The CN–CH Positive Correlation in the Globular Cluster NGC 5286

During the last two decades, an increasing number of observations have shown that most of the Milky Way globular clusters (GCs) host multiple stellar populations, each of which has different chemical properties (e.g., Lee et al. 1999; Carretta et al. 2009; Gratton et al. 2012; Piotto et al. 2015, and references therein). These GCs share similar characteristics, such as light element abundance variations and a central concentration of later-generation stars (e.g., Carretta et al. 2009; Lardo et al. 2011), although some exceptional cases have been reported. The abundance variations in the light elements, discovered in most GCs, are explained as an enrichment and/or pollution by intermediate-mass asymptotic giant branch (AGB) stars (D'Antona & Caloi 2004; D'Antona et al. 2016), massive interacting binary stars (de Mink et al. 2009; Bastian et al. 2013), and fast-rotating massive stars (FRMSs; Prantzos & Charbonnel 2006; Decressin et al. 2007). Several GCs with heavy element abundance variations, including ω-Centauri and M22 (Lee et al. 1999, 2009; Marino et al. 2009; Johnson & Pilachowski 2010), however, show evidence of supernova (SNe) enrichment, which suggests that they were massive enough in the past to retain SNe ejecta (Baumgardt et al. 2008; Silich & Tenorio-Tagle 2017). In the hierarchical merging paradigm, they would have contributed to the formation of the Milky Way, since these GCs could be former nuclei of dwarf galaxies (see Lee et al. 2007; Han et al. 2015; Da Costa 2016), and therefore, would help to solve the "missing satellites problem" (Klypin et al. 1999; Moore et al. 1999, 2006).

A direct way to find these unique GCs is a measurement of the heavy element abundance of stars in a GC using high-resolution spectroscopy (e.g., Da Costa et al. 2009; Yong et al. 2014; Johnson et al. 2015). However, our previous studies have demonstrated that the low-resolution spectroscopy for the calcium HK' index can be used to more effectively detect the heavy element variations in a GC (Han et al. 2015; Lim et al. 2015). Interestingly, we found that these GCs also show the CN–CH positive correlation among red giant branch (RGB) stars, unlike the CN–CH anticorrelation generally observed in "normal" GCs (Suntzeff & Smith 1991; Lee 2005; Kayser et al. 2008; Pancino et al. 2010; Smolinski et al. 2011). If confirmed, this would imply that the CN–CH positive correlation can be used as a probe for the GCs with heavy element variations.

In order to further confirm our conjecture, we have performed low-resolution spectroscopy of the RGB stars in NGC 5286. This GC is relatively poorly studied, although Marino et al. (2015, hereafter M15) recently showed some abundance variations in Fe and s-process elements among RGB stars from high-resolution spectroscopy. The purpose of this paper is to report that RGB stars in this GC are clearly divided into three subpopulations by CN index, and CN-strong stars are also enhanced in the calcium HK' and CH indices, indicating the CN–CH positive correlation.

Our observations were performed with the du Pont 2.5 m telescope at Las Campanas Observatory (LCO) during four nights from 2016 April to June. We have used the multi-object spectroscopy mode of the Wide Field Reimaging CCD Camera (WFCCD) with the HK grism, which provides a dispersion of 0.8 Å/pixel and a central wavelength of 3700 Å. Spectroscopic target stars are selected from the 2MASS All-Sky Point Source Catalog. In particular, we have included a number of stars observed by M15 for the comparison with high-resolution spectroscopy. For these observations, three multi-slit masks, each of which contains about 25 slits of 12 width, were designed. We had obtained four 1500 s science exposures, three flats, and an arc lamp frame for each mask. The data reduction was performed with IRAF¹ and the modified version of the WFCCD reduction package, following Lim et al. (2015) and Prochaska et al. (2006). The radial velocity (RV) of each star was measured using the rvidlines task in the IRAF RV package, and the signal-to-noise ratio (S/N) was estimated at ∼3900 Å. After the rejection of non-member stars (RV > 2.0σ of the mean velocity of the GC) and low S/N (<8.0) spectra, 44 stars were finally used for our analysis. The median RV for these stars is 52.5 km s⁻¹, which is comparable to but somewhat smaller than the values of 57.4 km s⁻¹ reported by Harris (2010) and 61.5 km s⁻¹ estimated by M15. Compared to the RVs measured from high-resolution spectroscopy, the typical uncertainty of those values from our low-resolution measurement appears to be quite large (∼20 km s⁻¹). Figure 1 shows the selected sample stars on the color–magnitude diagram (CMD), for which the photometry was obtained at the LCO 2.5 m du Pont telescope. Fifteen target stars, however, are outside the field-of-view (FOV) of the photometry (885 × 885) and are therefore not plotted in this CMD. A detailed description of this photometry can be found in Lim et al. (2015).

