© The Authors 2024

1. Introduction

The presence of multiple stellar populations in Galactic globular clusters (GCs) is now firmly established (see Bastian & Lardo 2018; Milone & Marino 2022, for recent reviews). This phenomenon has generated several questions concerning the origin of GCs, many of which are still to be answered.

The chromosome map (ChM; Milone et al. 2015) is a powerful tool for investigating stellar populations within GCs. It is a pseudo-two-color diagram extremely sensitive to the chemical composition of distinct stellar populations, thereby enhancing their separation. The standard Δ_{C F275W, F336W, F438W} versus Δ_{F275W, F814W} ChM (Milone et al. 2017) is derived from photometry in the F275W, F336W, F438W, and F814W bands of the Hubble Space Telescope (HST). This photometric diagram is particularly useful to separate stellar populations within the red-giant branch (RGB) or the upper main sequence (MS). The advent of the James Webb Space Telescope (JWST) allowed the construction of new ChMs, derived by combining optical and near-infrared bands of the HST and the JWST. As an example, recent works (Milone et al. 2023b; Ziliotto et al. 2023) introduced the Δ_{F115W, F322W2} verus Δ_{F606W, F115W} and Δ_{F090W, F277W} versus Δ_{F090W, F150W} pseudo-two color diagrams, optimized to separate distinct stellar populations in the M-dwarf regime at the bottom of the MS. Despite their differences, all ChMs exhibit two main groups of stars, called first-generation (1G) and second-generation (2G): 1G stars are composed of pristine material reflecting the chemical composition of their natal cloud, whereas 2G members exhibit various degrees of chemical enrichment.

In addition to their effectiveness in identifying and characterizing multiple stellar populations, ChMs also provide direct insights into the formation environment of GCs. In this context, a notable discovery based on RGB ChMs indicates that the 1G sequences in most Galactic GCs do not conform to simple stellar populations. Instead, they exhibit either a spread or a bimodality in the ChM (see Figs. 3–7 in Milone et al. 2017, for a comprehensive collection of ChMs in 57 GCs). This phenomenon is not limited to RGB stars; Dondoglio et al. (2021) observed extended 1G sequences along the red horizontal branch (HB) in 12 clusters. More recently, the 1G color extension has been detected even in the ChMs of unevolved MS stars (Legnardi et al. 2022; Milone et al. 2023b), suggesting that chemical variations are likely imprinted in the environment where 1G stars originated.

In the standard ChM, 1G stars exhibit a significant dispersion in the direction of the x-axis while maintaining a relatively constant position along the y-axis. This observation implies a connection between the extension of 1G stars and variations in the effective temperature among stars with similar F814W magnitude. The spread in helium content can impact the effective temperature of stars, as initially observed by Milone et al. (2015) to interpret the extended 1G sequence that they observed in the ChM of NGC 2808. To account for this phenomenon, Milone et al. (2015, 2018b) explored various physical mechanisms and found that none of them could produce pure helium enrichment without concurrent changes in light elements. Alternatively, helium inhomogeneities may come from events that occurred in regions of the Universe where the baryon-to-photon ratio was significantly enhanced (e.g., Arbey et al. 2020).

The spectroscopic investigation by Marino et al. (2019b) suggested that variations in metallicity, rather than helium content, may be responsible for the extended color range observed in 1G stars. In their study, Marino and collaborators analyzed 18 1G stars within the Galactic GC NGC 3201 and observed a dispersion in [Fe/H] of ∼0.1 dex. The photometric research by Legnardi et al. (2022) further supports this conclusion. To distinguish between helium and metallicity effects, Legnardi and collaborators introduced a pseudo-two-magnitude diagram aimed at maximizing the separation between stellar populations with varying iron content. Their analysis revealed a spread of [Fe/H], rather than helium (Y), among 1G stars in the two investigated targets, namely NGC 6362 and NGC 6838. Expanding this finding to a large sample of GCs, Legnardi et al. (2022) estimated that internal variations in iron content within the 1G of 55 Galactic GCs range from less than ∼0.05 dex to ∼0.30 dex and mildly correlate with GC mass.

As discussed by Marino et al. (2019b), unresolved binaries among 1G stars can provide extended sequences in the ChM and spurious lower metallicity abundances for stars with low Δ_{F275W, F814W} values. However, they showed that very high fractions of binaries are needed to reproduce the observations, in contrast with what is observed in Galactic GCs and intermediate-age Magellanic Cloud (MC) star clusters (Milone et al. 2012a; Mohandasan et al. 2024). Hence, they concluded that binaries very likely provide a minor contribution to the color extension of the 1G (see also Kamann et al. 2020; Martins et al. 2021).

The discovery of chemical abundance variations among 1G stars provides the opportunity to constrain the formation mechanisms of the stellar populations in GCs (e.g., D’Antona et al. 2016). The observed [Fe/H] variations could be indicative of either chemical inhomogeneities within the original pristine material where the 1G population formed or could be the result of internal stellar feedback processes taking place within the clusters themselves (McKenzie & Bekki 2021). Moreover, the evidence that 2G stars of NGC 6362 and NGC 6838 exhibit lower metallicity variations than the 1G is consistent with the scenarios where 2G stars formed in a high-density environment in the cluster center (Legnardi et al. 2022). Finally, the color extension of 1G stars in the ChM would constrain the late stages of stellar evolution in star clusters. The 1G stars of GCs with the same metallicity but different HB morphology span different ranges of [Fe/H] (Legnardi et al. 2022).

Despite extensive observational efforts, several aspects of the extended 1G sequence phenomenon remain unexplored. For instance, it is still uncertain whether the metallicity distribution of 1G stars depends on factors such as the radial distance from the cluster center or other parameters associated with the host GCs. Similarly, the status of extended 1G sequences as a unique characteristic exclusive to Galactic GCs with multiple populations, or their potential existence in Galactic open clusters or massive intermediate-age star clusters in the MCs, remains unknown. This work aims to shed light on these points, thus providing a comprehensive analysis of chemical inhomogeneities among 1G stars in Galactic and extragalactic stellar clusters.

First, we study metallicity variations among 1G stars of the Galactic GC NGC 104 (47 Tucanae). In this context, 47 Tucanae is an ideal target. The evidence that 2G stars are more centrally concentrated than the 1G stars (e.g., Milone et al. 2012b; Cordero et al. 2014; Dondoglio et al. 2021; Lee 2022) indicates that this GC retains information on the radial distribution of its multiple populations at formation. Moreover, a recent spectroscopic analysis conducted by Marino et al. (2023) uncovered that 1G stars in 47 Tucanae display variations in [Fe/H] on the order of ∼0.1 dex, thus corroborating previous findings obtained through photometry. The second part of the present research investigates whether clusters with no evidence of multiple populations exhibit an extended ChM sequence comparable to that of 1G stars in Galactic GCs. To this end, we focus on the Galactic open cluster NGC 6791, and the Large Magellanic Cloud (LMC) cluster NGC 1783. Finally, we come back to Galactic GCs looking for possible relations between the metallicity distribution of 1G stars and the main parameters of the host cluster.

