Statistics of wide pre-main sequence binaries in the Orion OB1 association
aa r X i v : . [ a s t r o - ph . S R ] O c t Draft version October 19, 2020
Preprint typeset using L A TEX style emulateapj v. 12/16/11
STATISTICS OF WIDE PRE-MAIN SEQUENCE BINARIES IN THE ORION OB1 ASSOCIATION
Andrei Tokovinin
Cerro Tololo Inter-American Observatory — NSF’s NOIRlab, Casilla 603, La Serena, Chile
Monika G. Petr-Gotzens
European Southern Observatory, Karl-Schwarzschild-Strasse, 2 D-85748 Garching bei M¨unchen, Germany andUniversit¨ats-Sternwarte, Ludwig-Maximilians-Universit¨at M¨unchen, Scheinerstr 1, D-81679 M¨unchen, Germany
Cesar Brice˜no
Cerro Tololo Inter-American Observatory, — NSF’s NOIRLab, Casilla 603, La Serena, Chile
Draft version October 19, 2020
ABSTRACTStatistics of low-mass pre-main sequence binaries in the Orion OB1 association with separationsranging from 0 . ′′ ′′ (220 to 7400 au at 370 pc) are studied using images from the VISTA Orionmini-survey and astrometry from Gaia. The input sample based on the CVSO catalog contains 1137stars of K and M spectral types (masses between 0.3 and 0.9 M ⊙ ), 1021 of which are considered to beassociation members. There are 135 physical binary companions to these stars with mass ratios above ∼ ± ± Subject headings: stars:binary; stars:young INTRODUCTION
Orbital parameters and mass ratios of binary starsdepend on their formation environment. It is knownthat star formation regions (SFRs) of low stellar den-sity, like Taurus-Auriga, spawn a rich binary popula-tion, including a substantial number of very wide pairs(Joncour et al. 2017). In contrast, in more dense SFRsthe binary fraction is lower, comparable to the fieldbinary population (Duchˆene & Kraus 2013; King et al.2012). It is generally accepted that most stars in thefield were formed in relatively dense environments andthat some young wide binaries were destroyed by dy-namical interaction with neighboring stars. However,Duchˆene et al. (2018) found an excess of close (10–60au) binaries in the dense Orion Nebula Cluster (ONC),compared to the field. These close binaries are not sus-ceptible to dynamical disruption (Parker & Meyer 2014).Critical examination of binary statistics in several nearbySFRs has led Duchˆene et al. (2018) to the disconcertingconclusion that none of those groups is compatible withthe binary statistics in the field. However, the excess ofbinaries with separations <
60 au in the ONC has laterbeen contested by De Furio et al. (2019).The ongoing debate on the origin of the field binarypopulation and the role of SFR density and dynamicalinteractions in shaping the binary separation distribu-tion stimulates further observational studies. Currentlyavailable data on multiplicity statistics suffer from largeerrors owing to the small size of available samples and [email protected] from various biases caused by observational constraintsor sample selection effects. Modern large–scale surveysand catalogs change the landscape by providing large andhomogeneous data sets. For example, the Gaia census ofnearby wide binaries gave new insights on the distribu-tion of their separations and mass ratios (El-Badry et al.2019). A sample of ∼
600 stars in the Upper ScorpiusSFR has been recently observed with high angular resolu-tion to refine the binary statistics (Tokovinin & Brice˜no2020).Here we use the opportunity to learn about young bi-naries offered by the combination of three modern sur-veys: CVSO, VISTA Orion, and Gaia. The CVSO(CIDA Variability Survey of Orion; Brice˜no et al. 2019)was an optical, multi-epoch imaging survey that pro-duced a large sample of pre-main sequence (PMS) starsacross ∼
180 deg in the Orion OB1 association, span-ning all the region between α J = 5 | rmh − h , and δ J = − ◦ to +6 ◦ (Figure 1), with an average reso-lution of 3 ′′ (equivalent to 1112 au at 370 pc) as mea-sured in the original CVSO images. The young starswere selected based on photometric variability and con-firmed by follow-up spectroscopy. These PMS stars aremostly located outside the Orion A and B molecularclouds (Maddalena et al. 1986) and we refer to them asoff-cloud PMS stars. The off-cloud stars have on av-erage low extinction ( A V . . ∼ . − . M ⊙ . The off-cloud CVSO PMS starsare localized in the two main sub-associations in which Fig. 1.—
Wide field optical image of the area encompassed bythe CVSO in the Orion OB1 association, showing as an irregularpolygon the approximate footprint of the VISTA Orion Survey(Petr-Gotzens et al. 2011). The Orion OB1b sub-association is theregion within the dashed line circle, as in Brice˜no et al. (2005). TheOB1a sub-association is the area to the west of the OB1b regionand the dashed lines north and south of it, which roughly markthe limits of the Orion A and B molecular clouds, as indicated bythe labels. We also indicate the location of the three Orion beltstars δ , ǫ , and ζ Ori, as well as the σ Ori OB1 was been traditionally subdivided, namely OriOB1a and Ori OB1b (Blaauw 1964; Warren & Hesser1977), with further sub–clustering within each group(Brice˜no et al. 2019).The VISTA Orion survey (Petr-Gotzens et al. 2011)covers a 30 square degree area toward the Orion beltand was designed to overlap in large parts with theCVSO footprint, although the total area covered is muchsmaller, as shown in Figure 1. Its near infra-red (nIR)images in the
Z, Y, J, H, K s photometric bands have atypical stellar point spread function FWHM (full widthat half maximum) resolution of 0 . ′′
9, and allow us to de-tect binaries down to 0 . ′′ ∼
220 au at 370 pc distance). Statistics of widebinaries can be studied after accounting for chance pairsof unrelated field stars (optical companions). However,distinguishing statistically true binaries from random as-terisms becomes progressively uncertain with increasingseparation and magnitude difference.The Gaia Data Release 2 (Gaia collaboration 2018),hereafter GDR2, contains parallaxes and proper motions(PMs) of most bright stars in the VISTA Orion catalog,allowing a much more reliable distinction between realand optical pairs. At the same time, it helps to cleanthe main CVSO sample. However, GDR2 has its ownproblems, mostly caused by close (unresolved) binaries.As a result, reliable GDR2 astrometry is available formost, but not all, stars and companions in Ori OB1.This difficulty can be partially circumvented by usingthe CVSO and VISTA Orion data. So, all three datasources are more powerful when used jointly. We define the CVSO-VISTA-Gaia sample of PMS starsin section 2 and discuss its properties such as distance,clustering, etc. Then in section 3 the data on binarystars derived from the combination of the three surveysare presented and characterized. The resulting binarystatistics are studied in section 4. We summarize ourfindings and present our conclusions in section 5. THE CVSO-VISTA-GAIA SAMPLE OF PMS STARS
Target sample selection
The CVSO catalog of young stars in Orion byBrice˜no et al. (2019) served as a starting point for ourtarget selection. The CVSO contains 2062 spectroscop-ically confirmed T Tauri stars widely distributed acrossthe Orion OB1 association, mostly in the off-cloud re-gions, covering well over 100 square degrees on the sky(Figure 1; also, Figure 21 of Brice˜no et al. 2019). Thoughnot complete, we consider the CVSO to be a represen-tative sample of the population of PMS K and M typedwarfs in the off-cloud regions of the Orion OB1 asso-ciation, spanning the OB1a and OB1b sub-associations.First, because of how the sample was selected, it is notbiased toward accreting stars with optically thick disks,as would be the case of surveys that select objects withstrong H α emission or near-IR excesses. Therefore, weexpect it to contain a reasonable representation of bothaccreting and non-accreting PMS stars, most impor-tantly because the later constitute the bulk of the popula-tion in the slightly more evolved off-cloud areas of the as-sociation. Second, the spatial distribution of the CVSOPMS stars across the OB1a and 1b sub-associations isuniform enough, and there should be no significant, un-expected biases due to sampling only a small area of oneor the other region. We point out that because of how itwas constructed, the CVSO does not represent the muchyounger, on-cloud population, which we do not addresshere.Positions reported in the CVSO were determined witha custom pipeline that measured an ( x, y ) weighted cen-troid for each object; these positions were translatedto coordinates on the celestial sphere using astromet-ric transformation matrices referenced to the USNO A-2.0 catalog (Monet 1998). A positional match betweenCVSO right ascension and declination coordinates andthe 2MASS catalog (Skrutskie et al. 2006), using a 1 ′′ radius, yields a Root-Mean-Square (RMS) difference of0 . ′′ ± . ′′
