Two Wide Planetary-Mass Companions to Solar-Type Stars in Upper Scorpius
Michael J. Ireland, Adam L. Kraus, Frantz Martinache, Nicholas M. Law, Lynne A. Hillenbrand
aa r X i v : . [ a s t r o - ph . S R ] N ov Draft version May 28, 2018
Preprint typeset using L A TEX style emulateapj v. 03/07/07
TWO WIDE PLANETARY-MASS COMPANIONS TO SOLAR-TYPE STARS IN UPPER SCORPIUS.
Ireland, M.J.
Sydney Institute for Astronomy (SIfA), School of Physics, University of Sydney NSW 2006, Australia
Kraus, A. ∗ Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, HI 96822, USA
Martinache, F.
National Astronomical Observatory of Japan, Subaru Telescope, Hilo, HI 96720, USA
Law, N.
Dunlap Institute for Astronomy and Astrophysics, University of Toronto, 50 St. George Street, Toronto M5S 3H4, Ontario, Canada andHillenbrand, L.A.
California Institute of Technology, Department of Astrophysics, MC 105-24, Pasadena, CA 91125
Draft version May 28, 2018
ABSTRACTAt wide separations, planetary-mass and brown dwarf companions to solar type stars occupy acurious region of parameters space not obviously linked to binary star formation or solar-system scaleplanet formation. These companions provide insight into the extreme case of companion formation(either binary or planetary), and due to their relative ease of observation when compared to closecompanions, they offer a useful template for our expectations of more typical planets. We presentthe results from an adaptive optics imaging survey for wide ( ∼ ∼
14 M J companion around GSC 06214-00210, and confirm that the candidate planetary mass companion 1RXS J160929.1-210524 detectedby Lafreni`ere et al. (2008b) is in fact co-moving with its primary star. In our survey, these twodetections correspond to ∼
4% of solar type stars having companions in the 6-20 M J mass and ∼ Subject headings: brown dwarfs, planetary systems, Infrared: planetary systems, INTRODUCTION
Over the past five years, direct imaging surveys forextrasolar planets have discovered a small but signifi-cant number of ultra-low mass companions (henceforthULMCs, masses . M J ) at &
50 AU separations fromtheir primaries. These objects have estimated massesthat are in the same range as radial velocity or tran-siting planets with < ∼ M J , then agap in mass until arguably star-like objects are foundat >
60 M J (Grether & Lineweaver 2006; Deleuil et al.2008; Bouchy et al. 2010; Anderson et al. 2010). For thisreason, ULMCs are often called “planetary mass” com-panions.The prototypical wide ULMC, 2M1207-3933, consistsof a 4–8 M Jup companion located ∼
50 AU away from a10 Myr old brown dwarf (Chauvin et al. 2004). Since itsdiscovery, half a dozen other ULMCs have also been re-ported, most of which orbit much higher-mass primaries( ∼ M ⊙ ;e.g. Lafreni`ere et al. 2008b; Kalas et al.2008; Marois et al. 2008). Some of these systems ap-pear to be genuinely scaled-up versions of our own solar ∗ Hubble Fellow system, with planets that are consistent with formationin a disk. One example is the companion to Fomal-haut, which is coplanar with its debris disk. Anotheris the HR 8799, which has multiple planetary compan-ions of similar mass. Other cases like CHXR 73 and1RXS J160929.1-210524 are more ambiguous since theirorbital radii are even wider, and it is unclear whetherthey lie in the original plane of planet formation. Suchcompanions could very well form like planets within acircumstellar disk or like binaries from the collapse of amolecular cloud.These ULMCs pose a significant challenge to exist-ing models of planet and binary formation. Their or-bital radii are so large that the core accretion timescale( >
100 Myr at 100 AU; Pollack et al. 1996) shouldbe much longer than the typical protoplanetary diskdissipation timescale ( ∼ &
100 AU, much less at ∼
300 AU. These compan-ions could represent the extreme end of the binary mass Ireland et al.function, which appears to be linearly-flat (with all com-panion masses being equally probable, Kraus et al. 2008;Raghavan et al. 2010) well into the substellar regime.However, it is unclear whether this trend could extendto planetary masses (by which we mean <
20 M J in thispaper). These companions have masses near the opacity-limited minimum mass (Hoyle 1953; Low & Lynden-Bell1976; Bate 2005), and unless their formation occurred ex-actly as the circumstellar envelope was exhausted, thenthey should have quickly accreted enough mass to be-come high-mass brown dwarfs or stars.In this paper, we report the discovery of two ULMCs inthe Upper Scorpius OB association, one of which was in-dependently discovered by Lafreni`ere et al. (2008b). Thesurvey consists of a subset of the aperture-masking inter-ferometry sample reported by Kraus et al. (2008). Wedescribe our observations and data analysis techniquesin Section 2, and in Section 2.2, we report the detectionsand detection limits from our survey. In Section 4, wereport the first measurement of a frequency for ULMCsaround young stars. Finally, in Section 5, we discussthe implications for possible formation mechanisms forULMCs. OBSERVATIONS
Discovery Observations
Nearby ( .
200 pc) young ( .
