The VVV Open Cluster Project. Near-infrared sequences of NGC6067, NGC6259, NGC4815, Pismis18, Trumpler23, and Trumpler20
K. Peña Ramírez, C. González-Fernández, A.-N. Chené, S. Ramírez Alegría
MMNRAS , 1– ?? (2020) Preprint 22 February 2021 Compiled using MNRAS L A TEX style file v3.0
The VVV Open Cluster Project. Near-infrared sequences of NGC 6067,NGC 6259, NGC 4815, Pismis 18, Trumpler 23, and Trumpler 20.
K. Peña Ramírez, ★ C. González-Fernández, A.-N. Chené, and S. Ramírez Alegría Centro de Astronomía (CITEVA), Universidad de Antofagasta, Av. Angamos 601, Antofagasta, Chile. Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK Gemini Observatory/NSF’s NOIRLab, 670 N. A‘ohoku Place, Hilo, Hawai‘i, 96720, USA
Accepted 2021 February 1. Received 2021 January 29; in original form 2020 October 13
ABSTRACT
Open clusters are central elements of our understanding of the Galactic disk evolution, as an accurate determination of theirparameters leads to an unbiased picture of our Galaxy’s structure. Extending the analysis towards fainter magnitudes in clustersequences has a significant impact on the derived fundamental parameters, such as extinction and total mass. We perform ahomogeneous analysis of six open stellar clusters in the Galactic disk using kinematic and photometric information from theGaia DR2 and VVV surveys: NGC 6067, NGC 6259, NGC 4815, Pismis 18, Trumpler 23, and Trumpler 20. We implement twocoarse-to-fine characterization methods: first, we employ Gaussian mixture models to tag fields around each open cluster in theproper motion space, and then we apply an unsupervised machine learning method to make the membership assignment to eachcluster. For the studied clusters, with ages in the ∼ ∼
45% new member candidateson average in our sample. The data-driven selection approach of cluster members makes our catalog a valuable resource fortesting stellar evolutionary models and for assessing the cluster low-to-intermediate mass populations. This study is the first of aseries intended to homogeneously reveal open cluster near-infrared sequences.
Key words: astronomical databases: miscellaneous − methods:data analysis − stars: evolution − open clusters and associations:individual: NGC 6067, NGC 6259, NGC 4815, Pismis 18, Trumpler 20, Trumpler 23. Open clusters are understood as groups of coeval stars that all have thesame chemical composition since, in principle, they formed in singleevents from a distinct molecular cloud. Besides being ideal laborato-ries for studying stellar structure and evolution by themselves, openclusters are spatially distributed throughout our Galaxy, making themexcellent tracers of the Galactic structural, dynamical and chemicalevolution. They are among the best candidates for providing preciseinformation on both the ages and chemical compositions at variousspatial positions in the disk (e.g. Dias & Lépine 2005; Piskunov et al.2006; Buckner & Froebrich 2014; Jacobson et al. 2016; Casamiquelaet al. 2017). To make significant progress, any observational endeavorin that domain need to fulfill at least two conditions: (1) accountingfor a reasonably large dataset of stellar clusters, assuring a uniformspatial distribution of stellar groups, and (2) it needs to be as homo-geneous as possible, in terms of observations and processing, and tohave a consistent method to derive cluster parameters, so to minimizesystematic uncertainties (Dias et al. 2014; Netopil et al. 2015).Large databases at optical wavelengths of open Galactic clusters inthe Solar neighborhood have been available for the last two decades(Dias et al. 2002; Kharchenko et al. 2013; Bica et al. 2019, amongothers), allowing various studies to identify new clusters and dismissfalse positives (e.g. Moitinho 2010; Camargo et al. 2016; Turner et al. ★ E-mail: [email protected] © a r X i v : . [ a s t r o - ph . GA ] F e b Peña Ramírez et al. photometry. Observations taken with VISTA are systematically re-duced at CASU as part of the VISTA Data Flow System (VDFS;Irwin et al. 2004). The latest data release uses v1.5 of the pipeline.More details about the performance and photometric properties ofthe data can be obtained in González-Fernández et al. (2018).We present the dataset and the methodology implemented to un-veil the open cluster near-infrared sequences of six open stellar clus-ters down to 𝐾 𝑠 ∼ . We selected a set of six clusters out of the list of clusters recoveredfrom the literature and revisited by Cantat-Gaudin et al. (2018) andCantat-Gaudin & Anders (2020) using an unsupervised membershipassignment code (see Section 3.2). The clusters of our sample havealso been recently studied by Jackson et al. (2020) who added opticalspectroscopic information to their procedure to determine member-ship. All these six clusters fall within the VVV footprint. To locatethe clusters, we first have used the clusters’ centers, distances, and 𝑟
