwoextremevaluesofIp.Bothstarsareplottedtwice indifferentcoloursatclosestandfarthestseparationfromtheplanet.The planet,atclosest(ζ )andfarthest(θ )separation.InFig.7bwe A,B A,B anglessubtendedbytheorbitonthecelestialsphereareonlyshownforthe zoomintoseehowtheanglesaredefined3 primarystar,toavoidclutter.Thehatchedregioncorrespondstoobservers whowouldseetheplanetintransitabilityatsomepointintime.In(b)we (cid:32)a sin∆I+R (cid:33) zoomintoseehowθandζaredefined. ζ =tan−1 A,B A,B (22) A,B a −a cos∆I p A,B and bandinFig.7ais (cid:90) π/2 (cid:32)a sin∆I+R (cid:33) PA,B= sinIbindIbin θ =tan−1 A,B A,B . (23) π/2−∆I−ζ A,B ap+aA,Bcos∆I =sin(∆I+ζA,B) Thelargerangleζiswhatcorrespondstothelimitoftransitability =sin(cid:32)∆I+ aA,Bsin∆I+RA,B(cid:33), (24) andhenceθdoesnotappearinanyfurtherequations. a −a cos∆I p A,B Allobserverswithinthehatchedbandwilleventuallyseetran- where because ζ is generally small we can apply the small angle sitability.Theprobabilityofanobserverbeingwithinthehatched approximation to remove the tan−1 function4. Equation 24 is the probabilityoftransitabilityontheprimaryand/orsecondarystars, 3 Thecurrentequationsarecorrectedfromtyposthatappearedinthepub- 4 Relatedtotheerrormentionedinfootnote3,Eq.24hasbeencorrected lishedversion:inEq.22thereusedtobea+signinthedenominatorand fromthepublishedversionwhichincorrectlycontainedan+signinthe inEq.23thereusedtobea−signinthedenominator.Wearesorryforthe denominator.Thiserrorwaspurelytypographicalandallresultsshownin error. thepaperwerecalculatedusingthecorrectequation. (cid:13)c 2014RAS,MNRAS000,1–?? 6 MartinandTriaud forabinaryofanyorientation.Theinclusionofζ addsaperiod- 1 Eq. 234 (A) dependencythatisabsentinEq.21. 0.9 Eq. 234 (B) InFig.8awedemonstrateEq.24onanexamplecircumbinary Eq. 245 (A) system,comprisedofabinarywith MA = 1M(cid:12),RB = 1R(cid:12), MB = 0.8 Eq. 245 (B) 0.5M(cid:12), RB = 0.5R(cid:12) and abin = 0.082 AU (Tbin = 7 days). The 0.7 Eq. 201 planetsemi-majoraxisisvariedfrom0.24AUto2AU.Thethree mutualinclinationsare0◦,5◦and10◦. 0.6 As∆Iisincreasedtheprobabilityoftransitabilityisincreased ,B A0.5 significantly. For the misaligned cases, PB > PA. In Fig. 8b we P 0.1 zoominonthecoplanarcase.Asacomparison,weshowthetransit 0.4 0.08 probabilityonasinglestarofradiusR ,calculatedusing A,B 0.3 0.06 0.04 R 0.2 P = A,B. (25) 0.02 A,B ap 0.1 0 0 2 4 Forcoplanarsystemstheprobabilityoftransitabilityreduces 0 0 10 20 30 40 50 60 to (cid:54) I (deg) R P = A,B , (26) Figure9.TheprobabilityoftransitabilityonstarsAandBcalculatedusing A,B ap−aA,B Eq. 24, and the approximation using Eq. 21. The bottom right image is zoomedintosmallmutualinclinations.Inthisplotthehorizontallinesare whichcomesfromsetting∆I=0inEq.24andusingasmallangle theequivalentsinglestartransitprobabilities(Eq.25). approximationtoremovethesinfunction.Thisequationmatches Welshetal.