Berkeley 51, a young open cluster with four yellow supergiants
Ignacio Negueruela, Maria Monguió, Amparo Marco, Hugo M. Tabernero, Carlos González-Fernández, Ricardo Dorda
aa r X i v : . [ a s t r o - ph . S R ] M a r MNRAS in press, 1–19 (2018) Preprint 22 March 2018 Compiled using MNRAS L A TEX style file v3.0
Berkeley 51, a young open cluster with four yellowsupergiants
I. Negueruela, ⋆ M. Mongui´o, , A. Marco, H. M. Tabernero, C. Gonz´alez-Fern´andez, and R. Dorda Departamento de F´ısica, Ingenier´ıa de Sistemas y Teor´ıa de la Se˜nal, Universidad de Alicante, Carretera San Vicente del Raspeig s/n,E03690, San Vicente del Raspeig, Spain School of Physics, Astronomy & Mathematics, University of Hertfordshire, College Lane, Hatfield, Hertfordshire AL10 9AB, UK Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, UK
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
The heavily obscured open cluster Berkeley 51 shows characteristics typical of youngmassive clusters, even though the few previous studies have suggested older ages.We combine optical (
UBV ) and 2MASS photometry of the cluster field with multi-object and long-slit optical spectroscopy for a large sample of stars. We apply classicalphotometric analysis techniques to determine the reddening to the cluster, and thenderive cluster parameters via isochrone fitting. We find a large population of B-typestars, with a main sequence turn-off at B3 V, as well as a large number of supergiantswith spectral types ranging from F to M. We use intermediate resolution spectra ofthe evolved cool stars to derive their stellar parameters and find an essentially solariron abundance. Under the plausible assumption that our photometry reaches starsstill close to the ZAMS, the cluster is located at d ≈ . kpc and has an age of ∼ Ma,though a slightly younger and more distant cluster cannot be ruled out. Despite theapparent good fit of isochrones, evolved stars seem to reside in positions of the CMDfar away from the locations where stellar tracks predict Helium burning to occur. Ofparticular interest is the presence of four yellow supergiants, two on the ascendingbranch and two others close to or inside the instability strip.
Key words: stars: evolution – supergiants – open clusters and associations: individ-ual: Berkeley 51
Evolved stars in open clusters represent the best test bedsfor theoretical evolutionary tracks. After the end of Hydro-gen burning in their cores, stars evolve towards lower ef-fective temperatures, T eff , and become, according to theirmasses, red giants (RGs) or supergiants (RSGs). For a lim-ited range of masses, loops in the HR diagram are ex-pected to bring the stars back to the yellow supergiantregion, where they can behave as classical Cepheids (e.g.Chiosi et al. 1992). As an example, in the most recentGeneva tracks (Ekstr¨om et al. 2012), stars of solar composi-tion with masses between 5 and M ⊙ experience these loopsboth for zero initial rotation and moderately-high initialrotation, while older isochrones showed this behaviour athigher masses (Schaller et al. 1992). The exact mass range ⋆ E-mail: [email protected] for which these loops happen depends on the physics of thestellar interior, generally modelled via poorly understood pa-rameters (e.g. Chiosi et al. 1992; Mowlavi & Forestini 1994;Salasnich et al. 1999; Meynet & Maeder 2000). In particu-lar, the extent of semi-convection and overshooting, whichare not well constrained, have crucial consequences on is-sues of fundamental importance in our understanding ofGalactic chemical evolution, such as the ratio of initial massto white dwarf mass (e.g. Jeffries 1997; Weidemann 2000)or the boundary between stars that leave white dwarfs asremnants and those that explode as supernovae (SNe; e.g.Poelarends et al. 2008).Unfortunately, due to the rarity of high-mass stars andthe short duration of the post-H-core-burning phase, mostyoung open clusters are only moderately useful as test bedsbecause of low number statistics (e.g. Ekstr¨om et al. 2013).For ages above 100 Ma, on the other hand, the number ofRGs increases for a given cluster mass, meaning that severalclusters are known to sport large populations of RGs (see © I. Negueruela et al.
Mermilliod et al. 2008). Finding young open clusters withlarge populations of evolved stars provides the laboratoriesthat can help constrain the inputs of models and hence ourunderstanding of stellar evolution (Negueruela 2016).As part of such an endeavour, we have been searchingthrough the databases of poorly-studied known open clustersto identify good candidates to massive young open clusters.Recent examples include the starburst cluster vandenBergh-Hagen 222, with a population of 13 yellow or red supergiantsat an age ∼ – 20 Ma (Marco et al. 2014), or the ∼
50 Maopen cluster Berkeley 55, with 6 – 7 supergiants or brightgiants (Negueruela & Marco 2012). Here we report on theidentification of another faint, Northern sky cluster as ayoung, massive cluster with a large population of evolvedstars.Berkeley 51 (Be 51) is a faint, compact clusterin the constellation Cygnus. The WEBDA database (Netopil et al. 2012) provides coordinates RA: 20h 11m54s, Dec: + ◦ ′ ′′ ( ℓ = . ◦ , b = + . ◦ ). Two pre-vious estimates by Tadross (2008) and Kharchenko et al.(2013), based on existing photometric catalogues, agreeon considering Be 51 an intermediate-age, distant cluster( τ = Ma, d = . kpc in Tadross 2008; τ = Ma, d = . kpc in Kharchenko et al. 2013, who estimate a red-dening A V = . mag). In contrast, a BV I
CCD study bySubramaniam et al. (2010) concluded that it was an old( τ = Gyr) cluster at only d = . kpc, despite its highreddening of E ( B − V ) = . . In this paper, we present amuch more complete study of Be 51, including comprehen-sive spectroscopy, that reveals it as a distant, moderately-massive young open cluster containing a large population ofevolved stars. UBV photometry of Be 51 was obtained in service mode us-ing ALFOSC on the Nordic Optical Telescope at the Roquede los Muchachos Observatory (La Palma, Spain) on thenight of 2008 September 19. In imaging mode, the cameracovers a field of . ′ × . ′ and has a pixel scale of . ′′ /pixel.Two standard fields from the list of Landolt (1992),MARK A and PG 2213 − . Aperture photometry using the PHOT packageinside DAOPHOT (IRAF, DAOPHOT) was developed onthese fields with the same aperture, 21 pixels, for each filter.Images of Be 51 were taken in two series of different At http://webda.physics.muni.cz IRAF is distributed by the National Optical Astronomy Obser-vatories, which are operated by the Association of Universities forResearch in Astronomy, Inc., under cooperative agreement withthe National Science Foundation
Table 1.
Log of the photometric observations taken at the NOTon June 2010 for Berkeley 51. There are two observations for eachexposure time. Exposure times (s)Filter Long times Short times U
900 250 B
200 60 V
40 10 exposure times to obtain accurate photometry for a magni-tude range. The log of observations is presented in Table 1.Photometry was obtained by point-spread function (PSF)fitting using the DAOPHOT package (Stetson 1987) pro-vided by IRAF. The apertures used are of the order of theFWHM, 5 pixels for all images in the U and B filters and 4pixels for the V filter images. We selected ≈ PSF stars ineach frame, from which we determined an initial PSF, whichwe allowed to be variable (in order 2) across the frame. Wethen performed aperture correction to obtain instrumentalmagnitudes for all stars. Using the standard stars and themedian extinction coefficients for the observatory, we car-ried out the transformation of the instrumental magnitudesto the standard system by means of the PHOTCAL packageinside IRAF.The number of stars that we could detect in all filters islimited by the long exposure time in the U filter. We iden-tify all stars with good photometry in all three filters on theimage in Figures 1 and 2. In Table B1, we list coordinates(obtained by a cross-match with 2MASS), their UBV pho-tometry, and their 2MASS
JHK S photometry, when avail-able. The values of V , ( B − V ) and ( U − B ) are given togetherwith the number of measurements and an error, which is thestandard deviation of all the measurements whenever severalmeasurements exist and the photometric error otherwise.The designation of each star is given by the number indi-cated on the images (Figures 1 and 2). We have three-bandphotometry for ∼
250 stars in the field, but in the analysiswe will only use 173 stars with photometric errors such thatthe error in their reddening-free Q parameter (see Sect. 3.3)is < . mag (roughly corresponding to an uncertainty ofone spectral type in photometric classification). We obtained
JHK S photometry from the 2MASS catalogue(Skrutskie et al. 2006). The completeness limit of this cata-logue is set at K S = . . We selected only stars with “good”quality flags in 2MASS (A or E), and photometric errors < . mag in all bands. This leaves out many stars close tothe centre of the cluster, where confusion becomes importantat the spatial resolution of 2MASS.We used the 2MASS data to carry out a preliminary se-lection of spectroscopic targets for an exploratory survey. Wetook a circle of radius ′ around the nominal cluster centreand built the corresponding K S /( J − K S ) diagram (see Fig. 3).We cleaned the diagram by making use of the reddening-free Q IR index, defined as Q IR = ( J − H ) − . × ( H − K S ) . Early-type (OBA) stars are easily separated, as they display Q IR ≈ . (cf. Comer´on & Pasquali 2005; Negueruela & Schurch2007). We selected stars with − . ≤ Q IR ≤ . (shownas large circles in Fig. 3). This range is intended to account MNRAS in press, 1–19 (2018) he young cluster Be 51 RA D E C Figure 1.