50 values from the literature. The latter is the radius (in degrees)from the center of the cluster that encompasses 50% of the membersidentified by Cantat-Gaudin & Anders (2020). While it is not meantto be a physically representative description of the cluster extension,it gives an idea of the cluster’s extension and density.We have conducted our study using five times the reported 𝑟 We aim to identify each open cluster sequence with the most extensivedynamical range possible in the near-infrared. Although there is somevariation with observational conditions, the dynamic range of theVVV survey goes from approximately (11.0, 11.0, 11.5, 12.0, 11.0)in ( 𝑍 , 𝑌 , 𝐽 , 𝐻 , 𝐾 𝑠 ) where detector non-linearity brings photometricerrors over 0.1 to approximately (18.5, 18.0, 17.5, 16.5, 16.0 mag)where sky noise has the same effect for a typical observing se-quence. To extend our analysis to brighter sources, we combine VVVdata with 2MASS photometry. We used a pre-computed nearest-neighbour, proper motion aware, cross-match between 2MASS andGaia DR2 to retrieve 𝐽 , 𝐻 , and 𝐾 𝑠 magnitudes from 2MASS and the5-parameter astrometric solution from Gaia DR2 (right ascension,declination, proper motions in right ascension and declination, andparallaxes).We used the photometric and astrometric information from theVVV survey toward the fainter near-infrared magnitudes. A typicalVISTA VVV tile was observed in between 50 to 80 epochs from 2010to 2015. Since 2016, an extended area in both the Galactic bulge anddisc is currently being surveyed as part of the VVVX survey, whichshould end by the time this paper is published. Both photometric andastrometric data was extracted for each source from the improved http://casu.ast.cam.ac.uk/ Using the Q3C software: Koposov & Bartunov (2006). version of the VVV Infrared Astrometric Catalogue (VIRAC V2.0).We refer the reader to Smith et al. (2017) and Clarke et al. (2019) formore details, although we outline below the main steps we followedto build the catalogs.In terms of astrometric information, the minimum spatial VISTAcoverage unit (pawprint) was constructed by cross-matching the tele-scope pointing coordinates within a 20 arcsec matching radius. Itresulted in a sequence of images of the same on-sky region at dif-ferent epochs. A VISTA tile has 2100 pawprint sets from whichindependent proper motions are measured. Within each of the paw-prints that fulfill the selection criteria outlined by Smith et al. (2017),a pool of proper motion reference sources that do not deviate sig-nificantly from the local average is extracted in an iterative process.All proper motions within a pawprint are calculated relative to thispool and corrected for any drift in 𝑙 and 𝑏 relative to Gaia DR2.As mentioned in Clarke et al. (2019), the difference in drift veloc-ity of the reference sources between pawprint sets within a VVVtile is smaller than the proper motion error obtained from a singlepawprint set. To calculate final proper motions for stars observed inmultiple pawprints, VIRAC 2.0 uses inverse variance weighting ofthe individual pawprint measurements.The final product is an astronomical dataset in eight dimen-sions (five astrometric parameters and three photometric ones, 𝐽 , 𝐻 , and 𝐾 𝑠 ) for each cluster covering an area of five times their pub-lished r50 value. In the 2MASS/VVV magnitude overlap, sourcespresent in both catalogues are combined with an optimal inversevariance weighting, both applied to magnitudes and proper mo-tions/parallaxes. Since the magnitude range is broad, there is a signif-icant difference in precision between the sources, from the brightestto the faintest. On the bright end, the nominal uncertainty reaches theGaia precision of 0.04 mas in parallax and 0.05 mas yr − in propermotions (Lindegren et al. 2018), while at 𝐾 𝑠 ∼ . ∼ − ) in parallax and proper motions(see Figure 1). As the clusters of our sample are in the direction of the inner Galaxy,background/foreground contamination is problematic. In order tojumpstart our membership analysis, we construct a 2D histogramof proper motions in an circular area around the nominal clustercenter of 𝑟
50 degrees, and an equivalent histogram of a ring withthe same area from 𝑟
50 to √ × 𝑟
50. The difference of these twohistograms, enhance the contrast of the maximum associated withthe cluster. Carrying over poissonian statistics from each histogram,we located and fitted this density maximum in the residuals witha 2D gaussian. We then used this gaussian to assign a preliminarmembership probability based on proper motions.Once we had a handle on the proper motion distribution of thecluster, we mapped this into ( 𝛼 , 𝛿 ) space, picking out sources withvery low membership probability. We modelled the projected densityof the remaining sources with a King profile (King 1966) added to aconstant background, assuming the coordinates of the cluster centerare unknown, but with a strong prior around the nominal coordinates.From this analysis we obtained refined central coordinates and ameasure of the tidal radius ( 𝑟 𝑡 ), that we used as an initial estimate ofthe cluster extension. MNRAS , 1– ?? (2020) he VVV Open Cluster Project Table 1.
Parameters of our cluster sample. Columns 2-10 contain the spatial location, proper motion, age, distance modulus, reddening, extinction, and thenumber of members with a membership probability above or equal the 90%, as defined in the literature. The number of sources identified in this work with amembership probability ≥
90% is presented in the last column. The number of recovered sources from the literature is indicated between parenthesis. Superscriptsrefer to references listed below the Table.
Name 𝛼 𝛿 𝜇 𝛼 ∗ 𝜇 𝛿 Age (M-m) (M-m) E(B-V) A v Number Number Number[deg] a [deg] a [mas yr − ] a [mas yr − ] a [Myr] b Literature c Literature b [mag] c [mag] b p ≥ a p ≥ c p ≥ d NGC 6067 243.299 − − − ±
58 11.79 ± ± ± ± − − − ±
124 11.91 ± ± ± ± − − − ±
128 12.87 ± ± ± ± − − − ±
199 12.43 ± ± ± ± − − − ±
244 12.29 ± ± ± ± − − ±
643 13.09 ± ± ± ± a Cantat-Gaudin & Anders (2020). b Cantat-Gaudin et al. (2020), Δ 𝑙𝑜𝑔𝑡 = .
20 (NGC 6067, NGC 6259), Δ 𝑙𝑜𝑔𝑡 = .
15 for the remaining clusters. c Jackson et al. (2020). d This work.
Figure 1.
