Galactic or extragalactic chemical tagging for NGC3201? Discovery of an anomalous CN-CH relation
Bruno Dias, Ignacio Araya, João Paulo Nogueira-Cavalcante, Leila Saker, Ahmed Shokry
aa r X i v : . [ a s t r o - ph . GA ] M a r Astronomy & Astrophysicsmanuscript no. dias_mar18_SAGA c (cid:13)
ESO 2018March 15, 2018
Galactic or extragalactic chemical tagging for NGC3201?
Discovery of an anomalous CN-CH relation ⋆ , ⋆⋆ B. Dias , I. Araya , , J.P. Nogueira-Cavalcante , , L. Saker , and A. Shokry , European Southern Observatory, Alonso de Córdova 3107, Santiago, Chilee-mail: [email protected] Instituto de Física y Astronomía, Facultad de Ciencias, Universidad de Valparaíso, Casilla 5030, Valparaíso, Chile Núcleo de Matemáticas, Física y Estadística, Facultad de Ciencias, Universidad Mayor, Chile Observatório do Valongo, Universidade Federal do Rio de Janeiro, Ladeira Pedro Antônio, 43, Saúde 20080-090, Rio de Janeiro,Brazil Observatório Nacional, Rua Gal. José Cristino 77, São Cristóvão 20921-400 Rio de Janeiro RJ, Brazil Observatorio Astronómico, Universidad Nacional de Córdoba, Laprida 854, Córdoba, CP 5000, Argentina National Research Institute of Astronomy and Geophysics (NRIAG), 11421 Helwan, Cairo, EgyptReceived ; accepted
ABSTRACT
Context.
The origin of the globular cluster (GC) NGC 3201 is under debate. Its retrograde orbit points to an extragalactic origin, butno further chemical evidence supports this idea. Light-element chemical abundances are useful to tag GCs and can be used to shedlight on this discussion.
Aims.
Recently it was shown that the CN and CH indices are useful to identify GCs that are anomalous to those typically found inthe Milky Way. A possible origin of anomalous clusters is the merger of two GCs and / or the nucleus of a dwarf galaxy. We aim toderive CN and CH band strengths for red giant stars in NGC3201 and compare these with photometric indices and high-resolutionspectroscopy and discuss in the context of GC chemical tagging. Methods.
We measure molecular band indices of S(3839) and G4300 for CN and CH, respectively from low-resolution spectra ofred giant stars. Gravity and temperature e ff ects are removed. Photometric indices are used to indicate further chemical information onC + N + O or s-process element abundances that are not derived from low-resolution spectra.
Results.
We found three groups in the CN-CH distribution. A main sequence ( S1 ), a secondary less-populated sequence ( S2 ), and agroup of peculiar ( pec ) CN-weak and CH-weak stars, one of which was previously known. The three groups seem to have di ff erentC + N + O and / or s-process element abundances, to be confirmed by high-resolution spectroscopy. These are typical characteristicsof anomalous GCs. The CN distribution of NGC 3201 is quadrimodal, which is more common in anomalous clusters. However,NGC 3201 does not belong to the trend of anomalous GCs in the mass-size relation. Conclusions.
The globular cluster NGC 3201 shows signs that it can be chemically tagged as anomalous: it has an unusual CN-CHrelation, indications that pec - S1 - S2 is an increasing sequence of C + N + O or s-process element abundances, and a multi-modal CNdistribution that seems to correlate with s-process element abundances. The non-anomalous characteristics are that it has a debatableFe-spread and it does not follow the trend of mass size of all anomalous clusters. Three scenarios are postulated here: (i) if thesequence pec-S1-S2 has increasing C + N + O and s-process element abundances, NGC 3201 would be the first anomalous GC outsideof the mass-size relation; (ii) if the abundances are almost constant, NGC 3201 would be the first non-anomalous GC with multipleCN-CH anti-correlation groups; or (iii) it would be the first anomalous GC without variations in C + N + O and s-process elementabundances. In all cases, the definition of anomalous clusters and the scenario in which they have an extragalactic origin must berevised.
Key words. (Galaxy:) globular clusters: individual: NGC 3201 – Stars: abundances – Galaxy: halo – Stars: Population II
1. Introduction
The paradigm that defines globular clusters (GCs) has slowlychanged since the first studies pointing to multiple popu-lations such as Cottrell & Da Costa (1981) and Norris et al.(1981). A substantial amount of work has been done based onhigh-resolution spectroscopic and photometric observations dur-ing the last two decades. It is well known that all globular ⋆ Observations done under programme 60.A-9501(B) at NTT / ESO,La Silla; and archival data from project 60.A-9700(D). ⋆⋆ Table 2 is also available in electronic form at the CDSvia anonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or viahttp: // cdsweb.u-strasbg.fr / cgi-bin / qcat?J / A + A / clusters present a star-to-star variation of light-element abun-dances, such as the Na-O anti-correlation (e.g. Gratton et al.2004, 2012) or the C-N anti-correlation (e.g. Cohen et al. 2002;Briley et al. 2004; Da Costa et al. 2004; Kayser et al. 2008;Pancino et al. 2010; Smolinski et al. 2011). The most plausi-ble scenario to explain the anti-correlations is self-enrichment(e.g. Prantzos & Charbonnel 2006), where the nature of the pol-luters is the current open question of this field. Asymptoticgiant branch (AGB) stellar ejecta (e.g. D’Ercole et al. 2008),fast-rotating massive stars (e.g. Decressin et al. 2007), and mas-sive binaries (e.g. de Mink et al. 2009) are among the possiblecandidates. In all cases, nucleosynthesis and stellar evolutionhave been discarded because un-evolved main sequence stars Article number, page 1 of 15 & Aproofs: manuscript no. dias_mar18_SAGA also present the anti-correlations above (e.g. Briley et al. 1994;Kayser et al. 2008; Pancino et al. 2010; Milone et al. 2013).Therefore the origin of the primordial chemical anomalies inGCs is environmental and still remains unknown (see e.g.Renzini et al. 2015). For an updated review on the recent sce-narios and comparisons with observational constraints we referto Bastian & Lardo (2018).A few GCs also present a star-to-star spread in metallicity,C + N + O, and s-process element abundances, where each grouphas a spread in p-capture element abundances, such as Na-Oand C-N anti-correlation, for all cases when abundances areavailable (more details in Da Costa 2015; Marino et al. 2015).They are called anomalous GCs and are usually associated toa peculiar formation and evolution history, possibly originat-ing in dwarf galaxies that were captured by the Milky Way(e.g. Da Costa 2015). They are important targets to be anal-ysed also in terms of light-element anti-correlations describedabove to characterise these clusters as typical globular clusters oranomalous. One of these targets is M 22 (Da Costa et al. 2009;Marino et al. 2009), although Mucciarelli et al. (2015b) arguedagainst a [Fe / H] spread for M 22. Marino et al. (2011) showedthat the CN index S(3839) traces [N / Fe] and the CH index G4300traces [C / Fe]; they split the stars into two groups in terms of s-process element abundances. The s-poor group has a bimodalC-N anti-correlation, as expected from CNO-cycle enrichment,and the s-rich group also has an anti-correlation but the sam-ple is small, therefore it is not possible to say whether the dis-tribution is bifurcated or not. Consequently M 22 presents twogroups with C-N anti-correlation and not a broad correlation asNorris & Freeman (1983) concluded (see also Lim et al. 2017).The anomalous cluster NGC 1851 is also potentially a resultof the merger of two globular clusters (Carretta et al. 2010). Asin the case of M 22, no CN-CH anti-correlation was found atfirst sight by Lardo et al. (2012). However, a more detailed anal-ysis revealed a strong anti-correlation of [C / Fe] and [N / Fe] byLardo et al. Another interesting result is that NGC 1851 does notpresent a bimodal distribution of CN no [N / Fe], but does presenta quadrimodal CN distribution as shown by Campbell et al.(2012). A closer look at the bimodal CN distribution of M 22also reveals a quadrimodal distribution (see Sect. 4.3.) A di ff er-ence between M 22 and NGC 1851 is that the latter presentsa di ff use stellar halo that could be a remnant of its host dwarfgalaxy after being captured by the Milky Way (Olszewski et al.2009; Bekki & Yong 2012). However, if M 22 had a stellar halo,it was possibly stripped o ff because the cluster is closer to theGalactic centre (Da Costa 2015). In conclusion, the formationscenario for both M 22 and NGC 1851 points to a merger ofglobular clusters, however the details may not be entirely thesame. A few other anomalous clusters can be added to this group:NGC 5286, M 2, NGC 5824, M 19, M 54, and M 75 (Da Costa2015; Marino et al. 2015).Another intriguing cluster is NGC 3201. AlthoughSimmerer et al. (2013) showed that this object is among theclusters with an intrinsic [Fe / H] spread, other works havestrongly suggested otherwise (Covey et al. 2003; Muñoz et al.2013; Mucciarelli et al. 2015a). Nevertheless, NGC 3201 hasa retrograde orbit (Gonzalez & Wallerstein 1998), which indi-cates a contentious history for this GC, with a possible extra-galactic origin, similar to the case of ω Cen (Bekki & Freeman2003). It is a good candidate to present a peculiar CN-CH re-lation as in the case of M 22. Smith & Norris (1982) founda bimodal distribution of CN for NGC 3201 red giant branch(RGB) stars, but not as marked as for other typical GCs, suchas NGC 6752 and M 4. They also found a mild CN-CH anti- correlation but with one star that was CN-weak and CH-weak,in agreement with Da Costa et al. (1981). A new analysis ofCN and CH for NGC 3201 in comparison with the latest high-resolution spectroscopy, photometric data, and CN and CH oftypical and anomalous GCs is needed. In this work we analyseCN and CH indices of RGB stars from NGC 3201, and discussthe origin of this GC as done for other anomalous GCs.The paper is organised as follows. In Sect. 2 we describe thephotometric and spectroscopic observations. The spectroscopicanalysis and calibrations are done in Sect. 3. The multiple stellargenerations of NGC 3201 are discussed in Sect. 4 and comparedto other anomalous clusters. Finally in Sect. 5 we discuss the ori-gin of NGC 3201. Summary and conclusions are given in Sect.6.