Finding chart for stars with photometry in the field ofBerkeley 51. The image is one of our deep U -band frames. Starsinside the rectangle (which approximately defines the cluster core)are marked in Fig. 2. Each star is identified by the nearest markerin the same colour as the circle around it (colours are assignedsimply for visibility). The size of the image is the full FoV ofALFOSC. for the typical errors in 2MASS (generally . − . mag ina given colour for stars with K S = − ) and also includeemission-line stars, which typically have Q IR < ∼ − . (e.g.Negueruela & Schurch 2007).In addition, we selected bright stars with K S ≤ . and . ≤ Q IR ≤ . , the range where Galactic red lumi-nous stars are generally found (see Negueruela et al. 2012;Gonz´alez-Fern´andez et al. 2015, for a discussion of these cri-teria). These objects are displayed as big squares in Fig. 3.We then aplied the same criteria to a circle of radius ′ around the nominal cluster centre (displayed as small sym-bols in Fig. 3). Comparison of both datasets shows a clearoverdensity of bright stars in the central cluster area (al-most half the bright stars in the large circle are inside thesmall circle, which has less than one fifth of the area). Inaddition, there is a strong overdensity of early type starswith K S = − in the central area, with most of thempresenting ( J − K S ) ≈ . . We interpreted this overdensityas the cluster main sequence, and selected targets for thespectroscopy runs among these objects. Spectroscopy of the brightest candidate members of Be 51was obtained with the red arm of the ISIS double-beam spec-trograph, mounted on the 4.2-m William Herschel Telescope(WHT) in La Palma (Spain) in three separate runs. An ini-tial survey was conducted with the R600R grating and the
Red+
CCD, a configuration that covers the 7600 – 9000 ˚Arange in the unvignetted section of the CCD with a nominaldispersion of 0.5 ˚A/pixel. Data were taken during a service night on 2007, July 26 and then completed during a run on2007, August 21. In July, the CCD was unbinned, and a . ′′ slit was used. In August, the CCD was binned by a factor2 in the spectral direction, and a . ′′ slit was used. For thisgrating and all slit widths > . ′′ , the spectral resolution isoversampled, and the resolution element is expected to be ∼ ≈ . ˚A for both con-figurations. The resolving power of our spectra is therefore R ∼ .The supergiants identified during this survey were thenre-observed at higher resolution, using the R1200R gratingin June 2012 and the unbinned Red+
CCD. The nominaldispersion is 0.26 ˚A/pixel. We used a . ′′ slit that providesa resolving power of R ∼
12 000 . Unlike in the previous runs,each spectrum was taken together with an arc exposure atthe same sky position, for accurate wavelength calibration. Alog of all the WHT/ISIS observations is presented in Table 2.The average number of counts per pixel in the continuumaround the Ca ii triplet is given to estimate the signal-to-noise (S/N) ratio.Candidate blue stars in Be 51, selected from our UBV photometry, were observed on the night of 2014 August 24thwith the Optical System for Imaging and low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS) instrument,mounted on the 10.4-m Gran Telescopio Canarias (GTC) inLa Palma (Spain). The instrument operated in the MOSmode, with . ′′ slitlets traced on a plate. The R2000Bgrism covers the nominal range 3950 – 5700 ˚A, but the ac-tual spectral coverage for a given object is a strong functionof its position on the plate. The nominal dispersion of theR2000B grism is 0.9 ˚A per binned pixel (the standard Mar-coni CCD42-82 mosaic is used in × binned mode). Theresolving power, measured on arc lamp spectra, is R ≈ .Only one plate was observed, as the effective field of . ′ × . ′ is much larger than the cluster size. Three 1895 sexposures were obtained. We reduced these data using the Starlink (Currie et al. 2014) software packages CCDPACK(Draper et al. 2011) and FIGARO (Shortridge et al. 2014)by following standard procedures. The spectra have beennormalized to the continuum using DIPSO (Howarth et al.2014).Finally, we observed one star ( R =
85 000 , and a spectral coverage from 377to 900 nm, though some small gaps exist beyond 800 nm(Raskin & Van Winckel 2014). Data were homogeneouslyreduced using version 4.0 of the HermesDRS automateddata reduction pipeline, which provides order merged spec-tra. In this case, the target is very faint and there are essen-tially no counts shortwards of ∼ ˚A. I. Negueruela et al. RA D E C Figure 2.
Finding chart for stars with photometry in the central part of Berkeley 51. Labelling as in Fig. 1.
Figure 4 shows the spectra of five bright red stars in thefield preselected as possible evolved members. Classifica-tion criteria for spectra at this resolution are discussed inNegueruela et al. (2012). The brightest star in the infrared, K S magnitudes and bluer colours. Com-parison to MK standards observed at similar resolution (e.g.Cenarro et al. 2001) suggests that they are all F-type super-giants. This can be confirmed by measuring the strength oftheir Ca ii triplet lines (e.g. Diaz et al. 1989; Mallik 1997).Two of the stars, . Thespectral types of all the cool stars are listed in Table 3. Notethat two objects lack optical photometry and have been la-belled as stars 301 (2MASS J20115344+3424427) and 302(2MASS J20114858+3424420).The GTC/OSIRIS observation provided classificationspectra for ten stars, which are listed in Table 4. Figure 6shows six of them. The S/N ratio varies greatly among them,but all allow spectral classification. We performed the clas-sification via comparison to spectra of MK standard starsdegraded to the same resolution. All the stars observed (cor-responding to the top of the cluster blue sequence) are earlyto mid B stars. Due to high-reddening, the S/N is quitelow around H γ and, when possible, we have used the set ofmetallic and He i lines in the 4 900 – 5100 ˚A range to improvethe luminosity classification. The spectral types derived arelisted in Table 4. The spectrum of star 143 shows very asym-metric Balmer lines, most likely indicating the presence oftwo stars (Fig. 6), one with spectral type ∼ B5 III, and anearlier, less bright companion. The spectrum of ∼ B3 V classification is likely.All the intrinsically blue stars that were observed in the In a forthcoming paper, Lohr et al. (submitted to A&A) showthat he young cluster Be 51 ( J - K S ) K S Figure 3.
Colour-magnitude diagram for 2MASS data in circlesof radius ′ (large symbols) and ′ (small symbols) around theposition of Be 51. Circles represent objects selected as possibleearly-type stars, while squares are candidate luminous stars (se-lected as described in the main text). Filled symbols representstars with spectra classified as likely members. Striped squaresare stars with spectra that we do not consider as members. Thetriangles are more distant objects whose connection to the clusteris unclear. Figure 4.
Intermediate-resolution spectra, taken with ISIS, inthe region around the Ca ii triplet of five bright red stars in thecentral concentration of Be 51. Z -band with the WHT were re-observed with GTC in theclassification region, except for We used the higher resolution ISIS spectra of the cool starsto compute effective temperature ( T eff ) and iron abundance Table 2.
Log of the WHT observations. The upper panel showsobservations of cool luminous stars confirmed as members, allof which have been observed at least twice. The middle panelcontains other cool luminous stars. The bottom panel includescandidate blue stars, observed only once.Star Exposure Date Counts/pixel a Time (s) (UT)126 30 21 Aug 2007 35 000350 28 Jun 2012 45 000200 2 Jul 2012 30 00070 200 26 Jul 2007 24 000450 28 Jun 2012 10 000134 200 26 Jul 2007 11 000400 2 Jul 2012 3 000301 200 26 Jul 2007 15 000500 28 Jun 2012 6 000172 200 26 Jul 2007 12 000350 2 Jul 2012 3 000162 200 26 Jul 2007 14 000600 2 Jul 2012 9 000302 200 26 Jul 2007 9 000350 2 Jul 2012 2 000146 200 21 Aug 2007 16 000350 2 Jul 2012 3 000105 200 21 Aug 2007 15 000450 28 Jun 2012 4 000501 200 22 Aug 2007 18 000502 400 22 Aug 2007 12 000503 200 22 Aug 2007 15 000350 2 Jul 2012 4 500901 200 2 Jul 2012 9 000114 900 21 Aug 2007 7 500103 1200 22 Aug 2007 4 000122 900 21 Aug 2007 6 000150 1200 22 Aug 2007 13 000143 900 21 Aug 2007 7 000147 900 21 Aug 2007 7 000 a Counts per pixel in the spectral direction after extraction ofthe whole slit width. Note that spectra from August 2007 werepre-binned by a factor of 2 to match the resolution element. For acorrect estimation of the SNR, all the other spectra should havetwo pixels added. by comparing them to a previously generated grid of syn-thetic spectra. We employed the new version of the auto-mated code
StePar (see Tabernero et al. 2018), which reliesupon spectral synthesis instead of equivalent widths (EWs)and uses a Markov-chain Montecarlo algorithm (emcee, seeForeman-Mackey et al. 2013) for optimisation. We exploredthe parameter space using 20 Markov-Chains of 1 250 pointsstarting from an arbitrary point. As objective function weused a χ -squared in order to fit any previously selected spec-tral features.The synthetic spectra were generated using two setsof one-dimensional LTE atmospheric models, based on M ⊙ and M ⊙ MARCS spherical atmospheric mod-els (Gustafsson et al. 2008). The radiative transfer codeemployed was spectrum (Gray & Corbally 1994). As linelist, we employed a selection from the VALD database(Piskunov et al. 1995; Kupka et al. 2000; Ryabchikova et al.2015), taking into account all the relevant atomic and molec-ular features (dominated by TiO and CN) that can ap-
MNRAS in press, 1–19 (2018)
I. Negueruela et al.
Figure 5.