Typical uncertainties in photometry, parallax and proper motions,as a function of 𝐾 𝑠 -band magnitudes for a random sample of 100000 sourcesin one of our initial cluster datasets. One of the main challenges when one attempts to spatially character-ize a stellar conglomerates is the contrast between a cluster and thefield stars, which is most of the time shallow. And it is a challenge thatthe motions of the stars due to proper motions and parallaxes onlymake harder. Nevertheless, it should be possible to identify clustermembers even among stars with significant astrometric uncertaintiesif other criteria such as photometric selections are employed. We revisited the near-infrared cluster sequences following amethodology as presented in Cantat-Gaudin et al. (2019). For eachcluster field, we first applied the Gaussian Mixture Model (GMM;e.g. Everitt et al. 2011) technique, which is based on the assumptionthat the stars distribution within an overdensity can be described by asuperposition of multivariate gaussian distributions (de Souza et al.2017). The GMM was implemented in R (R Core Team 2018) usingthe mclust library (Scrucca et al. 2016). For each cluster, we ap-plied the GMM technique to the overdensity detected in the propermotion space, using only the stars within the 𝑟 𝑡 radius (as describedin Section 3.1) and inside 1- 𝜎 of the distance of the cluster as es-timated with Gaia. We used the parallax inverse as a proxy to thedistance, since we implicitly assume that all the stars in the clustersare roughly at the same distance. We used ten multivariate gaussiancomponents per overdensity, yet the results are insensitive to theexact number of gaussians, as long as it is large enough to disen-tangle the background from the members. Any gaussian componentwith a covariance bigger than 0.04 mas yr − in both 𝜇 𝛼 ∗ and 𝜇 𝛿 (corresponding to a standard deviation of ∼ − ) and/or amixing probability higher than 15% was rejected. As expected, wewere left only with gaussian components centered on proper motionvalues within the uncertainties of the Gaia values. The total numberof sources populating the remaining gaussians ranged from 1616 forNGC 6259 to 211 for Pismis 18.Once the GMM analysis was completed, we employed the Un-supervised Photometric Membership Assignment in Stellar ClustersUPMASK (Krone-Martins, A. & Moitinho, A. 2014) on each clusterregion. This approach has successfully been applied to Gaia DR2data (Cantat-Gaudin et al. 2018, 2019) before we applied it to ournear-infrared data. The key assumptions are:(i) the cluster members share common properties (i.e., they areclustered in the magnitudes and colors spaces).(ii) their spatial distribution is not compatible with a uniform one.The coherent motion of the conglomerates did not need to be veri-fied at this point, since the input data was already clustered in the 2Dproper motion space ( 𝜇 𝛼 ∗ , 𝜇 𝛿 ) after the fist pass using the GMM.The first step is to group stars according to their spatial and pho-tometric distribution using the 𝐽 , 𝐻 , 𝐾 𝑠 photometry and the spatialpositions. To do so, we used the k-means clustering (Forgy 1965)partition algorithm with a large k to the dataset size guaranteeing atleast 25 objects per group. The second step is to test whether the dis-tribution of stars within each group is more concentrated than whatis expected for a random fluctuation in a uniform distribution. In thisimplementation, we use the total length of a minimum spanning tree(e.g. Graham & Hell 1985) and we iterate 100 times per cluster field. MNRAS , 1– ????
Typical uncertainties in photometry, parallax and proper motions,as a function of 𝐾 𝑠 -band magnitudes for a random sample of 100000 sourcesin one of our initial cluster datasets. One of the main challenges when one attempts to spatially character-ize a stellar conglomerates is the contrast between a cluster and thefield stars, which is most of the time shallow. And it is a challenge thatthe motions of the stars due to proper motions and parallaxes onlymake harder. Nevertheless, it should be possible to identify clustermembers even among stars with significant astrometric uncertaintiesif other criteria such as photometric selections are employed. We revisited the near-infrared cluster sequences following amethodology as presented in Cantat-Gaudin et al. (2019). For eachcluster field, we first applied the Gaussian Mixture Model (GMM;e.g. Everitt et al. 2011) technique, which is based on the assumptionthat the stars distribution within an overdensity can be described by asuperposition of multivariate gaussian distributions (de Souza et al.2017). The GMM was implemented in R (R Core Team 2018) usingthe mclust library (Scrucca et al. 2016). For each cluster, we ap-plied the GMM technique to the overdensity detected in the propermotion space, using only the stars within the 𝑟 𝑡 radius (as describedin Section 3.1) and inside 1- 𝜎 of the distance of the cluster as es-timated with Gaia. We used the parallax inverse as a proxy to thedistance, since we implicitly assume that all the stars in the clustersare roughly at the same distance. We used ten multivariate gaussiancomponents per overdensity, yet the results are insensitive to theexact number of gaussians, as long as it is large enough to disen-tangle the background from the members. Any gaussian componentwith a covariance bigger than 0.04 mas yr − in both 𝜇 𝛼 ∗ and 𝜇 𝛿 (corresponding to a standard deviation of ∼ − ) and/or amixing probability higher than 15% was rejected. As expected, wewere left only with gaussian components centered on proper motionvalues within the uncertainties of the Gaia values. The total numberof sources populating the remaining gaussians ranged from 1616 forNGC 6259 to 211 for Pismis 18.Once the GMM analysis was completed, we employed the Un-supervised Photometric Membership Assignment in Stellar ClustersUPMASK (Krone-Martins, A. & Moitinho, A. 2014) on each clusterregion. This approach has successfully been applied to Gaia DR2data (Cantat-Gaudin et al. 2018, 2019) before we applied it to ournear-infrared data. The key assumptions are:(i) the cluster members share common properties (i.e., they areclustered in the magnitudes and colors spaces).(ii) their spatial distribution is not compatible with a uniform one.The coherent motion of the conglomerates did not need to be veri-fied at this point, since the input data was already clustered in the 2Dproper motion space ( 𝜇 𝛼 ∗ , 𝜇 𝛿 ) after the fist pass using the GMM.The first step is to group stars according to their spatial and pho-tometric distribution using the 𝐽 , 𝐻 , 𝐾 𝑠 photometry and the spatialpositions. To do so, we used the k-means clustering (Forgy 1965)partition algorithm with a large k to the dataset size guaranteeing atleast 25 objects per group. The second step is to test whether the dis-tribution of stars within each group is more concentrated than whatis expected for a random fluctuation in a uniform distribution. In thisimplementation, we use the total length of a minimum spanning tree(e.g. Graham & Hell 1985) and we iterate 100 times per cluster field. MNRAS , 1– ???? (2020) Peña Ramírez et al.