The paper is organized as follows. Section 2 presents the datasets and the techniques employed to reduce them. Section 3 is dedicated to 1G stars of 47 Tucanae, whereas the two simple-population clusters investigated in this work and the Galactic GCs are the targets of Sects. 4 and 5, respectively. Section 6 provides the summary of this work together with our conclusions.

2. Observations and data reduction

In the following, we describe the observations and the photometric and astrometric catalogs employed for this work. For simplicity, we discuss the data on 47 Tucanae, on the simple-population star clusters NGC 6791 and NGC 1783, and on the other Galactic GCs, separately.

2.1. 47 Tucanae

We analyzed the chemical composition and the internal kinematics of 1G stars in the Galactic GC 47 Tucanae by using data of four distinct regions, located at progressively larger distances from the cluster center. Figure 1 illustrates the footprints of the four regions discussed in this paper, including the cluster center and three external fields, denoted as A, B, and C, respectively.

Fig. 1. Footprints of the regions of 47 Tucanae analyzed in this paper. The central region (blue) and field A (red) include only HST data, whereas field B (cyan) and C (orange) also cover JWST images.

The Central field (RA ∼ 00^h24^m06^s, Dec ∼ −72^d04^m53^s; blue footprints) covers the innermost region of ∼2.7  ×  2.7 arcmin² within the cluster. In this area, we used the catalog by Milone et al. (2017), which includes photometry in the F275W and F336W bands of the Ultraviolet and Visual Channel of the Wide Field Camera 3 (UVIS/WFC3) and in the F435W, F606W, and F814W filters of the Wide Field Channel of the Advanced Camera for Survey (WFC/ACS) on board the HST. To study the kinematics of 1G stars in the core of 47 Tucanae, we took advantage of stellar proper motions from Libralato et al. (2022).

Fields A (RA ∼ 00^h24^m37^s, Dec ∼ −72^d04^m06^s; red footprints) and B (RA ∼ 00^h22^m36^s, Dec ∼ −72^d09^m27^s; cyan footprints) are located at distances of ∼7 and ∼8.5 arcmin west from the cluster center, respectively. For those regions, we relied on the photometric catalogs and proper motions derived recently by Milone et al. (2023b). Observations of field A were taken with the near-infrared channel of the Wide Field Camera 3 (IR/WFC3) through the F110W and F160W bands and with the WFC/ACS in the F606W and F814W filters. For field B, instead, Milone and collaborators used exposures collected through the near-infrared camera (NIRCam) on board the JWST. This field has been observed also with the UVIS/WFC3 in the F606W filter and with the IR/WFC3 in the F110W and F160W bands. We refer to the aforementioned papers and their respective references for details on the data reduction and the techniques used to measure stellar proper motions.

Finally, field C (RA ∼ 00^h21^m16^s, Dec ∼ −72^d06^m16^s; orange footprints) is located at a distance of ∼11 arcmin west with respect to the cluster center. To investigate 1G stars in this outer region we used data obtained with the NIRCam/JWST as part of the GO-2560 program (PI: A. F. Marino), which includes images acquired with the F115W and the F322W2 filters. Moreover, we complemented these observations with HST exposures collected through the UVIS/WFC3 F606W and the IR/WFC3 F110W and F160W bands. The techniques adopted to reduce these observations, as well as to compute stellar proper motions, are described in Marino et al. (2024).

Figure 2 illustrates the m_F115W versus m_F115W − m_F322W2 color-magnitude diagram (CMD) of 47 Tucanae derived by Marino et al. (2024) for stars within field C. In close analogy with Milone et al. (2023b), this photometric diagram has been used together with the m_F115W versus m_F606W − m_F115W CMD to derive the Δ_{F115W, F322W2} versus Δ_{F606W, F115W} ChM, plotted in the inset of Fig. 2. Here, 1G stars lay in proximity of the origin, whereas 2G stars extend from (Δ_{F606W, F115W}, Δ_{F115W, F322W2})∼(−0.1, 0.05) to ∼(−0.35, 0.2).

Fig. 2. mF115W vs. mF115W − mF322W2 CMD of 47 Tucanae stars within field C (Marino et al. 2024). The inset shows the ΔF115W, F322W2 vs. ΔF606W, F115W ChM.

2.2. Simple-population clusters

To investigate the extended 1G sequence in simple-population clusters, we selected two star clusters, namely NGC 6791 and NGC 1783, with available multi-band HST photometry.

• NGC 6791 is an old (∼8.5 Gyr, Brogaard et al. 2012) and massive (∼10⁴ M_⊙, e.g., Cordoni et al. 2023b) Galactic open cluster with a super-solar metallicity ([Fe/H] ∼ 0.3 − 0.4, e.g., Boesgaard et al. 2009). In this work, we analyzed the innermost ∼2.7 × 2.7 arcmin² area of NGC 6791 by using HST images collected with the F275W, F336W, F438W, and F814W filters of the UVIS/WFC3 and the F814W band of WFC/ACS. The main properties of these observations are summarized in Table 1. Additionally, to extend our analysis outside the HST field of view, we combined the catalogs of proper motions and stellar positions provided by Gaia Data Release 3 (Gaia Collaboration 2021, DR3) with the ground-based photometry from Stetson et al. (2019). To derive the astro-photometric catalog for the central region of NGC 6791 we used the computer program KS2 (see, e.g., Sabbi et al. 2016; Bellini et al. 2017; Milone et al. 2023a, for more details), which was developed by Jay Anderson as an evolution of the kitchen_sync software (Anderson et al. 2008). Together with photometry, KS2 also provides a collection of diagnostic parameters that we exploited to select a sample of well-measured stars, as in Milone et al. (2023a, see their Sect. 2.4 for details). Finally, we corrected magnitudes for the effects of differential reddening following the method by Milone et al. (2012a) and Legnardi et al. (2023). The left panel of Fig. 3 shows the resulting m_F814W versus m_F275W − m_F814W CMD of NGC 6791 that we used to identify candidate binaries and blue stragglers. Similarly to photometric uncertainties, these objects contribute to the color broadening of evolutionary sequences in all CMDs. As an example, in the right panel of Fig. 3 we plotted the m_F814W verus C_{F275W, F336W, F438W} = (m_F275W − m_F336W)−(m_F336W − m_F438W) pseudo-CMD where the candidate binaries (red triangles) populate the reddest part of the MS, whereas blue stragglers (blue points) exhibit bluer colors than RGB stars. The C_{F275W, F336W, F438W} pseudo-color is sensitive to C, N, and O, thereby increasing the separation between stellar populations with different light-element abundances. The C_{F275W, F336W, F438W} color broadening is comparable to that of observational errors alone, indicated by the error bars on the right panel. This result allowed us to conclude that NGC 6791 is a prototype of a simple-population cluster, thus corroborating previous results based on spectroscopy (e.g., Bragaglia et al. 2014; Villanova et al. 2018) and on near-infrared photometry of M dwarfs (Dondoglio et al. 2022). To further investigate NGC 6791 over a wide field of view we combined the photometry in the U, B, V, and I bands derived by Peter Stetson (Stetson et al. 2019) together with the coordinates, proper motions, and parallaxes from the Gaia DR3 (Gaia Collaboration 2021). Photometry was obtained through an analysis conducted by Peter Stetson using images gathered from diverse ground-based telescopes. The procedures and computer programs outlined by Stetson (2005) were employed, and the derived photometric data were calibrated against the reference system established by Landolt (1992). We selected a sample of probable cluster members by using the proper motions and the stellar parallaxes from Gaia DR3 and the procedure by Cordoni et al. (2018). Photometry has been corrected for differential reddening (see Legnardi et al. 2023, for details).