20 Myr) stars have beenthe object of numerous high-resolution imaging cam-paigns over the past several decades. These observationshave included lunar occultation (e.g. Simon et al. 1995),speckle interferometry (e.g. Ghez et al. 1993), Adap-tive Optics (AO) imaging (e.g. Lafreni`ere et al. 2007;Masciadri et al. 2005; Chauvin et al. 2010), and mostrecently non-redundant masking (NRM) inteferometry(Kraus et al. 2008). The earlier techniques of lunar oc-cultaion and speckle were only sensitive to the presenceof bright, stellar-mass, companions, and only recently didAO imaging and NRM interferometry managed to probewithin the brown dwarf regime, going as far as samplingthe top of the planetary mass regime.In K08, we reported the results of one such survey ofyoung stars in the Upper Sco OB association that useda combination of conventional AO imaging and NRM-interferometry. In addition to the results for stellar andbrown dwarf companions, that paper also reported de-tection limits for ULMCs at small separations (within50 AU) where the probability of background star con-tamination was negligible. In this work, we report thecorresponding analysis for candidate companions at wideseparations, including multi-epoch follow-up imaging forthree candidates, of which two appear to be associatedand of approximately planetary mass.Table 4 of K08 lists the AO imaging observations con-ducted with the PHARO camera at the Palomar 200”telescope and the NIRC2 camera at the Keck-II 10mtelescope. We found that 10 out of these 62 targets hadstellar binary companions at separations of ∼ ′′ ,which should mean that the majority of additional faintcompanions are not dynamically stable in this range ofseparation; these were omitted from our sample. In ad-dition, we omitted the few targets with spectral types oflater than M2 ( M < . M ⊙ ) so as to work with a singlemass range of approximately “solar-type” stars, leaving49 targets in consideration. All observations used the smallest pixel scales (10mas/pix with NIRC2 and 25 mas/pix with PHARO)in order to achieve the best PSF sampling. We useda K s filter at Palomar and the Brγ filter at Keck,which yielded diffraction-limited resolutions of 100 masand 50 mas, respectively. The Keck observations usedthe narrowband filter despite the penalty in sensitivity,because the brighter primary stars in our sample wouldhave saturated the detector within the minimum expo-sure time. Much of this sensitivity was regained by us-ing more Fowler samples per frame, which significantlyreduces the high read noise of NIRC2’s detector. Theseobservations used relatively short total integration time,of the order of a minute, in comparison with other AOimaging surveys (e.g. Metchev & Hillenbrand 2009), re-sulting in limiting magnitudes of K lim ∼ ∼ . M Jup for 48 targets and . M Jup for 42targets).The detections and detection limits were derived usingmethods we previously described in Kraus (2009) andKraus & Hillenbrand (2010, in prep). Since the detec-tion limits at more than a few λ/D are driven by specklenoise, detections and detection limits were determinedfrom the coadded image stacks by placing a large numberof photometric apertures (with diameter λ/D ) aroundeach target, then measuring the mean and standard de-viation of the brightness distribution for all apertures inconcentric rings around the primary. For each ring, weidentified all candidate detections with significance & σ above the mean brightness, then compared those detec-tions to other stars taken on the same night to identifyand reject the quasistatic speckles and diffraction spikesthat are seen in common for many targets. All candi-date detections that could not be identified as PSF arti-facts were then adopted as genuine sources and hence ascandidate bound companions, while the 5 σ limits wereadopted as our formal detection limits. At large separa-tions, the detection limit from this algorithm convergesto the sky background limit, and at these separation ( > σ limit, because of the large numberof pixels in this regime. We found no candidate com-panions near the detection limits in the speckle-limitedregime, but there are many candidate companions in thesky-limited regime.We determined the photometry and astrometry forthese sources using the methods described in K08 andKraus & Hillenbrand (2009). To briefly summarize, wemeasured astrometry and aperture photometry for eachsource with respect to the known USco member usingthe IRAF task DAOPHOT (Stetson 1987); all measure-ments were conducted using apertures of 0.5, 1.0, and 2.0 λ/D , and then the optimal aperture was chosen to max-imize the significance of the detection (given the com-peting uncertainties from the sky background and thePoisson noise for the source itself). In order to estimatethe uncertainties from the data, we analyzed the mea-surements in individual frames and then combined thosemeasurements to estimate the mean and standard devi-ation. We then accounted for the systematic uncertaintyin the plate scales and distortion solutions of PHARO(0.3%; Section 2.3) and NIRC2 (0.05%; Ghez et al. 2008,ide-separation planets in Upper Scorpius 3 Fig. 1.—
An example image of GSC 06214-00210 from 2008July in the Kp filter, with a log stretch. The faint companioncan be seen as at an RA offset of ∼ -200 mas and a Dec offset of ∼ -2200 mas. Cameron 2008) by adding those terms in quadrature withthe observed scatter. Most of the new candidate compan-ions were identified with PHARO; as we describe below,its astrometric calibration is not yet well-understood andmight be further improved with more calibration obser-vations, but our systematic uncertainties should accountfor this effect. Many of the wider ( & ′′ ) companionsare also affected by anisoplanatism, yielding systemat-ically low brightness estimates for aperture photome-try. A correction of the photometry would require adetailed knowledge of the atmospheric turbulence pro-file and isoplanatic patch size, so it can not be accom-plished for our data. However, as we describe below, allof these sources have well-calibrated K magnitudes avail-able from UKIDSS. We choose instead to defer to thosemeasurements in determining colors (Section 3.2).Finally, all candidate companions with separations & ′′ have spatially resolved counterparts in theUKIDSS Galactic Cluster Survey (Lawrence et al. 2007),and many of the widest companions also have opticalcounterparts in the USNO-B1.0 digitization of the Palo-mar Observatory Sky Survey (Monet et al. 2003). As wedescribe further in Section 3.2, we have used those ob-servations to identify most of these candidate compan-ions as background stars. The rest require multi-epochastrometric monitoring to determine whether they areassociated. Follow-up Observations