17, sufficient for matching each source withother catalogs.We used TOPCAT (Taylor 2005) to match the CVSOcatalog star positions with the VISTA Orion source cat-alog, that provides accurate positions for ∼ ∼
80 mas forthe residual differential astrometry). The VISTA Orionsource positions are an average of the positions deter-mined in each of the photometric bands in which a sourcewas detected. A comparison of the VISTA source posi-tions with the UCAC 4.0 catalog (Zacharias et al. 2013)resulted in an RMS of ∼ . ′′
27 for the absolute astrometry,with no systematic offset (Spezzi et al. 2015). Runninga sky match with a 1 ′′ search radius between the CVSOand VISTA catalogs yielded 1216 matches, with an RMS http://casu.ast.cam.ac.uk/surveys-projects/vista of 0 . ′′ ± . ′′
17, which is dominated by the errors in theCVSO positions.We further restricted the matched sample as follows.We selected all stars that spatially belong to the popu-lations Ori OB1a or OB1b, as shown in Figure 1, butwe excluded a 0 . ◦ σ Ori, whichis the center of the eponymous stellar cluster. This ledto an initial list of 1137 stars, 405 in Ori OB1b, and 732in Ori OB1a including the 25 Ori and HR 1833 clusters.Table 1 contains all targets, numbered sequentially from1 to 732 and from 1001 to 1407 for the OB1a and OB1bgroups, respectively. These internal numbers N , alongwith the original CVSO numbers, are used throughoutthe paper. Cross-match with Gaia and characteristics of thesample
The next step was to match the sample of 1137 CVSOstars with VISTA catalog information against the GaiaData Release 2 catalog (GDR2). We first used Vizierto download all stars in GDR2 within a 30 ′′ radius ofeach of the 1137 CVSO target coordinates. Then, we dida cross-match between this temporary catalog and theCVSO positions, using a 5 ′′ radius and selecting the near-est Gaia source. Coordinate differences between CVSOand Gaia were small for single targets but offsets up to2 ′′ were found for binaries, because CVSO positions referto their unresolved (blended) images. When the offsetsbetween the CVSO positions and the actual positionsof primary components, determined from the VISTA im-ages as explained below, are accounted for, the rms coor-dinate difference with Gaia is 0 . ′′
06. Overall, 1078 CVSOstars have GDR2 astrometry. As for the other 59 ob-jects, 55 of them have no parallax and PM informationin GDR2, and 4 had no match at all in the GDR2 forno apparent reason (these stars are single and of averagebrightness: CVSO 685, 1097, 1283, 1819). Finally, we re-placed the CVSO equatorial coordinates with the GDR2coordinates (equinox J2000, epoch J2015.5), which weuse from now on.Having folded in GDR2 astrometry with the CVSO-VISTA sample, we can now take a look at the overallastrometric properties of our target sample. The toppanel of Figure 2 shows the location of the 1137 targetstars on the sky, where the symbols are colored accordingto the GDR2 parallax. The 55 stars with no parallaxesor PMs in the GDR2 and the 4 missing stars are markedwith crosses. The bottom panel of Figure 2 shows the PMdistribution of the 1078 targets having GDR2 astrometry.The closer stars (in red) are more tightly concentratedin PM space, while the more distant population (greenand blue) has a larger PM scatter.As already noted by Brice˜no et al. (2019), the GDR2astrometry shows that Ori OB1 stars are mostly locatedat distances from 300 to 450 pc, depending on the group.They also show, in accord with Figure 2, that the par-allaxes in Ori OB1b have a bi-modal distribution, indi-cating that closer stars, possibly belonging to Ori OB1a,project on the more distant Ori OB1b group.The detailed structure of the Orion star-formation re-gion is complex. It has been the subject of several stud-ies using GDR2 astrometry (Zari et al. 2019) and, ad-ditionally, radial velocities (Kounkel et al. 2018). Mostof our targets belong to the groups C and D identified
Fig. 2.—
Top: location of the 1137 CVSO-VISTA objects on thesky (squares for OB1a, triangles for OB1b). The points are coloredby parallax in the range from 2 to 3 mas, as shown by the colorbar. Black crosses are stars without GDR2 parallaxes. The dashedcircle indicates the approximate boundary of the Ori OB1b group.Bottom: distribution of the sample of 1078 objects with GDR2astrometry in proper motion space. The symbols and colors arethe same as in the top panel. by Kounkel et al.; these groups have different mean dis-tances (416 and 350 pc, respectively) and radial velocitiesbut spatially overlap on the sky. Since the structure ofthe Ori OB1 association is outside the scope of this paper,and our focus is on binaries, we use the traditional divi-sion into OB1a and OB1b groups based only on the skylocation, following the boundaries used by Brice˜no et al.(2005, 2019). Their mean parallaxes are 2.748 and 2.576mas respectively, corresponding to distances of 363 and388 pc. The mean PMs are close to zero and have adispersion of ∼ − . We point out that groupsOB1a and OB1b as considered here are, however, nothomogeneous in terms of their age and distance. OB1acontains clusters like 25 Ori and HR1833 within the morewidely spread ”field” PMS population. Though it seemsclear that OB1a as a whole is a population originat-ing in an earlier star-forming episode compared to OB1b(Kounkel et al. 2018; Brice˜no et al. 2019), the ages anddistances we use are only indicative.Using GDR2 astrometry, in the next section we willinvestigate the membership of our targets to Ori OB1aand Ori OB1b, and identify likely non-members. In fact,Brice˜no et al. (2019) note that their catalog of OrionPMS stars can still be slightly contaminated by activeforeground K- and M-type dwarfs with spectral signa-tures resembling those of PMS stars. For example,CVSO 569 is a 6 . ′′ − , −
7) mas yr − and parallaxes about4 mas; this is a physical binary, and its spectrum doesshow H α in emission and Li I 6707 in absorption, there-fore it is clearly a low-mass, PMS star but likely fore-ground and unrelated to the Orion OB1 PMS population.Finding young stars with motions discrepant from thosegenerally agreed to characterize the bona-fide Orion OB1population seems increasingly less surprising, since re-cent studies find that the structure of the stellar pop-ulation across Orion is richer and more complex thanpreviously thought (Chen et al. 2019; Kos et al. 2019). Analysis of the GDR2 astrometry
The large distance to Ori OB1 and its small PM meanthat very accurate astrometry is needed to discriminatetrue association members from foreground and back-ground stars. In addition, unresolved binaries degradethe quality of GDR2 astrometry. Therefore, we focus inthe following on filtering out from our targets the likelynon-members, but keep those that potentially have theirGaia astrometry compromised due to the presence of abinary companion. The latter can be evidenced in threedifferent ways.First, pairs with separations from 0 . ′′ . ′′ m often have undeter-mined astrometric parameters (parallax and PM) be-cause they were recognized as non-point sources. For ex-ample, all GDR2 stars without parallaxes were resolvedin the speckle interferometric survey of Upper Scorpius(Tokovinin & Brice˜no 2020). There are 55 of our targetsthat do not have GDR2 parallaxes (section 2.2). Sec-ond, the GDR2 astrometry of many close binaries, whenpresent, is often substantially biased because their mo-tion does not conform to the standard 5-parameter as-trometric model. Typically, these stars have large errorsof astrometric parameters, e.g. the parallax error σ ̟ .Our experience shows that the parameters of such starscan deviate from their true values (known, e.g., fromwide components of well-resolved physical triple systems)much larger than allowed even by those inflated errors.In short, the GDR2 astrometry of these stars is unre-liable. Third, even when the 5-parameter astrometricmodel is adequate, the PM can still be slightly biased bythe orbital motion in a long-period binary. A solar-massbinary with a semimajor axis a (in au) and a typical massratio of 0.5 would have the orbital PM on the order of1 . /a ) . mas yr − at a distance of 370 pc.In order to define a measure for the reliability of GDR2astrometry specific to our target sample, we plot in Fig-ure 3 the parallax error vs. G magnitude for all targetswith GDR2 astrometry and having G <
19. Note, forvery faint targets the GDR2 astrometry becomes very
Fig. 3.—
Parallax errors vs. G magnitude (crosses). The full lineis σ ( G ) defined by equation. 1, the dashed line represents 2 σ ( G ). uncertain and we therefore reject a priori 17 stars with G >
19, which means in the context of our analysis wedefine those as non-members. The distribution shown inFigure 3 follows a well-defined trend which we approxi-mate by the formula σ ( G ) ≈ .