At each iteration, the photometry data were randomly sampled fromthe probability distribution function of each star’s positional param-eters, while taking into account uncertainties for each variable. Theclustering score for a given source is derived directly from the num-ber of iteration during which it was member of a concentrated group,and can be interpreted as a membership probability.Our clustering score is then based on the 5D ( 𝛼 , 𝛿 , 𝐽 , 𝐻 , 𝐾 𝑠 ) in-formation of each star and associated nominal uncertainties. Figure 2shows the spatial distribution, near-infrared sequence and, propermotions for each cluster. Complementary, Figure A.1 shows the clus-ters near-infrared sequences with the different membership rangescolor-coded as well with the available literature data overplotted.The full membership list per cluster is available in electronic form(see Section 7). High precision characterization of star clusters fundamental param-eters, such as age, distance, reddening, and total mass, depends onthe quality of the cluster membership determination. And the moredimensions are considered in membership determination, the better;therefore the need for datasets that are as diverse as they are ac-curate. The decontamination of background interlopers must ensurethan non-additional biases are imposed, and that whatever clustermembers remains is indeed robust and representative of the clusteritself.Using only sources with a membership probability larger or equalthan 90%, we have re-derived the median spatial position, propermotion, and distance for each cluster. We used the mean absolutedeviation (from the median), or MAD values (Feigelson & Jogesh2012), to estimate the dispersion of those values. The results aregathered in the Table 2, together with the derived 𝑟 𝑡 value. The offsetbetween our recalculated central positions and those from literaturerange from 0.01 to 1.5 arcmin, the larger value corresponding anoffset in declination calculated for NGC 6259.We obtained average proper motion values that agree within theuncertainties with the values published in the literature. The sameapplies to the distances, all between 2.2 and 3.9 kpc, yet our cen-tral values are between ∼
100 and 500 pc larger. Interestingly, thedifference in distance that we get somehow scales with the distanceitself.Our dataset allows us to study the dispertion of proper motionsand parallax for each cluster. The first quartile of our data is highlydominated by the brightest sources with the Gaia DR2 astrometry,whereas the fainter near-infrared sources, those in the third quartile,rely on our VVV astrometry. The proper motion dispersion of theclusters in our sample ranges from ∼ − . At the clusterdistances, such a dispersion translates into a 1D velocity dispersionbetween ∼ − . Those are in agreement with the inter-nal 1D velocity dispersions below the 2.0 km s − determined fromspectroscopy for Pismis 18 (Hatzidimitriou et al. 2019), Trumpler23 (Overbeek et al. 2017) and Trumpler 20 (Donati et al. 2014). Theintrinsic parallax dispersion of cluster members must correspond toa physically plausible depth (Cantat-Gaudin & Anders 2020). How-ever, the parallax distribution of members of distant clusters is dom-inated by errors which, combined with the possible contaminationof field stars, can artificially increase the observed proper motiondispersion.Using the tidal radius 𝑟 𝑡 as a proxy to clusters concentration, wefind that the clusters of our sample are remarkably spatially concen-trated. The youngest cluster ( ∼
125 Myr) seems to be gravitationally- bound, while even Trumpler 20, which is old and dynamically-evolved, is not very much physically extended (see 𝑟 𝑡 values inTable 2).Due to the levity of interstellar extinction in the near-infrared, wecan in this study map the cluster sequences better than the previ-ous work based on optical data, and we report an increase of ∼ 𝐴 𝑉 from thatsame study. The reddening vector was transformed to 𝐴 𝐾 𝑠 using 𝐴 𝐾 𝑠 / 𝐴 𝑉 = . ), assuming the total-to-selective extinction ra-tios 𝑅 𝐽 and 𝑅 𝐾 𝑠 of González-Fernández et al. (2018) for VISTAdata. Since the resulting isochrones (plotted in orange in Figure 3)are not well aligned with our observed sequences, we also presentthe isochrones using our distance determination and adjusting thereddening. The adjustment was done using only the median of the50% of the highly probable members (p > Our membership determination relies on prior information on thecentral location, expected distance range, and an initial prior ap-parent size of the studied clusters. The radius we used around theclusters NGC 6067, NGC 6259 and, NGC 4815 (defined as five timestheir 𝑟
50 value from Cantat-Gaudin et al. 2018) includes other knownclusters and cluster candidates. The area around NGC 6067 includesFSR 1716, and Harvard 10, the area around NGC 6259 overlaps withNGC 6249, and the area around NGC 4815 includes sources identi-fied as members of Gulliver 59. Our cluster sequence determinationcould successfully weed out all the contaminants and re-identify allthe already known members of our clusters. The following presentsthe comparison of our results with those in the literature.For the broad comparison of our results with the literature, wehave focused on the most recent works on the studied clusters, andspecific notes on the clusters will be discussed in Section 5. Wemust first point out that our membership probabilities have a meanin the range 66–94%, which is sightly higher than that of Cantat-Gaudin & Anders (2020, with mean probabilites of 38–67%), orJackson et al. (2020, with mean probabilites of 31–54%). With our http://stev.oapd.inaf.it/cgi-bin/cmd MNRAS , 1– ?? (2020) he VVV Open Cluster Project MNRAS , 1– ????
50 value from Cantat-Gaudin et al. 2018) includes other knownclusters and cluster candidates. The area around NGC 6067 includesFSR 1716, and Harvard 10, the area around NGC 6259 overlaps withNGC 6249, and the area around NGC 4815 includes sources identi-fied as members of Gulliver 59. Our cluster sequence determinationcould successfully weed out all the contaminants and re-identify allthe already known members of our clusters. The following presentsthe comparison of our results with those in the literature.For the broad comparison of our results with the literature, wehave focused on the most recent works on the studied clusters, andspecific notes on the clusters will be discussed in Section 5. Wemust first point out that our membership probabilities have a meanin the range 66–94%, which is sightly higher than that of Cantat-Gaudin & Anders (2020, with mean probabilites of 38–67%), orJackson et al. (2020, with mean probabilites of 31–54%). With our http://stev.oapd.inaf.it/cgi-bin/cmd MNRAS , 1– ?? (2020) he VVV Open Cluster Project MNRAS , 1– ???? (2020) Peña Ramírez et al.
Figure 2.
Left:
Spatial distribution of the cluster members.
Middle: 𝐾 𝑠 vs. 𝐽 - 𝐾 𝑠 color-magnitude diagrams of the studied open stellar clusters. The point sizescales with the distance to the cluster center. Right:
Members proper motion distribution. The uncertainties are color-coded based on their distance value. Meanproper motions in right ascension and declination are plotted with dotted lines. Only sources with membership probabilities 𝑝 ≥
90% are shown in all thediagrams. Text on the plots shows their corresponding name.