Around three targets in the K08 survey we found faintvisual candidate companions with projected separationsof 2–3 ′′ and brightnesses of K ∼ K <
17 is approxi-mately 0.002 per square arcsecond (Kraus & Hillenbrand2010, in prep.), which gives a 86% chance of a chancealignment within 2.5 ′′ for at least one star in our sample,but only a 30% chance of all 3 candidates being chancealignments. It is impossible to use proper motions torule out chance alignments with other association mem-bers because the internal velocity dispersion of Upper Scois only ∼ − ( ∼ − , Kraus & Hillenbrand2008), meaning that they would be comoving even if theywere only seen in chance alignment. However, the sur-face density of young stars in Upper Sco is no more than100 deg − Kraus & Hillenbrand (2008) or 10 − arcsec − ,meaning that the probability of a chance alignment isnegligible.The proper motion of the Upper Scorpius associationis (-11.5,-23.5) mas yr − in equatorial coordinates, deter-mined from the members listed in de Zeeuw et al. (1999),and using updated proper motions from van Leeuwen(2007). This is relatively small when compared to nearbymoving groups, so an astrometric accuracy of order 2 masor 0.1% at 2 ′′ is required in order to clearly determinewhether a companion is co-moving or not on a 1 yeartime baseline.The NIRC2 camera has been shown to have a stabil-ity better than this in the precise galactic center as-trometric work of Ghez et al. (2008). Cameron et al.(2009) showed that the PALMAO adaptive optics sys-tem (Troy et al. 2000) and the PHARO Near Infrared(NIR) camera are stable in distortion to ∼ µ as overseveral months. However, PALMAO underwent severalupgrades between these common proper motion confir-mation observations, including preliminary work for areconfiguration of the output beam path to accommo-date new science instruments.We searched for PALMAO distortion solution changesduring this period using observations of the core ofthe M5 globular cluster from four nights (28/5/2007;29/5/2007; 17/07/2008 and 17/08/2008). The obser-vations consisted of 100 co-added 1.4 second exposuresat each epoch, were taken in the 25 mas PHARO platescale, used the Br- γ narrow-band filter to avoid differ-ential chromatic refraction, and were timed to be ob-served at very similar airmasses and hour angles. Thesame AO guide star was used in each case, and care wastaken to align each epoch to the same pointing within anarcsecond and to keep all other AO and camera param-eters consistent. We fitted 2D Gaussians to derive thepositions of 34 stars covering the 25” ×
25” field in eachdataset; we also checked the fitted positions using SEx-tractor (Bertin & Arnouts 1996) and found very similarresults.We first used a simple model to match the stellar posi-tions between epochs, allowing for a change in pointingposition, an arbitrary rotation, and a separate scalingin the X and Y axes. There was no detectable changein rotation or scale within 2007 or within 2008, but be-tween those years the field rotation changed by 0.18 de-grees, while the X and Y plate scales changed by -0.5% and +0.2% respectively. However, the simple rota-tion and plate scale change model leaves 3-5mas residu- Ireland et al.als in matching the 2007 and 2008 measured positions.The residuals are reduced to ∼ ∼
10 mas/pixel image scale and the K p filter, whichwe found did not quite saturate for these targets usingthe shortest exposure time with a minimum number ofFowler samples.The follow-up NIRC2 data were reduced using a cus-tom pipeline written in IDL. After standard image-processing tasks (background subtraction, flat-fieldingand bad pixel removal) were completed, the distor-tion in the image plane was then removed by the pro-gram nirc2dewarp available from the NIRC2 camerahome page, using the updated distortion solution ofB.Cameron (personal communication, 2007). Finally,aperture-photometry and centroids were computed withthe ImExam function of atv , version 2.0b4. Selected datasets were also analysed with the DAOPHOT function ofIRAF, giving results consistent well within the error bars.For centroiding, a centering box size of 5 pixels wasused, with the exception of one single observation inL-band, where a box size of 9 pixels was used. Theaperture-radius for photometry was 5 pixels for the Kpfilter, but 7 pixels for the J and H filters due to lowStrehls and dispersion, and 10 pixels in L-band due tothe larger diffraction-limited core. The sky annulus wasalways set at 10 to 20 pixels. Note that we did not at-tempt to calibrate absolute photometry, and have onlycomputed the relative photometry between primary andsecondary. RESULTS
The following subsections describe the multi-epoch as-trometry obtained on each of the three targets in Ta-ble 1, in order to confirm proper motion. Two clearlyassociated objects are discussed, as well as two objectsidentified as background stars.
New Companions
GSC 06214-00210 b
GSC 06214-00210 is a young M1 star near the east-ern edge of Upper Sco. It was originally identified as acandidate young star by Preibisch et al. (1998) basedon its X-ray emission, then confirmed to have stronglithium absorption ( EW = 0 . A ) and H α emission( EW = − . A ) as compared to stars of equivalent spec-tral type in several young clusters. It was also foundto have a proper motion µ = ( − . ± . , − . ± . − (Zacharias et al. 2010), consistent with the mo-tion of Upper Sco. Given its M1 spectral type, the 5 Myr mass-temperature relations of Baraffe et al. (1998)predict a mass of 0.60 M ⊙ . Note that this mass mustbe taken with caution, as an uncertainty of 1 subclassin spectral type should correspond to an uncertainty of ∼ M ⊙ in mass. One should keep in mind that themodels themselves carry an unknown uncertainty sincethe mass-luminosity and mass-temperature relations ofyoung stars are almost completely uncalibrated for . M ⊙ (e.g. Hillenbrand & White 2004).An example image of GSC 06214-00210b used for as-trometry is shown in Figure 1. Our photometric andastrometric observations of this companion are given inTable 2 and Figure 2. The candidate companion showeda relative motion of 7 ± ∼ ± − correspondsto 1.7 ± − assuming a 7 mas parallax for UpperScorpius, which is roughly equal to the circular orbitalvelocity of 1.7 km s − expected for a ∼
300 AU orbit.As can be seen from the larger contrast in bluer filters(cf. Table 2), GSC 06214-00210b is quite red comparedto its primary star. The primary has an observed 2MASScolor of J − K = 0 . ± .
04, and most M1 stars have atypical K − L ′ color of ∼ ± J − K = 1 . ± . K − L = 1 . ± .
10. The mean distance for UpperSco is 145 pc (de Zeeuw et al. 1999), and a ± ∼
14 pc error on thedistance of any individual star. This gives a a distancemodulus of m − M = 5 . ± .
2, so given the 2MASSmagnitudes of the primary, the absolute magnitudes ofthe companion are M J ∼ . M H ∼ . M K ∼ . M L ′ ∼ .