(2012),whoderivedananalyticestimateforthetran- sitprobability5 undertheassumptionof∆Ω = 0andstaticorbits, althoughitwasdulynotedthatcircumbinaryorbitsprecess.Equa- (cid:12)(cid:12)π (cid:12)(cid:12) R +R −2αR tpiaorntic2u6lairslyclaotssehtoortthpeersioindgsl.eTshtiasripsrboebcaabuilsietythbeutstsalrisghartleybhroiguhgehrt, sin(cid:12)(cid:12)(cid:12)2 −Ibin(cid:12)(cid:12)(cid:12)(cid:54) A aBbin B, (28) closertotheplanetsbytheirorbitalmotion. where α determines whether the criterion is for grazing eclipses ThecrucialdifferencetothesinglestarcaseisthatEq.25de- (α=0),fulleclipses(α=1)oranythinginbetween.Sinceeclipses creasestowardszeroforlargesemi-majoraxes,butforcircumbi- occurwhenI ≈π/2,wecanapplythesmallangleapproximation bin narysystemsthelimitofEq.24is inEq.28todeducethatthedistributionofI foreclipsingbinaries bin isuniformbetween lim P =sin∆I, (27) A,B ap→∞ π δ ± , (29) whichisequaltothefirst-orderderivationinEq.21.Thisapprox- 2 abin imateprobabilityisalsoapplicabletosystemswithalarge∆I be- wheretosimplifytheequationwehavedefined causetheangleζ,whichencompassestheperiod-dependence,be- δ=R +R −2αR . (30) comesrelativelysmall. A B B InFig.9wedemonstratehowtheprobabilityoftransitability Knowing that the binary eclipses with this uniform random varieswith∆I,usingthesamecircumbinarysystemasinFig.8a, distribution of Ibin, the probability of transitability is the fraction butfixingap=0.26AU.Inthebottomrightofthisfigurewezoom ofeclipsingbinarieswithIbin suchthattheinequalityinEq.17is innear∆I=0◦.Thecurveforthesecondarystarisseentoovertake satisfied. First, multiply Eq. 28 by PA,B and insert it into Eq. 17, thatoftheprimaryataround∆I=3◦. usingβA,BfromEq.19,toobtain (cid:32) (cid:33) δ a δ R ∆I=P −sin−1 P µ bin − A,B . (31) A,Ba A,B B,A a a a 6 CONSEQUENCESFORECLIPSINGBINARIES bin p bin p TheinequalityfromEq.17hasdisappearedsincewearecalculat- Eclipsingbinariesareonlyasmallfractionofthetotalbinarypopu- ing P for a given ∆I. Use the small angle approximation and lationbuttheeasiestbinariestodetectphotometrically.Itwassug- A,B rearrangetoobtain gestedbyBorucki&Summers(1984)thattheyarefavourabletar- gets for transit surveys because they positively bias the planetary (cid:32) (cid:33) 1 1 R orbit towards being aligned with the line of sight. In this section ∆I=P δ −µ − A,B. (32) wederivetheprobabilityoftransitability, PA,B,underassumption A,B abin B,Aap ap thatthebinaryisknowntoeclipse.Indoingsowequantifywhat BysolvingforP weget A,B wsc5nirotaiItatnseubfirsfiiialronicstgntyt,tnfthiohosaerttieearndanadsmieienerecis.Svltiaepttcisoote.nf4iuwsluafipslofinonrdttehhreeivpcirnaosgbeaEboqilf.it7eyc:oltifhpetsriacnnrgistietbarbiinoilanitryfi,eosdr.etsrTpahintee- PA,B= 1δ(cid:32)a∆1biIn+−RµaABp,,ABa1p(cid:33) iiff∆∆II(cid:62)<∆∆IIlim , (33) lim (cid:13)c 2014RAS,MNRAS000,1–?? Analyticcircumbinarytransitprobability 7 (a) (b) 0.25 0.025 Eq. 234 (A) Eq. 234 (A) Eq. 234 (B) Eq. 234 (B) 0.2 ΔI = 10° Eq. 21 0.02 Eq. 245 (A) Eq. 245 (B) 0.15 0.015 B B , , A A ΔI = 5° P P 0.1 0.01 ΔI = 0° 0.05 0.005 ΔI = 0° 0 0 0 0.5 1 1.5 2 0 0.5 1 1.5 2 a (AU) a (AU) p p Figure8.In(a)weshowtheprobabilityoftransitabilityonstarsAandB(Eq.24)asafunctionofap,forthreedifferentmutualinclinations.Thehorizontal dashedlinesarecalculatedusingthefirst-orderapproximateprobability(Eq.21).In(b)wezoominonthe∆I = 0◦ case.Asacomparison,weshowthe equivalentsinglestarprobabilityindashedlines(Eq.25). wherewedefine ismorelikelytofindplanetstransitingasinglestarwhereanother (cid:32) (cid:33) transiting planet has already been found, compared to around a 1 1 R ∆I =δ −µ − A,B, (34) randomstar.Thisisbecausethemutualinclinationdistributionof lim a B,Aa a bin p p