Table 2.
Center, proper motion, tidal radius, distance, reddening, extinction, and total mass values for the studied clusters.Name 𝛼 𝛿 𝜇 𝛼 ∗ 𝜇 𝛿 Age a 𝑟 𝑡 d E( 𝐽 − 𝐾 𝑠 ) 𝐴 𝐾 𝑠 M[deg] [deg] [mas yr − ] [mas yr − ] [Myr] [arcmin] [pc] [mag] [mag] [M (cid:12) ]NGC 6067 243.296 − − ± − ± ±
58 23.4 2230 ±
269 0.15 0.7 3030 ± − − ± − ± ±
124 37.8 2416 ±
299 0.35 0.2 4380 ± − − ± − ± ±
128 15.3 3791 ±
667 0.42 0.2 2230 ± − − ± − ± ±
199 12.0 2981 ±
407 0.35 0.2 6320 ± − − ± − ± ±
244 16.4 2820 ±
370 0.42 0.2 4960 ± − − ± ± ±
643 20.5 3955 ±
618 0.18 0.15 3930 ± a Adopted from Cantat-Gaudin et al. (2020), Δ 𝑙𝑜𝑔𝑡 = .
20 (NGC 6067, NGC 6259), Δ 𝑙𝑜𝑔𝑡 = .
15 for the remaining clusters. approach, we are not recovering most of the low-probable candidatesof those same studies. The Figure 4 presents the distribution of the 𝐾 𝑆 magnitude of the highest probable members (p ≤ 𝑟 𝑡 ), were excluded by the GMM procedure, or did notpass our distance restrictions. Most of them have large proper motionsand parallax uncertainties and are systematically skewed towardsfainter magnitudes in the literature. It is worth mentioning that thedistance ranges reported by Cantat-Gaudin & Anders (2020) are shallower than those of Jackson et al. (2020) within their systematicuncertainties.In NGC 6259, we identify a group of high probability members at 𝐾 ∼
11 mag that was missed by Cantat-Gaudin & Anders (2020), yetwe failed to recover most of the members they found at 𝐾 ∼ MNRAS , 1– ?? (2020) he VVV Open Cluster Project Figure 3.
Color-magnitude diagrams of the studied open clusters (absolute 𝐾 𝑠 , 𝐽 - 𝐾 𝑠 ). The PARSEC-COLIBRI isochrones are overplotted. The orangeisochrones are located following the values of distance and extinction of Cantat-Gaudin et al. (2020). The red isochrone was shifted in each case to reproducethe high-probable near-infrared sequence and the derived values are reported along with each cluster label. Figure 4. 𝐾 𝑠 magnitude distributions of the sources classified as members with probabilities 𝑝 ≥
90% in the Cantat-Gaudin & Anders (2020) and Jackson et al.(2020) studies, compared with our high probable members ( 𝑝 ≥ , 1– ????
90% in the Cantat-Gaudin & Anders (2020) and Jackson et al.(2020) studies, compared with our high probable members ( 𝑝 ≥ , 1– ???? (2020) Peña Ramírez et al.
In Figures 5 and A.2, we compare different datasets analyzed us-ing the same methodology. For efficiency and to avoid lengtheningthis section too much, we only detail here the case of Trumpler 20 –the most distant/oldest cluster of our sample – as an example, as itpresents the same general trends than those seen for all the clustersin this study. The improvements are twofold; firstly, the deep VVVphotometry provides a sharper clusters’ sequence, avoids the contam-ination of red giants, and reaches deeper magnitudes. Secondly, eventhough the near-infrared sequence shows the same general trends asthe optical one, there are differences introduced by our near-infrarednative, data-driven approach. For example, in the case of Trum-pler 20, the evolved population ( 𝐾 𝑠 (cid:46) . 𝐾 𝑠 ∼ Each cluster’s density profile was integrated to obtain the total num-ber of stars contained in the cluster’s main sequence. We determinedthe completeness magnitude for each cluster as the faintest magni-tude at which the number of sources per interval of magnitude doesnot deviate from an increasing distribution. Assuming the centralvalues of Kroupa mass function (Kroupa 2001) and the best-fittingPARSEC-COLIBRI isochrone, we extended the counting, adding allthe masses of stars down to the stellar/substellar frontier (0.08 M (cid:12) ).The evolved sources out from the main sequence were also linked tothe best-fitting position on the corresponding isochrone. The massvalues were weighted by the membership probability of each source.The six studied clusters have total masses in the range 2230 − (cid:12) , and the values are consigned in Table 2. The adopteduncertainties consider the density profiles with and without weightingthe mass values by each source’s membership probability. When weconsider those uncertainties, the masses extreme values become 2170 − (cid:12) . In the case of NGC 6259 the secondary group of sourcesat ( 𝐽 − 𝐾 𝑠 ) ∼ . 𝐾 𝑠 ∼
11 mag (see Section 5) was isolated forits total mass determination. The total mass cluster values availablein the literature show a significant dispersion. For NGC 6067, theextreme values goes from 893 to ∼ (cid:12) (Piskunov et al. 2008;Alonso-Santiago et al. 2017), while we report an intermediate value.In the case of Trumpler 20, we report a total mass deviating at a3 𝜎 level from lowest mass of the values reported in the literature:6700 ±
800 M (cid:12) (Donati et al. 2014) and 15000 M (cid:12) (Günes , et al.2017). In the case of NGC 4815, we present a total mass value thatduplicates more consolidated values based on optical photometrydata (Chen et al. 1998; Prisinzano et al. 2001). Finally, (Bonatto & Bica 2007) reported a total mass value of 3100 ± (cid:12) forTrumpler 23 which is at 1.4 𝜎 from the total mass value presentedhere. To the best of our knowledge, there are no previous total massestimates for the clusters NGC 6259 and Pismis 18. This cluster has a diverse population composed of B-type dwarfs andgiants, AF-type supergiants, and two Cepheids. Our catalog does notreport the G1 Iab Cepheid QZ Nor as member, since even though GaiaDR2 provides radial velocity and parallax that are consistent with itsmembership to the cluster, there is no proper motion informationavailable. Also, the metallicity discrepancy reported for that star by(Alonso-Santiago et al. 2017) remains unsolved.Regardless of its proximity, the planetary nebula Pn HeFa 1 isnot associated to the cluster (Moni Bidin et al. 2014), although thehigh-mass planetary nebulae BMP J1613-5406 is (Fragkou et