9. The interstellar extinction toward UpperSco is negligible ( A V . A K . . J − K color of 1.3 is consistent withthe M8-L4 spectral type range of field dwarfs fromLeggett et al. (2002). The K-L color of 1.05, however,would clearly place the star at the red end of this range,at L3-L4. The absolute K-magnitude of 9.1 is then about2 magnitudes brighter than for corresponding field ob-jects, providing clear evidence of a larger radius andyoung age; this height above the main sequence is similarto that observed for other late-type members of UpperSco (Lodieu et al. 2007) and TW Hya (Mamajek 2005;Teixeira et al. 2008).The colors and spectral type of GSC 06214-00210b arenot expected to match field dwarfs due to the low sur-face gravity. Allers et al. (2010) find that J-K and K-Lcolors are significantly redder for low gravity objects ata given spectral type. Although GSC 06214-00210b isa low-mass object, Allers et al. (2010) focusses on red-der object at J-K colors of ∼ & & TABLE 1Stellar Properties
Name RA DEC SpT Mass
R K (J2000) ( M ⊙ ) (mag) (mag)GSC 06214-00210 16 21 54.67 -20 43 09.1 M1 0.60 11.6 9.151RXS J160929.1-210524 16 09 30.30 -21 04 58.9 K7/M0 0.68-0.77 12.1 8.921RXS J160703.4-203634 16 07 03.56 -20 36 26.5 M0+? 0.68+0.59 11.3 8.10 Note . — R magnitudes are from USNO-B, while coordinates and K magnitudes are from 2MASS.The mass of the secondary star for 1RXS J160703.4-203634 is inferred from the mass ratio, whileother masses are directly inferred from the mass-temperature relations of Baraffe et al. (1998).Spectral types are taken from the discovery sources, Preibisch et al. (1998) and Kunkel (1999). TABLE 2Followup Observations
Julan Date Band Separation Position Angle Contrast(mas) (degrees) (mags)
GSC 06214-00210 b ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± a - a ± ± ± ± ± ± ± ±
14 219.5 ± ± ± ± ± ± ± ± ± ± ± ± ± ba Data not suitable for precision astrometry. b Data were taken through the corona400 coronograph at 1.04 ± ∼ (e.g. Scholz et al. 2007), with a trend of increasing disklifetime with decreasing mass. Spatially resolved spec-troscopy of the system would be required in order to de-termine the spectral type more accurately, and to searchfor signs of accretion onto the secondary.Given the observed brightness in the JHK filters, acomparison to the 5 Myr DUSTY models (Chabrier et al.2000) et al. 2000) suggests that the mass of the compan-ion is ∼ M Jup , with bluer filters suggesting slightlylower masses than redder filters. The COND models(Baraffe et al. 2003) never predict sufficiently red J − K colors to match our observations, and indeed are not ap-propriate for objects with T > M Jup . As we discussed above, the L ′ photometry could have an excess from a circumstellardisk, so we suggest that the L ′ magnitude should not beused directly in the mass estimate. Fig. 2.—
The observed position of the companion to GSC 06214-00210(diamonds), with the expected motion of the companion if itwere a background star with respect to the first epoch over-plotted(solid line, triangles at the times of observation). The dashed linesjoin the locations of the companion at each epoch with the expectedposition if it were a background star. This shows that GSC 06214-00210b is physically associated.
Fig. 3.—
J-K versus K-L’ color for our two confirmed com-panions, with the colors overplotted for M dwarfs from Leggett(1992) (triangles),M4 to L6 dwarfs from Golimowski et al. (2004)(asterisks), and for M giants from Fluks et al. (1994) (crosses).The L band photometry of 1RXS J160929.1-210524comes fromLafreni`ere et al. (2010). by Preibisch et al. (1998) and exhibits both stronglithium absorption ( EW = 0 . A ) and H α emission( EW = − . A ). The proper motion reported byUCAC3 (Zacharias et al. 2010) for this object is (-11.2,-21.9) ± − , which is also consistent with thevalue for Upper Sco. The spectral type reported byPreibisch et al. (1998) was M0, but, as was described byLafreni`ere et al. (2008b), newer measurements by D. C.Nguyen suggest a more likely spectral type of K7. The in- Ireland et al. Fig. 4.—
The observed position of the closer companion to1RXS J160929.1-210524, with symbols as in Figure 2. The twoobservations are at nearly indistinguishable locations in this plot,demonstrating that the companion is physically associated.. ferred mass would be 0.68-0.77 M ⊙ for the two estimates,with uncertainties similar to that for GSC 06214-00210.Our observations for 1RXS J160929.1-210524b arelisted in Table 2 and plotted in Figure 4. The compan-ion was first detected in Palomar images as early as May2007, but as the astrometric performance of the PHAROcamera was unverified on large timescales (cf. Section2.2), we were not able to convincingly demonstrate com-mon proper motion until 2009. This object was indepen-dently detected by Lafreni`ere et al. (2008b) and furthercharacterized by Lafreni`ere et al. (2010). Our astromet-ric results are consistent with those recently reported byLafreni`ere et al. (2010) at 2- σ in separation and at 1- σ in position angle, but have higher precision due to theaccurate astrometric characterization of NIRC2. Basedon a spectrum indicating a late spectral type and lowgravity, these authors argued that this faint compan-ion was a young object and therefore likely a physicalcompanion. Since that time, we have continued mon-itoring this object (cf. Table 2, and can now confirmthat 1RXS J160929.1-210524b is a physical companion of1RXS J160929.1-210524. The apparent proper motion of1.2 ± − , corresponding to 0.8 ± − . Thisis consistent with the ∼ − orbital motion ex-pected, especially considering possible projection effects.Since Lafreni`ere et al. (2008b) already reported NIRcolors, we only took K -band observations. These mea-surements are consistent with theirs, and indicate an ab-solute magnitude of M K ∼ .
4. The predicted massfrom the DUSTY and COND models is ∼ M Jup fromthe K band photometry or ∼ M Jup from J band pho-tometry, though again, these estimates are completelyuncalibrated by observations. Given the assumptionslisted above for the distance of Upper Sco, the projectedseparation is ∼
320 AU, similar to the value for GSC06214-00210 system.
Background Stars
Unassociated background stars typically are identifiedbased on multi-epoch astrometric monitoring (which in-dicates that they are not comoving) or multi-wavelengthobservations to measure colors (which indicate that theydo not fall along the same color-magnitude sequence).As we describe in the next several subsections, several of
Fig. 5.—
The observed position of the wider companion to1RXS J160929.1-210524, with symbols as in Figure 2. As the firstepoch has such large errors, we reference the apparent motion ex-pected from a background star to the weighted average of the twoepochs rather than to the first epoch. Motion of 1RXS J160929.1-210524with respect to this object is clearly detected, demonstratingthat the candidate companion is a background star. the closer companions will require astrometric monitor-ing with high-resolution imaging to confirm or disprovetheir association, but the wider companions can be iden-tified and rejected based on archival photometry fromseeing-limited all-sky surveys.