multi-planetsystemsisnotisotropicbutweightedtowardscopla- narity6. This is analogous to a circumbinary system, if one con- in order to truncate P at 1. As an example, for a binary with A,B solar and half-solar masses and radii and a = 0.3 AU orbited sidersthesecondarystarasthe“innerplanet”.Thereis,however, byaplanetata = 1.0AU,α = 0.5andcobipnlanarorbits,Eq.33 afundamentaldifferencebetweensingleandbinarystars:planets yields P = 0.3p3 and P = 0.19. Coplanar orbits correspond to orbitingsinglestarsdosooneffectivelystaticorbits. A B a minimum value of P . A slight increase in ∆I to 0.5◦ raises A,B theseprobabilitiesto0.96and0.89,respectively.For∆I = 1◦ the probability on both stars is 1. The circumbinary geometry is evi- 7 CONNECTINGTRANSITABILITYTOTRANSITS dentlyveryfavourablefortransitabilityoneclipsingbinaries.We noteforreferencethatthemeanmutualinclinationinthetransiting 7.1 Doestransitabilityguaranteetransits? circumbinary planets found so far is 1.73◦ (see Table 1) and that theSolarSystemmutualinclinationdistributionroughlyfollowsa Transitabilityaloneisnotdetectableviaphotometry,onerequires 1◦Rayleighprofilerelativetotheinvariantplane(e.g.Clemence& anactualtransit.Afundamentalelementofthedefinitionoftran- Brouwer1955;Lissaueretal.2011) sitability is that transits are possible but not guaranteed on any For∆I >∆I twothingsoccur:1)transitsareguaranteedon given passing of the binary orbit. By having a transit probability lim eclipsingbinariesofanyorientationand2)transitsbecomepossible between0and1foreachpassing,itisintuitivetothinkthatatran- onnon-eclipsingbinaries.InFig.10weplot∆I asafunctionof sitwilleventuallyhappenifobservedcontinuouslyforasufficiently lim a anda ,forabinarywithsolarandhalf-solarmassandradius longtime.ThisconclusionwassharedbySchneider(1994);Welsh bin p and a between 0.007 and 0.2 AU, where the lower limit corre- etal.(2012);Kratter&Shannon(2013);Martin&Triaud(2014). bin spondstoacontactbinary:a =R +R .Fortheeclipsecriterion Wetestedthishypothesisbynumericallysimulatingcircumbi- bin A B weusedα=0.5.Thewhiteemptyspaceontheleftistheunstable nary systems over 50 precession periods and looking for transits regionaccordingtoEq.4. incaseswheretransitabilityoccurred.Thiswasfirstdoneforcom- Thebinarysemi-majoraxisisthebiggestfactorinthecalcu- pletelyrandomsystemstakenfromthetestsinAppendixA.Thede- lation.ForthesystemsinFig.10wherecircumbinaryplanetshave tailsareprovidedinAppendixB.Lessthan0.3%ofsystemsman- beenfoundsofar(a >0.08AU),∆I islessthan3◦. agedtoevadetransit.Alloftheseexceptionalcasescorresponded bin lim For closer binaries ∆I rises sharply, reaching a maximum to the limit of transitability, where the planet only spends a very lim of38◦foracontactbinary.Transitsonveryshort-periodeclipsing short time in transitability, and hence the chance of transiting on binariesareofcoursepossiblebutEq.34showsthatnotallsuch binaries can be transited unless there is significant misalignment. InSect.8.4weapplythisworktotheKeplerdiscoveriessofar. 6 Ragozzine&Holman(2010)founditresemblesaRayleighdistribution. There are similarities between transits on eclipsing binaries Thedistributionof∆Iincircumbinarysystemsispresentlyunknown,be- and studies of multi-transiting systems orbiting single stars (e.g. cause the detections so far have been highly biased towards coplanarity Ragozzine&Holman2010;Gillonetal.2011).Geometrically,one (Martin&Triaud2014). (cid:13)c 2014RAS,MNRAS000,1–?? 