Intermediate-resolution spectra, taken with ISIS, inthe region around the Ca ii triplet of four bright yellow stars inthe central concentration of Be 51. The vertical dashes indicatethe positions of the Paschen lines, which weaken as we move alongthe spectral sequence (note that Pa 13, 15 & 16 blend with thestronger lines of the triplet). The O i Figure 6.
GTC classification spectra of B-type stars in Be 51. pear in cool luminous stars up to mid-M types. In addition,as Van der Waals damping prescription we employed theAnstee, Barklem, and O’Mara theory (ABO), when avail-able in VALD (see Barklem et al. 2000). Effective tempera-ture T eff ranges from K to
K with a step of
K forthe spectra generated using M ⊙ atmospheric models. Forthe M ⊙ MARCS synthetic models, T eff varies from Kto
K; the step is
K above
K and
K other-wise. Finally, the metalicity ranges from [M/H] = − . dex to[M/H] = . dex in . dex steps for M ⊙ models, whereas M ⊙ MARCS models cover only from [M/H] = − . dexto [M/H] = . dex, in . dex increments. Surface gravity( log g ) varies from − . to 2.0 dex in 0.5 dex steps, whenavailable in each MARCS grid. We convolved our grid of syn-thetic spectra with a gaussian kernel (FWHM ≈ km s − )to account for the instrumental broadening.The present version of StePar allows the derivationsof any set of stellar atmospheric parameters simultaneously.In this case, we restricted them to only two variables, T eff and metallicity ([M/H]). Given the strong degeneracy be- tween log g and metallicity, surface gravity was kept fixedto values compatible with the position of stars in the obser-vational HR diagram (see Sect. 4.2 and compare to Table 3);microturbulence ( ξ ) was adjusted according to the 3D modelbased calibration described in Dutra-Ferreira et al. (2016).Since our stars are more luminous than any of their cali-bration benchmarks we have checked the consistency of ourresults by obtaining parameters with ξ fixed to − and − (i.e. higher than any value used) for all stars. Thederived T eff ’s are quite similar in all cases. Increasing ξ hassome effect on the metallicity derived (higher ξ implies lower[M/H]), but even for the extreme case it is less than ∼ . dexlower. Our analysis employs some empirically selected linesof Mg, Si, Ti, and Fe (Dorda et al. 2016) in the spectralrange around the CaT, – ˚A. The results of theanalysis are listed in Table 3.The spectrum of star 70 cannot be properly fit. TheFourier transform indicates two components separated in ve-locity. Analysis of the spectrum reveals two similar objectsseparated by about − , i.e. barely resolved at our reso-lution. This multiplicity could explain the mild spectral vari-ability. Star 146 cannot be fit either; it is probably too hotfor the set of lines used, more suited for K/M spectral types,but fast rotation also helps to dilute the diagnostic lines. Wecan estimate its rotational velocity even if the resolution ofour spectrum is quite low for this task. As we have a goodidea of the physical parameters of the star, which must bequite similar to those of star 105, we can choose a suitablestellar model from the POLLUX database (Palacios et al.2010) and convolve it with a gaussian of width appropriateto our spectral resolution. We can then take this syntheticspectrum and convolve it with a rotation profile followingGray & Corbally (2009), and compare the result for differ-ent values of v rot and limb darkening with our spectrum. Thiswas done following a Bayesian framework, so that we couldmarginalize over the limb darkening parameter, arriving ata value v rot = . ± . − . This is a very high rotationalspeed for a supergiant, but we do not see any evidence for asecond component in the spectrum.For all the other stars, we obtain a convincing fit. Dueto the low number of lines used in the analysis, the for-mal uncertainties are moderately large. The values of T eff found, though, are appropriate for the observed spectraltypes. Likewise, the analysis of the two spectra of star 126shows rather better agreement than indicated by the formaluncertainties. Although some of our metallicity determina-tions have moderate uncertainties, the seven likely membersin the cluster core present consistent values. If we averagethen using the S/N of the spectra as weight, we obtain amean of + . ± . dex, slightly supersolar. This result im-plies that we can safely use solar-metallicity isochrones forthe analysis. We start the photometric analysis by plotting the V /( B − V ) and V /( U − B ) diagrams for all stars in the field. In Figure 7,we can observe that the cluster sequence (as defined by thespectroscopically identified early-B stars) is heavily contam-inated by what seems to be foreground population. All theconfirmed B-type stars have ( B − V ) ≈ . , and therefore wecan assume that this is the location of the cluster sequence. MNRAS in press, 1–19 (2018) he young cluster Be 51 Table 3.
Stellar parameters for cool stars in the field of Berkeley 51. The top panel lists stars in the central condensation that areconsidered likely members. The bottom panel presents two other stars at higher distances.Star Spectral Date T eff log g [M/H] ξ MARCS grid v hel v LSR type (K) (dex) (km s − ) (km s − ) (km s − )126 M2 Iab 2012 Jun 28 ± . ± . M ⊙ − −
126 2012 Jul 2 ± . ± . M ⊙ − − ∼ F8 Ib 2012 Jun 28 — — — — — − −
134 G8 Ib 2012 Jul 2 ± . ± . M ⊙ − −
301 G8 Ib 2012 Jun 28 ± . ± . M ⊙ − −
172 K0 Ib-II 2012 Jul 2 ± . ± . M ⊙ − −
162 F8 Ib † ± . ± . M ⊙ − −
302 G8 Ib-II 2012 Jul 2 ± . ± . M ⊙ − −
146 F4 Ib 2012 Jul 2 — — — — — − +
105 F5 Ib 2012 Jun 28 ± . ± . M ⊙ − −
503 K2 Ib 2012 Jul 2 ± . ± . M ⊙ − −
901 M1 Iab 2012 Jul 2 ± . ± . M ⊙ +1 +6 † Spectral type in the spectrum analysed. It appears decidedly later in the lower-resolution spectrum taken in 2007.
We proceed with a classical photometric analysis, followingJohnson & Morgan (1953) and Johnson (1958).For early-type stars, it is possible to achieve approxi-mate classification by means of the reddening-free Q param-eter, defined as Q = ( U − B ) − E ( U − B ) E ( B − V ) ( B − V ) , (1)where E ( U − B ) E ( B − V ) depends on the extinction law (taking astandard value of 0.72; Johnson & Morgan 1953) and alsoweakly on the spectral type. For a standard extinction law,we can use the expression E ( U − B ) E ( B − V ) = X + . E ( B − V ) , (2)where X depends weakly on spectral type (or, correspond-ingly, intrinsic colour; Johnson 1958), to calculate an accu-rate Q parameter. Since there are ten early-type membersfor which we have both classification spectra and UBV pho-tometry (see Table 4), we can use them to check if this is avalid approximation. From the spectra, we find that star 131is a Be star with heavy veiling and anomalous colours, whichcan thus not be used for this purpose. Star 147 has weak Becharacteristics, but still can be left out. For the other eightstars, we calculated E ( U − B ) and E ( B − V ) using the intrinsiccolours of Fitzgerald (1970). The average of the reddeningslope for them is . ± . , with the error reflecting thestandard deviation. On the other hand, using Eq. 2 with thevalue listed for a typical spectral type B3 in Johnson (1958),we find a slope of 0.74. This confirms that the extinction inthis direction is standard (see also Section 4.1 later) andthat we can use Q with this value of the slope.Now we use the expression ( B − V ) = . · Q (Johnson1958) to estimate the intrinsic colours and hence the colourexcess E ( B − V ) to each star. For the stars with spectra, wecan compare the values of ( B − V ) obtained from the spectraltype calibration with those derived from the Q method. Theaverage difference is − . ± . mag, and all the differencesare within the typical dispersion of the calibration. The pho- tometric classifications agree with our spectra in placing themain-sequence turnoff at B3.Application of this method reveals that a large frac-tion of the stars in our photometry are B-type stars (i.e. Q < ). However, many of them have to be foreground tothe cluster. For example, star 42 has Q = − . , correspond-ing to a mid-B star, but with V = . and ( B − V ) = . ,it is very far away from the position of the spectroscopicmembers. In addition there is a significant fraction of late-B stars ( > Q > − . ) that have ( B − V ) < ∼ . . These ob-jects are in all likelihood foreground interlopers. This highpercentage of early-type contaminants is not so surprisingas one could na¨ıvely think, because the line of sight goesthrough the Galactic plane, reaching a high distance, andwe are selecting only objects with good U -band photometry.To confirm this foreground character, we divided the B-typestars in two groups, those with E ( B − V ) > . and those with E ( B − V ) ≤ . . The former group contains most of the starswith spectral types B3 – B5 and is much more tightly con-centrated around the position of the cluster than the latter.So we can start our membership determination by discardingstars with values below this threshold as non-members.The eight non-emission stars with classification spectra(which had been selected because they seemed to belongto the top of the cluster sequence) have an average E ( B − V ) = . with a standard deviation σ = . . The standarddeviation indicates that the colour excess is only moderatelyvariable for all these likely members. In view of this, we canincrease our threshold by rejecting stars with E ( B − V ) morethan 3 σ away from this value as probable non-members.There are no stars with colour excesses more than 2 σ abovethis average, but many objects with values lower than 3 σ below the average.To make the final selection, we proceed with an iterativeapproach. Firstly, we divide the sample in two groups. Starswith E ( B − V ) ≥ . (less than 1.5 σ away from the averagefor stars with spectra) are taken as likely members, while therest are taken a possible members. For the first list, we derivephotometric spectral types by comparing their Q values tothe calibration of Johnson (1958) for main-sequence stars.We then calculate a first estimate of the extinction by takingthe intrinsic colours from Fitzgerald (1970) corresponding to MNRAS in press, 1–19 (2018)