024 + 0 . G −
13) + [0 . G − ] , (1)where the quadratic term is added only for G >
16 and σ ( G ) is in mas. We then use the ratio of the parallaxerror to its model, r ̟ = σ ̟ /σ ( G ), as a measure of theexcess astrometric noise indicative of biased GDR2 as-trometry. The Gaia errors depend on the source positionon the sky, and we caution against using our simplisticmodel (1) in a more general context; it is just suitablefor Orion.Figure 4 plots the parallax and total PM of our targetswith colors that correspond to r ̟ . Targets with reliableastrometry, defined here as r ̟ <
2, are tightly concen-trated at parallaxes between 2.2 and 3.2 mas and PMsbelow 3 mas yr − . A bi-modal distribution of parallaxescan be noted. Considering potential biases caused byunresolved binaries, we adopt the following relaxed cri-teria. Targets with G <
19 and parallaxes from 1.5 to4 mas and a total PM less than 5 mas yr − , irrespectiveof their excess noise r ̟ , are considered astrometricallyconfirmed members of Ori OB1. There are 934 stars thatcomply with these criteria and are assigned a member-ship flag 2 in Table 1. The high rate of astrometricallyconfirmed members of Ori OB1 validates the spectro-scopic and photometric selection of young stars adoptedin the construction of the CVSO sample. For compari-son, Kounkel et al. (2018) adopted a parallax range from2 to 5 mas and the PM limit of ± − in both co-ordinates as membership criteria.All targets with reliable astrometry, i.e. r ̟ <
2, but to-tal PM and parallax values outside our adopted selectionbox are considered astrometric non-members (member-ship flag 0 in Table 1, 89 stars). The remaining 41 starswith unreliable GDR2 astrometry (likely close binaries)outside the adopted parallax and PM limits (includingthree with negative parallaxes) are considered as mem-bers, unless their total PM is larger than 13 mas yr − .This PM threshold is chosen by examining the tail of thePM distribution and applies to only 6 stars, which means35 stars are finally considered as members. These mem- Fig. 4.—
Correlation between parallax and total PM for oursample. The symbols are colored according to the excess error r ̟ ,as shown by the color bar. The dotted box shows the limits ofparallax and PM adopted for the 934 astrometric members. bers are assigned the membership flag 1 to distinguishthem from astrometrically confirmed members. Mem-bership flag 1 is also assigned to 52 stars with missingGDR2 astrometry and G <
19. Admittedly, the thresh-old r ̟ < G = 19 mag, and the 6 stars with unreliable GDR2astrometry and total PM larger than 13 mas yr − areexcluded from the following statistical analysis togetherwith the astrometrically confirmed non-members (mem-bership flag 0). The numbers of targets with variousmembership status are reported in Table 2. Overall,there are 1021 members, 658 in Ori OB1a and 363 inOri OB1b. We provide data for all 1137 targets of theoriginal sample and their companions and use the mem-bership flag defined here only for evaluation of the mul-tiplicity statistics.The CVSO-VISTA-Gaia targets are listed in Table 1.They are numbered sequentially from 1 to 732 for starsin Ori OB1a and from 1001 to 1407 for those in OriOB1b (the latter group contains 405 stars). Withineach group, the targets are ordered in the right ascen-sion. These numbers N , along with the CVSO num-bers from Brice˜no et al. (2019), link the targets to thelists of double stars presented below. In the followingcolumns of Table 1 we give the information extractedfrom GDR2, namely the equatorial coordinates (equinoxJ2000, epoch 2015.5), parallax ̟ , its error, proper mo-tions µ ∗ α and µ δ , and the G band magnitude. The J mag-nitude from 2MASS and the spectral type are retrievedfrom the CVSO catalog. The last three columns con-tain the excess noise r ̟ (zero if parallax is not known),the membership flag, and, for binaries, the separation inarcseconds. Photometry and CMD
Figure 5 shows cumulative distributions of J magni-tudes for members of Ori OB1a and Ori OB1b. Themedians are 13.21 and 12.93 mag, respectively, consis-tent with Ori OB1a being slightly older than Ori OB1b; Fig. 5.—
Cumulative distributions of J magnitudes for all targetsclassified as members (flag 1 or 2). Fig. 6.—
Color-magnitude diagram. Known binaries with sepa-rations less than 5 ′′ are plotted by green triangles, other stars byblue crosses. Absolute magnitudes have been derived for each starbased on its parallax. Two PARSEC isochrones for solar metal-licity are plotted. The squares on the isochrones and numbersmark masses from 0.3 to 0.9 M ⊙ . The line marks the effect of an A V = 0 . the median G magnitudes in these groups are 16.20 and15.89 mag.The color-absolute magnitude diagram (CMD) in Fig-ure 6 shows only 934 astrometrically confirmed membersof the association with measured parallaxes. We plot the4 Myr and 10 Myr PARSEC isochrones for solar metallic-ity (Tang et al. 2014) using the 2MASS and Gaia colorsand mark corresponding masses; these ages are consistentwith those adopted by Brice˜no et al. (2019) for OB1b(5 Myr) and OB1a ( ∼
11 Myr); remember though thatthese groups are not strictly coeval, as noted before. Wedo not use here the VISTA Orion photometry becausefor brighter targets it is biased by saturation. The ex-tinction is not corrected for, but since these are off-cloudpopulations, the overall reddening is small. In fact, forour sample the median extinction A V determined in theCVSO catalog (Brice˜no et al. 2019) is 0.36 mag. About24% targets have A V = 0, and only 11% have A V > A V = 1 magcorresponds to A G = 0 .
47 mag and A J = 0 .
24 mag for astar of 4000 K effective temperature. The A V = 0 . TABLE 1CVSO-VISTA-Gaia sample (fragment) N CVSO α δ ̟ σ ̟ µ ∗ α µ δ G J
Spectral r ̟ Memb. ρ (deg) (deg) (mas) (mas) (mas yr − ) (mag) (mag) type ( ′′ )1 405 79.57240 − − − − − − − − − − TABLE 2Classification of the targets
Member flag N < r ̟ < r ̟ > r ̟ = 02 934 869 65 01 87 0 35 520 116 102 7 7 bluer (or fainter) compared to the isochrones, showingthat evolutionary models of PMS stars are still far fromperfect. This systematic deviation from the isochrones isconfirmed by our photometry of binaries, see section 3.5.Binary stars are located on the CMD above the single-star isochrone. Known binaries with separations lessthan 5 ′′ are distinguished in Figure 6 by green triangles.The G magnitudes of those 61 targets refer to the pri-mary components resolved by Gaia, while their J magni-tudes from 2MASS refer to the combined light, displacingthe points to the right by as much as 0.75 mag. How-ever, the majority of binaries are not recognized becausethey are closer than 0 . ′′
6, the resolution limit of our sur-vey. Binarity certainly contributes to the scatter in theCMD.The CMDs of various sub-groups of the Ori OB1 as-sociations are plotted and discussed by Brice˜no et al.(2019) and Kounkel et al. (2018). They derive model-dependent ages ranging from 4 to 13 Myr for varioussub-groups. However, even within one sub-group thespread of the CMD is substantial. One of the reasonsis that all CVSO stars are variable (this was one of theselection criteria in building the sample). The variabilityof low-mass PMS stars ranges from a median value of 0.5mag in the V band for accreting Classical T Tauri stars,caused by a combination of variable accretion, rotationalmodulation by hot/cold spots and possible disk obscu-ration, down to ∼ . M ⊙ , with 0.4 to 0.8 M ⊙ beingdominant, i.e., spectral types ∼ K2 to M4 (see Figure 1in Brice˜no et al. 2019). However, masses of PMS starsestimated from absolute magnitudes or colors are knownto be highly uncertain. The isochrones appear to deviatesystematically from the observed pre-main sequence, andthe problem is aggravated by the intrinsic variability of
Fig. 7.—
Surface density of companions vs. separation in OriOB1a and OB1b. The dotted line shows the companion density inTaurus according to Larson (1995). all CVSO stars that adds uncertainty of magnitudes andcolors. Moreover, ages for individual stars are not welldetermined and there appears to be a considerable agespread in both sub-associations. Given these intrinsicuncertainties, we refrain here from estimating individualmasses and mass ratios. A crude estimate of mass ratiosbased on the isochrones is used here only for the purposeof translating the limit of our survey from photometriccontrast into approximate mass ratio. The isochronessuggest that the magnitude difference in the J band isrelated to the mass ratio of a young binary q = M /M as q ≈ − . J (see section 4.1). In the following, we es-timate approximate mass ratios using this formula with-out insisting on its correctness or uniqueness. Accord-ing to this relation, binaries with ∆ J < q > . . J <
Clustering and chance projections
Companions belonging to the Ori OB1 group accordingto the astrometric and photometric criteria are not nec-essarily bound to the main targets. Instead, they couldbe random pairs of association members projecting closeto each other on the sky. To elucidate this issue, wecomputed the spatial density of CVSO stars around eachtarget in 4 annular zones with a logarithmic radius stepof 0.5 dex, from 9 ′′ to 900 ′′ (0 . ◦ − .