In addition to the companion listed as companion bin Table 2, an additional wider companion candidate,listed as companion c in Table 2 was found around 1RXSJ160929.1-210524 in both the 2007 Palomar images (seeSection 3.2.3) and the 2009 Keck images. This com-panion candidate was also reported by Lafreni`ere et al.(2008b). Although the Palomar astrometry had signifi-cant errors, the time baseline of 2 years was sufficient toclearly show that this object was a background star, asshown in Figure 5. ∼
27 AU) and flux ratio ∆ K = 0 .
15 (massratio q ∼ . ∼ ′′ .As we summarize in Table 2 and Figure 6, multi-epochastrometry for the candidate companion shows that it isnot comoving with 1RXS J160703.4-203634, but insteadappears to be nearly stationary, as would be expected fora background star. The relative motion of the primarywith respect to the companion is (7 ± ±
2) mas yr − ,consistent with the proper motion of Upper Scorpius.ide-separation planets in Upper Scorpius 7 Fig. 6.—
The observed position of the companion to 1RXSJ160703.4-203634 with respect to the center of light of the closebinary, , with symbols as in Figure 2. This candidate companionis consistent with a background star.
The contrast with respect to the brighter member ofthe close binary pair is ∆ K = 8 . ± .
08 and ∆ J =8 . ± .
05, indicating that the companion has a colorof J − K = 0 . ± .
10. Given that the total extinctionalong this line of sight is only A V ∼ E ( J − K ) ∼ J − K & & K5; Kraus &Hillenbrand 2007). Combined with its nonmotion (whichindicates it is likely to be quite distant), we thereforeconclude that the candidate companion is likely a back-ground K or early M giant, perhaps located in the MilkyWay bulge.
Wider Potential Companions
Multi-epoch observing campaigns can be observation-ally expensive, so where possible, it is best to use archivaldata to rule out possible companions. This is some-times impossible since many candidate companions ofinterest can only be distinguished from their candidateprimary using high-resolution imaging techniques. How-ever, wider companions can often be resolved in seeing-limited data, especially for new surveys that have verygood spatial resolution (i.e., . ′′ for UKIDSS images).Of the 25 companions with separations of & ′′ , 21were detected in both the H and K filters by UKIDSS, sowe can use their H − K colors to determine whether theymight be associated. UKIDSS observations of knownlow-mass members of Upper Sco by Lodieu et al. (2007)show that most members fainter than K ∼
13 (i.e. withspectral type & M7) have colors of H − K > .
5; thislimit is consistent with the typical H − K colors of & M7field dwarfs as compiled in Kraus & Hillenbrand (2007).As we show in Table 3, none of the candidate compan-ions with both H and K magnitudes meet this criterion,so we identify all of them to be unassociated field stars,most likely in the distant background behind Upper Sco.In addition, 17 of the wider candidate companions havecounterparts visible in the USNO-B1.0 digitization of thePalomar Observatory Sky Survey (Monet et al. 2003).Most are not present in the USNO-B1.0 source catalogsince its source identification algorithm was extremelyconservative in identifying faint neighbors to bright stars.However, these sources can be manually identified by vi-sual inspection. In Table 3, we list the bluest plate (B or R) at which each of the companions was visible. Aswe noted above, any true companions should have spec-tral types of & M7, so our compilation of field dwarf col-ors (Kraus & Hillenbrand 2007; Kraus & Hillenbrand,in prep) suggests that the expected colors for true com-panions are B − K > R − K >
6. The detectionlimits of the POSS survey were B ∼
21 and R ∼
20, so all17 sources with counterparts in the B or R plates mustbe background stars that are bluer than these limits. Inall cases, these identifications agree with the UKIDSSidentification.Three wider companions in Table 3 can not be elim-inated as Upper Scorpius members due to insufficientphotometry or astrometry. These are the companions toGSC 06793-00994, GSC 06794-00156 and ScoPMS 015.The closest of these systems is ScoPMS 015, with a com-panion just outside of 500 AU. This small incomplete-ness defines the outer limit of our survey until additionalfollow-up observations can be obtained, though some az-imuthal uncertainty remains at separations of <
500 AUdue to image boundaries (Section 3.3).
Detection Limits
In order to infer the properties of the distribution ofwide ULMCs, it is essential not only to establish the exis-tance of a small number of physical companions, but alsoto determine the magnitude limit as a function of sepa-ration for possible companions not confidently identifiedin the observations. This has not been general practicein previously reported wide ULMCs, in particular not forseveral of the .
40 M J companions in our separation andprimary mass range: GQ Lup b, CT Cha b and 1RXSJ160929.1-210524 b. In this section, we will describe thedetection limits for all stars in our sample.As we described in Section 2.1, we measured our detec-tion limits in a method that accounts for both spuriousdetections from speckle noise (at small separations) andthe sky background limit (at large separations). Theselimits could in principle be used as input for Monte Carloor Bayesian techniques to study the underlying popula-tion, though as we discuss in the Section 4, using theresults from our survey alone could yield a biased mea-surement.Table 4 lists and Figure 7 shows the companion detec-tion limits for each star in the sample. The last columnof Table 4 also gives the maximum separation in arcsec-onds where we surveyed all position angles for compan-ions. The few small ( . ′′ ) values of ρ in this columnare due to quick image sets taken in the camera sub-array mode we used for aperture-masking interferometry.The limits at separations smaller than 1 ′′ are set by theseparation-dependent speckle noise and extended PSFhalo of the primary star, while the wider limits at sepa-ration greater than 1.5 ′′ are constant and result from thesky background (for broadband K observations) or readnoise (for narrowband Brγ observations). For each star,we list the primary K magnitude (from 2MASS, withthe flux from any close binary companions subtracted),the detection limit in K sec at a range of angular separa-tion, and the corresponding detection limit in M Jup atthe corresponding projected orbital distances. The massdetection limits were derived from the K magnitudes ofthe 5 Myr DUSTY isochrone (Chabrier et al. 2000). Ireland et al.
