Astronomy&Astrophysicsmanuscriptno.paper c ESO2011 (cid:13) January13,2011 Mass constraints on substellar companion candidates from the re-reduced Hipparcos intermediate astrometric data: Nine ⋆ confirmed planets and two confirmed brown dwarfs SabineReffertandAndreasQuirrenbach ZAH-Landessternwarte,Ko¨nigstuhl12,69117Heidelberg,Germany Received¡date¿/Accepted¡date¿ 1 1 ABSTRACT 0 2 Context.The recently completed re-reduction of the Hipparcos data by van Leeuwen (2007a) makes it possible to search for the astrometric signatures of planets and brown dwarfs known from radial velocity surveys in the improved Hipparcos intermediate n astrometricdata. a Aims.Ouraimistoputmoresignificantconstraintsontheorbitalparameterswhichcannotbederivedfromradialvelocitiesalone, J i.e.theinclinationandthelongitudeoftheascendingnode,thanwaspossiblebefore.Thedeterminationoftheinclinationinparticular 1 allowstocalculateanunambiguous companion mass,ratherthanthelower masslimitwhichcanbeobtainedfromradialvelocity 1 measurements. Methods.Wefittedtheastrometricorbitsof 310substellarcompanions around258 stars,whichwerealldiscovered viatheradial ] velocitymethod,totheHipparcosintermediateastrometricdataprovidedbyvanLeeuwen. P Results.Eventhoughtheastrometricsignaturesofthecompanionscannotbedetectedinmostcases,theHipparcosdatastillprovide E lowerlimitsontheinclinationforallbut67oftheinvestigatedcompanions,whichtranslatesintoupperlimitsonthemassesofthe . unseencompanions.Forninecompanionsthederiveduppermasslimitliesintheplanetaryandfor75companionsinthebrowndwarf h massregime,provingthesubstellarnatureofthoseobjects.Twoofthoseobjectshaveminimummassesalsointhebrowndwarfregime p andarethusproventobebrowndwarfs.TheconfirmedplanetsaretheonesaroundPollux(βGemb),ǫErib,ǫRetb,µArab,υAndc - o andd,47UMab,HD10647bandHD147513b.TheconfirmedbrowndwarfsareHD137510bandHD168443c.In20cases,the r astrometricsignatureofthesubstellarcompanionwasdetectedintheHipparcosdata,resultinginreasonableconstraintsoninclination t andascendingnode.Ofthese20companions,threeareconfirmedasplanetsorlightweightbrowndwarfs(HD87833b,ιDrab,and s a γCepb),twoasbrowndwarfs(HD106252bandHD168443b),andfourarelow-massstars(BD–04782b,HD112758b,ρCrBb, [ andHD169822b).Oftheothers,manyareeitherbrowndwarfsorverylowmassstars.ForǫEri,wederiveasolutionwhichisvery similartotheoneobtainedusingHubbleSpaceTelescopedata. 1 v Keywords.Astrometry–binaries:spectroscopic–planetarysystems–Stars:low-mass,browndwarfs 7 2 2 1. Introduction calledplanetorbrowndwarfcandidatessincetheirtruemassis 2 notknown. 1. Most of the around 400 planets and planet candidates known In order to establish substellar companion masses, a com- 0 sofar1havebeendetectedwiththeradialvelocitymethod.Asis plementary technique is required which is sensitive to the in- 1 verywellknown,onlyfiveofthesevenorbitalparameterswhich clination.Forcompanionswithinclinationscloseto90 ,transit ◦ 1 characterize a binary orbit can be derived from radial velocity photometrycanprovidetheinclinationandthus,amongotherin- : observations(period,periastron time, eccentricity,longitudeof v formation,anaccuratecompanionmass.Formostsystemshow- periastronandeitherthemassfunction,theradialvelocitysemi- i everthisisnotthecase,andastrometryisthemethodofchoice. X amplitude, the semi-major axis of the companion orbit or the Astrometry provides both missing parameters, the inclination r semi-majoraxisoftheprimaryorbittimesthesineoftheincli- and the ascending node, so that the orbits are fully character- a nation). The remaining two orbital parameters, inclination and ized.Inprinciple,allorbitalparameterscanbederivedfromas- longitudeof the ascendingnode2, are related to the orientation trometric measurements, but since with current techniques the oftheorbitinspaceandcannotbederivedfromradialvelocities radialvelocityparametersareusuallymoreprecise,anotherop- alone.Unfortunately,thismeansthatthetruecompanionmassis tionistoderiveonlythetwomissingorbitalparametersfromthe usuallynotknownforplanetsdetectedviaradialvelocities;the astrometry. onlyquantitywhichcanbederivedisalowerlimitforthecom- The astrometric signatures of planetary companions are panion mass, which is the true companionmass times the sine rather small compared to the astrometric accuracies which are of the inclination. Strictly speaking, all substellar companions currentlyachievable.Theastrometricsignatureα canbecal- max detectedviatheradialvelocitytechniquewouldthushavetobe culated if the radial velocity semi-amplitude K , the period P, 1 theeccentricitye,theinclinationiandtheparallax̟areknown ⋆ BasedonobservationscollectedbytheHipparcossatellite (seee.g.Heintz1978oranyotherbookondoublestars): 1 seehttp://www.exoplanets.org/ 2 In the following, we denote this orbital element with ‘ascending K P √1 e2 ̟ node’,shortfor‘longitudeoftheascendingnode’ αmax = 11·AU· 2π−sin·i (1) · · 2 Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata α corresponds to the semi-major axis of the true orbit of 14 planetary companions, the derived upper mass limits imply max the photocenter around the center of mass. This is however in that the companionsare of substellar nature, but for the others generalnot the astrometric signature which is observable from nousefuluppermasslimitscouldbederived.Similarly,evenfor Earthinthecaseofeccentricorbits,duetoprojectioneffects.In thebrowndwarfstheHipparcosastrometryisingeneralnotpre- theworstcase,thesemi-majoraxisoftheprojectedorbitcorre- ciseenoughtoderivetightupperlimitsortoestablishtheastro- sponds to the semi-minor axis of the true orbit only, while the metricorbit.However,forsixofthe14browndwarfcandidates semi-minoraxisoftheprojectedorbitcouldbeidenticaltozero, itturnedoutthatthecompanionwasstellar,andtheastrometric so that no astrometric signal can be observed along that direc- orbitcouldbederived.ThisconfirmstheresultsofHalbwachset tion. The astrometric signature α which is actually observable al. (2000), who examined the Hipparcos astrometry for eleven fromEarthcan becalculatedwith theformulaeforellipse pro- stars harboringbrown dwarf candidates. Seven of those brown jectiongiveninAppendixA. dwarf secondaries turned out to be of stellar mass, while only Iftheinclinationisunknown,asisthecasewhentheorbitis