2. The data
The observations were carried out using the European South-ern Observatory (ESO) Faint Object Spectrograph and Camera(v.2), EFOSC2 (Snodgrass et al. 2008), mounted on the 3.6 mNew Technology Telescope (NTT) at the ESO-La Silla Obser-vatory, Chile. We used archival images in B and V filters centredat NGC3201 to select RGB stars and prepare the masks in thisregion. To obtain more RGB stars we observed additional point-ings to the north and to the south of the cluster; we observedalso in B and V filters and kept half of the cluster in the field ofview of 4 ′ x4 ′ (see Table 1). Data reduction was done using ESOpipeline esorex. To select RGB stars we carried out point spread function(PSF) photometry on B and V images using default procedureswith DAOPHOT at image reduction and analysis facility soft-ware (IRAF) (Stetson 1987). From the colour-magnitude dia-gram (CMD, see Fig. 1) we selected all RGB stars brighterthan V <
15 and identified them in the pre-image. We note thatthe broad RGB is due to di ff erential reddening (Kravtsov et al.2009; von Braun & Mateo 2001). We selected the best targetsthat would provide spectra with features of CN, CH, and Fe onthe detector without overlapping with neighbouring stars. In to-tal we selected 46 RGB stars in NGC3201. From Fig. 1b onestar is located at the AGB phase at about V ∼ ∼ = . ff er by about 0.2to 0.4 mag at the metallicity of NGC 3201 ([Fe / H] = -1.46, seee.g. Riello et al. 2003; Di Cecco et al. 2010). Therefore, it is fairto say that the RGB bump of NGC 3201 is at V ∼ . .Article number, page 2 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201? same photometric data. In panels c and d of Fig. 1 we show theluminosity function of all stars and RGB stars respectively, in-dicating the over densities of the horizontal branch and RGBbump, which agree with the expected values. B−V V (a) B−V(b) (c) all stars horizontal branch (d)RGB starsbump
Fig. 1.
Colour-magnitude diagram (CMD) of NGC 3201 and selectedstars for spectroscopic observations. (a)
Photometry from pre-imagesas described in the text. No calibration or de-reddening process wasdone, except by a zero point o ff set to match the calibrated CMD frompanel (b). Targets for spectroscopic observations were selected from thisCMD and are shown as orange circles. (b) Same as (a) but using pho-tometry from Kravtsov et al. (2009) that is calibrated and corrected bydi ff erential reddening. The selected targets are identified and reveal anoutlier from the RGB region that should be excluded from our analy-sis. Moreover, RGB stars are colour-coded. (c) Luminosity function ofall stars from (b) indicating the over density of the horizontal branch. (d)
Luminosity function only of RGB stars from (b) indicating the overdensity of the RGB bump.
After selecting the targets and preparing the masks, we per-formed the spectroscopic observations in Multi Object Spec-troscopy (MOS) mode of EFOSC2. For each pointing (centre,north, south) we took three observations of 20 minutes each toreach enough signal-to-noise (S / N) and to correct cosmic rays.We used grism ′′ , and length 8.6 ′′ , which means a spectralresolution of ∆ λ = ≈ ′′ , andwe chose binned readout mode 2x2 that makes the pixel scale be0.24 ′′ . Data reduction was done also using the ESO pipeline es-orex . Each spectra was normalised locally to measure each indexof CN and CH separately. For the CN index, we fitted a straightline to the two pseudo-continua defined by Pickles (1985), andfor CH we proceeded in the same way using the two pseudo-continua defined by Harbeck et al. (2003). Table 1 gives furtherinformation.
3. Spectroscopic analysis
From 48 spectra observed, we excluded six that did not haveenough coverage to measure a CN index. These include the AGBstar with V = Besançonmodel Member stars
CN−weak, Sequence 1 (S1)CN−strong, Sequence 1 (S1)CN−weak, Sequence 2 (S2)CN−strong, Sequence 2 (S2)CN−weak,CH−weak star, peculiar (pec) −100 0 100 200 300 400 500 600 v helio (km/s) F e4383 ( Å ) Fig. 2.
Heliocentric velocities versus Fe4383 index for all 28 valid spec-tra. Smoothed histograms show the distribution of the two parameters.We also show the velocities distribution from the Besançon model.Member stars are located in the lower right quadrant indicated by thedashed lines. Sequence 1 ( S1 ), sequence 2 ( S2 ), and peculiar stars ( pec. )defined in Fig. 5 are indicated by blue squares, orange diamonds, andgreen crosses, respectively. CN-strong ( δ S(3839) ≥
0) and CN-weakstars ( δ S(3839) <
0) are represented by filled and open symbols, re-spectively. These symbols and colours are used also in the followingfigures. Dashed lines are set by eye at the lower limit of v helio and theupper limit of Fe4383 distributions. lower S / N. From the 40 remaining spectra, 12 were excludedbecause they had S / N CN ≤
25 (i.e. S / N CH ≤ σ CN ≥ .
06 and σ CH ≥ .
03 as shown in Fig. B.1. Among themis the outlier at (V,B-V) = (15.1, 1.3). We ended up with a sampleof 28 valid spectra. The other three RGB tip stars are not biasedto any particular group in Fig. 2, two are S1 , and the other is pec .To ensure that we are studying stars only from NGC 3201,member stars were selected using the traditional plot of Fe abun-dance versus heliocentric velocity (Fig. 2). We derived radial ve-locities of the individual stars using the cross-correlation pythonpackage crosscorrRV. This package calculates the radial veloc-ity of stars shifting the rest-frame wavelength axis of a tem-plate. We used as template a synthetic spectrum (Coelho et al.2005 ) for a typical red giant star in NGC 3201 with parametersT e ff = g ) = / H] = − .
5, and [ α / Fe] = + .
4. Theresolution of the synthetic spectrum ( ∆ λ ≈ gauss to match the resolution of the stellar spectraof our sample ( ∆ λ ≈ helcorr . Errors were assumed to be the full width at half maxi-mum (FWHM) of the cross-correlation function, which is about http://pyastronomy.readthedocs.io/en/latest/ . http://specmodels.iag.usp.br/ . Article number, page 3 of 15 & Aproofs: manuscript no. dias_mar18_SAGA
Table 1.
Log of observations for pre-images and spectroscopy.Obs.Type Name RA DEC Obs. date Obs. time Filter / Grism Exp. time Airmass Seeing ( ′′ )IMA NGC3201 (centre) 10:17:37 -46:24:34 21-01-2011 06:46:20 B 20 1.05 0.78(archive) 06:41:10 V 10 1.05 0.87IMA NGC3201 (north) 10:17:36 -46:22:40 28-02-2016 05:15:25 B 20 1.06 0.6105:17:22 V 10 1.01 0.61NGC3201 (south) 10:17:36 -46:26:40 29-02-2016 05:21:17 B 20 1.06 0.5805:23:14 V 10 1.06 0.61MOS NGC3201 (centre) 10:17:36 -46:24:40 29-02-2016 03:33:00 Gr × Gr × Gr × Gr × / s for all spectra. Fe abundance is indicated by the Lick in-dex Fe4383 measured with the LECTOR code. We ran a Bensançon model (Robin et al. 2003) with a solidangle of 0.2 deg at the direction of NGC 3201 and selectingonly giant, bright giant, and supergiant stars from all Galacticcomponents with visual magnitude between 18 < V <
10. Thesimulation resulted in 309 stars with metallicities higher than[Fe / H] & -1.5, which is the metallicity of NGC 3201 (Dias et al.2016b) ; only three stars had a metallicity below that. Radialvelocities are indicated in Fig. 2 and show no overlap. There-fore, field stars would have v helio <
320 km / s and higher metal-licities, that is they would be in the upper left quadrant, wherethere are no stars. In other words, all 28 valid stars were selectedas members. The average velocity of the member stars is v helio = ±
63 km / s, which is compatible with 494km / s from Harris(1996, 2010 edition). We adopt here the modified index definition of CN (S3839)and CH (G4300) by Harbeck et al. (2003) to have a homoge-neous analysis and compare results with those from Kayser et al.(2008) who used Harbeck’s definition. The indices defined byHarbeck et al. (2003) have slightly di ff erent spectral regionswith respect to the classical ones defined by Norris & Smith(1981) in order to avoid strong hydrogen lines nearby in mainsequence stellar spectra. This was not an issue for the RGB starsobserved in the 1980s. The di ff erences in the indices using dif-ferent definitions are briefly discussed in Sect. 4.3.Indices from Equations 1 and 2 were measured using an Rcode written by B. Dias (see Fig. 3) resulting inS(3839) = − . · log R F λ d λ R F λ d λ (1)G4300 = − . · log R F λ d λ . · R F λ d λ + . · R F λ d λ . (2)Uncertainties are discussed in Appendix B. model.obs-besancon.fr (R Core Team 2015) CN Ca H+K no r m . f l u x CH a r b i t r a r y f l u x l (A) Fig. 3.