• NGC 1783 is an intermediate-age (∼1.5 Gyr, Milone et al. 2023a) LMC cluster which shows no evidence of multiple stellar populations (Milone et al. 2020). We chose to analyze this object because it is the only simple-population MC cluster for which photometry in the UVIS/WFC3 filter F275W is available in the HST archive. In this work, we employed the catalog of NGC 1783 by Milone et al. (2023a), which includes photometry in the F275W, F336W, F343N, and F438W bands of the UVIS/WFC3 and the F435W, F555W, and F814W filters of the ACS/WFC. Further details about the dataset and the procedures applied to reduce it can be found in the paper by Milone and collaborators.

Fig. 3. mF814W vs. mF275W − mF814W (left) CMD and mF814W vs. CF275W, F336W, F438W (right) pseudo-CMD of NGC 6791. In both panels red triangles and blue points correspond to candidate binaries and blue stragglers, respectively, whereas the error bars represent photometric uncertainties.

To determine the photometric uncertainties in the HST dataset of NGC 6791 and NGC 1783 we carried out artificial-star (AS) tests following the method by Anderson et al. (2008). In a nutshell, we generated a list including coordinates and fluxes for 100 000 stars. To do that, we assumed the same radial distribution for the ASs as the one measured for real stars. Moreover, we estimated the magnitudes of ASs by employing a set of fiducial lines derived from the observed CMDs.

To reduce ASs, we used once again KS2 and repeated the procedure described above. Subsequently, similar to our approach with real stars, we took advantage of the various diagnostic parameters provided by KS2 to select a sample of well-measured sources.

2.3. Galactic globular clusters

Legnardi et al. (2022) extensively investigated 1G stars in 55 Galactic GCs based on the ChMs of their RGB stars and inferred their internal metallicity variations. In this work, we further used the same dataset to derive the metallicity distributions of 1G stars and examine whether the color distribution of the 1G sequence is connected to some parameters of the host GCs.

3. An in-depth analysis of first-generation stars in 47 Tucanae

To study 1G stars at progressively larger distances from the center of 47 Tucanae we used the ChMs illustrated in Fig. 4. The Δ_{C F275W, F336W, F438W} versus Δ_{F275W, F814W} ChM of the central field is constructed for evolved RGB stars, whereas the Δ_{F110W, F160W} versus Δ_{F606W, F814W} and the Δ_{F115W, F322W2} versus Δ_{F606W, F115W} ChMs are derived for M dwarfs below the MS knee.

Fig. 4. ChMs of 47 Tucanae for the cluster center and the three external fields, A, B, and C. The ChM of the central region is based on RGB stars, whereas the ChMs of fields A, B, and C are derived for M dwarfs in the lower part of the MS. In all ChMs, 1G and 2G stars are marked with red and gray points, respectively, whereas photometric uncertainties are represented by the error bars in the bottom left corner. The top panels illustrate the kernel-density distributions of 1G stars and observational errors.

As discussed in previous works (Milone et al. 2017, 2023b), the stars of all ChMs are distributed in two main populations, 1G and 2G, marked with red and gray colors, respectively. To identify 1G and 2G stars in the ChMs of RGB and M-dwarf stars, we adopted a similar selection as Milone et al. (2017, see their Fig. 3) and Milone et al. (2023b, see panel b of their Figs. 14 and B2), respectively¹. In all ChMs, 1G stars lay in proximity of the ChM origin, whereas the 2G sequence extends toward high Δ_{C F275W, F336W, F438W} (Δ_{F110W, F160W}, Δ_{F115W, F322W2}) and low Δ_{F275W, F814W} (Δ_{F606W, F814W}, Δ_{F606W, F115W}) values, and exhibits hints of stellar overdensities, especially in the ChMs for M dwarfs.

A visual inspection at Fig. 4 reveals that the 1G stars of 47 Tucanae define an extended sequence along the x-axis of all ChMs. The width of the Δ_{F275W, F814W} (Δ_{F606W, F814W}, Δ_{F606W, F115W}) pseudo-color for 1G stars is much wider than observational errors alone and is associated with star-to-star metallicity variations (Legnardi et al. 2022; Marino et al. 2023). Additionally, the pseudo-colors of 1G stars exhibit complex distributions, as suggested by the corresponding kernel-density distributions plotted in the upper panels of Fig. 4. Clearly, the 1G kernel-density distribution of the central field shows two peaks located at Δ_{F275W, F814W} ∼ −0.18 and ∼ − 0.03 mag, respectively. In the M-dwarf ChMs, instead, the frequency of 1G stars reaches its maximum at Δ_{F606W, F814W} ∼ −0.06 and Δ_{F606W, F115W} ∼ −0.10 mag, gradually declining toward the origin. The Δ_{F110W, F160W} versus Δ_{F606W, F814W} ChM exhibits hints of an additional peak at Δ_{F606W, F814W} ∼ −0.01 mag.

In the following, we further investigate the 1G stars of 47 Tucanae. Specifically, in Sect. 3.1 we use the pseudo-color broadening observed along the x-axis of the ChMs to derive the metallicity distribution of 1G stars at distinct radial distances from the cluster center and investigate the radial behavior of metallicity variations within the 1G. In Sect. 3.2 we exploit the proper motion catalogs introduced in Sect. 2 to study the internal kinematics of 1G stars.

3.1. The radial distribution of metallicity variations among first-generation stars

To infer the [Fe/H] distribution of 1G stars at different radial distances from the core of 47 Tucanae we applied a similar approach to that described in Legnardi et al. (2022), which is based on the method originally developed by Cignoni et al. (2010; see also Cordoni et al. 2022) to derive the age of young star clusters through the MS turn-on.