al.2019). Using this object Fragkou et al. (2019) sets the cluster’s mainsequence turn-off point at the B6V-types, agreeing with Alonso-Santiago et al. 2017 and Mermilliod 1981). From our analysis, welocated it precisely at ∼ (cid:12) . There has been numerous attempts to characterize this cluster atoptical wavelengths (Hawarden 1974; Anthony-Twarog et al. 1989;Mermilliod et al. 2001; Clariá et al. 2008). It was also includedin the works of Magrini et al. (2018b,a) who were assessing itsabundance pattern of s-process elements and its N/O abundanceratio. Assuming an age of 0.2 Gyr, Lagarde et al. (2019) derived aturn-off mass of 3.7 M (cid:12) . From our analysis, we relocate the turn-offmass at ∼ (cid:12) . The cluster population has been studied looking forbinary systems with negative results Mermilliod et al. (2008), while(Ciechanowska et al. 2006) claims eclipsing variables are associatedto the cluster based on their color and magnitude. From our analysis,the source 2MASSJ17004708 − ( 𝐽 − 𝐾 𝑠 ) ∼ . 𝐾 𝑠 ∼
11 mag, that share common properties with the most probablecluster members. This group appears isolated in the CMD and is toofaint to fit the isochrone. Given their proximity to NGC 6259 thesesources can not be members of NGC 6249, although their spatialdistribution in right ascension is asymmetric. Their radial velocitiesobtained from the literature are in complete agreement with the clus-ter’s mean radial velocity. To assess the membership status of thisgroup of sources we have repeated the same membership determina-tion algorithms discussed in this work to the sources lying in an ringaround the cluster and beyond 𝑟 𝑡 (up to the equivalent in area). Wedid not recover a single source in the color-magnitude region wherethis group of sources linked to NGC 6259 lies. We also attemptedvarious isochrone fittings exploring the possibility of a differencein age but none of the values explored can link the faint group ofsources to the cluster main sequence. We can conclude that it is un-likely that these sources at ( 𝐽 − 𝐾 𝑠 ) , 𝐾 𝑠 ∼ MNRAS , 1– ?? (2020) he VVV Open Cluster Project Figure 5.
Comparison of 𝐾 𝑠 vs. 𝐽 − 𝐾 𝑠 color-magnitude diagrams for Trumpler 20. Left:
Our VVV data following the procedure outlined in Section 3.2.
Middle:
Members from Cantat-Gaudin & Anders (2020) correlated with 2MASS data.
Right : Members from Cantat-Gaudin & Anders (2020) directly correlated withour VVV photometry. The horizontal dashed line shows the depth of the cluster sequences at 𝐾 𝑠 =
15 mag. Only members with probabilities 𝑝 ≥
90% areshown.
This open cluster has been extensively studied in terms of its C, N,O, Na, Al abundances and abundance ratios of elements belongingto different nucleosynthetic channels (Tautvaišien˙e et al. 2015; Frielet al. 2014; Magrini et al. 2014) mostly based on Gaia-ESO data.Its spatial distribution was explored in the optical by Chen et al.(1998) and based on their members selection the authors report ahigher concentration of the cluster’s brighter stars. With our accessto fainter magnitudes, the effect seems less evident, especially giventhat our exploration reaches the cluster tidal radius. The luminosityfunction and stellar mass function for NGC 4815 have been studiedby Prisinzano et al. (2001); Sagar et al. (2001), finding it consistentwith a Salpeter slope.
The cluster has been recently revisited using the optical spectroscopicGaia-ESO survey by Hatzidimitriou et al. (2019). Their observationalsample is based on the membership analysis of Cantat-Gaudin et al.(2018). They illustrated the importance of using proper motions whenassigning cluster membership. Indeed, there are several objects withradial velocities consistent with their membership to the cluster, butwhich are scattered over large distances from the cluster’s center andhave inconsistent proper motion values. The authors found a meanparallax different from literature, but the same survey team solvedthat issue in Jackson et al. (2020).
The study by Bonatto & Bica (2007), based on 2MASS photom-etry only, shows the analysis of this cluster to be challenging dueto complications with crowding, differential reddening, and possible tidal effects. Later on, Overbeek et al. (2017) used Gaia-ESO datato isolate 70 sources members, based on their radial velocity. Theircluster parametrization agrees within the uncertainties with the val-ues presented in our work, although the authors locate the cluster ata closer distance. Overbeek et al. (2017) found an apparent spreadin the lower main sequence, which they explain with a broadeningfrom both differential reddening and binaries, as well as the possiblepresence of a blue and a red main sequences. Our study does notconfirm a double sequence, although there is an evident spread in thesequence (see Figure 3). From our membership assessment, there aretwo faint-red sources identified as probable members in Trumpler 23that can be potential interlopers given their high parallax errors.
This an old open cluster located in the inner disc (beyond the greatCarina spiral arm) that has been targeted by the Gaia-ESO Survey(Donati et al. 2014), which sample was only limited to the clus-ter’s core. As shown in Figure A.1, Trumpler 20’s sequence is broad,especially at faint magnitudes, and seems effected by differential red-dening, which is consistent with its position in the Galactic disc andits reddening (Donati et al. 2014; Platais et al. 2012). The source ofthe differential reddening could mainly be a patchy dust structuresin the field of view (see Figure 2), along the line of sight. Usingisochrone fitting of the turn-off, Carraro et al. (2014) derived clusterparameters that are in agreement with the most recent literature dataand our derivations.
This study unveils the near-infrared sequence of six open clusterslocated up to ∼ MNRAS , 1– ????