Definition of CN and CH indices S(3839) and G4300 from Eqs.1 and 2 using spectra of two NGC 3201 stars. Thin lines are fromthe CN-weak, CH-strong star N3201centre_08, thick lines are from theCN-strong, CH-weak star N3201centre_11, both S1 stars with V ≈ Both S(3839) and G4300 indices are sensitive to surface gravityand e ff ective temperature (see Norris & Smith 1981). Remov-ing this dependency is a sine qua non condition for detectingtheir sensitivity to nitrogen and carbon. The typical proxies usedto correct the indices are colour or magnitude (Norris & Smith1981; Harbeck et al. 2003; Kayser et al. 2008; Campbell et al.2012). We adopt here a similar method to Pancino et al. (2010)who traced a ridge line in the distribution of index versus mag-nitude. Instead of a ridge line we fitted a second-order polyno-mial to this distribution; Figure 4 displays the polynomial fit tothe data. The di ff erence between the index and the polynomialfor a given magnitude defines the excess indices δ S(3839) and δ G4300 that will be used in this paper from now on. We notethat the spread seen in Fig. 4 is real because all member starshave σ CN < .
06 and σ CH < .
03. The parameters for all mem-ber stars are given in Table 2.
Article number, page 4 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201?
Table 2.
Final parameters for all 28 good quality member stars of NGC 3201
Star RA DEC V B-V v helio
Fe4383 S / N CN S / N CH CN = S(3839) CH = G4300 δ CN δ CHhh:mm:ss.sss dd:mm:ss.sss mag mag km / s Å mag mag mag mag(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)N3201centre_01 10:17:49.057 -46:24:58.572 13.823 1.186 375 2.745 60.37 129.7 -0.314 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Notes. ( ) Star ID from our three masks. ( ) Coordinates of each star, equinox J2000.0. ( ) Magnitude and colour from Kravtsov et al. (2009)calibrated and corrected by di ff erential reddening. ( ) Heliocentric velocities. Errors are about 5 km / s, assumed as the FWHM of the cross-correlation function. ( ) Lick index of iron used as proxy for metallicity. ( ) Signal-to-noise ratio at the wavelength of CN and CH indices. ( ) CN and CH indices as defined in Equations 1 and 2. ( ) CN and CH indices corrected by surface gravity e ff ects as discussed in Sect. 3.3. −0.4 −0.2 0.0 . . . . . V S(3839)
G4300NGC3201 − this work
Fig. 4.
Correction of S(3839) and G4300 for stellar surface temperatureand gravity for 28 member and good quality (S / N CN ≥
25) RGB stars ofNGC 3201. We fitted a second-order polynomial to each dataset, whichis represented by the red dashed lines. The di ff erence between the in-dices and the fitted line for a given magnitude defines δ S(3839) and δ G4300. Symbols are the same as in Fig.2. If the peculiar stars indi-cated by the crosses are not considered, two groups of stars with anti-correlated CN and CH can be identified: S1 , indicated by squares, and S2 , indicated by diamonds (see also Fig. 5 and discussions). Error barsare of the order of the point size, and are thus omitted in the plots.
4. Multiple stellar generations in NGC 3201
We show in Fig. 5d the anti-correlation between the corrected in-dices δ S(3839) and δ G4300 for the 28 good quality member starsof NGC 3201. We split them into two CN-CH anti-correlationsequences, namely S1 and S2, and a group of CN-weak, CH-weak peculiar ( pec ) stars. We fitted a second-order polynomialto S1 and shifted the curve to match S2 and peculiars. The sep-aration between the sequences is ∆ δ G4300 = σ distance. Therefore the separation of our sampleinto these groups is significative and we adopt the same pointcolours and shapes defined in Fig. 5d to the other plots in this pa-per. One peculiar star was already indicated by Smith & Norris(1982) and Da Costa et al. (1981) and we now increase this sam-ple to four peculiar stars. The other panels of Fig. 5 show thatstars from di ff erent groups are indistinguishable in the Na-Oanti-correlation, at least with the limited sample we have. Theusual correlations Na-CN and O-CH are also shown, where first(1G, Na-poor, CN-weak) and second generation (2G, Na-rich,CN-strong) stars can be identified.The globular cluster M 22 also has two sequences of C-Nanti-correlation and the C-rich sequence is less populous thanthe C-poor sequence as in the case of S2 in comparison with S1 for NGC 3201. In the case of M 22, s-process element abun-dances were available and the conclusion of Marino et al. (2011)was that the C-poor / N-poor sequence (equivalent to S1 ) is s-poorand Fe-poor, while the C-rich / N-rich sequence (equivalent to S2 )is s-rich and Fe-rich. Oxygen is the same for the two groups, Article number, page 5 of 15 & Aproofs: manuscript no. dias_mar18_SAGA −0.20.00.20.40.60.8 [ N a / F e ] (a) −0.8 −0.4 0.0 0.4−0.10−0.050.000.050.10 [O/Fe] d G (c) −0.3 −0.1 0.1 0.3 d S(3839) (b) −0.1 0.0 0.1 −0.20.00.2 d G4300 d S ( ) (d) pec.S1 S2 Fig. 5. (a)
Na-O anti-correlation for all globular clusters fromCarretta et al. (2009) in grey dots superimposed by the 11 NGC 3201stars in common with our sample with available [Na / Fe] and [O / Fe];symbols are the same as in Fig. 2. Smooth spline function is shown asred dashed line to highlight the anti-correlation in our data. Peculiarstars indicated by green crosses are not considered in the fitted functionbased on the selection of panel (d). (b)
Correlation between sodium andcyanogen, which is a proxy for nitrogen abundances for the 18 starsin common with available [Na / Fe]. A straight line is fitted to S1 stars. (c) Same as (b) but for CH and oxygen for stars in common with avail-able [O / Fe]. The fitted line is shifted by 0.1 mag in δ G4300 to match theonly S2 star. (d) Distribution of CN and CH indices for all 28 NGC 3201good quality member stars, revealing three groups of stars: sequence 1( S1 ), sequence 2 ( S2 ), and peculiar ( pec ). A second-order polynomialwas fitted to S1 stars and the same curve was shifted by 0.1 mag in δ G4300 as in panel (c) to match the S2 stars, and by − but C + N + O is higher for the s-rich stars. High-resolution spec-troscopy is needed to confirm whether NGC 3201 has an increas-ing C + N + O and s-process element abundances in the sequence
S1-S2 as well. Gonzalez & Wallerstein (1998) found hints thatO-rich NGC 3201 stars are depleted in [Ba / Eu]. In Fig. 5 weshow that O-rich stars are C-rich, which would mean that thesequence pec - S1 - S2 has decreasing s-element abundances, notincreasing as is the case for M 22. However, 11 out of the 13stars studied by Gonzalez & Wallerstein have constant [Ba / Eu]abundance ratios, and only two O-rich stars are depleted in Ba.A larger sample, preferably including the pec - S1 - S2 stars, isneeded in order to derive abundances of C, N, O, and s- / r-processelements.Lim et al. (2017) analysed the double CN-CH anti-correlation of M 22 in a di ff erent way. They argued that thesestars show a positive CN-CH correlation, which is also the casefor other anomalous clusters such as NGC 1851, NGC 6273, andNGC 5286. They showed that the CN-strong stars are s-rich, andCN-weak stars are s-poor. NGC 288 was used as reference andit shows only a single anti-correlation. If the CN-CH relation ofNGC 3201 stars is interpreted following this line, it is another indication that the sequence pec - S1 - S2 may have increasing s-element abundances and possibly increasing metallicities.The CN-CH anti-correlation can be explained by CN-cycleprocessing. Some clusters show that C + N increases with de-creasing C (e.g. M 5; Cohen et al. 2002) and ON-cycle processedmaterial is required to explain that. During the first dredge-up,the convective envelope of an RGB star can move material pro-cessed during the CNO cycle up to the atmosphere. However,for metal-poor stars, the convective envelope does not go deepenough to reach the H-burning shell to mix enhanced (or de-pleted) CNO-cycle processed elements. NGC 3201 is relativelymetal-poor, but S1 , S2 , and peculiar stars have di ff erent CHstrength (proxy for C) and we can only speculate that these threegroups may have di ff erent C + N abundances. pec - S1 - S2 stars Monelli et al. (2013) defined a colour index based on Johnsonfilters U, B, I as given by the equationc U , B , I = (U − B) − (B − I) . (3)They found that the V-c U , B , I diagram reveals a split RGB for allglobular clusters analysed, after di ff erential reddening correc-tion. The multiple branches correlate with the chemical abun-dances of light-elements O, Na, C, N, and Al. −4 −2 0 2 4−4−2024 relative a (cid:215) cos d (’) r e l a t i v e d ( ’ ) Fig. 6.