I. Negueruela et al. ( B - V ) V ( U - B ) V Figure 7.
Raw CMDs for Berkeley 51, from our
UBV photometry. The location of the stars observed with GTC and confirmed as earlyB-type stars is marked by the filled circles. The location of stars observed with the WHT and confirmed as cool supergiants is shownwith filled diamonds. this photometric type, calculating the corresponding colourexcess and then applying A V = . × E ( B − V ) . With thisvalue, we deredden all the stars and obtain a dereddenedmagnitude m V = V − A V .For the second list (stars with . ≤ E ( B − V ) < . ),we carry out the same procedure separately. We then plotthe likely members and the possible members against theobservational ZAMS from Schmidt-Kaler (Aller et al. 1982)displaced at different distance moduli (as in Fig. 8). We ob-tain an initial cluster distance modulus ( µ ). After subtract-ing this µ , we compare the absolute magnitude of each starto the photometric spectral type to estimate whether thestar has to be a giant or a dwarf to belong to the cluster.Taking into account this luminosity classification, we thenproceed to re-estimate the spectral type for the objects thatmust be giants to be cluster members, using the correspond-ing Q calibration from Johnson (1958). The colour excessesare re-calculated and the process converges, because the dif-ferences in colour between giants and main-sequence starsare very small. We then re-check the fit to the ZAMS andproceed to reject as non-members objects that should havespectral types unexpected for a cluster with a turnoff at B3(for example, b8 iv) if they were at the cluster distance. Afew objects with spectral type b7 that are too bright to beon the main sequence are moved from the list of likely mem-bers to the list of possible members. This does not affectour ZAMS fit, which steadily returns µ = . . Figure 8 (toppanel) shows the fit of the ZAMS to the very likely and pos-sible members that remain after this procedure. The turnoffaround ( B − V ) ≈ − . seems to be well defined. The twodetected Be stars fall clearly to the left of this turnoff, butthis is not unusual for Be stars (and their spetral types showthem to be blue stragglers in any case). A third star, µ shown, is the best possible.Unfortunately, the position of the ZAMS seems to be de-termined mainly by stars in the list of possible members.This introduces the doubt of whether our photometry is re-ally reaching deep enough to touch the ZAMS. On the one -0.2 -0.1 0 ( B - V ) V -0.2 -0.1 0 ( B - V ) V Figure 8. Top panel:
Dereddened CMD for very likely (filledblue circles) and possible (magente circles with vertical stripes)members of Be 51 with the ZAMS of Schmidt-Kaler displaced atdifferent distance moduli ( µ = . , 13.7 and 13.9). The two Bestars are marked by orange circles with horizontal stripes. Bot-tom panel:
The same with the confirmed mebers of Melotte 20displaced to the same µ overploted as cyan diamonds.MNRAS in press, 1–19 (2018) he young cluster Be 51 hand, since stars with types B5 – 6 V are much fainter thanthe B3 V and B5 III objects that we have observed spectro-scopically and our photometry is not very deep, it makessense to assume that we reach only those ZAMS memberswith rather lower than average reddening (that have thusbeen included in the list of possible members). On the otherhand, the luminosity classes derived from the spectra areconsistently lower than those implied from the photometrictypes at µ = . . In fact, if we assume A V = . × E ( B − V ) and the intrinsic magnitude calibration of Turner (1980) toderive spectroscopic distances for the blue stars with spec-troscopy, we obtain an average µ = . ± . , leaving asideagain stars 131 and 143. The two values are just compati-ble within their respective errors (in view of the top panelof Fig. 8, a conservative error of ± . mag is assumed forthe visual fit of the ZAMS as a lower envelope, given theuncertainties under discussion), but the difference is quitesignificant.To investigate this issue further, in the lower panel ofFig. 8, we have added the members of Melotte 20 (the α Percluster), a very well studied cluster which also has a MSturn-off at B3 V. Taking the photometry from Harris (1956),we have displaced them from their
Hipparcos µ of 6.2 mag(van Leeuwen 2009) to µ = . by applying exactly thesame dereddening procedure. We notice that the B3 – 5 Vstars in Mel 20 (including the MK primary B3 V standard29 Per) are well separated from the ZAMS, while in Be 51 wehave B3 V stars all the way down to the ZAMS. This impliesthat Be 51 is somewhat younger than Mel 20. The confirmedlate-B members of Mel 20 trace very well the ZAMS, givingfurther support to the value that we take as definitive µ = . ± . , corresponding to d = . ± . kpc.Leaving out the two known Be stars, the averagereddening for stars in the list of very likely members is h E ( B − V ) = . i with a 1- σ dispersion of 0.08 mag. Thisis essentially identical to the values obtained for the spec-troscopic members alone, thus confirming the validity of theselection criteria. The objects in this list fulfill all the req-uisites to be cluster members. Their derived properties arelisted in Table 4. The list of possible members, given in Ta-ble 5, includes two b3 v stars that fall together with thevery likely members. The membership of these two objectsis rather likely, as B3 field stars are rare, and thus suggeststhat some stars with lower reddening do indeed belong tothe cluster population. On the other hand, a number of ob-jects with classifications b5 – 7 iii that are fainter than theB3 III – IV stars are most likely interlopers. In any event, wemust take into account that errors in Q are > ∼ . mag forthe fainter cluster members, and this can imply changes byalmost one whole subtype, which could move an object fromone list to the other.————————————————————————————————– Cluster members with spectroscopy give an average h E ( B − V )i = . ± . . There is a degree of variability inextinction, expected for a distant object in this region of Table 4.
Intrinsic parameters for all very likely photometricmembers, ordered by dereddened magnitude. Spectral types aregiven for the ten stars observed with GTC/OSIRIS and also for E ( B − V ) spectral type , (if available)71 b3 iv 1.86249 b3 iv 1.84252 b3 iv 1.8986 b6 iii 1.81195 b6 iii 1.66114 b3 iv B3 III 1.76147 b2v B2 IVe 1.78 ,
131 b1v B2 Ve 1.85 ,
175 b3 iv B3 IV 2.02143 b6 iii B5 III 1.88122 b3 iv B3 III–IV 1.7299 b2v 1.75103 b3 v B2 V , Spectral types are derived from the photometry following theprocedure described in the text, under the assumption of µ = . . Colour excess values for the confirmed Be star are not entirelyinterstellar. Derived from the Z -band spectrum, and so less certain. the Galactic Plane. Images of the area (e.g. DSS2) revealthe presence of extended nebulosity in this area, and a pos-sible uncatalogued H ii region about ′ NE from the cluster.WISE images also show substantial dust emission over thewhole area. In the optical images, there is a clear contrastbetween the high stellar density seen to the East and Southof the cluster and the much lower density to the West andNorthwest, suggestive of a foreground dark cloud. The ob-jects listed in Table 4 display a moderately broad range ofcolour excesses, but this is by design (a consequence of theselection procedure in Sect. 3.3), and so this range should betaken as a lower limit. The list of possible members in Ta-ble 5 includes two b3 v stars ( E ( B − V ) ≈ . , sig-nificantly below the range used to define certain members.On the other hand, a number of objects in Table 5 with clas-sifications b5 – 7 iii that are less bright than the stars aroundthe turnoff are unlikely to be members, even though someof them have E ( B − V ) > . . This confirms that E ( B − V ) MNRAS in press, 1–19 (2018) I. Negueruela et al.
Table 5.
Intrinsic parameters for stars that could be members,ordered by dereddened magnitude,Star Spectral type , E ( B − V )
174 b6 iii 1.5956 b5 iii 1.5830 b7 iii 1.84184 b7 iii 1.5321 b6 iii 1.639 b7 iii 1.632 b5 iii 1.50246 b7 iii 1.5431 b7 iii 1.7148 b7 iii 1.69239 b7 iv 1.77111 b3 v 1.62248 b3 v 1.61136 b5 v 1.64102 b5 v 1.53100 b6 v 1.5949 b5 v 1.5585 b7 iv 1.71144 b7 iv 1.6520 b5 v 1.5714 b7 v 1.5672 b7 v 1.5091 b7 v 1.50 Spectral types are derived from the photometry following theprocedure described in the text. The luminosity class indicated isthat needed to be a cluster member, but not necessarily the truevalue. alone is not enough to identify members, while the rangeof E ( B − V ) present in the cluster is likely wider than thatadopted to select certain members.To verify that the reddening law can be approximatedby the standard values, we can also calculate the infraredexcess for very likely members, using the calibration of in-trinsic colours of Winkler (1997), the photometric spectraltypes and ( J − K S ) values from 2MASS. We find h E ( J − K S )i = . ± . , in line with the dispersion expected from the op-tical value (taking into account the relatively large errors of2MASS for the faintest objects). We find h E ( J − K S )ih E ( B − V )i = . , (3)in good agreement with the expectations for a standard ex-tinction law. As the two averages are obtained with differ-ent samples , the average of the ratios E ( J − K S ) / E ( B − V ) for individual targets with good quality photometry, namely . ± . (where the error is the standard deviation), isprobably more informative, and again ratifies the standardreddening law .Despite this, an attempt to individually deredden mem-bers with good 2MASS photometry using the intrinsiccolours of Winkler (1997) results in positions on the CMDthat are too red compared to the isochrone. As this situa-tion has prevented us from individual dereddening of starsin the past (e.g. Marco et al. 2014), we have carried out an Because not all our photometric members have 2MASS pho-tometry passing the quality criteria. A value of 0.52 is expected for Rieke & Lebofsky (1985). -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 ( J - H ) -0.2-0.100.10.20.30.40.5 ( H - K S ) Figure 9.