62 at separations exceeding 10 au and − .
15 atcloser separations. Compared to Orion OB1, Taurus hasa much lower density and a stronger clustering inher-ited from the structure of molecular clouds. In contrast,in the older Orion OB1 association the stars are wellmixed at scales less than a parsec, although they retainclustering at larger scales (Brice˜no et al. 2019, see alsoFigure 2).The first bin shows a reduced stellar density in OriOB1a compared to larger scales, in strong contrast withTaurus. Taken at face value, this implies an anti-correlation, i.e. avoidance of close pairs relative to auniform distribution. Most likely, this is a selection effectthat arises from the construction of the CVSO sample. Itused multi-fiber spectroscopy for confirming the PMS na-ture of ∼
70% of the candidates. Because there is a min-imum distance on the sky between adjacent fibers (e.g.20 ′′ for Hectospec; Fabricant et al. 2005), close compan-ions (also PMS stars) would not have been observed forthis technical reason. Therefore, we ignore this effect andassume that the average density is 55 stars per squaredegree in both groups. This means that we expect tofind 1.4 and 5.4 random pairs of association memberswithin 10 ′′ and 20 ′′ , respectively, in a sample of 1021stars. However, this is only a lower limit because theCVSO does not contain a complete census of the asso-ciation members; this is further explored below usingGDR2. The expected number of random pairs is sub-tracted in the following analysis of the separation distri-bution. We restrict the statistical analysis to separationsbelow 20 ′′ and to moderate ∆ m to minimize the impactof random pairs. Extending these limits would aggravatethe uncertainty caused by random pairs of associationmembers. OBSERVATIONAL DATA AND THEIR ANALYSIS
Our primary source of data on binaries is the exami-nation of the images from the VISTA Orion mini-survey.We attempted to detect almost all companions within7 ′′ from all original 1137 CVSO stars visible in the im-ages and only later realized that the detection depth isexcessive for our survey that needs only a contrast up to3 mag. When studying the binary frequencies (section 4)we will restrict the analysis to the members of Ori OB1.We complemented the image analysis by searching forwider pairs in the VISTA Orion photometric catalog andby identifying all pairs in the GDR2. Joint analysis ofthis information allows us to discriminate real binariesfrom unrelated (optical) asterisms and sets the stage forthe statistical analysis presented in section 4. Detecting binaries in the VISTA images
The VISTA Orion mini-Survey (Petr-Gotzens et al.2011) provides seeing-limited images with a typicalFWHM resolution of 0 . ′′ . ′′ Z, Y, J, H, K s bands for one field were executed sequen-tially in all filters, spanning no more than 2–3 hours in CVSO 1195 ResidualsCVSO 1516 Residuals
Fig. 8.—
Modeling the postage-stamp images by Moffat func-tions. The images are shown on the left, the residuals on the right.Top: the image of target 1027 (CVSO 1195) in the H band withthree stars. Note another faint star in-between that has been ig-nored. Bottom: the image of target 1176 (CVSO 1516) in the J band, with an insert showing residuals for the main star indicatingthe elongation. The residuals after fitting three stars (on the right)are smaller. total. This way effects of variability on the stars’ colorsshould have been diminished.For each CVSO target, fragments of VISTA imagesof 43 ×
43 pixels, corresponding to a size of 14 . ′′ × . ′′ ′′ (up to 10 ′′ in the corners) can be found in thesepostage stamps. The VISTA/VIRCAM focal plane isa mosaic of detectors, with large gaps in between, thatmust be dithered in a 6-point pattern to contiguouslyfill the field of view. Furthermore, at Z and Y filterslong and short exposures were taken. This means tar-gets were imaged several times in all five filters. How-ever, images where targets fall near a detector edge arepartially truncated. Among the 37,464 postage stampsused in this project, 836 severely truncated ones are ig-nored. On average, there are 5 images per target in thefilters Z and Y , 10 images in J and H , and only 2.4 in K s ; some targets lack the K s –band images altogether.We used a custom IDL code to process these images,automating the work as much as possible. For each tar-get, the program selects all postage stamps and displaysthe chosen (usually the first) image in a graphical win-dow. The user defines the number of visible stars andtheir approximate positions and fits a model to deter-mine accurate relative positions and intensities of thesestars. Modeling of all other images of this same targetcan then be done by one command, using previous resultsas a first approximation.Several important comments are in order here. First,all CVSO targets are 4 to 6 mag brighter than thefaint magnitude limit of the VSTA Orion survey, assur-ing that well-resolved companions with a contrast un-der 3 mag are always securely above the noise-limiteddetection threshold. Second, no good estimates of thepoint spread function (PSF) are available because manypostage stamps contain just one star, the target itself.So, a decision on whether the PSF asymmetry is causedby a close semi-resolved companion or by a residual tele-scope jitter (the ellipticity of most images reported in theheaders is under 0.05, but in some cases reaches 0.1) isnot always straightforward. Unlike the situation in thesurvey of De Furio et al. (2019), where accurate mod-els of both PSF and noise were available, detections ofclose companions in the VISTA images cannot be auto-mated and their significance cannot be rigorously eval-uated by a metric like χ . On the positive side, how-ever, we have multiple images of each target obtainedunder different seeing conditions in five filters. There-fore, the companions are confirmed as many times asthere are images. Mutual agreement of binary-star pa-rameters derived from many independent postage-stampimages guarantees the reliability of detections; they areall secure and there are no false positives, as indicated bythe independent detection of all, except one, VISTA closebinaries (0 . ′′ < ρ < . ′′
2) by GAIA and/or high spatialresolution observations (cf. section 3.7). The detectionlimit is further discussed in section 3.4.The background level in each image is determined bythe median pixel value and further refined by excludingpixels around known stars within a radius of four timesthe FWHM resolution, typically about 12 pixels. Imagesof stars that do not overlap significantly can be modeledby fitting a symmetric Moffat profile F ( x, y ) = p [1 + r /a ] β , r = ( x − p ) + ( y − p ) (2)with five free parameters p , p , p , p = a, p = β . Thefirst two parameters are pixel coordinates of the center,the third is the maximum intensity, the parameters a and β define the width and shape of the point spreadfunction (PSF). The FWHM equals 2 a √ /β −
1. Thebackground level is subtracted prior to fitting and notincluded in the model. The PSF is fitted by minimizing R , the un-weighted normalized rms difference betweenthe image I i and its model M i over all pixels i within aradius of 10 from the center: R = sX i ( I i − M i ) / sX i I i . (3)The residuals for single stars are dominated by the dif-ference between the actual PSF shape and its model (2),rather than by the detector and photon noise. Hence R isthe appropriate goodness of fit metric and its minimiza-tion achieves the best approximation of the PSF shape.For modeling saturated stars, pixels near the center canbe excluded. We also tested elliptical Moffat models withtwo additional parameters, ellipticity and orientation,but found that the symmetric model (2) works well inmost cases; therefore, the elliptical Moffat function wasnot used.When the pair is well separated, we model the sec-ondary star by fixing the PSF parameters to those of theprimary and fit only the position and relative intensity.For partially overlapping stars, we have the option of ad-justing the common parameters a and β for all stars, i.e.2 + 3 n parameters for an image containing n stars. This method works very well even for close (blended) pairsand it was used for measuring all companions. Residualsafter fitting a triple source 1027 (CVSO 1195) are shownin Figure 8. The two companions are separated from themain star by 6 . ′′ . ′′ J of 2.3 and 2.9mag, respectively; both are unrelated field stars.Detection of close binaries with separation less than theFWHM is helped by modeling the PSF by a symmetricMoffat function and visual examination of the residu-als. A persistent asymmetry of multiple images of thesame target indicates a real companion, as opposed tooccasional PSF elongation. An a posteriori test of com-panion detection is furnished by comparison with Gaia(section 3.3). The lower panel of Figure 8 illustratesmodeling of the close binary star 1176 (CVSO 1516) inan image with a FWHM resolution of 0 . ′′
96. The residu-als after approximating the central star by a symmetricMoffat function (in the insert) have a “butterfly” shapeindicative of asymmetry and are large, R = 0 . . ′′
61 yields smaller residuals of R = 0 . . ′′
75 pair . ′′ m < θ and separation ρ are average values for all processed im-ages in all filters where the given companion is detected.The rms scatter σ ρ gives an idea of the internal agree-ment between these measurements. The following fivecolumns give the average magnitude differences in theVISTA Z, Y, J, H, K s bands. The remaining five columnscontain the rms scatter of ∆ m in each filter where two ormore measurements are available. For a single measure-ment, the scatter is zero. Some companions lack mea-surements in some filters either because these images areunavailable or because the companions were not detectedowing to noise or truncation. Table 3 contains 490 rows,i.e. unique companions to our 1137 targets, with all com-panions having a detection in at least two filters. Themajority of targets have one companion, and at mostfour. Most of these companions are unrelated field stars.The Moffat models also provide the FWHM resolutionin each image through parameters a and β . Its medianvalue is 0 . ′′
88, the mean is 0 . ′′
89, and the dispersion is0 . ′′
20. Ninety per cent of FWHM values are comprisedbetween 0 . ′′
74 and 1 . ′′ Wide pairs in the VISTA Orion photometriccatalog
The VISTA Orion photometric catalog contains equa-torial coordinates and
Z, Y, J, H, K s magnitudes of allpoint sources found in the survey. The catalog is typ-ically complete (at 10 σ significance) to Z = 21 . J =19 .