Spatial distribution of NGC 3201 stars from Kravtsov et al.(2009) in grey dots with a greyscale density pattern. Spectroscopic tar-gets are shown with the same symbols as in Fig. 2 for the 28 goodquality member stars. Dashed outer circle shows the half-light radius(r h = . ′ ), the internal dashed red circle is the core radius (r c = . ′ ),and the blue dashed square delimits the region of the HST observationsused here. We adopted the available UBVI photometry corrected by dif-ferential reddening by Kravtsov et al. (2009) to produce a V-c U , B , I diagram for NGC 3201. In Fig. 6 we show the sky distribu-tion of the stars from Kravtsov et al. (2009) and identify the coreradius and half-light radius, as well as our 28 targets. We plottedthe pseudo-CMD, aka V-c U , B , I diagram, using stars from Fig. 6 For a recent review, see Gratton et al. (2004) and references therein.Article number, page 6 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201? and displayed them on Fig. 7. We also fitted a second-order poly-nomial to all RGB stars within a box defined by 12 < V <
16 magand − . < c U , B , I < − . δ S(3839), we define δ c U , B , I as thedi ff erence between c U , B , I and the fitted line for a given V magni-tude.The corrected spectroscopic index δ S(3839) is plottedagainst the corrected photometric index δ c U , B , I on the upper pan-els of Fig. 7. Stars fainter than V >
14 mag have larger photo-metric uncertainties and generate dispersion on the plot. If onlybright stars are plotted, a correlation between these indices isclearly seen, as expected, and 1G and 2G stars are well sep-arated. However, pec - S1 - S2 stars are all mixed. In conclusion, δ c U , B , I cannot split the three groups. −2.0 −1.9 −1.8 −1.7 −1.6171615141312 c U,B,I (mag) V ( m ag ) −0.3−0.2−0.10.00.10.20.3 d S ( S ) All stars −0.3−0.2−0.10.00.10.20.3 −0.10 −0.05 0.00 0.05 d c U,B,I (mag) d S ( S ) Stars with V < 14
Fig. 7. Bottom panel:
V-c U , B , I diagram for NGC 3201 using photometryfrom Kravtsov et al. (2009) corrected by di ff erential reddening but with-out decontamination from field stars. We selected a rectangle around theRGB stars and fitted a second-order polynomial shown by a dashed redline. This line is used to define δ c U , B , I as the di ff erence between c U , B , I and the fitted line for a given V magnitude for the 28 stars analysed here.Symbols are the same as in Fig. 2. Average error bars are indicated atthe upper right corner. Middle panel:
Correlation between δ S(3839)and δ c U , B , I for all 28 stars. A linear fit to the S1 stars is shown by a reddashed line. Upper panel:
Same as middle panel but only for the 14stars brighter than V <
14 mag. Average error bars are indicated at theupper left corner.
UV filters are also useful to split the RGB of globular clus-ters. The Hubble Space Telescope (HST) Large Legacy Trea-sury Program (Piotto et al. 2015) had its first public data re-lease with the homogeneous photometric catalogues published recently (Soto et al. 2017). This release is preliminary, there-fore no di ff erential reddening correction was taken into account,which could cause some dispersion on the photometric indicesbut not as much as ground-based photometry. This team definedthe colour indexc F275W , F336W , F438W = ( m F275W − m F336W ) − ( m F336W − m F438W ) (4)that is very sensitive to nitrogen abundances and is useful to dis-entangle 1G and 2G RGB stars in globular clusters.The lower left panel of Fig. 7 shows the V- δ c F275W , F336W , F438W diagram where the S2 CN-strong starsfall in between the CN-strong and CN-weak stars of S1 , whichis a possible explanation for why the RGB is not clearly bimodalbut rather has a smooth transition. If S2 stars are indeed s-richand S1 s-poor, the RGB in this CMD would have s-poor starsin the left branch and mixed s-poor and s-rich stars towardsthe right branch. We notice that the S1 -1G star to the right islikely to be there because it has δ S(3839) very close to zeroand is therefore more sensitive to our definition of CN-strongand CN-weak with a cut at δ S(3839) =
0. This CMD seemsuseful to split S1 -2G from the others. As done for c U , B , I we alsofitted a line to the RGB stars and defined the di ff erential index δ c F275W , F336W , F438W as the di ff erence between c F275W , F336W , F438W and the fitted line for a given magnitude. The correlationbetween δ c F275W , F336W , F438W and δ S(3839) on the upper leftpanel is clear, as expected. It splits 1G and 2G stars but mixes S1 and S2 stars. Interestingly, peculiar stars seem to have lower δ c F275W , F336W , F438W values for the same δ S(3839) as the 1Gstars. This correlation seems useful to find peculiar stars.Milone et al. (2017) defined the so-called ‘chromosomemap’ that consists of a plot of the pseudo colour index δ c F275W , F336W , F438W versus the colour index δ c F275W , F814W . On thefirst data release of the HST UV treasury, the F814W magnitudewas not yet available, but the authors provide an I magnitudewhich is very similar. We replaced m F814W by I and producedour version of the ‘chromosome map’ as shown on the upperright panel of Fig. 8. The regions of 1G and 2G stars are clearlyidentified and split by a line inclined by θ = ◦ as defined byMilone et al.. Stars of type S1 follow the 1G-2G split (with theoutlier being explained above). Stars of type S2 are not consis-tent with an additional 1G or 2G sequence shifted from the main1G or 2G regions, as is the case for other anomalous clusters(type II in the nomenclature of Milone et al.) such as M 22 andNGC 1851. In fact, Milone et al. classified NGC 3201 as type I.In conclusion, the chromosome map is not able to split pec - S1 - S2 stars.A third photometric index was defined by Marino et al.(2015) based on the Johnson filters B, V, I as given byc B , V , I = (B − V) − (V − I) . (5)These authors say that c B , V , I is not very sensitive to Na or N, butit was useful to separate s-rich and s-poor stars on NGC 5286.Although s-process element abundances do not directly a ff ectbroad-band filters, at least indirectly this index was able to splitwell the two groups of stars. They further claim that C + N + Oabundances may also a ff ect this index. In fact, s-rich stars arealso rich in C + N + O in anomalous clusters. We use again thephotometry from Kravtsov et al. (2009) to compare the photo-metric index with our spectroscopic measurements.We show in Fig. 9 a similar CMD to the one done in Fig. 7but now with the colour defined by Eq. 5. We defined the dif-ferential index δ c B , V , I in a similar way to that for δ c U , B , I . Only Article number, page 7 of 15 & Aproofs: manuscript no. dias_mar18_SAGA c F275W,F336W,F438W (mag) V ( m ag ) −0.2−0.10.00.10.2 −0.2 −0.1 0.0 0.1 0.2 d c F275W,F336W,F438W d S ( ) F275W − I (mag) I ( m ag ) −0.4 −0.2 0.0 0.2 0.4 0.20.10.0−0.1−0.2 d c F275W,I d c F W , F W , F W Fig. 8.
Similar to Fig. 7 but using HST UV photometry from Soto et al.(2017), available only for 12 out of the 19 good quality member starswithin the HST field of view (See Fig. 6). The bottom panels show theCMDs with our spectroscopic targets identified using the same symbolsas in Fig. 2 . The index δ c F275W , F336W , F438W is plotted against δ S(3839) onthe upper left panel. A second-order polynomial was fitted to S1 stars,and it is shown as red dashed line. The index δ c F275W , F336W , F438W is plot-ted also against δ c F275W , I on the upper right panel, which is a versionof the so-called ‘chromosome map’ from Milone et al. (2017) that dis-tinguishes 1G and 2G stars, split by a line inclined by θ = ◦ . Ourspectroscopic targets are identified on the map. stars brighter than V <
14 are analysed as before. All S2 starsare to the right of the fitted line, the pec are to the left, and S1 stars are spread around the curve. Kernel density estimations(KDEs) were produced for these stars separated into pec - S1 - S2 ,and bandwidth as the average uncertainty. Although the KDEsare broad, it is possible to identify that the peaks follow the se-quence pec - S1 - S2 in terms of increasing δ c B , V , I . In conclusion, δ c B , V , I is able to split pec - S1 - S2 NGC 3201 stars, even if not asclearly as for NGC 5286. If this index is able to split groups ofstars with di ff erent C + N + O abundance and possibly s-processelement abundances, NGC 3201 could join the group of anoma-lous clusters.
Campbell et al. (2012) revealed a quadrimodal CN distributionfor the anomalous cluster NGC 1851. It is known that themore metal-rich the cluster the larger range its CN distributionhas (e.g. Kayser et al. 2008; Schiavon et al. 2017; Milone et al.2017), therefore Campbell et al. compared NGC 1851 withNGC 288 that has a similar metallicity ([Fe / H] = − . , Dias et al. 2016b). They concluded that both clusters indeedhave similar ranges on the CN distribution, but NGC 288, takenas a typical globular cluster, has a clearly bimodal distribution,while NGC 1851 has a quadrimodal distribution. Moreover, theyshowed that the two CN-strong peaks are also Ba-rich, and theother two peaks are Ba-poor. We discuss whether a quadrimodalCN distribution can be another indicator of anomalous clusters. −0.30 −0.25 −0.20 −0.15 −0.10171615141312 c B,V,I (mag) V ( m ag ) pec.S1 −0.10 −0.06 −0.02 0.02 0.06 0.10 S2 d c B,V,I (mag)
Fig. 9. Bottom panel:
Same as Fig. 7 but for the index c B , V , I definedin Eq. 5. Average error bars are indicated at the upper left corner. Toppanels:
The KDEs of δ c B , V , I are shown for each group of stars: S1 , S2 ,and peculiar only for stars brighter than V <
14. The KDE bandwidthis the average uncertainty of 0.03. Peak values are indicated by verticaldashed lines.
In order to compare NGC 3201 with other clusters, we pro-duced the KDE for each cluster shown in Fig. 10 using the in-dices from the RGB stars of Kayser et al. (2008) corrected fol-lowing Appendix A. The KDE smoothing bandwidth was con-sidered as the average index uncertainty of each cluster. For eachdistribution we fitted the minimum number of Gaussian func-tions that was needed to converge the fit using non-linear leastsquares. They are shown in the figure and represent well theKDEs. The sigma of each Gaussian was forced to be similar tothe average index uncertainty in each case. A possible check ofthe significance of the multi-gaussian fit is the number of degreesof freedom. Each Gaussian has two variables to be fit, thereforeat least three points per Gaussian are needed in order to give flex-ibility for the fit to proceed. For NGC 288, two Gaussians werefitted to 16 points, which is more than six, hence the result doesnot over-fit the KDE. The same applies for all other clusters.The CN distribution for all 28 good quality member RGBstars from our NGC 3201 sample shows four peaks (see Fig.10). This is not similar to the bimodal CN distribution of typical R script adapted from http://research.stowers.org/mcm/efg/R/Statistics/MixturesOfDistributions/index.htm .Article number, page 8 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201?