oneofthestudiedcompanionsis,withlowconfidence,abrown derivedfromradialvelocitiesalone,Eq.1providesalowerlimit dwarf.Fortheotherthreecandidates,nousefulconstraintscould fortheastrometricsignaturebysettingsini=1andmultiplying be derived. In a previous paper (Reffert & Quirrenbach 2006), with1−e2(convertingthesemi-majorintothesemi-minoraxis): wederivedmassesof37+3169MJupfortheoutercompanioninthe HD 38529 system, and−of 34 12 M for the outer compan- Jup K P (1 e2)3/2 ̟ ion in the HD 168443system±, based on Hipparcosastrometry. α = 1· · − · (2) min 1AU 2π Thisestablishedthebrowndwarfnatureofbothobjects.Mostre- · cently,Sozzetti&Desidera(2010)havefollowedthatsameap- We notethatamorestringentlowerlimitontheobservableas- proachtoderivemassesforthetwobrowndwarfcandidatesor- trometricsignaturecouldbeobtainedifthelongitudeofthepe- bitingHD 131664andHD 43848,respectively.Witha massof riastron is taken into account;the abovelower limit for the as- 23+26 M , the companion to HD 131664 is indeed a brown trometricsignaturecorrespondstothecasewherethelongitude dw−a5rf,wJhuiplethecompaniontoHD43848turnedouttobestel- oftheperiastronis±90◦. larwithamassof120+14637MJup. Instead of the semi-amplitude K1, one could also use the Using the Fine G−uidance Sensors (FGS) of the Hubble massfunctionorthesemi-majoraxisoftheprimaryorbit,mul- SpaceTelescope(HST),anastrometricprecisionsomewhatbet- tiplied by sini, as an alternative orbital element to express the ter than in the original Hipparcos Catalogue can be achieved. unique relation between astrometric signature and various or- Single measurement accuracies of around 1 mas and paral- bitalelements. laxes as accurate as about 0.2 mas have been determined Equation1 showsthatwe canalwaysderiveanuppermass (Benedictetal.2007, Benedictetal.2009) with HST; the most limit for the companion based on astrometry if the companion accurate parallaxes in the original Hipparcos Catalogue are masstimesthesineoftheinclination,m sini,isknownfrom 2 around 0.4 mas (for a few bright stars). An upper mass limit · radialvelocities.Theastrometryprovidesanupperlimitonthe ofabout30M ,ataconfidencelevelof99.73%,wasderived Jup astrometricsignatureof the companion,whichtranslatesintoa byMcGrathetal.(2002)forthecompanionto55Cncb,based lower limit on the inclination (Eq. 1). In combination with the on HST/FGS data. Benedict et al. (2002) reported the first as- minimumcompanionmass m sini, the lower limiton the in- 2 trometricallydeterminedmassofanextrasolarplanet.Theyde- · clination translates into an upper limit of the companion mass termined the mass of the outermost planet orbiting GJ 876 as m . 2 1.89 0.34M ,at68.3%confidence,basedalsoonHST/FGS Jup Forexample,theastrometricsignatureofacompanionwitha ± data.ForthecompaniontoǫEri,Benedictetal.(2006)obtained massof1M andaperiodoffiveyearsorbitingaroundasolar- Jup amassof1.55 0.24M ,againat68.3%confidenceandbased Jup mass star located at a distance of 10 pc amounts to 0.28 mas, on HST/FGS d±ata. For the planetcandidate aroundHD 33636, not taking projection effects into account. Although this value Beanetal.(2007)derivedamassof0.14 0.01M withthesame israthersmall,severalastrometricdetectionsofsubstellarcom- method,implyingthatthecompanionisa±low-ma⊙ssstarandnot panionshavebeenreportedinthepast,mostlyforsystemswith aplanetorabrowndwarf. rather long periods and relatively massive companions around Mostrecently,twobrowndwarfswereconfirmedwithHST nearbystars,whichallhelptoincreasetheastrometricsignal. astrometry:HD136118bhasamassof42+11M (Martioliet BasedontheHipparcosdata,Perrymanetal.(1996)derived 18 Jup al.2010),andHD38529chasamassof17.−6+1.5M (Benedict upper mass limits of 22 MJup for the substellar companion to 1.2 Jup et al. 2010). In the υ And system, the inclin−ationsoftwo com- 47UMaandof65M forthecompanionto70Vir,with90% Jup panionscouldbemeasured,whichnotonlyallowedforthede- confidence. This confirmed for the first time the substellar na- termination of their masses (13.98+2.3 M for υ And c and ture of these newly detected companions. For 51 Peg with its 5.3 Jup period of only a few days no useful upper mass limit could be 10.250.37.3 MJup for υ And d, McArth−ur2010), butalso the mu- establishedbasedonHipparcosdata. tual in−clination could be shown to be 29.9 1◦. This is the first ± Mazeh et al. (1999) and Zucker&Mazeh(2000) followed suchmeasurementandshowsthepotentialofastrometryforthe that same approach to derive masses or upper mass limits for measurement of the 3-dimensional orbit geometry in multiple the companionsto υ And and HD 10697, respectively. For the systems. outermostcompanionintheυAndsystem,Mazehetal.(1999) Themedianprecisionofpositionsandparallaxesintheorigi- deriveda massof 10.1+4.7 M ata confidencelevelof 68.3% nalversionoftheHipparcosCatalogueisjustbetterthan1mas, 4.6 Jup andamassof10.1+9.5M− ataconfidencelevelof95.4%.For which is rather good, but still not good enough to detect those thecompaniontoH−6D.0106Ju9p7,Zucker&Mazeh(2000)obtained typical planetary companions which have been identified by amassof38 13M ,whichimpliesthatthecompanionisac- radial-velocitysurveys. Jup ± tually a brown dwarf and not a planet. These studies were ex- However, a new reduction of the raw Hipparcos data has tended to all the 47 planetary and 14 brown dwarf companion beenpresentedbyvanLeeuwen(2007a).Throughanimproved candidates known at the time in Zucker&Mazeh(2001). For attitude modeling,systematic errorswhich dominatedthe error Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata 3 budgetforthebrighterstarsinparticularweremuchreduced,by is actually the one pertaining to the star, since this is the com- uptoafactoroffourcomparedtotheoriginalversionofthecat- ponentobserved(weverifiedthisforafewexamples).Thelon- alog.Theformalerroronthemostpreciseparallaxesintheorig- gitude of periastron of the two componentsdiffer by 180 ; the ◦ inalversionoftheHipparcosCatalogueisaround0.4mas,and distinctionisimportantforthecombinationwithpositionaldata. around0.1masinthenewreductionpresentedbyvanLeeuwen Our final list comprises 258 stars with 310 substellar com- (2007a).Thenewreductionhasbeenclearlyshowntobesupe- panionsandiscurrentasofApril2010.Allstarswhichwereex- riortotheoldreductionofthedatainvanLeeuwen(2007b).The aminedare listed in Table1, alongwith oneor morereferences smaller formal errors, in particular for the bright stars, greatly totheorbitalelements. improvetheprospectoffindingastrometricsignaturesofplanets andbrowndwarfsinthedata. 