First, we generated a grid of simulated ChMs, where all stars share the same helium and light-element contents but have different metallicities. In close analogy with Legnardi et al. (2022), we simulated N stellar populations with [Fe/H] ranging from [Fe/H]₀ − 0.5 to [Fe/H]₀ + 0.5, where [Fe/H]₀ = −0.72 (from Harris 1996, 2010 updated). Then, we combined the simulated ChMs to derive the Δ_{F275W, F814W} (Δ_{F606W, F814W}, Δ_{F606W, F115W}) histogram distribution. To do that, we multiplied the ChM of each population j for c_j, a scaling coefficient that ranges from 0 to 1 and tells how much the population j contributes to the final combined Δ_{F275W, F814W} (Δ_{F606W, F814W}, Δ_{F606W, F115W}) distribution. Finally, we compared the resulting simulated histogram with the observed one by minimizing the Poissonian χ² given by the equation

$$\begin{array}{c}\hfill {\chi }^2={\displaystyle \sum _{\mathrm{i}}^{N_{\mathrm{bins}}}}{n}_{\mathrm{i}}ln(n_{\mathrm{i}}/m_{\mathrm{i}})-n_{\mathrm{i}}+m_{\mathrm{i}}\end{array}$$(1)

where n_i and m_i are the bin values of the observed and simulated 1G stars, respectively. For the χ² minimization, as in Cordoni et al. (2022) and Legnardi et al. (2022), we used the geneticalgorithm Python public library².

As an example, the top panels of Fig. 5 compare the ChM of 1G stars in the central field (red points in panel a) with the ChMs of simulated 1G stars (panels b and c). All stars in the yellowish ChM of panel b share the same iron abundances of [Fe/H] = −0.94, while those of the reddish ChM have [Fe/H] = −0.32. The ChM of panel c provides the best match with the observed one. The comparison between the observed Δ_{F275W, F814W} and the best-fitting Δ_{F275W, F814W} pseudo-color distribution is illustrated in panel d. Similar comparisons for 1G stars within fields A, B, and C are provided in panels e, f, and g, respectively.

Fig. 5. Illustration of the procedure to derive the metallicity distributions of 1G stars in 47 Tucanae. Top panels: comparison between the observed ΔC F275W, F336W, F438W vs. ΔF275W, F814W ChM for 1G stars within the central field (red points in panel a) and the ChM of simulated 1G stars (panels b and c). The stars in the yellowish ChM in panel b have an iron abundance of [Fe/H] = −0.94, while those of the reddish ChM have [Fe/H] = −0.32. The ChM of panel c represents the best-fitting simulation. The number of simulated 1G stars is about five times larger than the observed stars. Bottom panels: histograms of the observed ΔF275W, F814W (ΔF606W, F814W, ΔF606W, F115W) distributions inferred from 1G stars within the cluster center (panel d), fields A, B, and C (panels e, f, and g, respectively). In each panel, the red line corresponds to the simulated 1G ΔF275W, F814W (ΔF606W, F814W, ΔF606W, F115W) distribution that provides the best match with the observed data.

Figure 6 illustrates the resulting histograms of [Fe/H] variations among 1G stars in the cluster center, fields A, B, and C, together with their corresponding kernel-density distributions. The [Fe/H] distributions for red giants and M dwarfs span similar intervals of ∼0.1 dex, ranging from δ[Fe/H]_1G ∼ −0.15 to ∼0.05, and exhibit a main peak at δ[Fe/H]_1G ∼ −0.07.

Fig. 6. Histograms of the 1G metallicity distribution inferred from RGB stars within the cluster center (top left) and M dwarfs of fields A (top right), B (bottom left), and C (bottom right). In each panel the corresponding kernel-density distribution is superimposed on the histograms with the continuous black line, whereas the red dashed line represents the kernel-density distribution derived for RGB stars in the cluster core.

To compare the cluster center with the external regions, in panels A, B, and C we plotted the kernel-density distribution of the relative iron abundances of 1G stars within the central field (red-dashed line). Clearly, the latter shows a secondary peak centered at δ[Fe/H]_1G ∼ −0.02 that gradually becomes less evident moving toward the outer fields.

To further investigate the 1G metallicity variations within the cluster, we calculated the average value of the δ[Fe/H]_1G distribution, ⟨δ[Fe/H]_1G⟩, for each of the four analyzed fields, that we plotted as a function of the radial coordinate in Fig. 7. On the bottom axis, we normalized the radial coordinate to the half-light radius from Baumgardt & Hilker (2018, r_hl = 2.78). Conversely, on the top axis, we converted the radial coordinate in parsec by assuming a distance of 4.41 kpc (Baumgardt & Hilker 2018).

Fig. 7. Average metallicity variations among 1G stars, ⟨δ[Fe/H]1G⟩, as a function of the radial distance from the cluster center. The gray points correspond to the average δ[Fe/H] of 1G stars within each analyzed field, whereas the thick and thin error bars indicate the uncertainties associated with ⟨δ[Fe/H]1G⟩ and the dispersions, respectively. On the bottom axis, the distance from the cluster center is normalized to the half-light radius from Baumgardt & Hilker (2018, rhl = 2.78 arcmin). On the top axis, instead, the radial coordinate is converted to parsec by adopting a distance of 4.41 kpc (Baumgardt & Hilker 2018). The horizontal bars mark the extension of each radial interval.

The radial distribution of ⟨δ[Fe/H]_1G⟩ suggests a concentration of more metal-rich 1G stars in the innermost cluster regions, whereas in the outer fields 1G stars have lower average iron content. To demonstrate that this result is not affected by the adopted 1G samples, we repeated the analysis based on 1G stars identified in ChMs that maximize the separation between distinct stellar populations. Our analysis confirmed that we obtained consistent results when deriving [Fe/H] variations from these new samples of 1G stars. As an example, for M dwarfs within field A, we selected 1G stars in the Δ_{C F606W, F814W, F322W2} versus Δ_{F606W, F814W} ChM (see Fig. 8 of Marino et al. 2024), finding that the average [Fe/H] variation differs just by ∼0.001 dex from the original value. Similar results have been found for field C, where we used the Δ_{F110W, F160W, F115W, F322W2} versus Δ_{F606W, F115W} ChM (see Fig. 9 of Marino et al. 2024) to identify a new sample of 1G stars.

However, it is important to note that the outcomes for the central field are derived from RGB stars, while in the other fields, we focused on analyzing M dwarfs. Consequently, it cannot be ruled out that the observed difference in metallicity between the central and external fields may be influenced by the varying stellar masses within these regions.