TABLE 3Faint Candidate Companions to Young Stars in Upper Scorpius
Known Member ρ PA ∆ K USNO-B1.0 H UKIDSS K UKIDSS H − K (mas) (deg) (mag) (color/epoch) (mag) (mag) (mag)RXJ1603.6-2245 10959 ±
34 206 ± ± ± ± ± ±
29 280 ± ± ± ± ± ±
22 62.6 ± ± ± ± ± ±
36 47.7 ± ± ± ± ± ± ± ± ± ± ± ±
24 323.8 ± ± ± ± ± ±
32 69.1 ± ± ± ± ± ±
27 274.2 ± ± ± ± ± ±
47 183.7 ± ± ± ± ± ±
27 216.6 ± ± ± ± ± a ±
14 219.5 ± ± ±
33 116.4 ± ± ± ± ± ±
38 223.3 ± ± ± ± ± ±
17 357.4 ± ± ± ±
36 313.0 ± ± ± ± ± ±
39 30.4 ± ± ± ± ± ±
49 81.5 ± ± ± ± ± ±
16 73.6 ± ± ± ± ± ±
18 338.7 ± ± ± ±
51 180.8 ± ± ± ± ± ±
11 94.9 ± ± ±
23 25.7 ± ± ± ± ± ±
38 47.0 ± ± ± ± ± ±
21 190.3 ± ± ± ± ± ±
34 173.5 ± ± ± ± ± a See section 3.1.2 for multi-epoch astrometry of this object.
Fig. 7.—
Detections and detection limits for our direct imagingobservations of young stars in Upper Sco. Top: Contrast lim-its (∆ K in mag) as a function of angular separation (in arcsec-onds). Bottom: Corresponding limits in terms of secondary mass(in M Jup ) and physical separation (in AU). The detection limitsare shown with b lack dashed lines, the two confirmed ULMCs areshown with red points, and all other candidate companions (mostof which are confirmed as background stars) are shown with blackpoints. We have truncated the plot at a maximum separation of5 ′′ since deep seeing-limited imaging from UKIDSS demonstratesthat all wider companions are unassociated field stars (Section3.2.2). THE FREQUENCY OF WIDE ULTRA-LOW MASSCOMPANIONS
The most basic step in assessing the ability of currentplanet formation theories (see Section 1) to explain thewidely separated and relatively massive candidate plan-etary companions that are observed is to measure theirfrequency. Most of the 49 stars in our sample have rela-tively uniform detection limits ( ∼ M Jup at &
300 AU),so a naive estimate of the frequency is approximately2/49=4.1 +4 . − . %. Here we have quoted the most likely frequency, and the Bayesian 68% confidence interval onthe frequency with a prior distribution where all frequen-cies are equally likely. The detection limits are not com-pletely uniform, so the exact frequency will depend onmore sophisticated analysis using Monte Carlo (K08) orBayesian (Allen 2007; Kraus and Hillenbrand, in prep.)techniques, and ultimately should depend on the separa-tion and mass distributions of ULMCs.There is also a more fundamental issue that must beconsidered: this survey is not the first to be sensitiveto the presence of ULMCs, and a full treatment shouldconsider all surveys that have properly reported nulldetections and detection limits (e.g. Sartoretti et al.1998; Massarotti et al. 2005; Tanner et al. 2007;Lafreni`ere et al. 2008a; Metchev & Hillenbrand 2009;Chauvin et al. 2010). Chauvin et al. (2010) reported 1detection (AB Pic) from 30-40 young solar-type targetssurveyed (depending on the definitions of “young”and “solar-type”). Indeed, the full sample of past nulldetections is larger than our observed sample, suggestingthat the true frequency could be lower by up to a factorof &
2, and that our survey had good fortune to discovertwo new companions. Alternatively, including the com-panions to CT Cha and possibly GQ Lup (likely moremassive than 20 M J ) may increase the true frequency ofwide ULMCs, if only the details of the survey samplesin which these companions were discovered were known.A large census of the literature is beyond the scope of adiscovery paper, and the results of this analysis will bereported in a companion paper (Kraus et al., in prep.). DISCUSSION: IMPLICATIONS FOR FORMATIONMECHANISM
The population of directly imaged wide ( &
40 AU)ULMCs poses a significant challenge to planet forma-tion models. Existing models of solar-system scale planetide-separation planets in Upper Scorpius 9
TABLE 4Detection limits for additional companions
Name K lim (mag) at ρ = (mas) M lim ( M Jup ) at ρ = (AU) ρ
300 400 500 750 1000 1500 ≥ ≥
300 ( ′′ )GSC 06205-00954 12.9 13.4 13.4 14.1 15.0 16.4 17.8 42 28 28 21 14 6.9 4.5 10.4GSC 06208-00834 12.6 13.5 13.6 14.0 14.8 16.4 17.5 57 27 26 22 16 6.8 4.7 10.7GSC 06209-01501 12.3 13.3 13.1 13.7 14.8 16.1 17.7 74 30 33 25 16 7.6 4.5 10.4GSC 06213-00194 12.1 12.7 13.3 13.6 14.0 15.7 17.1 78 52 30 26 22 11 5.4 10.2GSC 06213-01358 12.9 13.6 13.8 14.2 15.2 16.5 17.9 39 26 23 20 14 6.5 4.4 10.4GSC 06214-00210 12.5 13.3 13.8 14.6 14.8 16.0 17.4 62 29 24 17 16 8.4 4.8 9.4GSC 06214-02384 12.3 12.9 13.5 13.9 14.4 15.8 17.5 76 41 27 23 18 9.9 4.8 10.2GSC 06764-01305 12.4 13.1 13.6 14.1 14.8 16.2 16.9 71 33 27 21 16 7.4 5.7 11.1GSC 06793-00797 11.7 12.9 13.2 13.8 14.4 16.0 17.3 97 42 31 24 19 8.3 5.0 10.2GSC 06793-00994 12.1 12.8 13.4 13.5 14.1 15.7 17.2 78 43 29 28 21 10 5.2 10.2GSC 06794-00156 9.8 10.7 11.4 12.1 12.9 14.6 17.2 413 212 130 78 41 17 5.1 10.2GSC 