This study unveils the near-infrared sequence of six open clusterslocated up to ∼ MNRAS , 1– ???? (2020) Peña Ramírez et al. lored to disentangle the cluster population from the field stars. Thecluster members’ search is centered in two main steps: a density-aware exploration of the data in the positional and astrometric spaceand a cluster membership assignment based on positional and pho-tometric similarity.Our methodology has allowed us to disentangle the cluster popu-lations from their dense backgrounds. Our study increased by ∼ ACKNOWLEDGEMENTS
We are grateful to the referee, Giovanni Carraro, for help-ful comments which significantly helped improve the paper.This work was supported by MINEDUC-UA project codeANT1855, CONICYT − ANID FONDECYT Regular 1201490,CONICYT − ANID FONDECYT Iniciación 11201161, 1171025,and CONICYT PAI “Concurso Nacional Inserción de Capital Hu-mano Avanzado en la Academia 2017” project PAI79170089. Thispaper made use of the Whole Sky Database (wsdb) created bySergey Koposov and maintained at the Institute of Astronomy, Cam-bridge by Sergey Koposov, Vasily Belokurov and Wyn Evans withfinancial support from the Science & Technology Facilities Coun-cil (STFC) and the European Research Council (ERC). This workwas supported by the international Gemini Observatory, a programof NSF’s NOIRLab, which is managed by the Association of Uni-versities for Research in Astronomy (AURA) under a cooperativeagreement with the National Science Foundation, on behalf of theGemini partnership of Argentina, Brazil, Canada, Chile, the Re-public of Korea, and the United States of America. This work hasmade use of data from the European Space Agency (ESA) mis-sion
Gaia ( ), processed bythe Gaia
Data Processing and Analysis Consortium (DPAC, ). Fund-ing for the DPAC has been provided by national institutions, in partic-ular the institutions participating in the
Gaia
Multilateral Agreement.
The data underlying this article are available in the Open SoftwareFoundation, at https://dx.doi.org/10.17605/OSF.IO/359QD . REFERENCES
Alonso-Santiago J., Negueruela I., Marco A., Tabernero H. M., González-Fernández C., Castro N., 2017, MNRAS, 469, 1330Anthony-Twarog B. J., Payne D. M., Twarog B. A., 1989, AJ, 97, 1048Bastian N., Kamann S., Cabrera-Ziri I., Georgy C., Ekström S., CharbonnelC., de Juan Ovelar M., Usher C., 2018, MNRAS, 480, 3739Bica E., Bonatto C., 2005, A&A, 443, 465Bica E., Pavani D. B., Bonatto C. J., Lima E. F., 2019, AJ, 157, 12Bisht D., Elsanhoury W. H., Zhu Q., Sariya D. P., Yadav R. K. S., RangwalG., Durgapal A., Jiang I.-G., 2020a, arXiv e-prints, p. arXiv:2006.13618Bisht D., Zhu Q., Yadav R. K. S., Durgapal A., Rangwal G., 2020b, MNRAS,494, 607Bonatto C., Bica E., 2007, MNRAS, 377, 1301Bonatto C., Bica E., Ortolani S., Barbuy B., 2006, A&A, 453, 121Buckner A. S. M., Froebrich D., 2014, MNRAS, 444, 290Camargo D., Bica E., Bonatto C., 2016, A&A, 593, A95Cantat-Gaudin T., Anders F., 2020, A&A, 633, A99Cantat-Gaudin T., et al., 2018, A&A, 618, A93Cantat-Gaudin T., et al., 2019, A&A, 624, A126Cantat-Gaudin T., et al., 2020, A&A, 640, A1Carraro G., Villanova S., Monaco L., Beccari G., Ahumada J. A., Boffin H.M. J., 2014, A&A, 562, A39Casamiquela L., et al., 2017, MNRAS, 470, 4363Castro-Ginard A., Jordi C., Luri X., Julbe F., Morvan M., Balaguer-NúñezL., Cantat-Gaudin T., 2018, A&A, 618, A59Castro-Ginard A., Jordi C., Luri X., Cantat-Gaudin T., Balaguer-Núñez L.,2019, A&A, 627, A35Castro-Ginard A., et al., 2020, A&A, 635, A45Chen B., Carraro G., Torra J., Jordi C., 1998, A&A, 331, 916Ciechanowska A., Pietrzynski G., Wyrzykowski L., Szewczyk O., Kubiak M.,Udalski A., Szymanski M. K., 2006, Acta Astron., 56, 219Clariá J. J., Piatti A. E., Mermilliod J. C., Palma T., 2008, AstronomischeNachrichten, 329, 609Clarke J. P., Wegg C., Gerhard O., Smith L. C., Lucas P. W., Wylie S. M.,2019, MNRAS, 489, 3519Dias W. S., Lépine J. R. D., 2005, ApJ, 629, 825Dias W. S., Alessi B. S., Moitinho A., Lépine J. R. D., 2002, A&A, 389, 871Dias W. S., Monteiro H., Caetano T. C., Lépine J. R. D., Assafin M., OliveiraA. F., 2014, A&A, 564, A79Donati P., et al., 2014, A&A, 561, A94Emerson J., McPherson A., Sutherland W., 2006, The Messenger, 126, 41Everitt B. S., Landau S., Leese M., Stahl D., 2011, Cluster Analysis, 5 edn.John Wiley & SonsFeigelson E. D., Jogesh B. G., 2012, Modern Statistical Methods for Astron-omy. Cambridge University PressForgy E. W., 1965, Biometrics, 21, 768Fragkou V., Parker Q. A., Zijlstra A. A., Crause L., Barker H., 2019, NatureAstronomy, 3, 851Friel E. D., et al., 2014, A&A, 563, A117Gaia Collaboration et al., 2016, A&A, 595, A2Gaia Collaboration et al., 2018, A&A, 616, A1Gaia Collaboration Brown A. G. A., Vallenari A., Prusti T., de Bruijne J. H. J.,Babusiaux