3.2. Internal kinematics of first-generation stars

To investigate the 1G kinematics, we first identified two groups of 1G stars based on the direction of their color extension in the ChM, with the criteria that each sample includes approximately half of the 1G members. The two subgroups of 1G stars, namely 1G_A and 1G_B, are marked with orange and cyan points in the ChMs of Fig. 8. We then computed the radial (σ_RAD) and tangential (σ_TAN) velocity dispersions of 1G_A and 1G_B following the procedure described in Cordoni et al. (2023a; see also Cordoni et al. 2020a,b), which is based on the minimization of the log-likelihood function by Bianchini et al. (2018). We estimated the corresponding uncertainties through the MCMC algorithm emcee (Foreman-Mackey et al. 2013), as in Cordoni et al. (2023a).

Fig. 8. Reproduction of the ChMs of Fig. 4. The subsamples of 1GA and 1GB stars are highlighted in orange and cyan, respectively.

The resulting radial and tangential velocity dispersion profiles are illustrated in the upper and middle panels of Fig. 9, respectively. Additionally, we calculated the anisotropy parameter, β = σ_TAN/σ_RAD − 1, that we show as a function of the radial coordinate in the lower panel of Fig. 9, where the horizontal dashed line corresponds to an isotropic stellar system. In each panel, orange and cyan points correspond to 1G_A and 1G_B stars, respectively, while on the bottom axis, the distance from the cluster center is normalized to the half-light radius from Baumgardt & Hilker (2018, r_hl = 2.78 arcmin). On the top axis, instead, we converted the radial coordinate in parsec by assuming a distance of 4.41 kpc (Baumgardt & Hilker 2018). To ensure having enough statistics, we treated each of the analyzed fields as one circular bin so that the resulting dynamical profiles are populated by four points for each 1G group. The horizontal bars associated with each point indicate the extension of the corresponding radial interval.

Fig. 9. Velocity dispersion profile along the radial (top panel) and tangential (middle panel) direction, together with the corresponding anisotropy profile (bottom panel). In all the panels, orange and cyan points indicate 1GA and 1GB stars, respectively, while the gray dashed lines correspond to the core and the half-light radius. On the bottom axis, the distance from the cluster center is normalized to the half-light radius from Baumgardt & Hilker (2018, rhl = 2.78 arcmin). On the top axis, instead, the radial coordinate is converted to parsec by adopting a distance of 4.41 kpc (Baumgardt & Hilker 2018). The horizontal bars mark the extension of each radial interval.

Our analysis of 1G stars in 47 Tucanae reveals that 1G_A and 1G_B stars share similar radial and tangential velocity dispersion profiles, ranging from ∼0.6 mas yr⁻¹ (∼13 km s⁻¹) close to the cluster center to ∼0.25 mas yr⁻¹ (∼5 km s⁻¹) at a radial distance of ∼4.5 r_hl. The only exception is represented by the innermost region, where the radial velocity dispersion of 1G_A stars is slightly larger than that of 1G_B members. The anisotropy profile in the bottom panel of Fig. 9 demonstrates that 1G stars exhibit, on average, an isotropic motion, thus confirming previous results in the literature (Richer et al. 2013; Milone et al. 2018a; Cordoni et al. 2020b). Moreover, 1G_A and 1G_B stars show no significant difference in their level of anisotropy.

To test whether our selection of 1G subsamples affected the results, we identified two new groups of 1G_A and 1G_B stars according to the metallicity distribution derived in Sect. 3.1. Then, we recomputed the velocity dispersion and anisotropy profiles finding consistent results with the original ones. Overall, metal-rich and metal-poor 1G stars exhibit isotropic motions with no significant differences in their radial and tangential velocity dispersion profiles.

The fact that the 1G groups of 47 Tucanae have different radial distributions but similar kinematics would deserve further investigation. Stellar populations with different radial distributions typically exhibit different kinematics (see, e.g., Richer et al. 2013; Milone et al. 2018a; Cordoni et al. 2020b, for the case of 1G and 2G stars in 47 Tucanae). More data are needed to better constrain the kinematics of 1G stars with different metallicities, whereas dynamic simulations are crucial to shed light on the initial configuration and the dynamic evolution of metal-rich and metal-poor 1G stars in 47 Tucanae.

4. Simple-population clusters: NGC 6791 and NGC 1783

So far, the color broadening of the 1G sequence has been detected only in Galactic GCs with multiple stellar populations. To verify whether this phenomenon depends on the occurrence of multiple stellar populations, in the following we analyze two simple-population clusters, namely NGC 6791 and NGC 1783.

4.1. The extended stellar population of NGC 6791 and NGC 1783

To investigate whether chemical anomalies similar to the ones detected within the 1G of Galactic GCs are common also in simple-population clusters, we derived the Δ_{C F275W, F336W, F438W} versus Δ_{F275W, F814W} ChMs of NGC 1783 and NGC 6791 by extending the method by Milone et al. (2017) to the two analyzed targets.

As an example, to derive the ChM of NGC 1783 we first selected RGB stars with 16.5 ≤ m_F814W ≤ 18.5 mag. Then, we defined the blue and red boundaries of the RGB in the m_F814W versus C_{F275W, F336W, F438W} pseudo-CMD and the m_F814W versus m_F275W − m_F814W CMD. To do that, we divided the sample of selected stars in magnitude bins of size 0.2 mag and for each bin we calculated the 4th and the 96th percentile of the m_F275W − m_F814W color and C_{F275W, F336W, F438W} pseudo-color distribution. These quantities have been interpolated with the mean m_F814W magnitude of each bin to derive the fiducial lines that mark the blue and red RGB envelopes. Finally, we took advantage of Eqs. (1) and (2) of Milone et al. (2017) to compute the Δ_{F275W, F814W} and Δ_{C F275W, F336W, F438W} pseudo-color. A similar procedure has been used to derive the ChM of NGC 6791, but in this case we considered RGB stars with 13.4 ≤ m_F814W ≤ 16.4 mag.

Results are illustrated in the left and right panels of Fig. 10, where we show the ChMs of NGC 6791 and NGC 1783, respectively. In the bottom left corner of each panel, we plot the distribution expected for a simple stellar population, derived by using ASs, and the ellipse that includes the 68.27% of the simulated points. Additionally, the top panels of each figure report the Δ_{F275W, F814W} kernel-density distribution of observed (black line) and simulated (orange line) stars.

Fig. 10. ΔC F275W, F336W, F438W vs. ΔF275W, F814W ChMs for NGC 6791 (left) and NGC 1783 (right). In both panels the orange points in the bottom left corner indicate the distribution expected for a simple population derived from ASs, while the orange ellipses encircle the 68.27% of simulated points. The ΔF275W, F814W kernel-density distributions of observed and simulated stars are shown in black and orange, respectively, in the top panel of each figure. The insets show a comparison between the ChMs of the two simple-population clusters and that of the metal-rich GC NGC 6624 (Milone et al. 2017).