06794-00480 12.0 12.7 13.1 13.8 13.9 16.0 17.0 81 53 33 24 23 8 5.5 10.2GSC 06794-00537 11.9 12.7 13.1 13.6 14.0 15.7 17.7 89 54 34 26 22 10 4.6 10.2RXJ1550.0-2312 13.7 14.4 15.1 16.8 17.4 17.2 17.1 25 19 14 5.8 4.9 5.1 5.4 2.1RXJ1550.9-2534 11.4 12.8 13.8 15.6 16.8 17.1 17.3 129 44 24 11 5.9 5.4 4.9 1.3RXJ1551.1-2402 14.3 15.5 15.8 17.0 17.1 17.1 17.2 19 12 10 5.5 5.3 5.3 5.2 1.5RXJ1557.8-2305 13.7 15.0 15.5 16.9 17.3 17.2 17.1 25 15 12 5.7 4.9 5.1 5.4 3.7RXJ1558.1-2405 11.4 12.5 13.2 14.1 14.6 16.1 17.0 126 64 32 21 17 7.7 5.6 8.9RXJ1558.2-2328 11.4 12.1 12.7 13.3 13.9 15.4 17.1 128 78 54 29 23 12 5.2 10.2RXJ1600.7-2127 12.5 13.6 14.1 14.2 15.0 16.3 17.6 61 26 21 20 15 7.1 4.6 10.6RXJ1601.1-2113 12.7 13.4 13.5 13.7 14.8 16.5 17.4 56 28 27 25 16 6.6 4.8 9.6RXJ1601.9-2008 11.7 12.2 12.3 12.9 13.5 15.6 17.6 101 76 72 39 28 11 4.6 11RXJ1602.0-2221 13.1 14.4 15.1 16.8 16.9 17.0 17.1 35 19 14 5.8 5.7 5.5 5.3 3.7RXJ1602.8-2401A 11.1 11.8 12.4 14.1 15.0 16.0 17.0 156 95 71 20 15 8.4 5.6 3.6RXJ1602.8-2401B 10.1 10.6 11.3 12.4 13.3 14.7 15.1 332 226 131 67 30 16 14.0 10.2RXJ1603.6-2245 11.9 12.7 13.4 13.8 14.4 15.8 17.4 89 54 28 24 19 9.3 4.8 10.2RXJ1603.9-2031A 11.9 12.7 13.4 13.6 14.5 16.0 17.6 88 54 29 27 18 8.4 4.6 10.1RXJ1604.3-2130 11.9 12.3 13.0 13.7 14.3 15.9 17.1 83 73 37 25 19 8.7 5.2 8.9RXJ1606.2-2036 10.4 11.1 11.7 12.8 13.5 15.1 15.4 271 164 96 45 27 14 12.0 11.1RXJ1607.0-2036 9.1 12.4 13.3 13.9 14.5 15.6 17.9 648 69 30 23 18 11 4.3 8.7ScoPMS015 11.3 13.4 13.8 14.1 15.1 16.9 17.9 140 28 24 20 14 5.7 4.3 10.4ScoPMS017 13.9 14.7 15.0 16.9 17.6 17.4 17.2 23 17 15 5.7 4.6 4.9 5.2 1.5ScoPMS019 11.3 12.1 12.5 13.5 13.9 15.5 16.6 133 79 65 27 23 11 6.2 8.8ScoPMS022 14.3 15.5 15.5 17.4 17.4 17.3 17.3 19 12 12 4.9 4.9 4.9 5.0 1.4ScoPMS027 11.8 12.6 12.7 12.9 13.5 15.3 16.4 89 57 49 42 28 13 6.9 3.0ScoPMS028 13.1 15.0 15.9 17.1 17.2 17.0 16.9 34 15 9.3 5.3 5.1 5.4 5.7 3.7ScoPMS044 11.8 12.3 12.7 13.0 13.4 15.5 17.7 91 72 53 36 28 12 4.5 10.3ScoPMS045 12.8 13.4 13.8 14.2 14.6 15.8 18.0 48 28 24 20 17 9.6 4.2 10.7USco-160341.8-200557 14.0 15.4 15.7 17.3 17.4 17.4 17.3 22 12 10 5 4.8 4.9 5.0 2.2USco-160643.8-190805 10.1 10.8 11.6 13.2 13.8 15.6 15.8 336 203 110 30 24 11 9.4 8.5USco-160707.7-192715 14.6 16.1 16.6 17.9 17.9 17.7 17.6 17 7.6 6.2 4.4 4.3 4.5 4.7 1.5USco-160801.4-202741 11.1 11.9 12.5 13.7 14.1 15.8 16.1 158 86 64 25 20 10 7.8 8.5USco-160823.2-193001 14.4 15.5 15.7 17.3 17.4 17.3 17.3 18 11 11 5 4.9 5.0 5.0 3.7USco-160825.1-201224 14.3 15.8 16.3 17.3 17.5 17.5 17.5 19 10 7 4.9 4.8 4.8 4.8 1.5USco-160900.7-190852 10.8 11.5 12.2 13.6 14.1 15.6 16.0 196 115 76 26 21 11 8.3 6.0USco-160916.8-183522 14.3 15.8 16.0 17.4 17.7 17.6 17.4 19 9.6 8.1 4.9 4.5 4.7 4.9 2.1USco-160954.4-190654 14.5 15.5 16.0 17.4 17.4 17.5 17.5 18 12 8.4 4.9 4.9 4.8 4.7 2.2USco-161031.9-191305 10.8 11.5 12.2 13.3 13.9 15.5 16.1 200 114 77 29 23 12 7.7 8.5USco-161347.5-183459 13.4 13.5 13.5 13.4 13.3 13.2 13.1 28 27 27 29 29 31 34 2.1 Note . — Detection limits are 5- σ for all columns except for the ≥ ≥
300 AU columns, where they are 10- σ . The 5 Myrisochrones of the DUSTY models were used to compute K magnitudes. ρ refers to the maximum separation where our survey is100% complete for additional companions. formation (i.e., via core accretion or gravitational insta-bility in a Class II disk; Pollack et al. 1996) have notbeen succesful at forming planets at the orbital radiithey are observed here ( &
200 AU). It is also difficultto form wide ULMCs like a binary (i.e. via fragmenta-tion of the freefalling protostellar core or fragmentationin the massive protostellar disk) without subsequentlyaccreting sufficient mass to become a stellar or browndwarf companion. However, one of these mechanismsmust occur. Complicating this picture is the possibility,and indeed likelihood, of multiple planet scattering insome formation scenarios. Therefore, we cannot answerthe question of formation mechanism without examiningthe population of ULMCs in separation and mass spacealongside more well studied classes of companions. The core-accretion mechanism for planetary formationonly operates close to the host star ( ∼ ∼ ∼ AU separations (Kraus et al. 2008;Raghavan et al. 2010; Kraus et al 2010 in prep.). Asyoung solar-type stars have stellar companion fractionsin the range 12–22% (Brandner et al. 1996; K08; Krauset al, 2010, submitted) per decade of separation, thiswould mean that 0.2-0.3% of solar-type stars should have ∼ q =0.006 to 0.02 companions (i.e. 6–20 M J for solar-type stars. We see a clear surplus to this model,strongly suggesting that wide ULMCs follow a differentformation path to stars.If wide ULMCs were to form via fragmentation of acircumstellar disk when the primary is at the Class IIstage (disk masses 0.001-0.1 M ⊙ , e.g. Boss 2001) it wouldbe much easier to form wide companions than with coreaccretion. Although most observed disks have masses of ∼ J , a few disks around solar-type stars (e.g. DL Tau)have large enough linear dimensions and mass (up to ∼ M ⊙ , Andrews & Williams 2005, 2007)to fragment into observed ULMCs. Relatively little workhas been done examining in detail how these large disksmight fragment, with the bulk of disk-fragmentation lit-erature discussing the possible formation of solar-systemscale planets via fragmentation. Where models of largedisks have been computed (e.g. Meru & Bate 2010), itis clear that it is easier for them to become Toomre-unstable and fragment than ∼