Zoom In Zoom Out Reset image size

Finally, we measured the S(3839) index for the CN band, the HK' index for the Ca ii H and K lines, and the CH4300 index for the CH band of each target star in NGC 5286, following Lim et al. (2015). The definitions for these indices are

$$\begin{eqnarray*}\mathrm{HK}^{\prime} & = & -2.5\mathrm{log}\displaystyle \frac{{F}_{3916-3985}}{2{F}_{3894-3911}+{F}_{3990-4025}},\\ \mathrm{CN}(3839) & = & -2.5\mathrm{log}\displaystyle \frac{{F}_{3861-3884}}{{F}_{3894-3910}},\\ \mathrm{CH}4300 & = & -2.5\mathrm{log}\displaystyle \frac{{F}_{4285-4315}}{0.5{F}_{4240-4280}+0.5{F}_{4390-4460}},\end{eqnarray*}$$

where ${F}_{3916-3985}$, for example, is the integrated flux from 3916 to 3985 Å. All of these indices are defined as the ratio of the absorption strength to nearby continuum strength. The measurement error for each sample was estimated from Poisson statistics in the flux measurements (Vollmann & Eversberg 2006). In addition, we measured the delta indices (δCN, δHK', and δCH) to compare the chemical abundances of stars without the effect of magnitude in the same manner as done in previous studies (e.g., Norris et al. 1981; Harbeck et al. 2003). These δ-indices are calculated as the difference between the original index for each star and the least-squares fitting of the full sample (black solid lines in the left panels of Figure 2) in a GC. The measured indices and errors are listed in Table 1.

Zoom In Zoom Out Reset image size

Figure 2 shows the measured spectral indices of stars as functions of K magnitude, obtained from the 2MASS catalog. The CN, HK', and CH indices increase with decreasing magnitude because the brighter RGB stars have lower temperatures and the strengths of these molecular bands generally increase with decreasing temperature. Therefore, the chemical abundances of stars are compared on the δ-index versus magnitude diagrams. It is important to note that the observed stars show a large spread in δ-index that is at least several times larger than the measurement error. The standard deviations for all sample stars are 0.23 for CN, 0.07 for HK', and 0.07 for CH. In particular, the CN index distribution shows the largest spread. Note that a bimodality or a large spread in CN distribution is generally observed in most GCs (Norris et al. 1981; Norris 1987; Briley et al. 1992; Harbeck et al. 2003; Kayser et al. 2008; Martell et al. 2008).² Therefore, we have divided subpopulations of RGB stars in NGC 5286 on the histogram of the δCN index (see Figure 3). It is clear from this histogram that RGB stars are divided into three subpopulations: CN-weak (δCN < −0.2; blue circles), CN-intermediate (−0.2 ≤ δCN < 0.1; green circles), and CN-strong (0.1 ≤ δCN; red circles). The distribution of CN index into three or more subpopulations is similar to that reported in NGC 1851 (Campbell et al. 2012; Lim et al. 2015; Simpson et al. 2017). This is also consistent with the recent results from population models and spectroscopic observations that show that most GCs host three or more subpopulations (see, e.g., Jang et al. 2014; Carretta 2015). The presence of multiple populations is also observed from recent photometry using UV filters, which are mainly sensitive to N abundance (Milone et al. 2015; Piotto et al. 2015). In this regard, further observations are required to check that the trimodal CN distribution, observed in NGC 1851 and NGC 5286, is a ubiquitous feature in other GCs as well.