Similarly to the 1G sequences of Galactic GCs, which move from the origin of the ChM to high Δ_{C F275W, F336W, F438W} and low Δ_{F275W, F814W} values, the ChMs of NGC 6791 and NGC 1783 exhibit a single stellar population that starts in proximity of (Δ_{F275W, F814W}, Δ_{C F275W, F336W, F438W})∼(0.00, 0.00) and extends to (Δ_{F275W, F814W}, Δ_{C F275W, F336W, F438W})∼(−0.40, 0.15) and (Δ_{F275W, F814W}, Δ_{C F275W, F336W, F438W})∼(−0.30, 0.25), respectively. The insets of Fig. 10 reinforce this visual impression, as the ChMs of the two investigated targets overlap with the 1G sequence of NGC 6624, which is one of the most metal-rich Galactic GCs ([Fe/H] = −0.44 dex, Harris 1996, 2010 updated) with available ChM.

As suggested by the corresponding kernel-density distribution in the top panels of Fig. 10, the ChMs of both clusters display an extended sequence in the direction of the Δ_{F275W, F814W} pseudo-color. To estimate the contribution of observational errors to the observed color spread, we quantified the Δ_{F275W, F814W} color spread of observed and simulated stars in the ChMs of the two investigated targets. We found that the observed color spread of NGC 6791 (${\sigma }_{F275W,F814W}^{\mathrm{obs}}=0.129\pm 0.018$ mag) and NGC 1783 (${\sigma }_{F275W,F814W}^{\mathrm{obs}}=0.058\pm 0.003$ mag) are higher than the corresponding color elongation of a simulated simple stellar population (${\sigma }_{F275W,F814W}^{\mathrm{sim}}=0.030$ mag and 0.032 mag, respectively).

To increase the number of available stars in NGC 6791 and study 1G RGB stars within 10 arcmin from the cluster center, we derived the Δ_{C U, B, I} versus Δ_B, I ChM from ground-based photometry, which is based on the B − I color and the C_U, B, I = (U − B)−(B − I) pseudo-color (Jang et al. 2022). A visual inspection at panel a of Fig. 11, where we show the resulting ChM superimposed to the one of NGC 5927 ([Fe/H] = −0.49 dex, Harris 1996, 2010 updated) derived by Jang et al. (2022), reveals that NGC 6791 exhibits a single stellar population which aligns with the 1G of NGC 5927. This result allowed us to confirm that NGC 6791 is a simple-population cluster.

Fig. 11. Results from the ground-based photometry of NGC 6791. Panel a: ΔC U, B, I vs. ΔB, I ChMs of NGC 6791 (red points) and the metal-rich GC NGC 5927 (gray triangles, Jang et al. 2022). Panel b: reproduction of the ground-based ChM of NGC 6791; the two subsamples of 1GA and 1GB stars are shown as orange and cyan points, respectively. Panels c and d: I vs. B–I and I vs. U–V CMDs of RGB stars in NGC 6791, where the fiducial lines for 1GA and 1GB are plotted in orange and cyan, respectively. Panel e: Δ(X − I) (where X = U, B, and V) calculated at Iref = 15.0, against the central wavelength of the X filter. The colors inferred from the best-fitting isochrones are overplotted with blue star symbols (see text for details).

To further investigate the presence of metallicity variations in NGC 6791, we extended to these clusters the procedure used by Milone et al. (2010, see also Anderson et al. 2009) to demonstrate that NGC 6752 hosts multiple populations. As illustrated in panel b of Fig. 11, we first identified two groups of stars in the Δ_{C U, B, I} versus Δ_B, I ChM, namely 1G_A and 1G_B, with the criteria that each sample includes about half of the total number of stars. We then analyzed the distribution of these two groups of stars in CMDs constructed with combinations of filters that are not used to derive the ChM. Panels c and d of Fig. 11 show the I versus B–I and the I versus U–V CMDs where we used orange and cyan colors to identify 1G_A and 1G_B stars, respectively. As highlighted by the corresponding fiducial lines, the two groups of stars remain well separated in both diagrams, thus corroborating the conclusion that the color differences between 1G_A and 1G_B stars with similar luminosity are intrinsic.

To quantify the metallicity difference between the two selected groups of NGC 6791 stars, we compared the observed stellar colors with the solar-scaled isochrones from Dotter et al. (2008) that provide the best fit to the I versus B–I CMD. Specifically, we assumed metallicity corresponding to [Fe/H] = 0.4 dex, age = 8.5 Gyr, distance modulus, (m − M)₀ = 13.5 mag, and reddening, E(B − V) = 0.1 mag. We selected the two groups of 1G stars identified in panel b of Fig. 11 in the I versus X–I CMDs, where X = U, B, and V, and we derived the corresponding fiducial lines. Subsequently, we calculated the Δ(X–I) color difference between the 1G_A and 1G_B fiducials at a reference magnitude of I^ref = 15.0 mag. Results are illustrated in panel e of Fig. 11, where we compare the observed color differences between 1G_A and 1G_B stars with those derived from the best-fitting isochrones. We found that the selected groups of NGC 6791 stars are consistent with a difference in iron abundance of Δ[Fe/H] ∼ 0.06 dex, which provides a lower limit to the maximum iron variations within NGC 6791. In summary, results from ground-based photometry of NGC 6791 confirm the conclusions that we obtained from HST data that this cluster is composed of 1G stars alone, which are not chemically homogeneous.

Our analysis on NGC 6791 and NGC 1783 suggests that the extension of the sequence observed on the ChM of the two simple-population clusters has an intrinsic origin, which is related to star-to-star metallicity variations. Therefore, this phenomenon is not necessarily associated with the occurrence of multiple stellar populations in GCs.

4.2. Metallicity variations within NGC 6791 and NGC 1783

Recent work, based on the m_F275W − m_F814W color width of RGB stars ($W_{F275W,F814W}^{1\mathrm{G}}$) of 55 Galactic GCs, reveals that the maximum [Fe/H] variations among 1G stars, δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$, range from less than ∼0.05 to ∼0.30 dex (Legnardi et al. 2022). In the following, we estimate the maximum [Fe/H] spreads of NGC 6791 and NGC 1783 by adapting the procedure of Legnardi et al. (2022) to the two investigated targets.

For each cluster, we first computed the intrinsic Δ_{F275W, F814W} pseudo-color extension. To do that, in close analogy with Milone et al. (2017), we subtracted the color errors in quadrature to the observed Δ_{F275W, F814W} pseudo-color width, defined as the difference between the 90^th and the 10^th percentile of the Δ_{F275W, F814W} distribution. As suggested by the visual comparison of the ChMs in Fig. 10, the intrinsic width of NGC 6791 ($W_{F275W,F814W}^{1\mathrm{G}}=0.266\pm 0.032$ mag) is consistently higher than that of NGC 1783 ($W_{F275W,F814W}^{1\mathrm{G}}=0.149\pm 0.014$ mag).