8 MartinandTriaud 0.2 g n25 1 i t i s t n i a20 m r i t ) l (AU 0.1 lity 2 tems15 bin abi sys 25 a t 4 10 s f 20 o 15 e 8 g Numerical (A) 10 a 5 16 t Numerical (B) 5 32 n 00 0.5 1 1.5 2 rce EEqq.. 223344 ((AB)) 00 2 4 6 8 a (AU) e 0 p P 0 50 100 150 200 250 300 Observing time (yr) Figure10.3Dhistogramoftheminimummutualinclinationneededinde- greestoguaranteetransitsonEBsofanyorientation,atdifferentbinaryand Figure11.Thepercentageofsystemsseentransitingasafunctionoftime planetsemi-majoraxes(Eq.34). fromnumericalsimulationsof10,000circumbinarysystems,andtheana- lyticpredictions(Eq.24).Systemsarecountedastransitingafterthedetec- tionofasingletransit.Thezoomedfigureinthebottomrightcornershows thepercentageoftransitingsystemsoverKepler-likeobservingtimes.The anygivenorbitissmall.Transitsareexpectedtooccureventually, blackverticallinedenotestheprecessionperiod. butafteratimelongerthanwhatwassimulated. Second,weconstructedsystemswith4:1and5:1periodcom- mensurabilities,specificallydesignedtomakeplanetspermanently to know how long an observer must wait to see a transit. It is a evade transit. The evasion percentage increased slightly but re- strongfunctionoftheprecessionperiod,sincethatdetermineshow mainedlessthan1%.Itisprobablethatthisvaluewouldeventually spacedaparttheregionsoftransitabilityare. droptozero,butlongersimulationswouldberequired.Theplanets An analytic calculation of the time-dependent transit proba- inevitablytransitbecauseexactperiodcommensurabilitiesarenot bility is outside the scope of this paper, and has been previously sustainable,owingtoperturbationsfromthebinaryontheplane- labelledimpossible(Schneider&Chevreton1990).Weinsteaduse taryorbit(Sect.8.5.2).AsideforHD202206whichhasaperiod numericalN-bodysimulations. rationear5:1(Correiaetal.2005),periodcommensurabilitieshave InFig.11wedemonstratethepercentageofsystemsseentran- notbeenobserved.Thissystemmaynotberepresentative,sinceit sitingstarsAand/orBasafunctionoftime,using10,000simulated straddlestheborderbetweenacircumbinaryandatwo-planetsys- circumbinary systems. The primary and secondary stars are solar tem; the secondary “star” has a minimum mass of 15MJup. It has andhalf-solarinmassandradius,T =7days,T =40daysand alsobeentheorisedbyKley&Haghighipour(2014)thatcircumbi- ∆I = 10.Overtime,thepercentagebinoftransitingspystemsreaches naryplanetsshouldformbetweenintegerperiodratios,notinthem. thevaluepredictedbyEq.24,inagreementwiththeconclusions Whilst not an exhaustive proof, our tests indicate that in the ofSect.7.1.Mostofthetransitingsystemshavedonesowithina vastmajorityofcases,transitabilityindeedleadstotransit,albeitat singleprecessionperiod(here∼sevenyears). anunspecifiedpointintime. Asanextendedtest,wetookthesystemsfoundtransitinginin Sect.7.1andcalculatedthetimetakenforprimaryandsecondary transits to occur. The results are provided in Table C1. Whilst a 7.2 Transitsovertime largermutualinclinationleadstomoreplanetstransiting,theme- Theprobabilityoftransitabilityisequivalenttotheprobabilityof dianwaittimeisincreased.Generally,asignificantnumberofsys- transit, granted the observer has infinite