Colour-colour diagram for early-type stars in the cen-tral ′ of the field, selected as in Section 2.2. The thin straight linesare reddening vectors for a standard Rieke & Lebofsky (1985) ex-tinction law that go through the position of three spectroscopi-cally confirmed cluster members. The thick (red) wavy track isthe locus of luminosity-class V B-type stars according to the ob-servational calibration of Winkler (1997). The dotted line is theposition of stars with masses between M ⊙ and M ⊙ in a PAR-SEC (Bressan et al. 2012) 3 Ma isochrone (essentially, a ZAMSfor B-type stars). The dashed line is the corresponding positionfor 3 – M ⊙ in a 3 Ma Geneva (Georgy et al. 2013) isochrone.The green square is the colour of all O-type main-sequence starsaccording to the synthetic models of Martins & Plez (2006). Theorange circle is the Be star 147. The error bars indicate the typicaluncertainty in the photometry. analysis of the infrared reddening law for the field. For thispurpose, we take all the stars that had been selected in Sec-tion 2.2 as likely early-type stars with good-quality 2MASSphotometry within the central ′ and plot their ( J − H ) and ( H − K S ) colours, together with the reddening vector, com-pared to the expected position of early-type stars accordingto several references in Fig. 9. As can be seen, the observedposition of all the stars lies along the standard reddeningvector. It is clear that all stars deproject to positions com-patible with the observational colour calibration of Winkler(1997), even though they are in the JHK system and somesmall differences with the 2MASS system are expectable,but not with the theoretical positions predicted by eitherthe Geneva or Padova isochrones, which are in the 2MASSsystem. We have repeated this experiment with the muchyounger open cluster Berkeley 90, which shows strong differ-ential reddening (Ma´ız Apell´aniz et al. 2015), coming to anidentical conclusion: in the near-IR colour-colour diagram,early-type stars project along the reddening line to a locuscompatible with the empirical colours of B-type star, but notwith the colours found in the isochrones. This suggests thatthe transformations used to convert the isochrones from thetheoretical plane to magnitudes and colours do not repro-duce well the near-infrared photometry of early-type stars.Note that the issue does not concern the ( J − K S ) colour,which is quite well reproduced (as can be seen in Fig. 11),but the ( J − H ) and ( H − K S ) colours. The first is alwaystoo high and the second always too low. MNRAS in press, 1–19 (2018) he young cluster Be 51 -0.5 0 0.5 1 1.5 2 ( B - V ) V
60 Myr Ω / Ω C = 0.3; DM = 13.7Parsec 65 Ma + 13.740 Ma Ω / Ω C = 0.3; DM =14.4 Figure 10.
Dereddened CMD for very likely members (bluefilled circles and diamonds) and possible members (brown stripedcircles) of Be 51. The solid blue line is the Ma Geneva(Georgy et al. 2013) track displaced to the spectroscopic µ = . .The dash-dotted green line is the Ma Geneva track displaced to µ = . . The dashed red line is the Ma parsec (Bressan et al.2012) isochrone displaced to µ = . . The grey points are starsfrom the synthetic cluster mentioned in the text, generated withthe best-fit parameters and the tools of the Geneva group. Using the values from the previous analysis, we plot a dered-dened CMD for the cluster in Fig. 10. The supergiants havebeen dereddened following the procedure of Fernie (1963),which transforms the observed E ( B − V ) to an equivalent E ( B − V ) for early type stars to account for colour effects .To determine cluster parameters, we used the isochrones ofGeorgy et al. (2013) that cover the B-type range at a widerange of initial rotational velocities, v rot . The effects of ro-tation introduce an extra dimension that cannot be con-strained with the observed CMD. While the position of thesupergiants narrows down significantly the range of possibleages, isochrones with high initial average rotations are al-most indistinguishable from younger isochrones with lowerinitial rotation. For example, isochrones for 30 Ma withno rotation and 40 Ma with a moderately high rotation( Ω / Ω crit = . ) are almost identical. As a compromise, wewill use isochrones with moderate rotation, Ω / Ω crit = . .If we accept the value of µ = . that we obtainedfrom the ZAMS fitting, then the isochrone for 60 Ma pro-vides a good fit to the data. If, however, we accept the µ = . that we obtain from the spectral classification, andthus the implication that our photometry might not be deepenough to reach the ZAMS, then the isochrone for 40 Maprovides a similarly good fit. The position of the red super-giants around ( B − V ) = . may be considered as supportfor the older value, but we must note that the good agree-ment with the isochrones must be interpreted with greatcaution. The evolved stars occupy positions compatible withthe isochrones, but not those where the models predict that This procedure results in h E ( B − V )i = . ± . for sevensupergiants with photometry, in perfect agreement with the valuesfor the blue members. they should spend most of the time. To illustrate this, wehave generated an artificial cluster using the interactive toolsmade available by the Geneva group (Georgy & Ekstr¨om2017) and employing the same astrophysical parametersthat we use in the fit: Ω / Ω crit =0.3, solar composition and τ = Ma. The cluster initially has 20 000 intermediate-massstars, so that statistical sampling is not an issue. This arti-ficial cluster is overplotted on Fig. 10, displaying the mainproperties of the evolutionary tracks: stars spend the firstpart of the He-burning phase as rather cool red giants (withspectral types K4 – M0) and the second half as blue (A orF) supergiants. None of the evolved stars occupy the regionsof highest density in the synthetic cluster. In particular, themodel predicts that > ∼
30% of the He-burning stars shouldappear as post-RSG A-type supergiants, while we do not seeany. This discrepancy is not unique to Be 51, but widespread.The similarly-aged cluster Berkeley 55 contains one late-Fand five K (super)giants. Except for one, the spectral typesof the K super(giants) place them at the same position onthe isochrone as the Be 51 objects (Negueruela & Marco2012). There are other well-studied clusters with similarages, but they do not contain many evolved stars. For exam-ple, IC 4665 has no evolved stars, while Melotte 20 only in-cludes the F5 Ib supergiant α Per. Looking at the literature,we find NGC 4609, NGC 6546 and Trumpler 3 with one red(super)giant each; NGC 6520 and NGC 6649 contain both F-type and red (super)giants, while NGC 6834 and NGC 7654contain one F supergiant each. Only two clusters of similarage, NGC 5281 and NGC 2345, have known A Ib/II stars,but their locations on CMDs are compatible with stars mov-ing off the main sequence, not with a post-RSG nature. Evenwhen we add together all these clusters, the total populationis still small. However, the fraction of F-type stars is muchhigher than predicted by the Geneva models, while post-RSG A supergiants are almost absent. This suggests thatthe models predict a blue loop that extends to higher tem-peratures than supported by observations. In fact, if we plota Ma parsec isochrone (Bressan et al. 2012) in Fig. 10,we see that the only significant difference with the Mamoderate-rotation Geneva isochrone is the size of the blueloop, which seems more compatible with the observations.Even then, the two mid-F supergiants in Be 51 occupy posi-tions more consistent with the ”Hertzsprung gap” than withthe loop.As a further test, in Fig. 11 we plot the individually-dereddened 2MASS data for the supergiants and the fewblue members with high-quality photometric values togetherwith the same isochrones used to fit the optical data.For the cool supergiants, we use the average calibrationof Gonz´alez-Fern´andez & Negueruela (2012) for K and Mstars, and the intrinsic colours of F-type supergiants fromKoornneef (1983) transformed to the 2MASS system accord-ing to the equations of Carpenter (2001) . The match is quitegood, but we note that: • The red supergiants lie very far away from their ex-pected position. However, this is likely an artifact of the As updated in .MNRAS in press, 1–19 (2018) I. Negueruela et al. -0.5 0 0.5 1 1.5 ( J-K S ) K S
40 Ma Ω / Ω C = 0.360 Ma Ω / Ω C = 0.365 Ma Parsec Figure 11.
Dereddened CMD diagram for cluster very likelymembers with useful 2MASS photometry. The isochrones are asin Fig. 10. intrinsic colour calibration, which goes from ( J − K S ) = . at G8 I to ( J − K S ) = . at K1 I (while it is 1.05 at M1 I). • The position of star 70 is slightly different from its lo-cation in the optical diagram. However, this is entirely de-pendent on the assumption of an F8 Ib spectral type. The ( B − V ) = . of ( B − V ) = . of ( J − K S ) = . . With a spec-tral type F8 Ib, E ( J − K S ) = . , well above thecluster average. All this suggests that • The position of star 126 is completely inconsistent withthe rest of the cluster in both diagrams.We cannot give too much weight to the position of anyof the individual evolved stars, as post-MS evolution is fast,and stochastic effects must contribute strongly to the verydifferent evolved populations seen in the clusters mentionedabove. To quantify this effect, 100 synthetic clusters weregenerated, all with the same parameters: solar metallicity,an age of Ma and a distribution of initial rotational veloc-ities. Each cluster has an initial total mass around M ⊙ ,i.e. typical for a moderately massive Galactic young cluster.Details of the clusters are presented in Appendix A. Thenumber of supergiants in a given cluster ranges from zero (7out of 100 clusters) to six (5 out of 100 clusters). The distri-bution is shown in Fig. 12. Most clusters have between twoand four evolved stars (about 20% in each case) and the av-erage number is three supergiants per cluster. This suggeststhat Be 51 must have a higher initial mass, at least twice asmuch to be in the range of statistical probability and mostlikely three times more, i.e. between and × M ⊙ . Radial velocities were calculated following the procedureoutlined in Koposov et al. (2011). Since ISIS is attached tothe Cassegrain focus of the WHT, it moves with the tele-scope, and is subject to large flexures. The method employeduses sky emission to refine possible systematics remaining inthe wavelength calibration, so that all the spectra are in acommon system that can then be anchored by using velocitystandards. After this, it compares the observed spectra with
Figure 12.