NGC288(Milky Way GC)[Fe/H] = −1.3(16 stars) NGC362(Milky Way GC)[Fe/H] = −1.3(20 stars) M22(anomalous GC)[Fe/H] = −1.9(16 stars) NGC3201(all)[Fe/H] = −1.5(28 stars) NGC3201(peculiar)[Fe/H] = −1.5(4 stars) NGC3201(S1)[Fe/H] = −1.5(19 stars)−0.6 −0.4 −0.2 0.0 0.2 0.4 0.6NGC3201(S2)[Fe/H] = −1.5(5 stars) d S(3839)
Fig. 10.
KDE of the excess CN-index δ S(3839) for NGC 3201 in com-parison with two normal Galactic GCs (NGC 288, NGC 362) and oneanomalous (M 22), as indicated in the panels. Data for the referenceclusters were taken from Kayser et al. (2008) and put in a standardscale. The rug plot indicates the positions of all stars in each panel.The KDE bandwidth is the average uncertainty of each dataset, i.e., σ NGC288 = . σ NGC362 = . σ M22 = . σ NGC3201 = . pec and S2 stars, as discussed in the text. We also indicate cluster metallici-ties in the scale of Dias et al. (2016a). Milky Way GCs NGC 288 and NGC 362 , even though it coversthe same range and the stars have similar metallicities. We alsocompare with the CN distribution of the anomalous cluster M 22,that presents two peaks with an uneven distribution. Although itdoes not have four peaks like NGC 1851, the CN-strong starsare s-rich and CN-weak are s-poor (e.g. Marino et al. 2011). Thethree bottom panels show the KDEs of our sample divided into pec , S1 , and S2 . We argued before about the possibility of anincreasing s-process element abundance from pec to S2 . Herewe show that, in fact, pec stars are all CN-weak and S2 stars arepredominantly CN-strong. Stars of type S1 cover the full range.We also checked the stability of the δ S(3839) distributionshape considering another definition for this index applied bySmith & Norris (1982) that uses slightly di ff erent wavelengthlimits. They analysed the NGC 3201 CN distribution for RGBstars brighter than V <
14 mag and only had one peculiar star intheir sample. They found a bimodal distribution with no sharpseparation between the peaks. We applied the same strategyas above and scaled S(3839) to δ S(3839) from Smith & Norris(1982) in the same way we did for our data and for Kayser et al.(Appendix A). The resulting curves are shown in Fig. 11 assum-ing a constant error bar of 0.05 dex as the KDE bandwidth. Wereproduce the four bottom panels of Fig. 10 on the four bottompanels of Fig. 11, but now using the same S(3839) definition asin Smith & Norris for a fair comparison. Their KDE presents See also Milone et al. (2017) who classified NGC 362 as type II witha small extra group of 2G stars, di ffi cult to detect here with a sample of20 stars. NGC1851(Campbell et al. 2012)[Fe/H] = −1.2(17 stars) M22(Norris & Freeman 1983)[Fe/H] = −1.9(100 stars) NGC3201(Smith & Norris 1982)[Fe/H] = −1.5(43 stars) NGC3201(all)[Fe/H] = −1.5(28 stars) NGC3201 (peculiar) [Fe/H] = −1.5(4 stars) NGC3201 (S1) [Fe/H] = −1.5(19 stars)−0.6 −0.4 −0.2 0.0 0.2 0.4 0.6NGC3201 (S2) [Fe/H] = −1.5(5 stars) d S(3839)
Fig. 11.
Same as Fig. 10 but with S(3839) measured following theclassical definition used by Campbell et al. (2012), Norris & Freeman(1983), and Smith & Norris (1982) for a fair comparison with their re-sults on NGC 1851, M 22, and NGC 3201, respectively.The KDE band-with is the average uncertainty of each dataset, i.e., σ NGC1851 = . σ M22 = .
07, and σ NGC3201 = .
05, respectively. The KDE bandwith forour dataset is 0.031. δ S(3839) was calculated for all samples followingthe same strategy as done in Fig. 4. three peaks while ours have five. This di ff erence is probably be-cause we have less stars and smaller error bars, therefore ourresults are more sensitive to individual isolated points. Neverthe-less both samples cover the same CN range and have a dominant1G population and a multi-peak 2G population. If our curves forNGC 3201 are compared between Figs. 10 and 11, both presenta dominant 1G peak and a 2G multi-peak population. The sepa-ration in pec. , S1 , and S2 is also similar.Norris & Freeman (1983) observed 100 stars in M 22 and weshow their CN distribution in Fig. 11. It is comparable to the dis-tribution shown in Fig. 10 with data from Kayser et al. (2008).The main di ff erence is that Norris & Freeman have many morestars and their data reveal two additional extreme peaks withCN-stronger and CN-weaker stars. In other words, M 22 alsohas a quadrimodal CN distribution with a division of s-rich ands-poor stars as in the case of NGC 1851. In the same Fig. 11we also show the results of NGC 1851 analysed by Campbellet al. and scaled in the same way as in Fig. 4 for consistency.The quadrimodal distribution found by Campbell et al. (2012) isstill present in this plot if the right peak is ignored because it hasonly one star, and if the two central peaks are considered as onepeak. In fact, the peaks found for NGC 1851 in the original pa-per do not necessarily follow a Gaussian shape. This complexityresembles that of NGC 3201 from our data.The conclusion is that the shape of the distribution is sensi-tive to the definition of the index S(3839), sample size, and un-certainties. Not surprisingly, the clearly bimodal distribution ofNGC 362 from Fig. 10 does not look as smooth in Smith (1983)and Norris (1987). Nevertheless, both concluded that this clus-ter has a bimodal CN distribution. The complexity may come Article number, page 9 of 15 & Aproofs: manuscript no. dias_mar18_SAGA from the fact that photometric indices were used instead of spec-troscopic ones. More recently, Milone et al. (2017) have shownthat the pair of clusters NGC 288 and NGC 362 do have a verysimilar chromosome map, but the latter reveals a small popu-lation of redder RGB stars, which made this cluster a type-IIGC according to their classification. In other words, the CN dis-tribution alone cannot be used to split groups of clusters, butit is certainly a valid piece of information in the vast parame-ter space that is required to disentangle the multiple populationswithin a GC and eventually correlate to global properties andenvironmental e ff ects. In fact, the majority of the clusters anal-ysed by Norris (1987) with CN distributions, by Carretta et al.(2009) with Na-O anti-correlations, and by Milone et al. (2017)with photometric chromosome maps show two main populationsof stars. Any cluster that di ff ers from that behaviour requires spe-cial treatment. We have shown in Figs. 10 and 11 that the clustersM 22, NGC 1851, and NGC 3201 do present an odd CN distri-bution. S-process element abundances, such as Y, Zr, Ba, La, andNd in contrast with r-process Eu, and also C + N + O abundancesare needed in order to split these sub-populations.