Zoom In Zoom Out Reset image size

As shown in the right panels of Figure 2, the CN-strong subpopulation is also significantly enhanced in the δHK' and δCH indices. The differences between CN-strong and CN-weak subpopulations are 0.537 for δCN, 0.123 for δHK', and 0.094 for δCH, which are significant at the levels of 20.7σ, 7.4σ, and 5.1σ, respectively, compared to the standard deviation of the mean. The CN-intermediate stars, however, show no clear difference in the strength of δHK' index with CN-weak stars (ΔδHK' = 0.02). Therefore, three subpopulations in the NGC 5286 can be characterized as CN-weak/HK'-weak, CN-intermediate/HK'-weak, and CN-strong/HK'-strong. In particular, the difference in calcium abundance (HK') suggests that this GC also belongs to the group of unique GCs showing intrinsic dispersion in heavy element abundance, in agreement with a result by M15 based on Fe and s-process elements.³ The fact that CN-strong stars are also enhanced in the CH band implies a presence of CN–CH positive correlation in this GC (see Section 4 below).

In order to see whether the CN- and HK'-strong stars in our study are also enhanced in Fe and s-process elements, we have compared our results with high-resolution spectroscopy by M15. In Figure 4, our δCN and δHK' indices are plotted with [Na/Fe] and [Ca/Fe] abundances, respectively, for 33 common stars. In general, the strength of the CN band is correlated with the N and Na abundances, while the CH band is affected by C abundance (Sneden et al. 1992; Smith et al. 1996; Marino et al. 2008). The upper panel of Figure 4 also shows a strong correlation between [Na/Fe] and the δCN index, which is in good agreement with previous studies (Sneden et al. 1992; Lim et al. 2016).⁴ The CN-weak (blue) and CN-strong (red) subpopulations are almost identical to the s-poor (triangles) and s-rich (squares) groups, respectively. In addition, the δHK' index is understandably correlated with the [Ca/Fe] abundance with a few exceptions (see the lower panel of Figure 4). According to this comparison, the difference in δHK' index between CN-weak and CN-strong stars (∼0.094) is equivalent to 0.09 dex in Δ[Ca/Fe] and 0.15 dex in Δ[Fe/H]. These comparisons confirm that our results from low-resolution spectroscopy are consistent with those from high-resolution spectroscopy by M15. Consequently, the later-generation stars in NGC 5286 show the enhancements not only in light elements (CN) but also in heavy elements (Fe and Ca) and s-process elements, although the presence of Fe spread requires further investigations (see Mucciarelli et al. 2015; Lee 2016).

Zoom In Zoom Out Reset image size

As described above, the CN–CH anticorrelation is one of the typical features in the low-resolution spectroscopic studies of GCs (Suntzeff & Smith 1991; Kraft 1994; Harbeck et al. 2003; Lee 2005; Pancino et al. 2010; Smolinski et al. 2011, and references therein). This feature is most likely due to the anticorrelation between C and N abundances (see, e.g., Smith et al. 1996; Cohen et al. 2005). In the multiple-population paradigm, the mechanism responsible for the Na–O anticorrelation would also produce C–N anticorrelation (Ventura et al. 2013; Di Criscienzo et al. 2016). Our previous studies, however, found a significant CN–CH positive correlation, instead of an anticorrelation, among RGB stars in M22 and NGC 6273 (Han et al. 2015; Lim et al. 2015). Interestingly, both GCs are known to host multiple stellar populations with different heavy element abundances (see also Marino et al. 2011; Johnson et al. 2017). Since NGC 5286 is also one of the GCs showing spreads in the abundances of heavy elements, we would expect a similar positive correlation between CN and CH indices. As expected, Figure 5 shows that δCN and δCH indices are positively correlated, similar to the cases of M22 and NGC 6273.