2.2.Hipparcosintermediateastrometricdata In this paper we take a new look at the Hipparcos inter- mediate astrometric data, based on the new reduction of the ThenewreductionoftheHipparcosCataloguebyvanLeeuwen HipparcosrawdatabyvanLeeuwen(2007a),foralargenumber doesnotonlyincludeanewestimateofthestandardfiveastro- ofstarswithplanetaryorbrowndwarfcandidatesfromradialve- metricparameters(meanpositions,propermotionsandparallax) locitysurveys.Withtheimprovedastrometricaccuracy,itmight for every star, but also the improvedindividual measurements, bepossibletodetectacompanionorplaceatighterlimitonits theso-calledintermediateastrometricdata. massthanwaspossiblebefore. In contrast to the original solution, the given abscissa data Theoutlineofthispaperisasfollows:inSection2,wede- arenotaveragedoverallobservationswithinanorbitinthevan scribe the various input data for our study, including the stel- Leeuwenversion.Rather,oneabscissaresidualisgivenperfield lar sample with known planetary and brown dwarf companion transit,whichincreasestheaveragenumberofindividualabscis- candidates as well as the astrometric data from Hipparcos. In saeavailableforanobject.Thenoiselevelofthenewabscissae Section 3, we explain how we fitted astrometric orbits to the is, after averaging, up to a factor of four smaller than before. new Hipparcos intermediate astrometric data. Results are pre- The errors of the averaged abscissae (per epoch accuracies) in sentedinSection4(uppermasslimitsforthecompanionstoall the new version of the Hipparcos Catalogue range from better examinedstars) and in Section 5 (a few stars for whichthe as- than 0.7 mas for a few really bright stars up to around 10 mas trometricorbitcouldbe detected).We concludethe paperwith forthefainteststars;moststarshaveabscissaerrorsbetween1.5 notesonindividualstarsinSection6andasummaryanddiscus- and5mas(vanLeeuwen2007b). sioninSection7.Intheappendix,weshowhowtocalculatethe Everythingelseisverysimilarasbefore,althoughinsteadof astrometricsignatureandorientationoftheapparentorbitfrom thepartialderivativeofeachabscissaresidualwithrespecttothe thetrueorbit,takingprojectioneffectsintoaccount. five standard astrometricparameters,the new versiongivesthe instantaneous scan orientation and the parallax factor instead. Butthisisjustadifferentparameterizationofthesameinforma- 2. Data tion;allrelevantquantitiescanbederivedfromthat,asexplained 2.1.StellarSample invanLeeuwen(2007a). In an effort to be as exhaustive as possible, we put together a sample of all known planetary and brown dwarf compan- 3. OrbitFitting ions to Hipparcos stars detected via radial velocities. We 3.1.Method started with the list of planetary companions compiled by Butleretal.(2006a).Weaddedthosestarswithsubstellarcom- WehavefittedastrometricorbitstothenewHipparcosabscissa panion candidates which were detected after 2006, as well as residualsforallstarslistedinTable1,simultaneouslywithcor- starswithbrowndwarfcompanionswhichwerenotincludedin rectionsto the five standardastrometric parameters,via a stan- the list by Butler. We removed stars which were either not in dard least squares minimization technique. The only two or- theHipparcosCatalogueorforwhichnoorbitalelementswere bitalparametersfittedforweretheinclinationandtheascending available,andweupdatedtheorbitalelementsforthosestarsfor node; all other five orbital parameters (period, periastron time, whichnewsolutionswerepublishedinthemeantime. eccentricity, longitude of periastron and mass function) were We calculated periastron times for a few stars for which keptfixedattheliteraturevaluesfoundviafitstotheradialve- those were not given in the original table, namely transiting locitydataofeachstar. planetswithcircularorbits3(HD189733)andplanetarysystems IfthestarhadanorbitalsolutionintheHipparcosCatalogue withsignificantinteractionbetweenthecompanions(HD82943, with a period of the same order of magnitude as the spectro- HD202206andGJ876).Forthelatter,onlyosculatingelements scopic one, we removed the astrometric signature of the orbit can be given,and periastrontimes were calculatedclose to the from the abscissa data using the astrometric orbital elements, epochtowhichtheelementsreferred.Forallthreestarsthiswas andthenfittedforthefullorbitagainusingspectroscopicvalues abouta decade later than the Hipparcosepoch of J1991.25,so asinput.Inotherwords,wedidnotfitforcorrectionstotheorbit thattheorbitalelementsavailablemightnotberepresentativefor asappliedintheHipparcosCatalogue,butforthefullorbitfrom thetimeatwhichtheHipparcosmeasurementsoccurred. scratch (this step is only necessary for the abscissa data pro- Furthermore, we assume that the longitude of periastron videdbyvanLeeuwen(2007a),sinceintheoriginalHipparcos whichisgivenforspectroscopicallydetectedextrasolarplanets Cataloguetheabscissaealwayscorrespondedtoasinglestarso- lution,evenifanorbitwasprovided).Thisappliestothreestars 3 Formally, theperiastronisnot defined for companions incircular in our sample: HD 110833 (HIP 62145), ρ CrB (HIP 78459), orbits,butonecanextendtheusualdefinitionbysettingthelongitude and HD 217580 (HIP 113718). Likewise, we removed the ac- oftheperiastronto0 ,sothattheperiastrontimewillrefertothetime ◦ when the observed stellar radial velocity curve reaches itsmaximum. celerationtermsfromtheHipparcosabscissavaluesifthesolu- Forcircularorbits,thisoccursexactlyaquarterofaperiodbefore(hy- tionwasa7or9parametersolutionbeforefittingfortheastro- pothetical)midtransittime. metricorbit.Thisappliestothefollowingfivestars:HD43848 4 Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata (HIP 29804), 55 Cnc (HIP 43587), HD 81040 (HIP 46076), largeconfidenceregionwouldindicatenorealdetection.Inor- γ1 Leo A (HIP 50583), and HD 195019 (HIP 100970).Please der to be conservative, we mostly use 3σ confidence regions, also note that the version of the van Leeuwen catalog which correspondingtoaprobabilityof99.73%thatthetruevaluefalls is available in VizieR4 is different from the one provided