time. Unfortunately, due temsarefoundtransitingwithinKepler-likemissiontimes. to limitations in technology, funding and human life-expectancy, onemuststrivetocapturetransitswithinafinitetime.Wecalcu- latesomeexampleobservationtimesneededinordertoobservea 8 DISCUSSIONANDAPPLICATIONS transit. Inthecaseofasinglestar,aftercontinuousobservationsofa 8.1 Thecircumbinaryplanetsdiscoveredsofar timeequaltoT ,theplaneteitherwillorwillnothavetransited. p Our first application is to calculate the transitability probabilities Thisisnotthecaseforcircumbinaryplanets,fortworeasons: for the Kepler discoveries so far, assuming of course that we do • Theplanetmaycurrentlybeoutsideoftransitability,butwill not have a priori knowledge of transits and eclipses7. In Table. 1 precessintotransitabilityatalatertime. wecalculatetheprobabilityoftransitabilityonbinariesofanyori- • Theplanetmaycurrentlybeinsidetransitability,butthecon- entation(Eq.24)andoneclipsingbinaries(Eq.33),whereforthe junctionrequiredforatransithasnotyetoccurred. latterweusedα=0.5todefineeclipses.Theequivalentsinglestar The fraction of circumbinary planets transiting therefore in- creaseswithtime,uptoavaluespecifiedbyEq.24.Itisimportant 7 Otherwiseyouwouldhaveaboringtablefullof100%’s. (cid:13)c 2014RAS,MNRAS000,1–?? Analyticcircumbinarytransitprobability 9 Table1.ProbabilitiesoftransitforthecircumbinaryplanetsdetectedsofarbyKepler. Name MA MB RA RB abin ap ∆I PA,B%(all) PA,B%(EBs) PA,B%(single) (M(cid:12)) (M(cid:12)) (R(cid:12)) (R(cid:12)) (AU) (AU) (deg) A B A B A B Kepler-16 0.69 0.20 0.65 0.23 0.22 0.71 0.31 1.04 0.91 75.5 66.1 0.42 0.15 Kepler-34 1.05 1.02 1.16 0.19 0.23 1.09 1.86 4.18 4.16 100 100 0.50 0.47 Kepler-35 0.89 0.81 1.03 0.79 0.18 0.60 1.07 3.11 2.94 100 100 0.78 0.61 Kepler-38 0.95 0.27 1.78 0.27 0.15 0.46 0.18 2.28 0.79 41.2 14.4 1.80 0.27 Kepler-47b 1.04 0.46 0.84 0.36 0.08 0.30 0.27 1.93 1.26 39.5 25.8 1.30 0.56 Kepler-47c 1.04 0.46 0.84 0.36 0.08 0.99 1.16 2.48 2.32 50.8 47.6 0.39 0.17 Kepler-64 1.50 0.40 1.75 0.42 0.18 0.65 2.81 6.54 6.66 100 100 1.25 0.30 Kepler-413 0.82 0.52 0.78 0.48 0.10 0.36 4.02 8.98 9.18 100 100 1.01 0.62 KIC9632895 0.93 0.19 0.83 0.21 0.18 0.93 2.30 4.68 5.09 100 100 0.49 0.12 Refs:Doyleetal.(2011);Welshetal.(2012);Oroszetal.(2012a,b);Schwambetal.(2013);Kostovetal.(2013,2014) Welshetal.(2014) Note:Kepler-47disexcludedbecauseithasnotyetbeenpublishedandlacksavaluefor∆I. probabilitywascalculatedusingEq.25.Inthetableweincludeall EBs on which there would be transitability by a putative planet, necessary variables for the calculations. In more than half of the with∆Ibetween0◦and10◦anda /a =3,10,20.Theresultsare p bin cases,transitsareguaranteedoneclipsingbinariesofanyorienta- only shown for the primary star, since the plot for the secondary tion. starisindistinguishable. Consistentwithearliersections,thebiggestfactoristhemu- tual inclination, with a higher ∆I leading to a greater chance of 8.2 Multi-planetcircumbinarysystems transitability.Transitabilityisfavouredforsmallervaluesofa ,but p thisdependencydiminishesatlargermutualinclinations. Onlyonemulti-planetcircumbinarysystemhasbeendiscoveredso The Kepler EB catalog would benefit from extended photo- far(Kepler-47,Oroszetal.2012b).Kratter&Shannon(2013)con- metric observations