Subsequently, we used the relation between the m_F275W − m_F814W color and metallicity from Dotter et al. (2008) to transform the Δ_{F275W, F814W} color extension in [Fe/H] variations, in the assumption that the color spread observed on the ChM is entirely caused by iron spreads. We found that NGC 6791 exhibits moderate [Fe/H] variations (δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}=0.080\pm 0.011$), whereas NGC 1783 is characterized by a lower metallicity spread (δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}=0.037\pm 0.008$).

To compare NGC 6791 and NGC 1783 with the 1G stars of Galactic GCs, we plot in the left and right panels of Fig. 12 the 1G [Fe/H] variations as a function of the logarithm of the cluster mass (from Baumgardt & Hilker 2018) and metallicity (from Harris 1996, 2010 updated), respectively. For NGC 6791 we used the present-day mass and the metallicity computed by Cordoni et al. (2023b) and Frinchaboy et al. (2013), respectively, whereas for NGC 1783 we exploited the values derived from Milone et al. (2023a) and Song et al. (2021).

Fig. 12. Maximum iron variations of 1G stars, δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$, as a function of the logarithm of the cluster mass (left) and metallicity (right). Gray dots correspond to the 55 Galactic GCs analyzed by Legnardi et al. (2022), whereas NGC 6791 and NGC 1783 are marked with red and blue triangles, respectively.

In both panels gray dots correspond to the 55 Galactic GCs studied by Legnardi et al. (2022), whereas the filled triangles indicate NGC 6791 (red) and NGC 1783 (blue). The δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$ values of Galactic GCs exhibit a mild correlation/anticorrelation with the cluster mass/metallicity. Both clusters follow the general trend of both relations, despite the high [Fe/H] variation observed among clusters with similar mass or metallicity.

5. The color distribution of first-generation stars in 51 Galactic globular clusters

The 1G color broadening is a widespread phenomenon among Galactic GCs. However, the collection of ChMs by Milone et al. (2017, see Figs. 3–7) and Jang et al. (2022, see Figs. 12 and 13) reveal that the Δ_{F275W, F814W} pseudo-color extension of 1G stars exhibit substantial variations among different clusters. By exploiting the color-metallicity relations by Dotter et al. (2008), Legnardi et al. (2022) converted the Δ_{F275W, F814W} values into [Fe/H] variations for a large sample of 55 Galactic GCs. Similarly to the Δ_{F275W, F814W} pseudo-color, the resulting values of δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$ changes dramatically from one cluster to another, ranging from ∼0.00 dex to ∼0.30 dex (see Fig. 12 of Legnardi et al. 2022, for more details).

Additionally, we noticed that the ChMs of the 55 GCs studied by Legnardi et al. (2022) reveal different distribution of the 1G Δ_{F275W, F814W} pseudo-color. This fact is illustrated in panels a1–a3 of Fig. 13, where we present the ChMs of NGC 5272, NGC 6637, and NGC 6723 that we consider as test cases. The red points represent 1G stars, while the vertical dotted lines indicate the 10th and the 90th percentiles of the 1G Δ_{F275W, F814W} distribution. To properly compare the pseudo-colors of 1G stars in the different clusters, we normalized the Δ_{F275W, F814W} pseudo-colors by the distance between the two vertical lines shown in the top panels of Fig. 13. Results are illustrated in panels b1–b3 of Fig. 13, where we show the Δ_{C F275W, F336W, F438W} versus ${\mathrm{\Delta }}_{F275W,F814W}^N$ diagrams for 1G stars alone.

Fig. 13. Comparison between the color distribution of 1G stars in NGC 5272, NGC 6637, and NGC 6723. Top panels: ChMs of NGC 5272 (a1), NGC 6637 (a2), and NGC 6723 (a3) where the 1G stars are colored red. Middle panels: zoomed-in image of the ChMs shown in the top panels for 1G stars of NGC 5272 (b1), NGC 6637 (b2), and NGC 6723 (b3). Here, the ΔF275W, F814W pseudo-color is normalized to the width of the 1G sequence. Bottom panels: cumulative (c) and kernel-density (d) distributions of the ${\mathrm{\Delta }}_{F275W,F814W}^N$ pseudo-color for 1G stars of NGC 5272 (dark blue), NGC 6637 (purple), and NGC 6723 (orange).

A visual inspection at the top and middle panels of Fig. 13 reveals that most 1G stars in NGC 5272 and NGC 6273 exhibit low and high values of ${\mathrm{\Delta }}_{F275W,F814W}^N$, respectively. Consequently, the δ[Fe/H] distribution of their 1G stars is dominated by metal-poor and metal-rich stars, respectively. NGC 6637 represents an intermediate case. The differences among the pseudo-color distributions of 1G stars in the ChMs of NGC 5272, NGC 6637, and NGC 6723 are highlighted by the cumulative and the kernel-density distributions plotted in the bottom panels of Fig. 13.

In the following, we further investigate 1G stars in a large sample of 51 Galactic GCs. Specifically, in Sect. 5.1 we parameterize the Δ_{F275W, F814W} pseudo-color distribution of 1G stars, while in Sect. 5.2 we investigate possible relations with the main parameters of the host GC.

5.1. Parameterization of the first-generation color distribution

To parameterize the ${\mathrm{\Delta }}_{F275W,F814W}^N$ distribution of 1G stars, we calculated the area under the cumulative curve, denoted as $A_{1\mathrm{G}}^N$, for 51 Galactic GCs³. Results are listed in Table 2. As illustrated in the left panel of Fig. 14, where we show the histogram distribution of $A_{1\mathrm{G}}^N$, the values of $A_{1\mathrm{G}}^N$ range from ∼0.6 to ∼0.8 mag, with high and low values indicating a predominance of stars with low and high ${\mathrm{\Delta }}_{F275W,F814W}^N$ values, corresponding to metal-poor and metal-rich stars, respectively. The distribution exhibits a main peak around $A_{1\mathrm{G}}^N=0.65$ mag plus a tail toward high values of $A_{1\mathrm{G}}^N$. When we limited the analysis to the GCs with wide 1G sequences and where the 1G stars are clearly separated by the 2G in the ChM⁴, we obtained the same qualitative result. This conclusion is supported by the similarity between the $A_{1\mathrm{G}}^N$ histogram distributions derived for all clusters and the selected GCs alone (left and right panels of Fig. 14, respectively).

Fig. 14. Histograms of the $A_{1\mathrm{G}}^N$ parameter for all of the 51 GCs (left panel) and the selected 37 GCs with wide 1G sequences and well-separated groups of 1G and 2G stars in the ChM (right panel) (see text for details).