to- withinthisparameterinterval. gether with the bookfrom van Leeuwen (2007a). E.g., 55 Cnc Inprincipleitwouldalsobepossibletofitforallorbitalel- (HIP45387)hasa7parameteraccelerationsolutioninthebook ementssimultaneouslyusingradialvelocitiesaswellasastrom- version,butastochasticsolutionintheonlineversion.Sincethe etry.Wright&Howard(2009)havedevelopedanefficientalgo- abscissae are only available based on the book version, this is rithmwhichcanfitseveralorbitalcompanionstoastrometricand whatweusedhere. radialvelocitydatajointly.Inclinationandascendingnodewill For stars with more than one substellar companion candi- always be derivedfrom the astrometricdata alone, since radial date, we did not fit for these multiple companions simultane- velocities are not sensitive to those parameters. The advantage ously, but individually. This should be a reasonable approach, ofjointlyfittingbothkindsofdataarebetterconstraintsonthe sincein thevastmajorityofthe casesonlyoneofthe compan- fivespectroscopicorbitalparameterswhichnowcomefromthe ionswilldominatetheastrometricsignalofthesystem. Thisis two differentdata sets, as well as explicitcovariancesbetween especially true for systems detected via radial velocities, since all orbital parameters. However,since in virtually all the cases forthoseallcompanionstendtohavesimilarradialvelocitysig- investigatedherethecompanionsignatureismuchmoresignif- nals (i.e. the more massive companionswill be located further icantlydetectedinradialvelocitiesthaninastrometry(ifatall), out).The astrometrywould then be dominatedby thatmassive theweightingoftheastrometricdatainthecombinedfitwould outercompanion,whereaslessmassiveinnercompanionshavea be much lower than that of the radial velocity data, and in the muchsmallerastrometricsignalnotdetectableintheHipparcos end the astrometry would not influence the values of the spec- data. troscopicparameters.Thereforenoattempthasbeenmadehere We explicitly used all five spectroscopicorbital parameters tocombinepreciseradialvelocitieswithHipparcosastrometry. in the fitting process and kept them fixed. The radial velocity Thissituationshouldchangeoncemoreaccurateastrometrybe- signal is in all the cases much more significant than the astro- comesavailable. metricsignal,sothatorbitalparametersderivedfromtheradial velocities should also be more precise and accurate than those 3.2.ǫEri derivedfromtheastrometry.Otherauthorshavechosentodisre- gardsomeofthespectroscopicorbitalparameters,e.g.theradial AnexampleforsuchanorbitalfitisgiveninFig.1forthecase velocitysemi-amplitude(Mazehetal.1999)orthelongitudeof of ǫ Eri. The panelsin the top row illustrate the 1-dimensional the periastron (Pourbaix & Jorissen 2000) and then later com- Hipparcos abscissa residuals (colored dots) in a 2-dimensional paredthe astrometricallyderivedvalueto theoriginal,spectro- plotofthesky.Thelinesrunningthroughthecoloreddotsindi- scopic one, as a consistency check. However, we prefer not to cate the direction which was not measured; the dots (measure- disregardany spectroscopicinformationin ourastrometric fits, ments)arethusallowedtoslideonthatline.Theactualmeasure- since this would likely compromise the accuracy of the orbital mentisperpendiculartotheindicatedline.Theupperleftpanel parameterswhich we are most interested in and notobtainable shows the abscissa residuals with respect to the standard solu- otherwise,theinclinationandtheascendingnode. tionasgiveninthecatalog,i.e.afterfittingforthefivestandard Likewise,wedidnottaketheapproachoffittingforthesemi- astrometricparameters,whilethepanelontheupperrightshows majoraxisoftheastrometricorbitandonlylaterlinkingthisto theabscissaresidualsforthecasewheretheorbitingcompanion the spectroscopic values via the parallax, but applied all those hasbeentakenintoaccount.Thus,intheleftpaneltheabscissa constraintssimultaneouslyand implicitly in the fitting process, residualsarereferredtothemeanpositionofthestarat(0,0)in whichisamoredirectapproachandshouldyieldthehighestac- thatdiagram,whileintherightpaneltheabscissaresidualsrefer curacy in the inclination and ascending node. We note that the tothecorrespondingpointsontheorbitwhichareindicatedby criterionby Pourbaix& Jorissen (2000) which is often used to small black dots. The solid line illustrates the astrometric orbit linkspectroscopicandastrometricquantitiesisnotexactlycor- oftheprimarystarasseenonthesky.Thecolorindicatesorbital rect,asnoallowanceismadeforprojectioneffects.Asdetailed phase;anabscissaresidualofagivencolorreferstothepointon intheappendix,thesemi-majoraxisoftheapparentorbitispos- theorbitwiththesamecolor.Thetimewhenameasurementwas sibly smaller than the semi-major axis of the true orbit, espe- takenis anadditionalconstraintin the fit, andthe colorcoding ciallyforeccentricorbitswithsmallinclinations. isanattempttovisualizethatadditionalconstraint. Also,wedidnotcomparetheχ2valuesofthestandardsolu- Alternatively, the two panels in the bottom row each show tion(fiveparameters)andtheoneincludingtheastrometricorbit one dimensiononlyas a functionof time. The solid lines indi- (twoadditionalorbitalparameters)inanFtest,aswasdoneby cate thesame orbitasin thepanelsin thetoprow.Thesmaller Pourbaix & Jorissen (2000) to decide whether the astrometric grey points indicate the individual abscissa residuals (for the orbitwasdetectedintheHipparcosdata.Thereasonisthatwe case wherethe orbitalmodelhasbeen taken into accountas in donotwanttoevaluatewhetherthestandardmodelortheorbital theupperrightpanel),whilethebiggerblackdotsshowthemean modelis thebetteronesincewe assumethattheexistenceofa ofallabscissaresidualstakenverycloselytogetherintime.This companion has been established already by the radial velocity is for illustration purposes only; the fit was done using the in- data.Thequestionwewouldliketoaskhereishowwelltheas- dividual, not the averaged abscissa residuals. When looking at trometric orbit can be constrained with the Hipparcosabscissa the two panelsin the bottom row please note thatagain notall data,andthemostsuitablecriterionforthiskindofquestionis constraintscouldbe visualizedat the same time;in these illus- the jointconfidenceregionallowedfor thetwo orbitalparame- trations the information about the orientation of the abscissae ters. A small confidence region, compared to the total allowed is lost. The measurements as indicated in right ascension and parameter range, indicates a detection of the orbit, whereas a declinationare notstatic; due to the one-dimensionalnatureof theabscissaresidualsthevaluesplottedinthelowerpanelscan 4 http://vizier.u-strasbg.fr/viz-bin/VizieR changeintheadjustmentprocess. Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata 5 Fig.1.IllustrationofabscissaresidualsfromHipparcosalongwiththeastrometricorbitasfittedforǫEri.Thetwopanelsinthetop rowshowtheabscissa residuals(coloreddots)withrespectto thestandardastrometricmodelwithoutorbitingcompanion(upper left)andwithrespecttoamodel(solidline)whichincludesthecompanion(upperright).Thecoloredlinesindicatethedirection whichwasnotmeasuredandalongwhichthedotsareallowedtoslideintheadjustmentprocess;theactualabscissameasurement isperpendicularto thatline.Color indicatesorbitalphase.Thetwopanelsinthe bottomrowshow onlyonedimensioneachas a function of time. The solid line is the orbital modelas in the upper panels, while the small grey dots are the individualabscissa residuals.Thebigblackdotsareaveragesofabscissaresidualstakenverycloselyintime.Theinformationabouttheorientationof theabscissaresidualsismissinginthepanelsinthebottomrow. One important complication which affects some of our or- 3.3.Verification bitalfitsisimmediatelyapparent:theHipparcosmeasurements Forverificationpurposes,weattemptedtoreproducetheinclina- donotcoverthefullorbitalphaserange,butlessthanhalfofthat tionsandascendingnodesofspectroscopicbinariesincludedin forthe6.9yearperiodofǫErib.Wewillcomebacktothatpoint the HipparcosCatalogue. Thereare 235 spectroscopicbinaries whenwediscusstheǫ ErisysteminmoredetailinSection6. forwhichanorbitislistedintheoriginalHipparcosCatalogue, andfor194ofthose,theinclinationandascendingnodeareac- tually obtained from a fit to the Hipparcos data (and not taken fromsomeotherreference).Forthose194starswetriedtorepro- ducethefittedinclinationsandascendingnodes,usingallother 6 Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata orbitalelementsasfixedinputvalues(eveniftheywerefittedfor and d, 47 UMa b, HD 10647b, and HD 147513b. For ǫ Eri b inHipparcos),sothatourapproachresemblesmostcloselythe and υ And c and d, the planetary nature was demonstrated al- onewhichwearefollowingwithsubstellarcompanionsdetected ready by HST astrometry, and for υ And d also by Hipparcos withtheradialvelocitymethod.Weusedtheoriginalversionof astrometry; see Sections 6.1 and 6.2. For 47 UMa b, an upper the HipparcosCataloguefrom1997,becauseherethe interme- mass limit in the brown dwarf regime at 90% confidence was diateastrometricdatacorrespondtothesinglestarsolutionand obatainedbefore,basedontheoriginalHipparcosdata,butthe becausetheorbitalparametersareavailableelectronically,which planetary nature could not be demonstrated unequivocally;see bothhelpsinthefittingprocess. Section6.3formoredetails. Wederivedinclinationsandascendingnodeswitherrorsfor For a further 75 companions the derived upper mass limit all194systemsandcomparedthesevaluestotheoneslistedin lies in the brown dwarf mass regime and thus confirms at theoriginalHipparcosCatalogue.In78%ofthecases, thetwo least the substellar nature of these companions. Two of those solutions agreed to within 0.3 σ, and in 97% of the cases to (HD 137510 b and HD 168443 c) have minimum masses also within 1 σ (althoughwe note that the errorson the two angles in the brown dwarf mass regime.These companionsare estab- can sometimes become rather large). Still, we consider this a lishedbrowndwarfsnowinsteadofjustbrowndwarfcandidates. very satisfactory result, and are thus confidentthat our method HD168443cwasalreadyconfirmedtobeabrowndwarfbased workscorrectly. ontheoriginalHipparcosastrometry;seeSection6.6. Fig. 2 illustrates the results for those 84 companions from Table1 for which the substellar nature could be established. It 4. UpperMassLimitsforPlanetsandBrownDwarfs can be seen that for many companionsthe allowed mass range While radial velocities provide m sini (where m is the com- betweenthelowermasslimitfromradialvelocitiesandtheup- 2 2 panionmassandiistheinclination),alowerlimitforthecom- permasslimitfromastrometryisstillratherlarge.However,for panion mass, astrometry can provide an upper mass limit for afewofthecompanionsthisrangeisreasonablysmall(seee.g. the companion. This is true even for stars where the astromet- ǫ Erib,υAnddorιDrab),andonemightevenspeakofade- ric signal of the companionis too small to be detectable, since terminationoftherealcompanionmassratherthanjustplacing inclinationsapproaching0 or180 (face-onorbits)wouldyield alowerandupperlimitonthemass,althoughthereisacontinu- ◦ ◦ companionswhicharesomassiveatsomepointthattheywould oustransitionbetweenthetwo.InSection5wewilltakeacloser show up in the astrometry. As the inclination of the orbit ap- look at those systems for which one might actually speak of a proachesaface-onconfiguration,thereisalwaysalimitatwhich detectionoftheastrometricorbit,ratherthanjustaconstrainton theinclinationisnotcompatibleanymorewiththeradialveloc- theinclinationandthusthemass. ities and the astrometry. This means that even for companions which are not detected in the astrometry, there is usually still 4.1.Limitations a (possibly weak) constrainton the inclination and thus on the massofthecompanion. Forsomestarstheformallyderived3σuppermasslimitofthe In order to derive upper mass limits, we fitted astrometric companion is larger than 5M , which is not useful anymore. orbitstotheHipparcosdataforallstarsonourlist.Theresults Infact,forastrometricupperm⊙asslimitslargerthanabout0.1- are given in Table 1. The first two columnsgive the usualdes- 0.5M there mightbe more usefuluppermass limits based on ignationandtheHipparcosnumber,respectively.Thefollowing photom⊙ etryand/orspectroscopy.Astellarcompanionthatmas- columnindicatesifmoreinformationonaparticularstarcanbe sive would show up eventually in such data (see e.g. Ku¨rster, foundinSection6.Thereferencecolumngivesanumericalcode Endl&Reffert2008).Ithasnotbeenattemptedtoderivethose to the reference from which the spectroscopic orbital elements otheruppermasslimitshere;thevaluesgivencorrespondtothe