by the future PLATO telescope (Rauer et al. sideredaneclipsingbinarywithaknowntransitingplanet,andcal- 2014),inordertofindnewcircumbinaryplanetsthathavemoved culatedthelikelihoodofasecondplanetbeingseentransiting.They intotransitabilityduringthe∼eightyearsbetweenmissions.This derived an analytic probability for whether or not the binary and mayincludeadditionalplanetsinknowncircumbinarysystems. planetorbitswouldoverlaponthesky,undertheassumptionthat thebinaryisperfectlyedge-on(I =π/2).Infact,whattheycal- bin culatedwastheprobabilityoftransitability.Theirderivationdoes notincludeprecession,andconsequentlyunderestimatestheprob- 8.4 Onthedearthofplanetsaroundshort-periodbinaries ability. Anobservedtrendhasbeenthelackofcircumbinaryplanetsaround BasedontheworkinSect.6,anyadditionalplanetswith∆I theclosestbinaries;theshortest-periodbinaryhostingaplanetis greaterthanthefirsttransitingplanetareguaranteedtoentertran- Kepler-47 with T = 7.4 days (Orosz et al. 2012b). This is de- sitabilityatsomepoint. bin spite the median10 period of the EB catalog being 2.8 days. This raisesvariousquestionsabouttheabilitytoformplanetsinsuchan 8.3 Kepler’seclipsingbinarycatalog environment,particularlyinthepresenceoftertiarystellarcompan- ion,asisoftenthecaseforverytightbinariesaccordingtotheory The Kepler telescope, with its four years of continuous observa- (Mazeh&Shaham1979;Fabrycky&Tremaine2007)andobser- tions and exquisite precision, has provided the most comprehen- vations(Tokovininetal.2006). sive catalog of EBs to date (Slawson et al. 2011). We used the The reason why EBs are preferentially found at short peri- online beta version of this catalog8 to test our transitability crite- ods despite a smaller natural occurrence (Tokovinin et al. 2006) riononhypotheticalorbitingplanets.Fromthecatalogweobtained isbecausethereisagreaterrangeof I thatallowforaneclipse bin M1,M2,R1,R2 and abin, which were derived from stellar temper- (Eq.29).WhentheEBishighlyinclined,however,theplanetitself atures calculated in Armstrong et al. (2013) using a method ex- needs a greater misalignment in order for transitability to occur. plainedinArmstrongetal.(2014).Onlysystemswithamorphol- WedemonstratethisinFig.13,wherewecalculatetheminimum ogyparameterlessthan0.5wereused,correspondingtodetached ∆IneededtoseetransitabilityontheprimarystarofeachEBinthe EBs(seeMatijevicetal.2012fordetails).Thebinaryinclination Keplercatalog,takinga /a = 3.5.Asareference,weshowthe p bin wasrandomisedbetweentheboundsdefinedinEq.299. meanandmaximummutualinclinationsfromtheKeplerdiscover- The remaining quantities needed for Eq. 17 are ∆I and ap. iessofar,althoughthesearebiasedtowardsbeingsmall(Martin& Giventhedistributionofcircumbinaryplanetsispresentlypoorly Triaud2014). knownandsubjecttostrongbiases,weconsideredawiderangeof ForT <7.4days,amisalignmentof1.73◦resultsin60%of bin potentialvalues.InFig.12wecalculatedthepercentageofKepler planetsmissingtransitability.Forthosesystemsmisalignedenough for transitability on the shortest period binaries, there should be 8 http://keplerebs.villanova.edu/maintainedbyAndresPrsaetal. 9 ItispossibletoobtainatruevalueofIbin,howevertheonlypublished versionisinthenow-outdatedcatalogofSlawsonetal.