5. Galactic or extragalactic?
Anomalous clusters like M 22, NGC 1851, M 2, and NGC 5286all present Fe / s-rich and Fe / s-poor stars, even though the Fe-spread is still under discussion for some of them. In all caseswhere abundances are available, s-rich and s-poor stars alsopresent di ff erent abundances of C + N + O. These groups corre-late well with an SGB and RGB split if appropriate coloursare used in the CMD. Each group has its own Na-O and C-N anti-correlations that are typical signatures of GCs. Whilethe explanation for the anti-correlations seems to be related toself-pollution of second generation stars by the primordial pop-ulation, the split into Fe / s-rich and Fe / s-poor stars for a fewanomalous clusters is not explained by the same mechanisms.Bekki & Yong (2012) proposed that such clusters could be theresult of a merger of two clusters. This is likely to happen inthe nucleus of dwarf galaxies where relative velocities and thevolume are smaller than that of the Milky Way halo. Shouldthis scenario be true, then the anomalous clusters would havean extragalactic origin from dwarf galaxies captured by theMilky Way (but see also Marino et al. 2015; Da Costa 2015;Bastian & Lardo 2018). We use the expression ‘extragalacticchemical tagging’ to refer to this small group of anomalous clus-ters, although the accuracy of this term is still debatable and ispart of a complex topic of study that has been developed in thelast decade. In this context, we try to tag NGC 3201 as a anoma-lous GC belonging to this group or not.In the previous section we showed that NGC 3201 presentsthree groups on the CN-CH correlation plot, namely pec , S1 ,and S2 , split by 7 σ in CH. An indication that these groups havedi ff erent abundances of C + N + O and s-process elements was re-vealed by photometric indices. Another mild constraint is theCN distribution that seems to be multi-modal for anomalousclusters and this is also the case for NGC 3201. These threepoints indicate that NGC 3201 may be included in this selectedgroup of globular clusters, with the note that it is possibly a tran-sient object between typical and anomalous GCs, because thefeatures listed above are not extreme for NGC 3201. In fact,even the star-to-star Fe spread has been discussed in the liter-ature (see Simmerer et al. 2013; Covey et al. 2003; Muñoz et al.2013; Mucciarelli et al. 2015a). We further note that Da Costa(2015) included M 22 to this group because it has a spread in s-process element abundances, in spite of its uncertain Fe-spread. NGC 3201 provokes a similar discussion on the Fe-spread butDa Costa did not consider this cluster anomalous because no de-tailed study on C + N + O and s-process elements was availableyet. We have presented some indication from photometry in thisdirection to be confirmed by high-resolution spectroscopy. versus environmentaleffects
Environmental e ff ects such as GC merger and extragalactic ori-gin are not the only possible explanation for the more com-plex GCs with unusual multiple populations. One example isNGC 2808 that presents three to five populations as shownby di ff erent analyses (e.g. Carretta 2015; Marino et al. 2017;Milone et al. 2017, and others). Stars in NGC 2808 all havethe same metallicity as opposed to M 22 and NGC 1851.The five populations of NGC 2808 are characterised by largevariations in light-element abundances, such as Na, O, Mg,and Al. Moreover, it was found that He abundance variationsplays an important role in characterising NGC 2808 popula-tions (e.g. Bragaglia et al. 2010; Marino et al. 2017). For refer-ence, we note that NGC 6121 (M 4) has a similar metallicityof [Fe / H] ≈ -1.1, but shows a bimodal distribution of Na and CN(Marino et al. 2008).A comparison cluster for NGC 3201 with similar metal-licity is NGC 6752, which presents a bimodal CN distri-bution (Norris et al. 1981). However, high-resolution multi-band photometry and spectroscopy revealed a third populationfor NGC 6752 (e.g. Milone et al. 2013; Nardiello et al. 2015;Gruyters et al. 2014). The explanation is that the three popula-tions have di ff erent light-element abundance ratios and also he-lium abundances (Milone et al. 2013; Nardiello et al. 2015). Infact, NGC 6752 has a complex horizontal branch morphologythat may be related to He abundances (Momany et al. 2002).Therefore He, and proton-capture element abundance varia-tions could be enough to explain multiple populations, howeverheavier elements such as Fe and neutron-capture (and the sumC + N + O) tend to vary within a GC only for anomalous casessuch as M 22 and NGC 1851 discussed above. Yong et al. (2013)found correlations of s-process elements with Na (that corre-lates with CN) for NGC 6752, although very small variationsof s-process element abundances were measured. No bimodal-ity or clear separation between s-rich and s-poor stars was re-ported (as is the case for M22 and NGC1851). Yong et al. alsospeculated that this should be the case for all GCs if they un-derwent extremely accurate spectroscopic analysis. Yong et al.(2015) found no spread on the C + N + O for NGC 6752. In sum-mary, NGC 6752 does not have two populations of s-rich (highC + N + O) and s-poor (low C + N + O), as is the case for M22 andNGC1851 (see also Carretta et al. 2005). The three populationsfound for NGC 6752 with photometry and spectroscopy are re-lated to light-element and helium abundances. NGC 3201 has asimilar metallicity and mass to NGC 6752 but di ff ers in manyother aspects: (i) it does not show a bimodal CN distribution,but a complex KDE resembling NGC 1851; (ii) CN-CH anti-correlation reveal two trends similar to the findings for M 22,as well as a group of peculiar stars; (iii) it has a much redderhorizontal branch (HB) morphology that covers red and bluecolours similar to but not as separated as the HB of NGC 1851 (Mackey & van den Bergh 2005). The di ff erence in HB index isconsistent with NGC 3201 being 1 Gyr younger than NGC 6752(Rey et al. 2001; VandenBerg et al. 2013; Dias et al. 2016a). Snapshots of CMDs can be retrieved from here: http://groups.dfa.unipd.it/ESPG/ground.html .Article number, page 10 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201?
In conclusion, if the three groups we found for NGC 3201on our CN-CH analysis have di ff erent s-process-element andC + N + O abundances, we may consider it a study case togetherwith the other anomalous GCs in view of the formation sce-nario of the merger of GCs in the nuclei of dwarf galaxies. Ifit only possesses a very small variation of s-process-elementand C + N + O abundances, and a possible mild trend or corre-lation of Na(CN) with s-process-element abundances (as forNGC 6752; Yong et al. 2015), then it is either a typical GC withan odd CN-CH relation or an anomalous GC without C + N + Oand s-element abundance variations. Another possibility is thatNGC 3201 could be a transition cluster in the sequence of CN-CH anti-correlation to CN-CH correlation shown by Lim et al.(2017).High-resolution spectroscopic analysis is needed at least tomeasure s-process-element and C + N + O abundances of the 28NGC 3201 stars analysed here. Another analysis that may help isa second-parameter analysis between NGC 6752 and NGC 3201to check whether only metallicity and age (plus the complex Hecontents of the former) are enough to explain the HB morphol-ogy di ff erences between the clusters or if NGC 3201 needs tohave He abundance variations to produce such HB. Anomalous clusters are among the most luminous in the MilkyWay. Those with M V . − . & × M ⊙ ) account for41 out of 157 from the catalogue of Harris (1996, 2010 edition).Da Costa (2015) said that 13 of those are unstudied or poorlystudied. In other words, if nine anomalous massive clusters werefound among the 28 well-studied massive clusters, we can ex-trapolate this ratio of 32% to the total sample from the Harriscatalogue and predict that about 13 massive clusters should beanomalous in the Milky Way. This is already about 50% morethan the eight clusters predicted by Da Costa. Instead, if weconsider clusters of all masses, the first approximation is to as-sume that the 70% of the catalogued clusters that have metal-licity measured spectroscopically (Dias et al. 2016b) also havedetailed enough studies to detect Fe-spread. In this case 8% ofall clusters should be anomalous, which leads us again to a to-tal of about 13 clusters. Assuming that for half of these clus-ters there was minimal spectroscopic information, then the num-ber of anomalous GC could be doubled to 26 of all masses.Some e ff ort has been done to find more candidates to host Fe-spread (e.g. Saviane et al. 2012). We are still far from havinga complete view and a explanation for these peculiar clusters(Bastian & Lardo 2018).We compare the structural parameters of NGC 3201 with theanomalous and to the non-peculiar massive GC NGC 6752 inFig. 12. Mass and size were taken from Norris et al. (2014), andwhen not available for the GC we interpolated from the fittedrelation between M V from Harris (1996, edition 2010) and themasses derived by Norris et al. M ∗ / M ⊙ = e . − . · M V . (6)We note that all anomalous clusters follow a tight mass-size re-lation in a region on the plot where objects could be the result ofa disruption of a nucleated dwarf galaxy (e.g. Bekki & Freeman2003; Pfe ff er & Baumgardt 2013). As we discussed before, notall massive clusters have the same characteristics and NGC 6752is plotted to illustrate that not all GCs following this trend areanomalous. Surprisingly, NGC 3201 does not belong to the trend of anomalous clusters, despite the other common characteristicsdiscussed in this paper. Should this trend of anomalous clustersin Fig. 12 be the locus of anomalous clusters with a possibleextragalactic origin, it could mean that NGC 3201 is the firstanomalous GC with a di ff erent formation scenario, or that theexplanation for the origin of all anomalous clusters should berevised, including NGC 3201. M (M
Sun ) R e ( p c ) dSphs dEs/dSOs Es/SOscEsNuclear SCsUCDsMW GCs M31 GCsAnomalous GCsNGC 6752NGC 3201 Fig. 12.
Mass-size distribution of compact stellar systems with datafrom Norris et al. (2014). We highlight the nine anomalous GCslisted by Da Costa (2015) and Marino et al. (2015). We also featureNGC 3201 and the massive GC with no dispersion in Fe or s-processelements, NGC 6752.
The small group of CN-weak and CH-weak peculiar stars wefound in NGC 3201 is intriguing. We evaluate here a possibleexternal origin for them. One possibility is to look at the G4300indices of NGC 3201 in Fig. 4 and compare them with thosefrom NGC 362, which has a similar metallicity, in Fig. A.2. Starsfrom NGC 362 are systematically weaker in CH ( ∆ = . ff erence between pec and S1 -1G stars in NGC 3201. We added the peculiar stars ofNGC 3201 to the plot of NGC 362 in Fig. A.2 after correctingthe magnitudes by the distance modulus di ff erence of 0.63 mag(Harris 1996, 2010 edition), for a better visualization. In fact, thepeculiar stars are in good agreement with 1G stars of NGC 362;they are CN-weak and CH-strong. Therefore, the Milky Way hasa GC with 1G stars matching the metallicities and CN-CH con-tents of pec stars in NGC 3201. Should this scenario of mergerbe true, there is already a pair of real clusters as constraints forthe models. As a result, CN-CH is useful to find potential merg-ers. For Na-O anti-correlation the di ff erences between clustersare less noticeable (Carretta et al. 2009) and Fig. 4 confirms thatpeculiar, S1 , and S2 stars all follow similar trends of Na-O.Another point of view is that there are only 1G peculiar starsand not 2G stars, that are typical signatures of GCs. Globularcluster 2G stars are more abundant than 1G stars, so if they werepresent in the peculiar group, we should have detected them. Article number, page 11 of 15 & Aproofs: manuscript no. dias_mar18_SAGA
This characteristic is typical in field stars, that are all 1G stars.In conclusion, we may be looking for an early GC-GC merg-ing of one typical GC with another with only first generationstars, both with the same metallicity or the merging of the pro-genitor of NGC 3201 with the building blocks of the Milky Wayhalo (in a single metallicity environment) that almost exclusivelypresents 1G stars (e.g. Fernández-Trincado et al. 2016). If thisscenario is correct, some questions still need to be answered.What is the probability of a merger of two globular clusters withsimilar metallicity? Is a merger enough to produce a retrogradeorbit? Was NGC 3201 formed in the nucleus of a dwarf galaxy?If so, why do we not detect a stellar stream or halo in such anempty region of the sky?