weretaken;thecodeisexplainedattheendofthetable.Thepe- limitssetbyastrometry,andonlyundertheassumptionthatthe riodcolumngivestheperiodindaysaccordingtothereference companion does not contribute a significant fraction to the to- in the previouscolumn;it is always derived from radial veloc- talfluxintheHipparcospassband.Ifthecompanionwasbright ities. The following column gives the minimum mass m sini enough to affect the photocenter of the system, our method 2 which corresponds to the orbital elements from the given ref- would fail since we assume that the photocenteris identical to erence;the star mass used in the conversionis also taken from the primary component in the system, which we assume to be thesamereference.Thelastfourcolumnsgivetheresultofour the Hipparcos star. If both componentscontribute significantly astrometric orbit fitting. The first three of those give the mini- to the total flux, the observed orbit of the photocenter around mumandthemaximuminclinationcorrespondingtothe3σcon- the center of mass of the system dependson the differencebe- fidence interval,i and i , and the resulting3σ uppermass tweenfluxratioandmassratioofthetwocomponents.Inorder min max limit for the companion. In a few cases, an inclination of 90 tomodelsuchasystem,onewouldneedtomakeadditionalas- ◦ couldbeexcludedinthefitting.Ifthisisthecase, anew,more sumptionsaboutthecomponents,whichisbeyondthescopeof stringent lower mass limit could be derived, too, which is de- thispaper. noted as m and given in the last column, if applicable. In Similarly,thereareanumberofstarsforwhichthe3σcon- 2,min somecases,especiallywhenaninclinationof90 isnotpartof fidenceintervalininclinationextendsfrom0.1to179.9 .Thisis ◦ ◦ the inclination confidence region, there are two minima in the oftenthecaseforcomponentswithverysmallminimummasses χ2 map, and the confidence region for the second minimum is and/or very small periods derived from Doppler spectroscopy, giveninasecondlineforthatstar.Thetableissortedaccording forwhichtheastrometricsignalisverysmallevenforsmallin- toHipparcosnumber(orrightascension,respectively). clinations. For those systems the derivation of the confidence For nine companions, the derived upper mass limit lies in intervalininclinationissometimescomplicatedbythefactthat the planetarymass regime, and thus unambiguouslyprovesfor thefittingprocessdoesnotconvergeforallpossibleinclinations, thefirsttimeformostofthemthatthesecompanionsarereally aswellasbythelimitedresolutionofourχ2mapintheinterval ofplanetarymassandnotjustplanetcandidates.Theconfirmed between0and0.1 or179.9 and180 ,respectively.Ofcourse ◦ ◦ ◦ planetsareβGemb(Polluxb),ǫErib,ǫRetb,µArab,υAndc itwouldbepossibletoincreasetheresolutionforthoseinclina- Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata 7 companion mass range [M_Jup] tionintervals(wealreadyusedhigherresolutionfortheintervals 0 10 20 30 40 50 60 70 80 from0to1◦ andfrom179◦to180◦thanfortherestoftheincli- nationrange),buttheresultswouldnotbeverymeaningfulsince themethodisjustnotsuitedverywellforcompanionswithex- ε Eri b υ And d tremelysmallastrometricsignatures.Asaconsequence,thede- 47 UMa b riveduppermasslimitsforthosecompanionsarenotasaccurate HD 10647 b µ Ara b asthosewhichcorrespondtoinclinationswhicharenotasclose HD 62509 b Hε DR e1t4 b7513 b to0or180◦. υ And c For67companionsnouppermasslimitcouldbederivedat HD 69830 d allfromtheHipparcosastrometry.Thereasoniseitherthatthe ι Dra b ι Hor b astrometric fits did not converge(in 5000 iterations, which we GJ 832 b 61 Vir d set as a limit) over large parts of the inclination and ascending HD 60532 c nodeparameterspace,orthatthe3σconfidencelimitsinincli- HD 87883 b GJ 649 b nation are very close to 0 or 180 , respectively, as mentioned κ CrB b ◦ ◦ 16 Cyg B b above.Ineithercasenomeaningfulconfidencelimitsininclina- HD 128311 c HD 190360 b tionoruppermasslimitscouldbederived.Forthosestarswelist ρ Ind b thewholeinclinationintervalfrom0 to180 as3σconfidence 14 Her b ◦ ◦ HD 19994 b limits, and the column giving the 3σ upper mass limit is left GJ 849 b γ Cep b emptyinTable1.54ofthe67companionswithoutuppermass HD 164922 b HD 128311 b limits have periods smaller than 20 days, and 64 companions π Men b haveperiodssmallerthan50daysandthusverysmallexpected 61 Vir c HD 210277 b astrometricsignatures. HD 33564 b τ1 Gru b mu Ara e HD 114783 b HD 154345 b 5. AstrometricOrbits HD 23079 b HD 70642 b For a few of the stars in Table 1, not only upper mass limits HD 111232 b HD 50554 b couldbederived,butalsotheastrometricorbitcouldbefurther HD 30562 b HD 60532 b constrained or fully determined. While for most companionsa HD 10697 b 55 Cnc f limitontheinclinationcouldbederived(evenifthelowerlimit HD 114729 b issmall),thisisnotnecessarilytruefortheascendingnode,the HD 82943 c 70 Vir b onlyotherorbitalelementwhichisnotdeterminedbytheradial HD 1237 b HD 12661 c velocityfit.Thisisduetothefactthataninclinationapproaching HD 169830 c HD 114386 b 0 degreewill increase the size of the orbit,untileventuallythe HD 11964 b orbit would not be compatible any more with the small scatter HD 141937 b HD 196885 b of the Hipparcos measurements. The ascending node however HD 37124 d HD 213240 b describes the orientationof the orbit in space. Thus usually no HD 11977 b µ Ara c or very weak constraints can be placed on the ascending node HD 155358 c if the orbit is not really detected. However, for 20 systems of HD 89307 b HD 4208 b Table1,atleastaweakconstraintontheascendingnodecould HD 23596 b HD 3651 b be derived,indicatinga (possibly weak) detection of the astro- HD 196050 b ε Tau b metricorbit.IncontrasttoSection4,wenowusetheconfidence HD 106252 b levels for two parameters (inclination and ascending node) si- HD 168443 c HD 47536 b multaneously, since we are now interested in the orbit, i.e. we HD 82943 b HD 47186 c wanttoderiveconstraintsonbothorbitalparametersatthesame HD 150706 b time.Again,weusedthecontourspertainingtoaprobabilityof HD 81040 b HD 137510 b 99.73%(3σ).Thiswillalwaysresultinalesstightconstrainton HD 50499 b GJ 876 b theinclinationthangiveninTable1. HD 181433 c HD 202206 c Those20systemswithdetectedastrometricorbitsareshown HD 175167 b in Fig. 3 and listed in Table 2. The table gives the best-fit in- HD 12661 