(2011),andcontains 10 Inthepublishedversionofthepaperweaccidentallywrotethatthiswas errors. the“mean”period. (cid:13)c 2014RAS,MNRAS000,1–?? 10 MartinandTriaud Percentage with transitability on A11234567890000000000000 5 ∆ I 1(0deg) aaa ppp1 5=== 312 00a aab ibbniinn2 0 ∆Minimum I for transitability (deg)110505 1 P P∆1 0bI i=(n4d=.a7∆0y. 2s4I°) = d1(a.Ky7es3p1 °l 0(e(0Krme−ep4al1ne3)r)−47) bin Figure12.ThepercentageofEBsfoundbyKepleronwhichtherewould Figure13.Theminimummutualinclinationneededfortransitabilityon betransitabilitybyaputativeplanetwithdifferentvaluesof∆Iandap/abin. eachoftheEBsfoundbyKepler. transitswithintheKeplertimeseries,fortworeasons:1)thepre- cessionperiodisonlyacoupleofyearslong11,sotheplanetand 1 binary orbits would have intersected at least once during the Ke- plermissionand2)theseareverytightsystems,sowearelikelyto 0.8 haveobservedoneorprobablymoretransitswhilstintransitability. Thedearthofplanetsmayalsobeexplainedbystellarnoise 0.6 inthelightcurves.Binarieswithperiodsthisshortareexpectedto 0.4 betidallylocked,whichleadstofasterrotationandincreasedstar π / spots,whichmayinhibitdetections. ) 0.2 Ω Thecurrentnulldetectionlikelyremainssignificantforplan- ∆ 0 etsthataremisalignedbyatleastafewdegrees,buttheremayre- ( n mainsomecoplanaronesthatareundetectablebyKepler.Transits i −0.2 s oncontactbinariesrequireanevenhigherlevelofmutualinclina- I tion.Discoveriesaroundcontactbinariesmayalsobehinderedby ∆ −0.4 Kepler’s30-minutecadence,whichispotentiallytoolongtoade- −0.6 quatelysampleitsorbitinthesearchfortransits. −0.8 8.5 Limitations −1 −1 −0.5 0 0.5 1 8.5.1 EccentricSystems ∆Icos(∆Ω)/π Theadditionofeccentricity,toboththebinaryandplanetorbits,in- troducestwocomplexities.First,thegeometryiscomplicatedsince Figure14.ThesamecircumbinarysystemsasinFig.3butwithebin=0.5. welosecircularsymmetry,andtherearetwoadditionalanglesto consider:ω andω (theargumentsofperiapse).Furthermore,the bin p orbitaldynamicscauseω tobetime-dependent,furthercomplicat- p The simulations show that the error is only of order ∼ 2% (Ap- ingthesituation. pendixD).Furthermore,theresultssuggestthateccentricityactu- Second,theprecessioncycleismorecomplexwhenthebinary allymakestransitabilitymorelikely,althoughamoredetailedstudy iseccentric.InFig.14wedemonstratetheprecessionofthesame systemasinFig.3,butwithe = 0.5.Themutualinclinationis isneededtoconfirmthis. bin nolongerconstant.Therearetwoislandsoflibration,centredon ∆Ω = 0 and ∆I = π/2 (red) and ∆I = −π/2 (magenta), within 8.5.2 Additionaldynamicaleffects which ∆Ω does not circulate through 0 to 2π. We therefore lose twooftheassumptionsmadeinSect.4. Acircumbinarysystemisathree-bodyproblemandhencenotsolv- Werannumericalsimulationstotesttheabilityofourcriterion ableanalytically.OuranalytictreatmentofitasapairofKeplerians inEq.17topredicttransitability,inthecaseofeccentricsystems. plusorbitalprecessionencompassesthemajorityofthephysics,but neglectssomesmalleramplitudeeffects. Thesemi-majoraxisandperiod,whichweassumedtobecon- 11 IfweassumetheobservedoverdensityofplanetsatTp∼5Tbinextends stant, experience slight variations over time because of perturba- toveryclosebinaries. tions from the binary. There are also small variations in the ec- (cid:13)c 2014RAS,MNRAS000,1–??

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.