6, and K s = 17 . J < J ∼
16 mag.The catalog was also used to study the stellar density
TABLE 3Companions found in the images (fragment) N CVSO θ ρ σ ρ ∆ Z ∆ Y ∆ J ∆ H ∆ K s σ ∆ Z σ ∆ Y σ ∆ J σ ∆ H σ ∆ Ks (degr) ( ′′ ) ( ′′ ) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag)1 405 340.1 3.653 0.201 5.61 5.59 5.06 4.89 . . . 0.19 0.35 0.16 0.00 . . .4 425 154.6 3.457 0.019 5.06 5.12 5.24 5.23 4.93 0.15 0.24 0.00 0.00 0.005 427 12.1 7.249 0.026 2.64 3.23 3.57 3.88 3.76 0.02 0.16 0.16 0.30 0.086 432 101.9 3.440 0.019 2.15 1.91 1.76 1.89 1.76 0.03 0.04 0.00 0.04 0.0114 455 228.5 7.596 0.147 5.58 5.59 5.73 5.70 . . . 0.19 0.31 0.00 0.00 . . .19 461 40.5 7.402 0.041 4.62 4.64 4.14 3.94 3.16 0.13 0.24 0.15 0.07 0.07 as a function of magnitude in the area of the Ori OB1association, to account statistically for the backgroundcontamination. The number of stars n per square de-gree brighter than a certain J magnitude is log n ( J ) ≈ .
44 + 0 . J −
15) in both groups of Ori OB1. Thesemodels are no longer needed in the light of Gaia, butcould be used to compute the density of unrelated com-panions.We retrieved as companions all catalog stars that areseparated between 2 ′′ − ′′ from the CVSO target po-sition and have ∆ J < ∼ ′′ . The comparison revealed that coordi-nates of single stars in the CVSO catalog match theirVISTA Orion coordinates with a median offset of only0 . ′′
047 and the maximum offset of 0 . ′′
23. However, forbinaries the spatial match was much worse because theCVSO positions refer to the centroids of blended imagesowing to its 3 ′′ typical FWHM resolution. Our imageanalysis gives the offsets of the main component from thepostage-stamp center (which is at the CVSO position).The CVSO coordinates corrected for these offsets matchthe VISTA positions with a median difference of 0 . ′′ ′′ are found in the VISTA Orioncatalog, without the need to examine the images. Weadded 411 wide pairs with ∆ J < ′′ < ρ < ′′ to the companion list, thus extendingthe separation range to 20 ′′ . The majority of these widecompanions are unrelated field stars, as shown below insection 3.5. Companions in GDR2
The GDR2 catalog was queried within 30 ′′ radius ofeach target. The coordinate offsets determined from theimage analysis helped to match securely the main targetswith GDR2 (the coordinates agree within ∼ . ′′ G are plotted in Figure 9. We matched the VISTA Orionpairs to the list of companions in GDR2 and thus re-trieved the GDR2 astrometry and photometry for allpairs, except for four close ones with separations below ∼ . ′′ Fig. 9.—
Separation versus ∆ G of companions found in Gaiawithin 20 ′′ . The line shows the detection limit ∆ G < ρ − . . that approximates the detection limit at 50% probability found byBrandeker & Cataldi (2019). in the G band). Conversely, six close pairs in GDR2 werenot recognized in the VISTA images (targets No. 253,458, 697, 1176 1257, 1258). All except two have sepa-rations below 0 . ′′
6. Sources 697 (0 . ′′
75) and 1176 (0 . ′′ > . ′′ r ̟ < p astro = 1. Otherwise, p astro = 0 and thepair is considered optical. For 78 companions with miss-ing or unreliable astrometry, we set p astro = 0 . p astro = 1, ρ < ′′ , and reliable GDR2 data for both components is0.6 mas yr − . We suspect that unrecognized inner sub-systems contribute to the scatter of relative motions inthese wide pairs.We compared the relative astrometry derived fromthe VISTA images with the presumably more accurateGDR2 astrometry (Figure 10). The agreement is excel-lent for astrometrically confirmed members with p astro =1 and ρ < ′′ , measured by us in the images. The meanoffsets between the relative companion’s positions in theVISTA images and in Gaia in the radial and tangentialdirections are +1 and +2 mas, respectively, while the rmsscatter of these offsets is 8 mas in both directions. On the0 Fig. 10.—
Comparison of double-star astrometry between VISTAOrion and Gaia. Squares are astrometrically confirmed associationmembers, pluses are other (mostly optical) companions with ∆
G < ρ sin ∆ θ . The axis scale is in arcseconds. Fig. 11.—
Location of companions in the ( ρ, ∆ J ) plane. Squaresdenote pairs found in our image analysis, crosses plot companionsfound both in the images and in the VISTA Orion catalog. Thedashed line is the empirical detection limit (eq. 4), the dotted rect-angle marks the limits of our statistical analysis. other hand, the positions of pairs wider than 7 ′′ rely onthe coordinates from the VISTA Orion catalog and areaccurate only to a fraction of an arcsecond. Detection limit
Figure 11 plots the location of all companions in the( ρ, ∆ J ) plane. One can appreciate the advantage of ourimage analysis in terms of resolution and contrast, com-pared to using solely the VISTA Orion catalog.The companion detection in the VISTA images de-pends on the variable FWHM resolution and on the sig-nal to noise ratio. Here we use the simplified optimisticempirical detection limit (dashed line in Figure 11) de-scribed by the formula∆ J < ρ + 0 . . , ρ > . ′′ . (4)This formula is chosen “by eye” to fit the envelope of thepoints. We also studied the empirical detection thresholdas a function of FWHM resolution, but, considering thelimited range of FWHM variation and multiple imagesavailable for each target, decided on a simpler alternative(4). According to this formula, all pairs with separation ρ > . ′′ J <
CVSO 1490 1281 CVSO 17431.25", 2.9 mag1.06", 2.5 mag1168
Fig. 12.—
Postage-stamp H -band images of two triple systemswith close faint companions, in negative rendering. Separationsand ∆ J of close inner pairs are indicated. Left: target 1168 (CVSO1490), FWHM resolution 0 . ′′
73. Right: target 1281 (CVSO 1743),FWHM resolution 0 . ′′ In the following, we restrict the statistical analysis topairs with ∆
J < ρ > . ′′ J of physical companionsin the 0 . ′′ . ′′ . ′′ . ′′ ρ < . ′′ J > . ′′
50, 1.4 mag) was undetected byGaia. This pair is measured from 2 to 4 times in eachVISTA band (12 measurements in total) with consistentparameters (rms separation scatter of 0 . ′′ ′′ < ρ < . ′′ J > . ′′ high-contrast pairs are accompanied by wider andbrighter companions at ∼ ′′ separation, also membersof the association according to the GDR2 astrometry.These young triple systems with comparable separationsmay be interesting in their own right. They are shownhere to prove that their close high-contrast inner pairsare quite obvious and hard to miss.An external test of our detections is furnished by Gaia.We selected all GDR2 companions to our targets with0 . ′′ < ρ < ′′ that conform to our astrometric criteriaof membership in Ori OB1 and cross-matched them withour list of companions derived from the VISTA Orionsurvey. No missed pairs closer than 7 ′′ were found, apartfrom the two close ones mentioned above. We concludethat the number of missed pairs (false negatives) is verysmall or zero. The joint use of two independent surveys,VISTA-Orion and GDR2, produces high-confidence re-sults.1 Fig. 13.—
Distribution of companions in the ( ρ, ∆ J ′ ) plane(top) and in the (∆ G, ∆ J ′ ) plane (bottom). Astrometrically con-firmed companions are plotted by red squares, non-members byblue crosses, and uncertain companions with p astro = 0 . M ⊙ mass. TABLE 4Meaning of the p phys flag p phys p astro N Comment0 0 397 Astrometric non-members0.1 0.5 27 Photometric non-members0.2 1 13 ρ > ′′ photometric non-members0.3 0.5/1 8 ∆ J < − . ρ > ′′ ρ < ′′ , ∆ J <
Discrimination between physical and optical pairs
The GDR2 astrometry of 78 companions with p astro =0 . p phys . For the majorityof other companions with good astrometry, p phys = p astro takes the values of either 1 or 0.Figure 13 plots parameters of the companions dividedby the p astro flag into three groups: physical, optical, anduncertain. The upper panel shows the expected behav-ior, where physical pairs concentrate at small separationsand small ∆ J , optical pairs show the opposite trend, andmost uncertain pairs are close, lacking Gaia astrometry for this reason. Note there are several wide ( ρ > ′′ ) pairsin which the secondary components are brighter than theprimary, ∆ J <
0. Some of those companions are astro-metrically confirmed members of Ori OB1, meaning thatthese CVSO targets are the secondary components tobrighter stars. Such pairs should not be considered inthe statistics. However, in pairs of comparable stars itis difficult to distinguish primary and secondary compo-nents, especially considering their variability. We includewide pairs with ∆
J > − . J ≈ . G (dashed line) is confirmed. However,the empirical slope of 0.75, chosen to match the trend,is steeper than the slope deduced from the isochrones;in other words, secondary components are slightly bluerthan predicted. This trend matches the CMD in Fig-ure 6 where most low-mass stars are located to the leftof the isochrones, i.e. have bluer colors. This suggestsa potential bias in masses and mass ratios derived fromthe isochrones.On the other hand, most optical pairs (blue crosses)are elevated above the ∆ J = 0 . G line by at least 0.6mag and satisfy the condition∆ J > . . G (5)(dotted line in the Figure); most optical companions arebluer than the astrometrically confirmed physical com-panions. This allows us to classify the pairs that lackgood astrometry. We set p phys = 0 . p phys = 0 . p phys = 0 .