b HD 41004 A b clinations and ascending nodes, along with the 3σ confidence HD 167042 b HD 13445 b limits. Note thatthe 3σconfidencelimits (denotedasimin, imax 18 Del b HD 45350 b and Ωmin, Ωmax, respectively) are now confidence regions in two parameters,inclination and ascending node, as opposed to Table1,whereconfidencelimitsinonlyoneparameter(theincli- 0 10 20 30 40 50 60 70 80 nation) were considered; the correspondingconfidence regions companion mass range [M_Jup] in Table 2 are therefore slightly larger than those in Table 1. Please also note that for some stars, two minima are visible in Fig.2. Illustration of the allowed 3σ range of the companion theχ2 maps,often(butnotalways)withcomparablelocalmin- massforallstarsfromTable1forwhichthederiveduppermass imum χ2 values. Those solutionscorrespondto orbits of about limitfromHipparcoscorrespondsto lessthan85 M , i.e. for Jup the same size (similar inclinations), but with ascending nodes all 84 stars for which the substellar nature could be confirmed whichdifferbyabout180 ,sothattheorbithastheoppositeori- byHipparcos.The lowermasslimitusuallycorrespondsto the ◦ entation.Forthosestars affected,we giveinTable2 thevalues lowermasslimitm sinifromtheradialvelocityfit,exceptfor 2 pertinentto the second minimum in a second line for that star. the few cases where the astrometryyields a tighterlower mass Forstarsforwhichtheconfidenceregionswraparoundthe360 limitbyexcludingtheinclinationof90 fromthe3σconfidence ◦ ◦ region(HD87883b,HD114783bandγCepb).Theuppermass limitalwayscomesfromtheHipparcosastrometry,accordingto Table1. 8 Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata HD 1844 5 b - HIP 137 69 BD -04 7 82 b - HI P 19 832 HD 4384 8 b - HIP 298 04 6 Lyn b - H IP 3 103 9 HD 8788 3 b - HIP 496 99 360 360 360 360 360 eg] 270 eg] 270 eg] 270 eg] 270 eg] 270 d d d d d Ω [ Ω [ Ω [ Ω [ Ω [ e e e e e d d d d d o o o o o n 180 n 180 n 180 n 180 n 180 g g g g g n n n n n di di di di di n n n n n e e e e e c c c c c as 90 as 90 as 90 as 90 as 90 0 0 0 0 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] γ1 Leo A b - HIP 505 83 HD 1062 52 b - HI P 59 610 HD 1108 33 b - HI P 62 145 HD 1127 58 b - HI P 63 366 HD 1316 64 b - HI P 73 408 360 360 360 360 360 eg] 270 eg] 270 eg] 270 eg] 270 eg] 270 d d d d d Ω [ Ω [ Ω [ Ω [ Ω [ e e e e e d d d d d o o o o o n 180 n 180 n 180 n 180 n 180 g g g g g n n n n n di di di di di n n n n n e e e e e c c c c c as 90 as 90 as 90 as 90 as 90 0 0 0 0 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] ι Dra b - H IP 7 5458 ρ CrB b - H IP 7 845 9 HD 1568 46 b - HI P 84 856 HD 1644 27 b - HI P 88 531 HD 1684 43 c - HI P 89 844 360 360 360 360 360 eg] 270 eg] 270 eg] 270 eg] 270 eg] 270 d d d d d Ω [ Ω [ Ω [ Ω [ Ω [ e e e e e d d d d d o o o o o n 180 n 180 n 180 n 180 n 180 g g g g g n n n n n di di di di di n n n n n e e e e e c c c c c as 90 as 90 as 90 as 90 as 90 0 0 0 0 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] HD 1698 22 b - HI P 90 355 HD 1848 60 b - HI P 96 471 HD 1902 28 b - HI P 98 714 HD 2 1758 0 b - HIP 113 718 γ Cep b - H IP 1 1672 7 360 360 360 360 360 eg] 270 eg] 270 eg] 270 eg] 270 eg] 270 d d d d d Ω [ Ω [ Ω [ Ω [ Ω [ e e e e e d d d d d o o o o o n 180 n 180 n 180 n 180 n 180 g g g g g n n n n n di di di di di n n n n n e e e e e c c c c c as 90 as 90 as 90 as 90 as 90 0 0 0 0 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 0 4 5 9 0 13 5 18 0 inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] inclination i [deg] Fig.3. Illustration of the χ2 maps for those 20 stars for which the astrometric orbit could be detected in the Hipparcos data, as a function of inclination and ascending node. The contours show the 1σ, 2σ and 3σ confidence regions in both parameters jointly.Please notethatforρCrB b theallowedrangein inclinationis so smallthatitis barelyvisiblein thisrepresentation;the correspondingdatacanbefoundinTable2. Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata 9 Table2.Bestfitinclinations,ascendingnodesandcompanionmassesforstarswheretheastrometricorbitcouldbedetected.i , min i ,andΩ ,Ω correspondtothelimitsofthe3σconfidenceregioninbothparametersjointly.Thesecondrowforastarrefers max min max to a second local minimumwhere present. The followingcolumnsgive the reducedχ2 value of the fit, the numberof individual Hipparcos abscissae available for that star, n , and, for reference, the astrometric signature α for the fit. Finally, the last four obs columnsgivethe minimummass m sini, derivedfrom radialvelocities, and the actualmass of the companionm derivedhere, 2 2 with3σconfidenceregions(m andm ).Ifthevalueinthecolumnindicatingthelowerlimitofthe3σconfidenceregionin 2,min 2,max massisnotgiven,thismeansthataninclinationof90 cannotbeexcludedandthusthatthelowermasslimitderivedfromradial ◦ velocitiesapplies. designation HIPno. i imin imax Ω Ωmin Ωmax χ2 n α m2sini m2 m2,min m2,max [◦] [◦] [◦] [◦] [◦] [◦] red obs [mas] [MJup] [MJup] [MJup] [MJup] HD18445b 13769 147.6 31.3 160.2 285.7 252.2 20.4 1.34 110 4.32 44. 84.5 139.1 BD 04782b 19832 12.4 10.3 15.1 307.9 297.9 318.4 0.17 104 17.51 47. 261.6 207.4 329.2 − HD43848b 29804 19.6 11.5 79.4 30.0 341.5 78.7 1.31 111 6.46 24.3a 75.2 24.6 130.7 158.3 98.2 167.1 99.2 48.5 152.3 1.32 6Lynb 31039 2.0 1.0 81.0 75.5 350.4 149.3 1.18 77 1.30 2.21 64.4 2.2 128.7 172.6 152.6 175.6 310.9 259.8 12.1 1.37 HD87883b 49699 8.5 4.8 57.7 254.9 182.2 316.5 0.79 77 2.75 1.78 12.1 2.1 21.4 168.9 133.7 173.9 96.4 40.7 164.0 0.86 γ1LeoAb 50583 172.1 73.5 175.9 157.4 75.2 231.0 8.39 77 1.57 8.78 66.2 9.2 130.9 13.6 7.5 59.3 359.9 298.4 66.7 9.50 HD106252b 59610 166.7 6.5 174.0 154.5 13.2 330.9 0.79 81 1.93 6.92 30.6 68.9 HD110833b 62145 10.4 7.7 16.8 120.9 96.4 149.4 2.03 171 5.97 17. 101.8 61.5 141.0 HD112758b 63366 8.9 6.5 15.2 150.5 126.9 174.9 1.77 81 4.85 34. 248.5 136.6 366.5 HD131664b 73408 167.1 149.3 171.9 320.8 276.4 10.0 0.76 154 4.13 18.15 85.2 36.3 139.7 ιDrab 75458 69.9 26.6 141.8 182.9 10.7 324.9 0.78 137 0.24 8.82 9.4 19.8 ρCrBb 78459 0.4 0.4 0.7 266.4 242.3 290.6 2.67 173 1.96 1.093 169.7 100.1 199.6 HD156846b 84856 177.3 2.0 178.8 185.3 91.4 265.6 1.73 89 3.52 10.45 263.0 660.9 HD164427b 88531 12.2 7.1 44.7 337.5 294.7 29.9 1.15 92 2.27 46.4 244.2 66.8 458.5 HD168443c 89844 36.8 15.2 164.7 134.3 29.2 334.1 1.49 50 2.02 18.1 30.3 71.0 HD169822b 90355 175.1 172.5 176.1 249.8 236.1 267.5 0.79 165 8.49 27.2 388.7 237.4 527.9 HD184860b 96471 160.2 86.5 169.4 339.0 258.8 22.0 1.14 104 5.51 32.0 99.7 195.3 HD190228b 98714 4.5 2.4 174.1 71.0 219.4 164.2 1.08 254 1.63 5.93 76.8 147.2 HD217580b 113718 43.3 31.2 75.3 158.8 124.5 190.1 2.28 68 7.65 67. 