6. Summary and conclusions
We have obtained low-resolution spectra of RGB stars in the GCNGC 3201, derived CN and CH indices, and characterised itsmultiple stellar generations. Three groups were found in the CN-CH relation for the first time for NGC 3201: the main sequence S1 , a secondary less-populated sequence S2 , and a peculiar group pec with only CN-weak and CH-weak stars. The three groups areseparated by 0.1 mag in CH, which is at 7 σ distance. Althoughthe stars are located around the RGB bump or above, no di ff er-ences were found in the sequences due to extra mixing, possiblybecause NGC 3201 is relatively metal-poor and the convectivecell does not reach the H-burning shell to mix up the CNO-cycleprocessed material.Photometric indices were calculated from publicly avail-able catalogues from observations with ground-based and space-based telescopes to find one that splits the groups pec - S1 - S2 .The key is the use of UV filters that are sensitive to light chem-ical elements. The index δ c U , B , I was only able to separate 1Gfrom 2G stars. The V- δ c F275W , F336W , F438W diagram revealed that S1 stars are well split into 1G and 2G along the RGB, and thatthe smooth transition between the two arms of the RGB havs S2 and pec. stars, which could explain why this cluster doesnot show a totally bifurcated RGB. The δ CN- δ c F275W , F336W , F438W correlation was useful to find peculiar stars, and the chromosomemap δ c F275W , I - δ c F275W , F336W , F438W separates 1G and 2G stars wellonly for S1 , while S2 and pec stars are concentrated. Proba-bly the most interesting index is δ c B , V , I that was able to splitthe groups in this increasing sequence pec - S1 - S2 . Although s-process elements do not directly a ff ect broad-band magnitudes,for the case of NGC 5286 this index splits s-Fe-rich from s-Fe-poor stars very well. These groups might also correlate with dif-ferences in C + N + O, like other anomalous clusters. In fact, it wasshown that anomalous clusters have a correlation in CN-CH withincreasing abundances of Fe, C + N + O, and s-process elements.High-resolution spectroscopic analysis is needed to confirm thisfinding for NGC 3201.Kernel density estimation of CN indices of NGC 3201 starswere compared to other clusters. We note that the distributionsare sensitive to the size sample, uncertainties, and definition ofthe index, therefore the evidence found in this analysis is mild.We conclude that anomalous GCs like M 22 and NGC 1851show a quadrimodal CN distribution while typical Milky WayGCs, like NGC 288 and NGC 362, reveal a bimodal CN distribu-tion. We uncover a quadrimodal CN distribution for NGC 3201,resembling that of anomalous clusters.NGC 3201 cannot be classified as a typical Milky Way GC,but its inclusion in the anomalous group needs confirmationby high-resolution spectroscopic analysis. So far, all anoma-lous clusters present star-to-star variation of Fe, C + N + O, and s- process elements, where each group has its own anti-correlationsof p-capture elements (Na-O, CN-CH), whenever these abun-dances are available. NGC 3201 is an exceptional case with mul-tiple sequences of CN-CH that seems to correlate with C + N + Oand s-process-element abundances, but with no Fe-spread (or atleast debatable as in the case of M 22).A possible scenario for the origin of anomalous clusters isin the nucleus of dwarf galaxies. Anomalous GCs seem to fol-low a trend on the mass-size relation. Some models have shownhow larger galaxies may evolve to ultra compact dwarfs and howdwarf galaxies can evolve to globular clusters. NGC 6752 illus-trates that not all clusters in this trend are anomalous. NGC 3201does not follow this trend. If the locus of an anomalous cluster isonly in this trend of mass-size, either the scenario needs revisionto incorporate NGC 3201, or NGC 3201 is not anomalous but arather abnormal cluster.Four peculiar stars were discovered presenting CN-weak andCH-weak indices. One of such stars was already indicated bySmith & Norris (1982) and Da Costa et al. (1981). The natureand origin of this group of stars needs further study with high-resolution spectroscopy and detailed chemical abundance analy-sis. Our proposal scenario to explain these stars is that they wereaccreted to NGC 3201 during a GC-GC merger, where one ofthem had only 1G stars, or because NGC 3201 was born in thenucleus of a dwarf galaxy and accreted 1G field stars. All starsshould have the same metallicity and similar distribution of CNindices, but with a systematic di ff erence in CH indices.We compared NGC 3201 with NGC 6752 that has similarmass and metallicity. The latter follows the mass-size relationof anomalous clusters but it has its complexities in the horizon-tal branch (HB) driven by He abundances. The di ff erence in HBmorphology can be explained by the age di ff erence of 1 Gyr.Despite the similarities in global parameters, NGC 6752 doesnot show complexities similar to the anomalous GCs. In otherwords, a combination of mass, size, metallicity, age, and nucle-osynthesis does not mean that a GC is anomalous. The originof the anomalies of NGC 3201 and anomalous GCs seems to beexternal.Finally, we envisage three possibilities for NGC 3201:1. CN-CH pec-S1-S2 is an increasing sequence of C + N + O ands-process-element abundances, which means that NGC 3201would be the first anomalous GC out of the mass-size rela-tion.2. pec-S1-S2 have the same contents of C + N + O and s-process-element abundances, which means that NGC 3201 wouldbe the first non-anomalous GC with multiple CN-CH anti-correlation.3. pec-S1-S2 have the same contents of C + N + O and s-process-element abundances, and NGC 3201 would be the firstanomalous GC without star-to-star C + N + O and s-process-element abundances.In all cases, the definition of anomalous clusters and the sce-nario in which they have an extragalactic origin must be revised.
Acknowledgements.
This work was only possible because of ESO approval forthe ESO / NEON Observing School at La Silla on February 2016. The authorsacknowledge M. Dennefeld and F. Selman for leading the organisation of theschool. JPNC thanks E. Pancino for the discussion on the error estimation. Theauthors are thankful to Prof. G. Da Costa for a careful reading and commentson the manuscript, and also to the anonymous referee for useful suggestions thathelped improve the paper.
Article number, page 12 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201?
References
Bastian, N. & Lardo, C. 2018, ArXiv e-prints [ arXiv:1712.01286 ]Bekki, K. & Freeman, K. C. 2003, MNRAS, 346, L11Bekki, K. & Yong, D. 2012, MNRAS, 419, 2063Bragaglia, A., Carretta, E., Gratton, R. G., et al. 2010, ApJ, 720, L41Briley, M. M., Harbeck, D., Smith, G. H., & Grebel, E. K. 2004, AJ, 127, 1588Briley, M. M., Hesser, J. E., Bell, R. A., Bolte, M., & Smith, G. H. 1994, AJ,108, 2183Campbell, S. W., Yong, D., Wylie-de Boer, E. C., et al. 2012, ApJ, 761, L2Carretta, E. 2015, ApJ, 810, 148Carretta, E., Bragaglia, A., Gratton, R. G., et al. 2009, A&A, 505, 117Carretta, E., Gratton, R. G., Lucatello, S., Bragaglia, A., & Bonifacio, P. 2005,A&A, 433, 597Carretta, E., Gratton, R. G., Lucatello, S., et al. 2010, ApJ, 722, L1Coelho, P., Barbuy, B., Meléndez, J., Schiavon, R. P., & Castilho, B. V. 2005,A&A, 443, 735Cohen, J. G., Briley, M. M., & Stetson, P. B. 2002, AJ, 123, 2525Cottrell, P. L. & Da Costa, G. S. 1981, ApJ, 245, L79Covey, K. R., Wallerstein, G., Gonzalez, G., Vanture, A. D., & Suntze ff , N. B.2003, PASP, 115, 819Da Costa, G. S. 2015, ArXiv e-prints [ arXiv:1510.00873 ]Da Costa, G. S., Cannon, R., Croke, B., & Norris, J. 2004, Mem. Soc. As-tron. Italiana, 75, 370Da Costa, G. S., Frogel, J. A., & Cohen, J. G. 1981, ApJ, 248, 612Da Costa, G. S., Held, E. V., Saviane, I., & Gullieuszik, M. 2009, ApJ, 705, 1481de Mink, S. E., Pols, O. R., Langer, N., & Izzard, R. G. 2009, A&A, 507, L1Decressin, T., Meynet, G., Charbonnel, C., Prantzos, N., & Ekström, S. 2007,A&A, 464, 1029D’Ercole, A., Vesperini, E., D’Antona, F., McMillan, S. L. W., & Recchi, S.2008, MNRAS, 391, 825Di Cecco, A., Bono, G., Stetson, P. B., et al. 2010, ApJ, 712, 527Dias, B., Barbuy, B., Saviane, I., et al. 2016a, A&A, 590, A9Dias, B., Saviane, I., Barbuy, B., et al. 2016b, The Messenger, 165, 19Fernández-Trincado, J. G., Robin, A. C., Moreno, E., et al. 2016, ApJ, 833, 132Gonzalez, G. & Wallerstein, G. 1998, AJ, 116, 765Gratton, R., Sneden, C., & Carretta, E. 2004, ARA&A, 42, 385Gratton, R. G., Carretta, E., & Bragaglia, A. 2012, A&A Rev., 20, 50Gratton, R. G., Sneden, C., Carretta, E., & Bragaglia, A. 2000, A&A, 354, 169Gruyters, P., Nordlander, T., & Korn, A. J. 2014, A&A, 567, A72Harbeck, D., Smith, G. H., & Grebel, E. K. 2003, AJ, 125, 197Harris, W. E. 1996, AJ, 112, 1487Kayser, A., Hilker, M., Grebel, E. K., & Willemsen, P. G. 2008, A&A, 486, 437Kravtsov, V., Alcaíno, G., Marconi, G., & Alvarado, F. 2009, A&A, 497, 371Lardo, C., Milone, A. P., Marino, A. F., et al. 2012, A&A, 541, A141Lim, D., Hong, S., & Lee, Y.