1. Of the 36 pairs with ρ < ′′ lacking reliable GDR2 astrometry, 30 are physical andone optical according to the photometric criterion; fivepairs with ∆ J < p phys = 0 .
8, basedon the low probability of chance projections at close sep-arations. For a typical J = 13 mag target, we expect tofind 2.7 optical companions within 2 ′′ with ∆ J < p phys = 0 . ρ > ′′ ) opticalcompanions, and we note in Figure 13 several red squaresabove the dotted line, i.e. photometric non-members;13 such pairs are assigned p phys = 0 . p astro = 1 flag. Individual examination of their GDR2astrometry indeed shows that the parallaxes and/or PMsof both components disagree significantly, although bothsatisfy our loose astrometric membership criteria. Someremaining wide pairs with p astro = p phys = 1 may still beoptical, and we address this issue in the following statis-tical analysis.2The meaning of the p phys flag and the number of com-panions in each group are summarized in Table 4. Over-all, we consider 142 pairs with p phys > . ∼ ′′ to 30 ′′ from our targets that passthe adopted astrometric criteria, ignoring their r ̟ . The J -band magnitudes of these companions were retrievedfrom the VISTA Orion catalog by positional match (typ-ically within 0 . ′′ J <
J <
J <
List of binaries in Ori OB1
Multi-band photometry provided by the VISTA Orionsurvey can potentially help in distinguishing physical andoptical pairs. However, magnitudes of a given low-massyoung star in all VISTA nIR bands are very similar. Thetop panel of Figure 14 compares magnitude differencesof physical binaries at the shortest and longest VISTAwavelengths. The slope is barely different from one, whilethe large scatter results from a combination of variabil-ity, photometric errors, and/or circumstellar extinction.We base our statistical analysis on the magnitude differ-ences in the J band, which are very close to ∆ m in theadjacent photometric bands Y and H . For physical com-panions, the rms differences between ∆ m in the band J compared to the bands Z, Y, H, K s are 0.38, 0.28, 0.29,and 0.19 mag, respectively. Hence, in the following tablewe replace ∆ J by ∆ J ′ — the median ∆ m in the Y, J, H bands — in order to reduce random errors and the im-pact of variability and taking advantage of the fact thatthere is no systematic difference between ∆ m in thesethree bands. We use ∆ J ′ in the following in place of∆ J .Figure 14 illustrates the distributions of ∆ m in six pho-tometric bands for physical binaries by plotting mediansand quartiles of the distributions. The distributions inall nIR bands are remarkably similar, with median ∆ m around 0.8 mag. Remember that we keep only pairswith ∆ J < m . The median∆ G = 1 . Fig. 14.—
Top: comparison of magnitude differences in the Z and K s bands for 86 physical companions with ρ > ′′ ; the dottedline marks equality. Bottom: Magnitude difference vs. wavelengthfor physical binaries. Squares plot the median values in each band,the bars show the first and third quartiles. The list of 587 companions with ∆
J < J ′ (median over Y, J, H bands) and the Gaia ∆ G . Thelast columns contain the flags p astro and p phys intro-duced above. There are 142 likely physical pairs with p phys > .
5. To distinguish the 7 physical pairs where themain star is not a member of Ori OB1 (e.g. CVSO 569,see above), their p phys are listed as negative. Close binaries observed at SOAR
To probe binary frequency at smaller separations, weobserved 123 relatively bright stars in Ori OB1 selectedfrom the CVSO catalog with the speckle camera at the4.1 m Southern Astrophysical Research (SOAR) tele-scope in 2016 January. The instrument, data reduction,and results are published in Tokovinin et al. (2019). Wedetected 28 pairs, including some wide ones also foundin the VISTA Orion images (Figure 15). The closestpair has a separation of 0 . ′′
09; 12 pairs have separationsbetween 0 . ′′
15 and 0 . ′′
6, and all these detections are re-liable. Some close CVSO pairs discovered in 2016 wereconfirmed by further speckle observations in 2017–2019.The SOAR data are published and we do not duplicatethem here.Only 74 stars observed at SOAR overlap with the3
TABLE 5Companions with ∆ J < mag (fragment) N CVSO θ ρ ∆ J ∆ G p astro p phys (degr) ( ′′ ) (mag) (mag)5 427 304.3 16.730 − − Fig. 15.—
Separations ρ and magnitude differences ∆ I of CVSOpairs resolved at SOAR (note: 2 pairs with respective separationof 3 . ′′ . ′′ . ′′
15 and 0 . ′′ present CVSO-VISTA-Gaia sample, the rest are lo-cated outside the sky region studied here. We can usethe higher spatial resolution of SOAR observations todouble-check VISTA detection sensitivity at small sepa-rations, although restricted to the overlap sample. TheSOAR observations confirmed that no companions withseparations > ′′ were missed in the VISTA companionsearch, thereby affirming the assumed completeness forseparations > . ′′ J < . ′′ . ′′
2, there is only one binary resolved atSOAR that was missed by VISTA (No. 214, CVSO 2001,0 . ′′
64, ∆ I = 4 . different sample only for reference. The SOAR sam-ple is, on average, brighter than the CVSO-VISTA-Gaiasample. More massive stars are expected to have an in-creased binary frequency and, indeed, the data hint at alarger companion fraction in the SOAR sample, althoughthe difference is not statistically significant. Moreover, ahigher companion fraction among the SOAR sample isnaturally expected due to the general shape of the sepa-ration distribution that has a peak at <
100 au. BINARY STATISTICS
In this section, we study the statistics of real (physical)binaries with separations from 0 . ′′ ′′ and ∆ J <
Fig. 16.—
Relation between log q and ∆ J derived from thePARSEC isochrones for primary stars of 0.4, 0.7, and 0.9 M ⊙ massand companions more massive than 0.075 M ⊙ . Squares correspondto the linear formula log q = − . J . Fig. 17.—
Cumulative distribution of the mass ratio q for 58binaries with separations between 1 . ′′ ′′ (line and crosses).The dashed and dotted lines correspond to the field binaries (seethe text) and to the uniform distribution, respectively. Ori OB1 contains 135 physical companions, including 4triples; 131 members have at least one companion. Weignore wide secondary companions that are brighter thanour targets by more than 0.5 mag in the J band. Distribution of the mass ratio
As noted above, we do not attempt to derive massesand mass ratios. The empirical distribution of mass ra-tios is evaluated here only to access the fraction of missedcompanions at close separations and to quantify the sur-vey depth. Fortunately, this fraction and the associatedincompleteness correction are small and hence have little4influence on our results.First, we study the distribution of the magnitude dif-ference ∆ J for 57 binaries in the 1 . ′′ ′′ separation range,where the detection is complete to ∆ J < J , as can be inferred by lookingat Figure 13. Only 10/58=0.17 fraction of pairs have2 < ∆ J <
J < ∼
17% of binaries.The PMS stars and their companions in Orion OB1 arevariable. Therefore, there is no unique relation betweenmagnitude difference and mass ratio q ; for any given bi-nary, different values of q can be inferred from obser-vations at different moments or in different photometricbands. Different circumstellar extinction (so-called infra-red companions) and accretion rates can further compli-cate translation of relative photometry into q . Here weadopt the relation q ≈ − . J derived from the 4 and10 Myr isochrones (Figure 16). There is little dependenceon the age, except maybe for the youngest and lowestmass stars in Ori OB1b, which represent at most ∼ M ⊙ primary components, roughly equivalentto a spectral type K7-M0, while for 0.9 M ⊙ stars it worksless well, with an rms error of 0.06 dex in log q . For0.4 M ⊙ stars at 4 Myr the derived mass ratios could beoverestimated by 0.1 dex in log q , but the number ofbinaries in this parameter space is small; their hydrogen-burning companions have ∆ J < J appear to be distributed uniformly (Figure 17). Reas-suringly, a uniform distribution of q is known to hold forfield solar-type binaries (Raghavan et al. 2010; Tokovinin2014). El-Badry et al. (2019) studied the mass ratio dis-tribution of wide field binaries, grouping them by theprimary mass and separation. They modeled the dis-tribution by a broken power law plus some twins with q > .