Number of supergiants found in a sample of 100synthetic clusters with masses ∼ M ⊙ . The plot shows thenumber of clusters containing a given number of supergiants (seeAppendix A). a battery of models using a Bayesian framework. This hasthe advantage that it is possible to marginalize over any pa-rameter in which we are not interested, removing it from theanalysis while at the same time taking it into account whenderiving uncertainties. In our case, we marginalize over thecontinuum normalization, since continuum determination isalmost impossible at these resolutions for late type stars,and over the stellar model, so that the derived velocities arenot model dependent. According to the ISIS manual, an in-ternal accuracy of km s − can be achieved with R1200R,though a small systematic shift could be present. Two ob-servations of star 126 taken on two separate nights (4 nightsapart) differ by km s − , suggesting that the manual doesnot overestimate greatly the accuracy achievable. The accu-racy is likely to be lower for the F-type stars, which have less(and generally broader) lines in the range used. The radialvelocities measured are listed in Table 3.At first sight, the radial velocities do not seem consistentwith a single population. The average of all the values (in-cluding the two measurements of v LSR = − km s − ,with a standard deviation of km s − . However, we can seethat there are only three stars with velocities moderatelydifferent from the rest. Of these, two are the objects iden-tified as spectroscopic variable, stars 70 and 162. If theseobjects are on the instability strip or binary (as seems tobe the case of v LSR = − km s − , witha standard deviation of km s − , perfectly compatible withthe expected internal accuracy of the measurements.We calculated the Galactic rotation curve with respectto the local standard of rest (LSR) in the direction of Be 51using the fit of Reid et al. (2014), with R = . kpc and φ = . km s − . A cluster velocity of v LSR = − km s − corresponds to a kinetic distance of ≈ . kpc, in verygood agreement with the photometric distance. A longerdistance of . kpc would correspond to a radial velocity MNRAS in press, 1–19 (2018) he young cluster Be 51 v LSR ≈ − km s − . Therefore the observed radial velocityfavours the distance derived from the photometric solution.At a distance of ∼ . kpc and with ℓ ≈ ◦ , Be 51 would beplaced behind the whole extent of the Local arm, which ex-plains its high and patchy extinction. Xu et al. (2013) showthat the Local arm is a major structure, consisting of sev-eral important star-forming regions and extending until atleast 4 kpc away from the Sun in this direction. However,at this distance Be 51 would already be in the Perseus arm.Indeed, for a distance > ∼ . kpc, Galactic structure mod-els would place the cluster beyond the Perseus arm, in theinter-arm region. This again favours the distance derivedfrom photometry and the older age.Xu et al. (2013) assume that the Local arm bendsslightly, and branches from the Perseus arm at ℓ ∼ ◦ ,but we note that a shallower arm curvature would meana branching point not far from the position of Be 51. Thereare no known tracers of the Perseus arm between ℓ ≈ ◦ and ℓ ≈ ◦ (Choi et al. 2014), and so Be 51 could representan important anchoring point for this arm. Even if our photometry is deep enough to sample the ZAMS,we are only confidently reaching a spectral type B5 V, cor-responding to a M V ≈ − . , and still are unlikely to be com-plete for the faintest (i.e. more reddened) objects. Even so,we find close to 40 photometric members and a few morepossible members down to slightly fainter M V . If we assumethe older age of Ma, these objects are in the range between ∼ . M ⊙ and ∼ . M ⊙ , i.e. they contribute > ∼ M ⊙ to thecluster mass. We cannot attempt to derive an initial massfunction or a total mass, but by integrating a simple Salpeterlaw under the assumption that there are 40 stars between . M ⊙ and . M ⊙ , we find a current mass of ∼ M ⊙ instars more massive than the Sun only, simply from this lowerlimit. This would imply an initial cluster mass > ∼ M ⊙ .Integration of a standard (Kroupa 2001) IMF would result ina cluster mass ≈ M ⊙ . This is a lower limit that does nottake into account the incompleteness of our photometry, theeffects of binarity or dynamical ejections from the cluster. Itis clear that Be 51 is at least a moderately massive cluster.As discussed in Sect. 4.2, comparison to synthetic clusterssuggest that a total mass above × M ⊙ is needed for thepresence of nine evolved stars.Even though the cluster appears strongly concentrated,there are some very likely members over the whole field cov-ered by our photometry. In addition, even within small field,there is an indication of increasing reddening towards theNorth. As mentioned in Sect. 4.1, the wide field DSS and2MASS images strongly confirm this impression. As a fur-ther check on the extent of the cluster, we took spectra ofsome of the luminous red star candidates found outside thecluster core (see Fig. 3 and Table 2). Since we could not ob-serve all objects in the diagram, we selected two stars thatlie close ( r < ′ ) to the cluster but have colours differentfrom those of cluster supergiants, and two stars found atlarger distances, with colours similar to cluster members.The two nearby stars, 501 (2MASSJ20114667+3423097) and 502 (2MASSJ20114564+3422422), have spectral types M4 Ib andM4 II, respectively. Since they were observed only in the 2007 run, we have no measurement of their radial velocities.The position of star 502 in the CMD clearly rules out anassociation with the cluster, as it is fainter in K S than theearly-K supergiants. Star 501 also appears too faint, andhas a colour excess E ( J − K S ) = . , lower than clustermembers. We thus conclude that these two objects arechance projections.Star 503 (2MASS J20121264+3420366) lies about ′ southeast of the cluster. It has exactly the same coloursand magnitudes as the clump of early K supergiants, andits spectral type K2 Ib is typical of this group. Its observedradial velocity, however, v LSR = − km s − , is quite differ-ent from the cluster average, even though its metallicity iscompatible with the cluster average. We note that there is amoderate number of objects with similar colours and magni-tudes within ′ of the cluster that would be worth checkingfor radial velocities.Finally, star 901 (2MASS 20115472+3427464) is a verybright infrared source (IRAS 20099+3418) located ∼ . ′ north of the cluster. It is an M1 Iab supergiant, with a veryhigh colour excess, E ( J − K S ) = . , very likely due to thedark cloud discussed above. Its WISE colours have a verylarge uncertainty, e.g. (W1 − W3) = . ± . , but do notsuggest a strong intrinsic reddening, associated with heavymass loss. Its nature will be discussed in Sect. 4.7. We find a younger age than the few previous works on Berke-ley 51. Even though our age estimate is fully supported bythe spectral types of the stars observed, it is interesting toconsider the reasons why Berkeley 51 had not been identifiedas a young open cluster before. In the case of Tadross (2008),he obtains an age of 150 Ma, based on 2MASS data. Sincethe global shape of the isochrones at these ages is similar, thedifference is likely due to the distance/age/extinction degen-eracy. However, we note that the ages of all young open clus-ters seem to be overestimated in Tadross (2008). For exam-ple, he correctly identifies Berkeley 90 as the youngest clus-ter in his sample, but he assigns an age of ∼ Ma, whileit is in fact a very young open cluster containing early-Ostars (Marco & Negueruela 2017). This is probably becausethe only young cluster in his calibration set, Berkeley 55, isgiven a rather old age of 300 Ma (cf. Negueruela & Marco2012).A direct comparison of our photometry with that ofSubramaniam et al. (2010) is difficult, as they do not pro-vide coordinates for their objects. However, we can compare ∼ bright objects that are easily identifiable in the clusterchart available in the WEBDA database. There are signif-icant systematic differences between the two photometricdatasets. Their V magnitudes are between 0.4 and 0.6 magbrighter than ours, while their ( B − V ) colours are larger, withtheir B magnitudes only 0.1 or 0.2 mag brighter than ours.Despite this large difference, inspection of their fig. 22 sug-gests that the much older age that they give to the clusteris mainly due to an incorrect identification of cluster mem-bers, as none of the bright supergiants falls on the isochronechosen. MNRAS in press, 1–19 (2018) I. Negueruela et al.