20 AU disks. The remain-ing questions to be answered about this fragmentationinclude how many fragments are expected, and if thefragments can accrete enough of the disk mass to becomeobjects like GSC 06214-00210 b and 1RXS J160929.1-210524 b. The question of where each formation mechanism op-erates is certainly not answered yet, but could be in thenext few years. Radial velocity techniques will providereal constraints on the ∼ ∼ ACKNOWLEDGMENTS
M.I. would like to acknowledge support from the Aus-tralian Research Council through an Australian Postdoc-toral Fellowship. We would like to express thanks toundergraduate students Alison Hammond and MatthewHill from the University of Sydney who made a first-passastrometric analysis of the data. ALK was suported bya SIM Science Study and by NASA through Hubble Fel-lowship grant 51257.01 awarded by the Space TelescopeScience Institute, which is operated by the Association ofUniversities for Research in Astronomy, Inc., for NASA,under contract NAS 5-26555. Some of these observationswere obtained at the Hale Telescope at Palomar Obser-vatory, as part of a collaborative agreement between theCalifornia Institute of Technology, JPL and Cornell Uni-versity. Some of the data presented herein were obtainedat the W.M. Keck Observatory, which is operated as ascientific partnership among the California Institute ofTechnology, the University of California and the NationalAeronautics and Space Administration. The Observa-tory was made possible by the generous financial supportof the W.M. Keck Foundation. The authors wish to rec-ognize and acknowledge the very significant cultural roleand reverence that the summit of Mauna Kea has alwayshad within the indigenous Hawaiian community. We aremost fortunate to have the opportunity to conduct ob-servations from this mountain.
REFERENCESAllen, P. R. 2007, ApJ, 668, 492Allers, K. N., Liu, M. C., Dupuy, T. J., & Cushing, M. C. 2010,ApJ, 715, 561Anderson, D. R., et al. 2010, submitted to ApJL, arXiv:1010.3006Andrews, S. M., & Williams, J. P. 2005, ApJ, 631, 1134—. 2007, ApJ, 659, 705Baraffe, I., Chabrier, G., Allard, F., & Hauschildt, P. H. 1998,A&A, 337, 403Baraffe, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt,P. H. 2003, A&A, 402, 701Bate, M. R. 2005, MNRAS, 363, 363Bate, M. R., Bonnell, I. A., & Bromm, V. 2002, MNRAS, 332, L65Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393Boss, A. P. 2001, ApJ, 563, 367Bouchy, F., Deleuil, M., Guillot, T., Aigrain, S., Carone, L., &Cochran, W. D. 2010, accepted for A&A, arXiv:1010.0179Brandner, W., Alcala, J. M., Kunkel, M., Moneti, A., & Zinnecker,H. 1996, A&A, 307, 121Cameron, B. P. 2008, PhD thesis, California Institute ofTechnology, United States – CaliforniaCameron, P. B., Britton, M. C., & Kulkarni, S. R. 2009, AJ, 137,83Chabrier, G., Baraffe, I., Allard, F., & Hauschildt, P. 2000, ApJ,542, 464 Chauvin, G., Lagrange, A., Dumas, C., Zuckerman, B., Mouillet,D., Song, I., Beuzit, J., & Lowrance, P. 2004, A&A, 425, L29Chauvin, G., et al. 2010, A&A, 509, A52+Currie, T., Lada, C. J., Plavchan, P., Robitaille, T. P., Irwin, J.,& Kenyon, S. J. 2009, ApJ, 698, 1de Zeeuw, P. T., Hoogerwerf, R., de Bruijne, J. H. J., Brown,A. G. A., & Blaauw, A. 1999, AJ, 117, 354Deleuil, M., et al. 2008, A&A, 491, 889Fluks, M. A., Plez, B., The, P. S., de Winter, D., Westerlund, B. E.,& Steenman, H. C. 1994, A&AS, 105, 311Ghez, A. M., Neugebauer, G., & Matthews, K. 1993, AJ, 106, 2005Ghez, A. M., et al. 2008, ApJ, 689, 1044Golimowski, D. A., et al. 2004, AJ, 127, 3516Grether, D., & Lineweaver, C. H. 2006, ApJ, 640, 1051Haisch, Jr., K. E., Lada, E. A., & Lada, C. J. 2001, ApJ, 553, L153Hern´andez, J., et al. 2007, ApJ, 662, 1067Hillenbrand, L. A., & White, R. J. 2004, ApJ, 604, 741Hoyle, F. 1953, ApJ, 118, 513Kalas, P., et al. 2008, Science, 322, 1345Kratter, K. M., Murray-Clay, R. A., & Youdin, A. N. 2010, ApJ,710, 1375Kraus, A. L. 2009, PhD thesis, California Institute of Technology,United States – CaliforniaKraus, A. L., & Hillenbrand, L. A. 2008, ApJ, 686, L111 ide-separation planets in Upper Scorpius 11ide-separation planets in Upper Scorpius 11