99.9 69.0 135.8 γCepb 116727 5.7 3.8 20.8 37.5 352.9 86.0 0.90 125 1.62 1.77 17.9 5.0 26.9 173.1 166.6 174.8 356.1 330.2 25.0 0.97 a Minnitietal.(2009)giveavalueof25MJupform2 sini;aniterativesolutionofthemassfunctionhowevergives24.3MJup. limitintheascendingnode,continuingat0 ,welistthehigher then this value correspondsexactly to the minimum secondary ◦ valueasΩminandtheloweroneasΩmaxinordertoindicatethe massderivedfromradialvelocities.Inordertoindicatethatthis correctorientationoftheconfidenceregion.Thefollowingthree is the case, no mass is givenhere if appropriate.Itcan be seen columnsgivethereducedχ2 values,thenumbern ofindivid- that for twelve systems an inclination of 90 can be excluded obs ◦ ual1-dimHipparcosmeasurements(abscissae,orfieldtransits) with99.73%confidence;forallothers,theminimumsecondary availableforthatstar,and,forreference,theastrometricsigna- mass derived from radial velocities, m sini, is within the 3σ 2 tureαforthefit. massconfidenceregionandthusalsothelowerlimitforthecon- fidenceregionofthemass. The last four columnsof Table 2 give the resulting masses fortheastrometricsolution.Theminimummassm siniderived 2 from the radial velocities is listed in the first of the four mass Some orbits are rather well determined, even if using the columns. This is also the minimum mass for any of the astro- conservative 99.73% confidence regions. The inclinations of metricsolutions,so itisgivenhereforreference.Itwouldcor- BD–04782b,ρCrBbandHD169822bcouldbedetermined respondtoaninclinationof90 .Thenextmasscolumnlabeled to better than 5 , while the inclinations of HD 110833 b and ◦ ◦ m givesthemasswhichcorrespondstothebestfitinclination. HD112758bcouldbedeterminedtobetterthan10 .Forthose 2 ◦ There is only one best fit mass for each system, so the second systems, the ascending node is also rather well constrained. lineisemptyeveniftherearetwominima.Thenexttolastcol- HD 168443b couldbe confirmedas browndwarf, evenif tak- umn,labeledm ,givesthelowerlimitforthe3σconfidence ingtheconservative3σconfidencelevelsintoaccount.ιDrabis 2,min regionin mass. If the inclinationvalueof 90 is partofthe 3σ mostlikelyahigh-massplanet,whileHD106252bandγCepb ◦ confidenceregion(seealsoFig.3orthecolumnsi andi ), aremostlikelybrowndwarfs.Thecompanionsforwhichthein- min max 10 Reffert&Quirrenbach:Massconstraintsonknownsubstellarcompanionsfromthere-reducedHipparcosdata Table 3. Corrections(firstrows) andfinalresultingvalues(secondrows)of the five standardastrometric parameters(meanposi- tions,parallax,propermotions)forthecase wheretheastrometricmodelingoftheobservationsincludestheeffectofanorbiting companion. ∆α⋆[mas] ∆δ[mas] ∆̟[mas] ∆µ [masyr 1] ∆µ [masyr 1] designation HIPno. α[rad] δ[rad] ̟[mas] µα⋆[masy·r−1] µδ[masy·r−1] α⋆ · − δ · − HD18445b 13769 1.61 2.95 1.65 0.63 0.99 − − 0.7732664947 0.4358967094 40.00 14.45 30.59 − − BD 04782b 19832 8.83 2.94 5.67 4.43 8.50 − − − − − 1.1133364645 0.0771102807 42.40 85.93 87.74 − − HD43848b 29804 4.06 4.63 1.40 3.73 0.59 − − − − 1.6428834704 0.7074229223 25.03 125.77 197.73 − 6Lynb 31039 0.04 0.12 0.12 0.66 0.56 − − 1.7051242326 1.0151438236 18.04 29.50 339.24 − − HD87883b 49699 0.14 0.41 0.30 1.82 0.00 − − − 2.6560409615 0.5976427335 55.23 62.23 60.51 − − γ1LeoAb 50583 2.40 0.56 2.11 1.89 1.01 − − 2.7051266777 0.3463057931 27.11 302.52 155.30 − HD106252b 59610 0.96 1.72 0.02 1.29 0.99 − − − 3.2004608067 0.1752714978 26.53 25.05 278.51 − HD110833b 62145 2.04 7.04 0.66 0.54 0.18 − 3.3346626142 0.9033780697 66.54 378.54 183.51 − − HD112758b 63366 0.63 2.50 0.50 1.55 2.05 − − 3.3991779486 0.1716455024 48.36 826.79 198.20 − − HD131664b 73408 0.14 2.23 0.93 1.54 2.37 − − − − 3.9274303384 1.2834350195 18.68 13.08 26.49 − ιDrab 75458 0.14 0.03 0.04 0.04 0.04 − − − 4.0357673046 1.0291512587 32.27 8.40 17.12 − ρCrBb 78459 0.27 0.76 0.65 1.49 0.07 − − 4.1933570759 0.5812886759 58.67 195.15 772.93 − − HD156846b 84856 3.43 3.83 0.94 0.33 0.24 − − − 4.5403574027 0.3374314890 20.05 137.00 144.00 − − − HD164427b 88531 0.38 1.51 0.56 0.08 0.21 − 4.7329564103 1.0333998216 25.76 199.22 51.31 − − − HD168443c 89844 0.71 0.50 0.37 1.34 0.10 − − − 4.7999453113 0.1674674410 26.35 92.47 222.94 − − − HD169822b 90355 8.35 4.24 3.31 2.39 2.24 − − − 4.8265774036 0.1532170533 31.32 196.50 462.44 − − HD184860b 96471 2.71 5.95 0.36 1.20 0.59 − − 5.1345996462 0.1822607121 36.15 292.58 273.12 − − − HD190228b 98714 0.41 1.37 0.22 1.16 0.30 − − − 5.2491288882 0.4940508070 16.03 104.05 69.54 − HD217580b 113718 2.39 3.99 6.29 1.19 1.85 6.0294808555 0.0671642787 65.00 397.39 205.83 − − γCepb 116727 0.27 0.10 0.18 0.69 0.36 − 6.1930813876 1.3549334224 71.10 47.27 126.95 − clinationcouldbedeterminedtobetterthan5 arealllow-mass Hipparcos Catalogue and the new resulting values for these in ◦ stars. Table 3. These astrometric parameters were fitted simultane- ously with the orbital parameters, and the changes are a direct Altogether, using the best fit masses, we find two planets consequenceofthenewfittingmodelwhichnowincludestheas- (HD 87883 b and ι Dra b), eight brown dwarfs and ten stars trometricorbit.Asusual,α⋆denotesavaluewherethecosδhas among our 20 systems with astrometric orbits. Note that this beenfactoredinalready.Notethatthepositionalcorrectionsare is by no means representative for our whole sample listed in given in milli-arcseconds, whereas the coordinates themselves Table1,sinceitismuchmorelikelytofindtheastrometricsig- aregiveninradians,justlikeintheHipparcosCataloguebyvan nature of a star or a brown dwarf than the tiny signature of a Leeuwen(2007a).Thecorrectionstotheastrometricparameters planetintheHipparcosdata. are usually rather small, typically of the order or smaller than Forcompleteness,wealsolistthecorrectionstothefivestan- 1masor1masyr 1,respectively. − dard astrometric parameters (mean positions at the Hipparcos · epoch of J1991.25, parallax and proper motions) in the new