Berkeley 51 is remarkable in containing four F-type su-pergiants. Only the starburst cluster Westerlund 1 has alarger population (6 yellow supergiants), but these are muchmore massive stars, with ∼ M ⊙ (Clark et al. 2010).NGC 7790, which is somewhat older than Be 51 ( ∼ Ma;Majaess et al. 2013), hosts three, all of which are Cepheidvariables, the two components of the binary CE Cas, andCF Cas. NGC 129, with an age similar to NGC 7790, con-tains the Cepheid DL Cas and the non-variable F5 Ib super-giant HD 236433. The Cepheid V376 Cas could be a halomember as well (Anderson et al. 2013).We can use the strength of the O i ± . ˚A, owing to the definition of the continuum. For star105, we measure an EW = . ˚A, which would correspond to M V = − . . For . ˚A, i.e. the same valuewithin errors, in good agreement with the fact that the twostars have approximately the same magnitudes and coloursin all photometric diagrams. This luminosity is a bit higherthan those of α Per or HD 236433, which are quite similarto each other according to Kovtyukh et al. (2012), and haveabout the same spectral type as the Be 51 objects.For star 162, however, we measure an EW= . ˚A. Ac-cording to the calibration, this implies an M V of only − . .For . ˚A, but this objectis likely a binary. Given the position of the two late-F starsin the CMDs, we would expect them to be slightly brighterthan the mid-F supergiants, instead of fainter. The mostlikely explanation resides in the fact that the calibration ofArellano Ferro et al. (2003) does not take into account thedependence of EW74 with effective temperature. Accordingto Kovtyukh et al. (2012), supergiants of later types haveweaker EWs at a given luminosity . Moreover, as seen infig. 3 of Kovtyukh et al. (2012), for temperatures ∼ Kand lower, i.e. as we move into G-types, EW74 seems to showa weaker dependence on luminosity. Star 162 had a spectraltype G1 Ib in the spectrum on which EW74 was measured,and therefore falls in this regime. So we have reason to be-lieve that its M V has been underestimated. We can thus con-clude that the values found are consistent, within the uncer-tainties typical of the calibration (around 0.7 mag), with theisochrones. For the short distance that we have preferred, M V ≈ − . , with Gaia distances exist for all of them. Note that we have not used the calibration of Kovtyukh et al.(2012) because it has a strong dependence on log g and especially ξ , which we have had to fix at assumed values. Star 126 is by far the brightest cluster member in the nearinfrared. It is also the counterpart of the mid-infrared sourceIRAS 20100+3415. This object, located in the inner core ofthe cluster, has a radial velocity fully compatible with othercluster members. However, its late spectral type and positionin the CMDs is incompatible with the best-fit isochrones. Asan M2 Iab supergiant, it is expected to be a moderately mas-sive star (typically, of ∼ M ⊙ ). Such objects are extremelyrare and a chance coincidence with a young open clusteris very unlikely. Moreover, its luminosity is fully consistentwith the expectations for an object of this spectral type atthe cluster distance, while the observed v rad makes chancecoincidence even more unlikely.Its WISE colours are poorly constrained, probably dueto saturation, but with (W1 − W3 ) = . ± . , there isclear evidence of strong mass loss. The interpretation of thisobject is further compounded by the detection of a secondluminous supergiant, star 901 discussed above, only ∼ . ′ to the North. While K-type supergiants in the Milky Wayare generally of luminosity class Ib and can be descendedfrom stars of only 7 or M ⊙ (c.f. Negueruela & Marco 2012;Alonso-Santiago et al. 2017), M-type supergiants of lumi-nosity class Iab are high-mass stars. Given their short life-times, they are rare objects (the known Galactic popula-tion runs into the several hundred, with estimates of a totalpopulation of a few thousand). Except in regions of intenserecent star formation, the chance detection of two such ob-jects within . ′ is very unlikely (see Negueruela et al. 2016,for estimates based on actual observations). If we take intoaccount that dered-dened colours and magnitudes, the possibility that they arenot physically related, in spite of a difference in radial ve-locity of km s − seems very unlikely.If the cluster age lies in the young half of the rangeconsidered, the most evolved stars are expected to have, ac-cording to isochrones, > ∼ M ⊙ . We should then consider thepossibility that an object like i i bona fide red supergiants ofthe same spectral type that we had observed at the sameresolution. In view of this, we consider that the most likelyexplanation for the presence of Gaia final release willbe able to ascertain this hypothesis.
MNRAS in press, 1–19 (2018) he young cluster Be 51 We have carried out a comprehensive spectroscopic and pho-tometric study of the highly reddened open cluster Berke-ley 51. Our analysis conclusively shows that this is a youngopen cluster with an important population of evolved stars.We identify a main-sequence turn-off at spectral type B3 Vand at least two Be stars with earlier spectral types. In addi-tion, we find four yellow and five red supergiants in the cen-tral overdensity. Two of the yellow supergiants show spectralvariability, displaying spectral types F8 Ib and later, a be-haviour typical of Cepheid variables.A fit to the main sequence indicates a distance of5.5 kpc, although we may be missing some of the faintest(most heavily reddened) members. The spectral types ofsome of the brightest members may favour a higher dis-tance, but both the cluster average radial velocity (whencompared to the Galactic rotation curve in this direction)and the strength of the O i triplet in the four yellow super-giants identified support a distance not much higher than5.5 kpc, which is compatible with a location in the Perseusarm according to most models, even though no other tracersare known in this direction.Isochrone fits would suggest an age of ∼ Ma formoderate-rotation Geneva models or Ma for parsec mod-els. Although the isochrones reproduce well the overall dis-tribution of evolved of stars in the CMDs, the supergiantsare not located at positions where the models predict thatstars should spend a substantial amount of time. The redsupergiants (with spectral types G8 or K0) appear all some-what warmer than the predictions for the first part of Hecore burning. Geneva models predict that, after this phase,stars will move to high temperatures and will spend the restof He core burning as A-type supergiants. The populationsobserved in a number of cluster with ages ∼ Ma do notseem to agree with this prediction, as they contain prefer-entially F-type supergiants, either Cepheid variables or ob-jects with stable spectral type close to F5 Ib. The parsec isochrones predict shorter blue loops at a given metallicity,which perhaps are in better agreement with observations.The mid-infrared source IRAS 20100+3415, locatednear the centre of the cluster, is an M2 Iab supergiant,probably the descendant of a blue straggler formed via bi-nary interaction. Its WISE colours suggest heavy mass loss.Searches for associated maser emission would be of high in-terest to exploit the availability of Be 51 as a tracer of thePerseus arm in a poorly known region of the Milky Way.
ACKNOWLEDGEMENTS
We are very thankful to Dr. Sylvia Ekstr¨om for generatingthe synthetic clusters used in the analysis. We also thank theanonymous referee for constructive comments that improvedthe paper.During part of this work, IND was a visitor atthe Institute of Astronomy, University of Cambridge,whose warm hospitality is heartily acknowledged. Thisvisit was funded by the Conselleria de Educaci´on, Cul-tura y Deporte of the Generalitat Valenciana under grantBEST/2014/276. This research is partially supported by theSpanish Government Ministerio de Econom´ıa y Competiti- vad (MINECO/FEDER) under grant AYA2015-68012-C2-2-P. HMT acknowledges support from MINECO under fel-lowship FJCI-2014-23001. MM acknowledges the support ofa research grant funded by the STFC (ST/M001008/1).The photometric observations were obtained with thethe Nordic Optical Telescope, operated by the Nordic Op-tical Telescope Scientific Association. The spectroscopic ob-servations were obtained with the WHT, which is operatedon the island of La Palma by the Isaac Newton Group, theGran Telescopio Canarias (GTC), and the Mercator Tele-scope, operated by the Flemish Community. All these tele-scopes are installed in the Spanish Observatorio del Roquede Los Muchachos of the Instituto de Astrof´ısica de Ca-narias. The Starlink software (Currie et al. 2014) is currentlysupported by the East Asian Observatory.This research has made use of the Simbad, Vizier andAladin services developed at the Centre de Donn´ees As-tronomiques de Strasbourg, France. This research has madeuse of the WEBDA database, operated at the Department ofTheoretical Physics and Astrophysics of the Masaryk Uni-versity. It also makes use of data products from the TwoMicron All Sky Survey, which is a joint project of the Uni-versity of Massachusetts and the Infrared Processing andAnalysis Center/California Institute of Technology, fundedby the National Aeronautics and Space Administration andthe National Science Foundation. This paper makes use ofdata obtained from the Isaac Newton Group Archive whichis maintained as part of the CASU Astronomical Data Cen-tre at the Institute of Astronomy, Cambridge.
REFERENCES
Aller L. H., et al., eds, 1982, Landolt-B¨ornstein: Numerical Dataand Functional Relationships in Science and Technology -New Series “ Gruppe/Group 6 Astronomy and Astrophysics ”Volume 2 Schaifers/Voigt: Astronomy and Astrophysics / As-tronomie und Astrophysik “ Stars and Star Clusters / Sterneund SternhaufenAlonso-Santiago J., Negueruela I., Marco A., Tabernero H. M.,Gonz´alez-Fern´andez C., Castro N., 2017, MNRAS, 469, 1330Anderson R. I., Eyer L., Mowlavi N., 2013, MNRAS, 434, 2238Arellano Ferro A., Giridhar S., Rojo Arellano E., 2003, Rev. Mex.Astron. Astrofis., 39, 3Barklem P. S., Piskunov N., O’Mara B. J., 2000, A&AS, 142, 467Bressan A., Marigo P., Girardi L., Salasnich B., Dal Cero C.,Rubele S., Nanni A., 2012, MNRAS, 427, 127Carpenter J. M., 2001, AJ, 121, 2851Cenarro A. J., Cardiel N., Gorgas J., Peletier R. F., Vazdekis A.,Prada F., 2001, MNRAS, 326, 959Chiosi C., Bertelli G., Bressan A., 1992, ARA&A, 30, 235Choi Y. K., Hachisuka K., Reid M. J., Xu Y., Brunthaler A.,Menten K. M., Dame T. M., 2014, ApJ, 790, 99Clark J. S., Ritchie B. W., Negueruela I., 2010, A&A, 514, A87Comer´on F., Pasquali A., 2005, A&A, 430, 541Currie M. J., Berry D. S., Jenness T., Gibb A. G., Bell G. S.,Draper P. W., 2014, in Manset N., Forshay P., eds, Astro-nomical Society of the Pacific Conference Series Vol. 485, As-tronomical Data Analysis Software and Systems XXIII. p. 391Diaz A. I., Terlevich E., Terlevich R., 1989, MNRAS, 239, 325Dorda R., Gonz´alez-Fern´andez C., Negueruela I., 2016, A&A,595, A105Draper P. W., Taylor M., Allan A., 2011, Starlink User Note, 139Dutra-Ferreira L., Pasquini L., Smiljanic R., Porto de Mello G. F.,Steffen M., 2016, A&A, 585, A75MNRAS in press, 1–19 (2018) I. Negueruela et al.