95. We average the model parameters from theirTable F1 in the mass range from 0.4 to 0.8 M ⊙ and theseparation range from 600 to 2500 au, roughly matchingour survey, and adopt γ smallq = 0 . γ largeq = − .
8, and f twin = 0 .
02. The corresponding cumulative distributionis over-plotted in Figure 17 by the dashed line; it barelydiffers from the uniform distribution.Adopting a uniform distribution of q , we evaluate thefraction of missing binaries with small separations us-ing the detection limit ∆ J ( ρ ) in equation (4). This cor-rection, relevant only at separations below 1 . ′′
2, is small(factor ∼ Separation distribution and companion fraction
The distribution of separations and the companion fre-quency are determined in five logarithmic bins of 2 × width covering the separation range from 0 . ′′ . ′′ J <
J < N i are used to computethe companion frequency per decade of separation f i =( N i − N rand ) / ( N tot log N tot is the sample size.The errors of f i assume the Poisson statistics. The firstbin is corrected for undetected companions relatively to Ori OB1aOri OB1b
Fig. 18.—
Companion frequency vs. separation (fraction perdecade) in Ori OB1a (top) and Ori OB1b (bottom). The full linecorresponds to binaries with ∆
J <
J < other bins; however, this correction is minor, 1.12 and1.04 times for the two ∆ J thresholds. The expected num-ber of random pairs N rand is estimated from the densityof potential contaminants (interlopers) deduced in sec-tion 3.5 by selecting companions in the 20 ′′ -30 ′′ separa-tion range and applying the same astrometric and photo-metric filters as used for the closer companions: 226 and187 stars per square degree for ∆ J <
J < f i .The separation distribution for the full sample and forthe OB1a and OB1b groups is given in Table 6, where N i and N ′ i are the numbers of pairs with ∆ J <
J < f i are the frac-tions per decade for ∆ J < N rand of contaminantswith ∆ J < f i error. Considering thisuncertainty, the last line gives the total number of bina-ries and the companion frequency in the first four binsonly covering 1.2 dex in separation.The separation distributions in Ori OB1a and OriOB1b are plotted in Figure 18. Projected separationsare translated into au using distances of 363 and 388 pc,respectively (section 2.2); a common distance of 370 pc isused for the full sample. The distributions appear to bedifferent, especially in the second bin, where there are al-most twice as many binaries in Ori OB1b as in Ori OB1a.5 TABLE 6Multiplicity on Ori OB1
Separation Full sample Ori OB1a Ori OB1b ′′ au N i N ′ i N rand f i N i N ′ i f i N i N ′ i f i ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± However, bear in mind that the projected separation s equals the semimajor axis a only statistically and theirratio s/a varies by a factor of 2 (i.e. the bin width) bothways owing to projections and random orbital phases.Therefore, even if the distribution of semimajor axes hada sharp feature, it would be spread over adjacent bins inthe distribution of s . Modeling shows that the distribu-tion of s/a depends on the eccentricity distribution: itsmedian is 0.9 when the average eccentricity is around 0.5and 0.98 if the eccentricity distribution is linear (ther-mal), f ( e ) = 2 e (see the Appendix). The latter is ap-propriate for wide binaries considered here (Tokovinin2020). Although various correction factors on the or-der of 1 have been proposed in the literature to convert s into a , no correction is actually needed and the statisticaldistributions of log s and log a can be compared directly,provided they are smooth on a > ± ± ± . σ level. Thecorrected multiplicity fractions in the last line of Table 6differ by 4.4 ± ± ∼ J -band magnitude as a proxy for mass, in order toexplore the dependence of multiplicity on stellar mass.We split the members of Ori OB1 into three equal setsgrouped by J magnitude: brighter than J = 12 .
63, in-termediate, and fainter than J = 13 .
56, with 340 starsin each group. The numbers of non-single stars in thesegroups (61, 48, and 26, respectively, or multiplicity frac-
Fig. 19.—
Distribution of separation in the full sample. Thedash-dot line shows the result for the SOAR sample. The dia-mond is the companion frequency in the ONC from Reipurth et al.(2007). A distance of 370 pc is used. tions 0.18 ± ± ± Close binaries
Figure 19 presents the distribution of projected sepa-rations in the full CVSO-VISTA-Gaia sample, includingthe observed frequency of closer pairs derived from theSOAR observations (the wide dash-dot bar). The latteris 12 /
123 = 0 . ± .
047 in the separation range from0 . ′′
15 to 0 . ′′
6. The faintest companion in this range has amagnitude difference ∆ I = 3 . . ± . I ∼ . ′′ . ′′
7, as demonstrated, e.g., byTokovinin & Brice˜no (2020). Moreover, stars with ex-cess parallax error are also likely close binaries. The total6
TABLE 7Close binaries in Gaia
Group N tot No ̟ r ̟ > f close OB1a 658 25 65 0.137 ± ± Fig. 20.—
Spatial distribution of single stars (blue crosses), closeGaia binaries (magenta triangles), and wider VISTA Orion binaries(green squares). The size of the squares reflects binary separationranging from 0 . ′′ ′′ . The two black asterisks show the lo-cations of 25 Ori and HR 1833. The dashed circle depicts theboundary of Ori OB1b. number of these close binary candidates can be used toestimate the frequency of close binaries f close , with thecaveat that the exact range of separations and mass ra-tios of these close binaries is not defined and the numbersare not directly comparable to the frequency per decadecomputed above. The numbers are reported in Table 7.We note the increased fraction of candidate close bina-ries in Ori OB1b compared to Ori OB1a. This echoes thedifference between these groups found for wider pairs, al-though the difference between the frequency of close Gaiabinaries, 3.6 ± ± ∼
200 ausuggested by their study is certainly refuted by our data.Even in the ONC, Jerabkova et al. (2019) found a sub-stantial number of binaries with separations from 1 to 3kau. The 14 low-mass binaries in the ONC with sepa-rations from 30 to 160 au discovered by De Furio et al.(2019) match, within errors, the frequency of M-type bi-naries in the field.
Spatial distribution of binaries
Fig. 21.—
Cumulative distributions of the density of associationmembers around single stars, wide, and close binaries, computedfor 0 . ◦
25 radius.