Zoom In Zoom Out Reset image size

In order to establish how the heavy element abundance variations would affect the CN–CH relation of GCs, we have plotted in Figure 6 the δCN versus δCH diagrams for six GCs (NGC 288, NGC 6723, NGC 1851, NGC 6273, M22, and NGC 5286), together with the δHK' distributions. The spectroscopic and photometric data are taken from Han et al. (2015) and Lim et al. (2015, 2016). The slope of the CN–CH relation for each GC is estimated by maximum likelihood, and the values are listed in the upper panels of Figure 6. We can see from this figure that normal GCs without a difference in Ca abundance, NGC 288 and NGC 6723, show the conventional CN–CH anticorrelation. On the contrary, GCs with a difference in HK' index between the two subpopulations, NGC 6273, M22, and NGC 5286, show the CN–CH positive correlation. In the case of NGC 1851, the difference in HK' index is relatively small, and the CN–CH relation seems to be flat.⁵ In order to test the significance of the correlation, we have calculated the Spearman's rank correlation coefficient for each GC. The obtained correlation coefficients are −0.52 and −0.62 for NGC 288 and NGC 6723, −0.05 for NGC 1851, and +0.37, +0.61, and +0.52 for NGC 6273, M22, and NGC 5286, respectively. The p-values are very small (7.3 × 10⁻⁷, 1.4 × 10⁻⁴, 7.3 × 10⁻⁴, 2.9 × 10⁻¹¹, and 2.9 × 10⁻⁴ for NGC 288, NGC 6723, NGC 6273, M22, and NGC 5286), confirming that the correlations are statistically significant, except for NGC 1851 (p-value = 0.72). Because the negative Spearman coefficient (−1.0 ∼ 0.0) indicates anticorrelation while the positive coefficient (0.0 ∼ +1.0) is for positive correlation, this result confirms the systematic variation in the CN–CH correlation among sample GCs. Therefore, the origin of the CN–CH positive correlation appears to be explicitly relevant to the heavy element abundance variations. In this respect, it would be interesting to measure [C/Fe], [N/Fe], and the C+N+O sum from high-resolution spectroscopy.

Zoom In Zoom Out Reset image size

As discussed in Lim et al. (2015), the origin of the CN–CH positive correlation, as well as that of the heavy element abundance variations, is most likely because the later-generation stars are enriched by some SNe in addition to the intermediate-mass AGB stars and/or FRMSs. Unlike the intermediate-mass AGB stars, which are suggested to mainly be responsible for the N-enhancement and C-depletion of later-generation stars, the SNe ejecta would supply both N and C elements together with other heavy elements. Our result for the CN–CH positive correlation in NGC 5286 (Figure 5) appears to be similar to those of M22 and NGC 6273. Interestingly, inspection of Figure 2 shows that the observed stars in NGC 5286 can be divided into three subpopulations: CN-weak/HK'-weak stars (first generation; G1), CN-intermediate/HK'-weak stars (second generation; G2), and CN-strong/HK'-strong stars (third generation; G3). These differences in chemical properties imply that the SNe enrichment played a role only in the formation of G3 stars, whereas it had almost no impact on the formation of G2 stars. Although the origin of this complex chemical enrichment requires further investigation, one possibility is a time-dependent gas removal of SNe ejecta in a proto-GC. For example, a recent hydrodynamical simulation by Caproni et al. (2017) shows that the gas removal in a dwarf galaxy was more efficient in the first 600 Myr, while most of the gas could be retained later when the SNe II rate was significantly decreased. Similar to this, the SNe ejecta could have fully escaped from the proto-NGC 5286 in the early phase, while some of them could have been retained later with decreasing SNe rate. This would explain the absence and presence of some SNe enrichment in G2 and G3, respectively.

We are grateful to the anonymous referee for a number of helpful suggestions. We also thank Sang-Il Han for providing photometric data for Figure 1. Support for this work was provided by the National Research Foundation of Korea to the Center for Galaxy Evolution Research, and through the grant programs No. 2017R1A6A3A11031025 and No. 2017R1A2B3002919.