Ekstr¨om S., et al., 2012, A&A, 537, A146Ekstr¨om S., Georgy C., Meynet G., Groh J., Granada A., 2013,in Kervella P., Le Bertre T., Perrin G., eds, EAS Publi-cations Series Vol. 60, EAS Publications Series. pp 31–41( arXiv:1303.1629 ), doi:10.1051/eas/1360003Fernie J. D., 1963, AJ, 68, 780Fitzgerald M. P., 1970, A&A, 4, 234Foreman-Mackey D., Hogg D. W., Lang D., Goodman J., 2013,PASP, 125, 306Garc´ıa-Hern´andez D. A., Garc´ıa-Lario P., Plez B., Manchado A.,D’Antona F., Lub J., Habing H., 2007, A&A, 462, 711Georgy C., Ekstr¨om S., 2017, in Charbonnel C., Nota A., eds,IAU Symposium Vol. 316, Formation, Evolution, and Survivalof Massive Star Clusters. pp 355–356 ( arXiv:1509.02779 ),doi:10.1017/S1743921315008868Georgy C., Ekstr¨om S., Granada A., Meynet G., Mowlavi N.,Eggenberger P., Maeder A., 2013, A&A, 553, A24Gonz´alez-Fern´andez C., Negueruela I., 2012, A&A, 539, A100Gonz´alez-Fern´andez C., Dorda R., Negueruela I., Marco A., 2015,A&A, 578, A3Gray R. O., Corbally C. J., 1994, AJ, 107, 742Gray R. O., Corbally J. C., 2009, Stellar Spectral ClassificationGustafsson B., Edvardsson B., Eriksson K., Jørgensen U. G.,Nordlund ˚A., Plez B., 2008, A&A, 486, 951Harris III D. L., 1956, ApJ, 123, 371Howarth I. D., Murray J., Mills D., Berry D. S., 2014, DIPSO:Spectrum analysis code, Astrophysics Source Code Library(ascl:1405.016)Jeffries R. D., 1997, MNRAS, 288, 585Johnson H. L., 1958, Lowell Observatory Bulletin, 4, 37Johnson H. L., Morgan W. W., 1953, ApJ, 117, 313Kharchenko N. V., Piskunov A. E., Schilbach E., R¨oser S., ScholzR.-D., 2013, A&A, 558, A53Koornneef J., 1983, A&A, 128, 84Koposov S. E., et al., 2011, ApJ, 736, 146Kovtyukh V. V., Gorlova N. I., Belik S. I., 2012, MNRAS,423, 3268Kroupa P., 2001, MNRAS, 322, 231Kupka F. G., Ryabchikova T. A., Piskunov N. E., Stempels H. C.,Weiss W. W., 2000, Baltic Astronomy, 9, 590Landolt A. U., 1992, AJ, 104, 340Ma´ız Apell´aniz J., et al., 2015, A&A, 579, A108Majaess D., et al., 2013, A&A, 560, A22Mallik S. V., 1997, A&AS, 124, 359Marco A., Negueruela I., 2017, MNRAS, 465, 784Marco A., Negueruela I., Gonz´alez-Fern´andez C., Ma´ız Apell´anizJ., Dorda R., Clark J. S., 2014, A&A, 567, A73Martins F., Plez B., 2006, A&A, 457, 637Mermilliod J. C., Mayor M., Udry S., 2008, A&A, 485, 303Meynet G., Maeder A., 2000, A&A, 361, 101Mowlavi N., Forestini M., 1994, A&A, 282, 843Negueruela I., 2016, IAU Focus Meeting, 29, 461Negueruela I., Marco A., 2012, AJ, 143, 46Negueruela I., Schurch M. P. E., 2007, A&A, 461, 631Negueruela I., Marco A., Gonz´alez-Fern´andez C., Jim´enez-Esteban F., Clark J. S., Garcia M., Solano E., 2012, A&A,547, A15Negueruela I., Clark J. S., Dorda R., Gonz´alez-Fern´andez C.,Marco A., Mongui´o M., 2016, in Skillen I., Barcells M., TragerS., eds, Astronomical Society of the Pacific Conference SeriesVol. 507, Multi-Object Spectroscopy in the Next Decade: BigQuestions, Large Surveys, and Wide Fields. p. 75Netopil M., Paunzen E., St¨utz C., 2012,Astrophysics and Space Science Proceedings, 29, 53Palacios A., Gebran M., Josselin E., Martins F., Plez B., BelmasM., L`ebre A., 2010, A&A, 516, A13Piskunov N. E., Kupka F., Ryabchikova T. A., Weiss W. W.,Jeffery C. S., 1995, A&AS, 112, 525 Poelarends A. J. T., Herwig F., Langer N., Heger A., 2008, ApJ,675, 614Raskin G., Van Winckel H., 2014, Astronomische Nachrichten,335, 32Reid M. J., et al., 2014, ApJ, 783, 130Rieke G. H., Lebofsky M. J., 1985, ApJ, 288, 618Ryabchikova T., Piskunov N., Kurucz R. L., Stempels H. C.,Heiter U., Pakhomov Y., Barklem P. S., 2015, Phys. Scr.,90, 054005Salasnich B., Bressan A., Chiosi C., 1999, A&A, 342, 131Schaller G., Schaerer D., Meynet G., Maeder A., 1992, A&AS,96, 269Shortridge K., Meyerdierks H., Currie M. J., Davenhall C., Jen-ness T., Clayton M., 2014, Starlink Figaro: Starlink versionof the Figaro data reduction software package, AstrophysicsSource Code Library (ascl:1411.022)Skrutskie M. F., et al., 2006, AJ, 131, 1163Stetson P. B., 1987, PASP, 99, 191Subramaniam A., Carraro G., Janes K. A., 2010, MNRAS,404, 1385Tabernero H. M., Dorda R., Negueruela I., Gonz´alez-Fern´andezC., 2018, MNRAS,Tadross A. L., 2008, MNRAS, 389, 285Turner D. G., 1980, ApJ, 240, 137Weidemann V., 2000, A&A, 363, 647Winkler H., 1997, MNRAS, 287, 481Xu Y., et al., 2013, ApJ, 769, 15van Leeuwen F., 2009, A&A, 497, 209
APPENDIX A: SYNTHETIC CLUSTERS
One hundred synthetic clusters were kindly generated by Dr.Sylvia Ekstr¨om by using the interactive tools made availableby the Geneva group (Georgy & Ekstr¨om 2017). Each clus-ter contains 200 stars with masses ranging between . ⊙ and the highest mass in the cluster. This highest mass de-pends on the stochastic sampling of the IMF and the initialrotational velocity distribution; it ranges between . ⊙ and . ⊙ , with a typical (both median and mode) valueof . ⊙ . The 200 stars (a fraction of which are binaries)have a mass ranging from 635 to
760 M ⊙ , with an average of
695 M ⊙ . For a standard IMF, this translates into an initialtotal mass of ∼ ⊙ for each cluster.To analyse the synthetic CMDs, we defined four classesof post-MS stars: • Blue giants:
These are stars just above the turnoff,but probably still burning H in their cores. We select objectswith ( B − V ) ≤ − . and − ≤ M V ≤ − . , typical of B2 – B8giants. • Blue supergiants:
We select objects with ( B − V ) ≤− . and M V ≤ − . (B-type supergiants) and objects with ( B − V ) > − . , ( U − B ) < − . and M V ≤ − . (A-typesupergiants). • Yellow supergiants:
We select objects with ( U − B ) ≥− . , ( B − V ) < . and M V < − . . The first condition im-plies a division between blue and yellow supergiants aroundspectral type A8; the second condition divides yellow fromred supergiants at G5. • Red supergiants:
These are selected as having ( B − V ) ≥ . , M V ≤ − . .The total number of evolved stars found in the 100 syn-thetic clusters is 304, i.e. a cluster of ∼ M ⊙ is typically MNRAS in press, 1–19 (2018) he young cluster Be 51 Figure A1.
Distribution of evolved stars in the synthetic clus-ters. The plot represents the number of clusters in the samplecontaining a given number of evolved stars: blue giants (cyan);blue supergiants (blue); yellow supergiants (yellow); red super-giants (red); total number of supergiants (black). The verticalscale has been cut for clarity. The number of clusters containingzero blue giants is 89, while the number of clusters containingzero blue supergiants is 69. expected to have 3 evolved stars. There are only 11 clusterswith blue giants, with a total of 13 blue giants. Among thesupergiants, the numbers are 33/99/159 for blue/yellow/red(as a fraction of the total number, 0.11/0.34/0.55 in goodagreement with the 20 000 star synthetic cluster discussed inSect. 4.2). The distribution of evolved stars in the clustersis shown in Fig. A1.
APPENDIX B: PHOTOMETRIC DATA
MNRAS in press, 1–19 (2018) I . N eg u e r u e l a e t a l . Table B1: Coordinates,
UBV photometry, 2MASS identification and2MASS
JHK S photometry for all stars in the field of Berkeley 51. Errorsin the UBV photometry represent the standard deviation of n measure-ments or the photometric error when n = . Errors in 2MASS data areas given by the catalogue. The flags identify the quality of 2MASS pho-tometry, with U indicating upper limits. Only the first ten entrances areshown. The whole table will appear as material on-line.