The difference in the binary statistics between the twosubgroups of Orion OB1 is intriguing. To clarify it fur-ther, we plot in Figure 20 the spatial distribution of single(i.e. unresolved) stars, wide binaries with separations be-tween 0 . ′′ ′′ , and potential close binaries inferredfrom Gaia. Binaries with separations > ′′ are ignoredbecause several of them are likely random pairs of asso-ciation members.One notable feature of Figure 20 is the apparent ab-sence of wide binaries in the two stellar over-densities ofOri OB1a near the stars 25 Ori and HR 1833 (markedby asterisks). The latter group lacks wide binaries com-pletely. In the 25 Ori cluster, wide binaries are locatedon the periphery, at a distance of ∼ . ◦ . ◦ . ′′ ′′ , and closebinaries inferred from Gaia (763, 97, and 127 stars, re-spectively). The distributions do not differ significantly(the two-sided Kolmogorov-Smirnov test gives a proba-bility of 0.60 for wide binaries and single stars having thesame parent distribution). Therefore, we cannot claimthat the frequency of wide binaries in Ori OB1 dependson the stellar density.On the other hand, the region has dynamically evolvedand the density structures seen today most likely differfrom the actual birth configuration. Yet, stars in OriOB1a appear more clustered than those in Ori OB1b,despite being roughly twice the age. Adopting the ve-locity dispersion in a subgroup of 0.5 mas yr − , whichcorresponds to 0.8 km s − , consistent with typical ve-locity dispersion of young stellar groups and clusters inOrion OB1 (Brice˜no et al. 2007; Kounkel et al. 2018), wefind that in 7 Myr stars can move from their birthplaceby ∼ ◦ . Stars within ∼ . ◦ Fig. 22.—
Comparison of multiplicity in different populations(see text). cluster. If the 198 stars located within 0 . ◦ . ′′ ∼ DISCUSSION AND SUMMARY
In Figure 22 we put our results in the context of mul-tiplicity in other regions, with the obvious caveat relatedto the differences in the stellar mass range and complete-ness of various surveys. Assuming a uniform distribu-tion of q between 0.05 and 1, we correct our estimateof multiplicity fraction in Ori OB1 by 0.95/0.9=1.055.The band shows the 1 σ statistical errors. The canoni-cal companion frequency of 0.6 for the field solar-typedwarfs (Raghavan et al. 2010) is assumed. The multi-plicity in the Upper Scorpius OB association measuredby Tokovinin & Brice˜no (2020) is similarly corrected by0.95/0.7=1.36, considering their lower mass ratio limitof q > .
3. Their data in the mass range from 0.4 to 1.5 M ⊙ are averaged because no clear dependence of mul-tiplicity on mass was found. We plot the multiplicity inTaurus using the data from Joncour et al. (2017), with-out any correction. Data from Reipurth et al. (2007) forthe ONC are also left uncorrected because the lower limitof the mass ratio in their survey is not known.The diversity of the multiplicity statistics in young stel-lar populations, noted already by King et al. (2012) fromthe scarce data available at the time, is emerging witha stronger confidence from the modern large multiplicitysurveys, including this one. In the overall Orion OB1association the binary fraction is somewhat less than inthe field; the opposite is true in Taurus, where the ex-cess of binary fraction over the field is well documented(Duchˆene & Kraus 2013). In the separation range of 1.2dex from 222 to 3552 au that corresponds to the firstfour bins in Table 6, the sample of 142 stars in Tau-rus counts 31 companions (Joncour et al. 2017), hence f = 0 . ± .
04 (the same number can be deduced bysumming the last 4 bins of Fig. 4 in Kraus et al. 2011).In Ori OB1, the multiplicity in the same separation rangecorrected by 1.055 is 0 . ± . . ± .
04 is statistically significant. Deacon & Kraus(2020) found a significant deficit of binaries with sepa- rations 300–3000 au in open clusters compared to thefield and moving groups, confirming the critical role ofenvironment in the binary population statistics.Considering Ori OB1a and OB1b individually, we notethat OB1b shows a binary fraction comparable to thefield and is more similar to Upper Scorpius, at least forthe wide binaries that we probe in our study. Thereis accumulating evidence that Upper Scorpius, as wellas other OB associations, most likely were formed in aconfiguration similar to how they appear today — i.e.as an assembly of loose stellar groups with moderate tolow stellar density (Wright & Mamajek 2018; Lim et al.2019). If the initial stellar density of a star forming en-vironment largely dictates the formation or dynamicaldestruction of wide binaries, one might speculate that adifferent stellar birth density causes the observed differ-ence in binary fraction between Ori OB1a and Ori OB1b.In this picture Ori OB1a would have formed from a densecluster, while Ori OB1b would stem from a wide-spreadpopulation with only some sparsely clustered substruc-tures. Future kinematic studies with Gaia providinghigher precision than GDR2 will hopefully allow to fur-ther explore this issue. Regarding the origin of the fieldbinaries, a mix of Orion-type and Taurus-type binarypopulations in a suitable proportion would resemble thefield.The main results of our study are as follows: • Double stars in a sample of 1021 low-mass PMSstars of the Orion OB1 association, selected fromthe CVSO catalog, have been identified by the anal-ysis of the nIR images from the VISTA Orion mini-survey. Using Gaia astrometry and photometry, werejected unrelated pairs and arrived at a list of 135most likely real (physical) companions, arranged in127 binaries and 4 triples, with projected separa-tions from 0 . ′′ ′′ (222 – 7400 au at 370 pc) andmagnitude difference ∆ J < • The distribution of magnitude difference ∆ J ofthese binaries is compatible with a uniform massratio distribution. Our survey is almost completefor wide binaries with mass ratios above 0.13. • We found that the two sub-groups, Ori OB1a andOri OB1b, likely have a different multiplicity rate:0.078 ± ± . ′′ . ′′ • The frequency of wide binaries in Ori OB1 dependson mass (more companions around more massivestars similarly to the field) but is independent ofthe currently observed surface density of stars. Lo-cation of wide binaries on the sky suggests thatthey avoid cluster centers. • Our survey highlights the differences in multiplic-ity properties between star-forming regions. Thebinary population in the field could result from amixture of these diverse populations.Based on observations made with ESO Telescopesat the La Silla Paranal Observatory under programme8
Fig. 23.—
Left: cumulative distribution of the ratio s/a (line) and its analytic model (squares) for thermal eccentricity distribution.Right: histogram of the same data and their model (green squares and line); vertical dotted line marks the median.
ID 60.A-9285(B). We acknowledge the great work doneby the VISTA consortium who built and commissionedthe VISTA telescope and camera. This work usedbibliographic references from the Astrophysics DataSystem maintained by SAO/NASA. We used the datafrom the European Space Agency (ESA) mission Gaia( processed by theGaia Data Processing and Analysis Consortium (DPAC, ).Funding for the DPAC has been provided by national institutions, in particular the institutions participatingin the Gaia Multilateral Agreement. This research hasmade use of the VizieR catalogue access tool, CDS,Strasbourg, France (DOI: 10.26093/cds/vizier). Theoriginal description of the VizieR service was publishedin A&AS 143, 23. The work of Tokovinin and Brice˜no issupported by NOIRLab, which is managed by Associa-tion of Universities for Research in Astronomy (AURA)under cooperative agreement with the USA NationalScience Foundation.
APPENDIX
RELATION BETWEEN PROJECTED SEPARATION AND SEMIMAJOR AXIS
On Referee’s request, we include a short discussion of the statistical relation between projected separation s usedthroughout this paper and the true semimajor axis a of the binary orbit. The ratio x = s/a is computed for 10 simulated binaries with random orbit orientation and random phase. The resulting distribution of x slightly dependson the adopted eccentricity distribution; here we assume the thermal distribution f ( e ) = 2 e appropriate for widebinaries. Figure 23 (left) shows the cumulative distribution of x for simulated binaries and its analytic model F ( x ) = 0 . . π/ x/x ) α ] , (A1)where x = 0 .
98 is the median and α = 0 .
94 encodes deviation from the pure sine curve. These parameters were fittedto the simulated distribution. The analytical model (A1) is remarkably good, with rms deviation of 0.0005 and themaximum deviation of 0.005. If a uniform eccentricity distribution is assumed, the fitted parameters are x = 0 . α = 1 .
05, but the difference between distribution and its model is ∼ × larger than for thermal eccentricities.For a thermal eccentricity distribution, the median projected separation s is an accurate measure of the true mediansemimajor axis, no correction is needed. In a sample of binaries, 82.8% of projected separations differ from the truesemimajor axis by a factor less than two, and the remaining 17.2% have s < . a .Multiplicative factors slightly larger than one have been proposed in the literature to convert s into a (e.g.Raghavan et al. 2010). Different values of scaling factors are obtained depending on the assumed eccentricity dis-tribution and on the metric used to compute the factor (median, mean s/a , mean a/s , mean log a/s , etc.). Thedistribution of s/a in Figure 23 is almost symmetric, its mean is very close to the median. However, the distribution ofthe logarithm is skewed, and h log( s/a ) i = − .
073 might suggest a ≈ . s , while h a/s i = 1 .