C., Biermann M., 2020, arXiv e-prints, p. arXiv:2012.01533Galli P. A. B., et al., 2020, A&A, 643, A148González-Fernández C., et al., 2018, MNRAS, 474, 5459Graham R. L., Hell P., 1985, Annals of the History of Computing, 7, 43Günes , O., Karatas , Y., Bonatto C., 2017, Astronomische Nachrichten, 338,464Hatzidimitriou D., et al., 2019, A&A, 626, A90Hawarden T. G., 1974, MNRAS, 169, 539Hoppe R., Bergemann M., Bitsch B., Serenelli A., 2020, A&A, 641, A73Irwin M. J., et al., 2004, in Quinn P. J., Bridger A., eds, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series Vol. 5493,Optimizing Scientific Return for Astronomy through Information Tech-nologies. pp 411–422, doi:10.1117/12.551449Jackson R. J., et al., 2020, MNRAS, 496, 4701Jacobson H. R., et al., 2016, A&A, 591, A37MNRAS , 1– ?? (2020) he VVV Open Cluster Project Kharchenko N. V., Piskunov A. E., Schilbach E., Röser S., Scholz R. D.,2013, A&A, 558, A53Kharchenko N. V., Piskunov A. E., Schilbach E., Röser S., Scholz R. D.,2016, A&A, 585, A101King I. R., 1966, AJ, 71, 64Koposov S., Bartunov O., 2006, in Gabriel C., Arviset C., Ponz D., EnriqueS., eds, Astronomical Society of the Pacific Conference Series Vol. 351,Astronomical Data Analysis Software and Systems XV. p. 735Krone-Martins, A. Moitinho, A. 2014, A&A, 561, A57Kroupa P., 2001, MNRAS, 322, 231Lagarde N., et al., 2019, A&A, 621, A24Lindegren L., et al., 2018, A&A, 616, A2Magrini L., et al., 2014, A&A, 563, A44Magrini L., et al., 2018a, A&A, 617, A106Magrini L., et al., 2018b, A&A, 618, A102Marigo P., et al., 2017, ApJ, 835, 77Mermilliod J. C., 1981, A&A, 97, 235Mermilliod J. C., Clariá J. J., Andersen J., Piatti A. E., Mayor M., 2001,A&A, 375, 30Mermilliod J. C., Mayor M., Udry S., 2008, A&A, 485, 303Minniti D., et al., 2010, New Astron., 15, 433Miret-Roig N., et al., 2019, A&A, 631, A57Moitinho A., 2010, in de Grijs R., Lépine J. R. D., eds, IAU Sym-posium Vol. 266, Star Clusters: Basic Galactic Building BlocksThroughout Time and Space. pp 106–116 ( arXiv:0911.1459 ),doi:10.1017/S1743921309990949Moni Bidin C., et al., 2014, A&A, 561, A119Netopil M., Paunzen E., Carraro G., 2015, A&A, 582, A19Olivares J., et al., 2019, A&A, 625, A115Overbeek J. C., et al., 2017, A&A, 598, A68Piatti A. E., 2017, MNRAS, 466, 4960Piskunov A. E., Kharchenko N. V., 2018, in Shustov B. M., WiebeD. S., eds, Stars and Satellites, Proceedings of the Memorial Con-ference Devoted to A.G. Masevich 100th Anniversary. pp 216–221,doi:10.26087/INASAN.2018.2.2.034Piskunov A. E., Kharchenko N. V., Röser S., Schilbach E., Scholz R. D.,2006, A&A, 445, 545Piskunov A. E., Schilbach E., Kharchenko N. V., Röser S., Scholz R. D.,2008, A&A, 477, 165Platais I., et al., 2012, ApJ, 751, L8Prisinzano L., Carraro G., Piotto G., Seleznev A. F., Stetson P. B., Saviane I.,2001, A&A, 369, 851R Core Team 2018, R: A Language and Environment for Statistical Comput-ing. R Foundation for Statistical Computing, Vienna, Austria,
Rangwal G., Yadav R. K. S., Durgapal A., Bisht D., Nardiello D., 2019,MNRAS, 490, 1383Sagar R., Munari U., de Boer K. S., 2001, MNRAS, 327, 23Santos-Silva T., Gregorio-Hetem J., 2012, A&A, 547, A107Scrucca L., Fop M., Murphy T. B., Raftery A. E., 2016, The R Journal, 8,205Skrutskie M. F., et al., 2006, AJ, 131, 1163Smith L. C., Lucas P. W., Kurtev A., 2017, MNRAS, 842, 65Tautvaišien˙e G., et al., 2015, A&A, 573, A55Turner D. G., Carraro G., Panko E. A., 2017, MNRAS, 470, 481de Juan Ovelar M., et al., 2020, MNRAS, 491, 2129de Souza R. S., et al., 2017, MNRAS, 472, 2808
APPENDIX A: DETAILED COLOR-MAGNITUDEDIAGRAMS.
MNRAS , 1– ????
MNRAS , 1– ???? (2020) Peña Ramírez et al.
Figure A.1.
Detailed 𝐾 vs. 𝐽 - 𝐾 color-magnitude diagrams of the studied open stellar clusters. Cluster members with membership probabilities 𝑝 ≥
90% arerepresented as red dots. Sources with membership probabilities between 0 . > p ≥ . < . ≥ .
9) from Cantat-Gaudin & Anders (2020) and Jackson et al. (2020) are represented as open circles and openboxes, respectively. Field contaminants are presented as light gray dots. The two faint-red sources identified as probable members in Trumpler 23 are not shownin previous figures. Text on the plots shows their corresponding name.MNRAS , 1– ?? (2020) he VVV Open Cluster Project Figure A.2.
Comparison of 𝐾 𝑠 vs. 𝐽 − 𝐾 𝑠 color-magnitude diagrams for NGC 6067, NGC 6259, NGC 4815, Pismis 18, Trumpler 23. Left:
Our VVV datafollowing the procedure outlined in Section 3.2.
Middle:
Members from Cantat-Gaudin & Anders (2020) correlated with 2MASS data.
Right : Membersfrom Cantat-Gaudin & Anders (2020) directly correlated with our VVV photometry. The horizontal dashed line shows the depth of the cluster sequences at 𝐾 𝑠 =
15 mag. Only members with probabilities 𝑝 ≥
90% are shown. MNRAS , 1– ????