-W. 2017, ApJ, 844, 14Mackey, A. D. & van den Bergh, S. 2005, MNRAS, 360, 631Marino, A. F., Milone, A. P., Karakas, A. I., et al. 2015, MNRAS, 450, 815Marino, A. F., Milone, A. P., Piotto, G., et al. 2009, A&A, 505, 1099Marino, A. F., Milone, A. P., Yong, D., et al. 2017, ApJ, 843, 66Marino, A. F., Sneden, C., Kraft, R. P., et al. 2011, A&A, 532, A8Marino, A. F., Villanova, S., Piotto, G., et al. 2008, A&A, 490, 625Milone, A. P., Marino, A. F., Piotto, G., et al. 2013, ApJ, 767, 120Milone, A. P., Piotto, G., Renzini, A., et al. 2017, MNRAS, 464, 3636Momany, Y., Piotto, G., Recio-Blanco, A., et al. 2002, ApJ, 576, L65Monelli, M., Milone, A. P., Stetson, P. B., et al. 2013, MNRAS, 431, 2126Muñoz, C., Geisler, D., & Villanova, S. 2013, MNRAS, 433, 2006Mucciarelli, A., Lapenna, E., Massari, D., Ferraro, F. R., & Lanzoni, B. 2015a,ApJ, 801, 69Mucciarelli, A., Lapenna, E., Massari, D., et al. 2015b, ApJ, 809, 128Nardiello, D., Milone, A. P., Piotto, G., et al. 2015, A&A, 573, A70Norris, J. 1987, ApJ, 313, L65Norris, J., Cottrell, P. L., Freeman, K. C., & Da Costa, G. S. 1981, ApJ, 244, 205Norris, J. & Freeman, K. C. 1983, ApJ, 266, 130Norris, J. & Smith, G. H. 1981, in IAU Colloq. 68: Astrophysical Parameters forGlobular Clusters, ed. A. G. D. Philip & D. S. Hayes, Vol. 68, 109Norris, M. A., Kannappan, S. J., Forbes, D. A., et al. 2014, MNRAS, 443, 1151Olszewski, E. W., Saha, A., Knezek, P., et al. 2009, AJ, 138, 1570Pancino, E., Rejkuba, M., Zoccali, M., & Carrera, R. 2010, A&A, 524, A44Pfe ff er, J. & Baumgardt, H. 2013, MNRAS, 433, 1997Pickles, A. J. 1985, ApJS, 59, 33Piotto, G., Milone, A. P., Bedin, L. R., et al. 2015, AJ, 149, 91Prantzos, N. & Charbonnel, C. 2006, A&A, 458, 135R Core Team. 2015, R: A Language and Environment for Statistical Computing,R Foundation for Statistical Computing, Vienna, AustriaRenzini, A., D’Antona, F., Cassisi, S., et al. 2015, MNRAS, 454, 4197Rey, S.-C., Yoon, S.-J., Lee, Y.-W., Chaboyer, B., & Sarajedini, A. 2001, AJ,122, 3219Riello, M., Cassisi, S., Piotto, G., et al. 2003, A&A, 410, 553Robin, A. C., Reylé, C., Derrière, S., & Picaud, S. 2003, A&A, 409, 523Salaris, M., Cassisi, S., & Weiss, A. 2002, PASP, 114, 375 Saviane, I., da Costa, G. S., Held, E. V., et al. 2012, A&A, 540, A27Schiavon, R. P., Zamora, O., Carrera, R., et al. 2017, MNRAS, 465, 501Simmerer, J., Ivans, I. I., Filler, D., et al. 2013, ApJ, 764, L7Smith, G. H. 1983, AJ, 88, 410Smith, G. H. & Norris, J. 1982, ApJ, 254, 149Smolinski, J. P., Martell, S. L., Beers, T. C., & Lee, Y. S. 2011, AJ, 142, 126Snodgrass, C., Saviane, I., Monaco, L., & Sinclaire, P. 2008, The Messenger,132, 18Soto, M., Bellini, A., Anderson, J., et al. 2017, AJ, 153, 19Stetson, P. B. 1987, PASP, 99, 191VandenBerg, D. A., Brogaard, K., Leaman, R., & Casagrande, L. 2013, ApJ, 775,134Vollmann, K. & Eversberg, T. 2006, Astronomische Nachrichten, 327, 862von Braun, K. & Mateo, M. 2001, AJ, 121, 1522Yong, D., Grundahl, F., & Norris, J. E. 2015, MNRAS, 446, 3319Yong, D., Meléndez, J., Grundahl, F., et al. 2013, MNRAS, 434, 3542 Appendix A: Scaling indices from the literature
AppendixA.1: Kayser et al.(2008)
We transformed the indices S(3839) and G4300 and their un-certainties from Kayser et al. (2008) from natural logarithm tocommon logarithm. Using the new indices we followed the pro-cedure described in Sect. 3.3 to correct the indices from stellarsurface temperature and gravity e ff ects. The goal is to put all theresults in the same scale to allow direct comparisons betweenour results and those from Kayser et al. (2008). −0.8 −0.4 0.0 0.4 . . . V S(3839)
G4300NGC288 − K08
Fig. A.1.
Similar to Fig. 4 but for NGC 288 with parameters fromKayser et al. (2008). Linear fit because bright stars are only CN-weak.Filled circles are CN-strong, and empty circles are CN-weak stars. Av-erage error bars are displayed in the bottom right corner of each panel.
AppendixA.2: Indices defined by Norris &Smith(1981)
In Section 4.3 we also use indices following the definition ofNorris & Smith (1981). We follow the same procedure as inFig. 4 to put into the same scale the CN indices of NGC 1851(Campbell et al. 2012), M 22 (Norris & Freeman 1983), andNGC 3201 (Smith & Norris 1982). Figures are shown below.
Appendix B: Errors of the indices
In order to derive the equations for the errors of the indices,we adapted the work realised by Vollmann & Eversberg (2006),who obtained these equations for equivalent widths. The index
Article number, page 13 of 15 & Aproofs: manuscript no. dias_mar18_SAGA −0.8 −0.4 0.0 0.4 V S(3839)
G4300NGC362 − K08
Fig. A.2.
Same as Fig. A.1 but for NGC 362. Green crosses are thepeculiar stars of NGC 3201 for discussion in Sect. 4. Magnitudes ofNGC 3201 stars are shifted by 0.63 mag to match the distance modulusof NGC 362. −0.8 −0.4 0.0 0.4 . . . V S(3839)
G4300M22 − K08
Fig. A.3.
Same as Fig. A.1 but for M 22. ( I ) definition for a general case, taking into account one or n continua regions, is expressed by I = − . R λ λ F ( λ ) d λ n P ni = R λ , i λ , i F c ( λ ) d λ (B.1)with F ( λ ) the flux in the band and F c ( λ ) the flux in the continuumat the wavelength λ . Then, defining ∆ λ = λ − λ and ∆ λ c i = λ , i − λ , i , and applying the mean value theorem, the expressioncan be written as I = − . ¯ F ∆ λ n P ni = ¯ F c i ∆ λ c i , (B.2)where ¯ F and ¯ F c i are the arithmetic mean within ∆ λ and ∆ λ c i ofthe band and continua fluxes, respectively. . . . . V S(3839)NGC1851 . . . . . S(3839)M22 −0.1 0.1 0.3 0.5 . . . . . S(3839)NGC3201
Fig. A.4.
Similar to Fig. 4 but only for CN of NGC 1851(Campbell et al. 2012), M 22 (Norris & Freeman 1983), and NGC 3201(Smith & Norris 1982). Filled circles are CN-strong and empty circlesare CN-weak stars.
Following the principle of error propagation, asVollmann & Eversberg (2006) did, we expand the last equationin a Taylor series, I = I ( ¯ F , ¯ F c , ..., ¯ F c n ) + ∂ I ∂ ¯ F (cid:16) F − ¯ F (cid:17) + n X i = ∂ I ∂ ¯ F c i (cid:16) F c i − ¯ F c i (cid:17) , (B.3)where F and F c i are random variables. The variance of the ex-pansion is σ ( I ) = ∂ I ∂ ¯ F · σ ( F ) ! + n X i = ∂ I ∂ ¯ F c i · σ ( F c i ) ! , (B.4)with σ ( F ) and σ ( F c i ) the standard deviation in the band and con-tinua, respectively. If we assume a Poisson statistic, the standarddeviations are defined by σ ( F c i ) = ¯ F c i S / N (B.5)and σ ( F ) = s ¯ F ¯ F c i · σ ( ¯ F c i ) = p ¯ F ¯ F ¯ c S / N , (B.6)where ¯ F ¯ c is the arithmetic mean of every ¯ F c i . The partial deriva-tives, using Equation B.2, are ∂ I ∂ ¯ F = − . F ln(10) (B.7)and ∂ I ∂ ¯ F c i = . ∆ λ c i P nj = ¯ F c j ∆ λ c j . (B.8) Article number, page 14 of 15. Dias et al.: Galactic or extragalactic chemical tagging for NGC3201?
Finally, we can write Equation B.4 as σ ( I ) = . / N ¯ F ¯ c ¯ F + n X i = ¯ F c i ∆ λ c i P nj = ¯ F c j ∆ λ c j / , (B.9)which is the error of an index with n continua regions.The errors of indices S(3839) and G4300 as a function ofsignal-to-noise ratio are given in Fig. B.1. This relation is usefulto plan observations for other objects estimating the minimumS / N to reach the desired precision. In this work we used onlyspectra with S / N CN >
25 (i.e. S / N CH >
55) in our analysis.
S/N index
10 40 70 100 130 160 1900.000.020.040.060.080.100.120.14 s i nde x CNCH