5.2. Relations between the first-generation color distribution and the parameters of the host cluster

To test whether the distribution of metal-rich and metal-poor 1G stars can be influenced by the global parameters of the host GC, in the following we investigate the relation between $A_{1\mathrm{G}}^N$ and the main GC parameters. Our analysis includes metallicity ([Fe/H]), reddening (E(B − V)), absolute visual magnitude (M_V), central surface brightness (SB₀), central stellar density (ρ₀), ellipticity (ϵ), concentration (c), and specific RR Lyrae density (S_RR) from the 2010 version of the Harris (1996) catalog. Additionally, we used various parameters from Baumgardt & Hilker (2018), including total cluster mass (M), mass-to-light ratio in the V band (M/L), core density (ρ_c), half-mass-radius density (ρ_hm), half-mass relaxation time (T_RH), slope of mass function (MF slope), mass fraction of the remnants (F_remn), central velocity dispersion (σ₀), central escape velocity (v_esc), and mass segregation parameter inside the core (η_c) and half-mass–radius (η_hm).

GC ages have been taken from Marín-Franch et al. (2009, hereafter MF09), Dotter et al.(2010, hereafter D10), VandenBerg et al. (2013, hereafter V13), and Tailo et al.(2020, hereafter T20), whereas the fraction of binaries in GCs comes from Milone et al. (2012a), as measured within the core (f_bin, c), between the core and the half-mass radius (f_bin, hm), and beyond the half-mass radius (f_bin, ohm).

Moreover, we took advantage of the initial mass of the cluster (M_i), the distance from the Galactic Center (R_GC), the mean heliocentric velocity (RV), the apogalacticon and the perigalaticon radius (R_apog and R_perig) from Baumgardt et al. (2019). HB parameters, such as the 1G and extreme-2G mass loss (μ_1G and μ_2Ge), the difference between the two (δμ_e), and the 1G and 2Ge average HB mass (${\overline{M}}_{1\mathrm{G}}^{\mathrm{HB}}$ and ${\overline{M}}_{2\mathrm{Ge}}^{\mathrm{HB}}$), have been computed by Tailo et al. (2020), while the RGB width in m_F275W − m_F814W (W_F275W) and in C_{F275W, F336W, F438W} (W_C, F275W), the RGB width of the 1G in m_F275W − m_F814W ($W_{F275W}^{1\mathrm{G}}$), the RGB width in C_{F275W, F336W, F438W} without [Fe/H] dependency (ΔW_C, F275W), and the fraction of 1G stars (N_1G/N_T) have been derived from Milone et al. (2017).

We used the mean and maximum internal helium enhancement (δY_2G, 1G and δY_max) together with the internal helium variation within the 1G (δY_1G) from Milone et al. (2018a). The remaining parameters, namely the HB ratio (HBR), the F606W–F814W color distance of the red HB from the RGB (L₁), the HB F606W–F814W color extension (L₂), the RGB width in C_{F336W, F438W, F814W} with and without [Fe/H] dependency (W_C, F336W and ΔW_C, F336W), the internal metallicity variations within the 1G (δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$), the mean differential reddening variation (σ_AF814W) have been taken from Mackey & van den Bergh (2005, HBR), Milone et al. (2014b, L₁ and L₂), Lagioia et al. (2019, W_C, F336W and ΔW_C, F336W), Legnardi et al. (2022, δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$), and Legnardi et al. (2023, σ_AF814W), respectively. Finally, we also included the mass of 1G stars (M_1G) and the dynamical age of the cluster (Age/T_RH), computed by using the fraction of 1G stars (Milone et al. 2017) and the half-mass relaxation time (Baumgardt & Hilker 2018), respectively.

To estimate the statistical correlation between $A_{1\mathrm{G}}^N$ and the parameters listed above, we calculated the Spearman’s rank coefficient (R_S) and the corresponding p-value, which indicates the significance of the correlation. Results are listed in Table 3, where we provide R_S, the p-value, and the number of freedom degrees for each couple of quantities. Intriguingly, $A_{1\mathrm{G}}^N$ exhibits no significant correlation with the considered parameters. We repeated this analysis after dividing the investigated targets in M 3- and M 13-like clusters according to the morphology of their HBs⁵. We found no strong correlation between $A_{1\mathrm{G}}^N$ and the global GC parameters neither for the M 3-like nor the M 13-like GC groups. While the maximum iron variation within 1G stars correlates with GC mass (Legnardi et al. 2022), the result of this analysis implies that the relative abundance of metal-rich and metal-poor 1G stars within a cluster is not influenced by the intrinsic properties of the host GC.

6. Summary and discussions

It is now widely accepted that the 1G stars within GCs commonly display variations in metallicity from star to star, leading to prolonged sequences in the ChMs of MS and RGB stars (e.g., Legnardi et al. 2022; Lardo et al. 2022; Marino et al. 2019a,b, 2023; Tailo et al. 2019). While the metal content of 1G stars has been thoroughly examined in the central regions of an extensive sample encompassing 55 GCs, several unresolved questions persist. As an example, the radial behavior of the metallicity variations within the GCs is still unexplored. Moreover, the sample of studied clusters is entirely composed of old Galactic GCs with multiple stellar populations and ages of more than ∼12 Gyr (e.g., Dotter et al. 2010; Tailo et al. 2020). Hence, we do not know whether metallicity variations are a peculiarity of old Milky Way GCs or are also present in Galactic or extragalactic star clusters younger than 12 Gyr. Finally, the relations between the metallicity distributions of 1G stars and the main parameters of the host clusters are poorly studied. In this work, we investigated these open issues by providing an in-depth analysis of 1G stars in different environments, including Galactic and extragalactic star clusters.

Our first analysis was focused on 47 Tucanae. This dynamically young Galactic GC is an ideal target to investigate the radial behavior of 1G stars with different metallicities. The fact that its 2G stars are significantly more centrally concentrated than the 1G suggests that the stellar populations of 47 Tucanae retain a memory of their initial distribution during the cluster formation. Specifically, we used HST and JWST data to study 1G stars in the cluster core and three external fields, located at ∼7, ∼8.5, and ∼11 arcmin from the cluster center, respectively. The results can be summarized as follows:

• We derived the ChMs of RGB stars in the central field and of M-dwarf stars in the external fields. We detected a well-elongated 1G sequence in each ChM, which is not consistent with a chemically homogeneous stellar population.

• We took advantage of the pseudo-color elongation along the x-axis of the ChM to infer the [Fe/H] distribution of 1G stars. All the distributions span similar intervals, included between δ[Fe/H]_1G ∼ −0.15 dex and ∼0.05 dex, and display a major peak at δ[Fe/H]_1G ∼ −0.07 dex. The [Fe/H] distribution of the cluster center shows an additional peak centered at δ[Fe/H]_1G ∼ −0.02, which gradually disappears moving outward. This fact suggests that 1G stars richer in metals are more centrally concentrated than the metal-poor ones.

• We identified two main groups of metal-rich and metal-poor 1G stars in the four studied fields and used stellar proper motions to investigate their internal kinematics. We found that both groups of 1G stars share similar radial and tangential velocity dispersion profiles and are characterized by isotropic motions.

McKenzie & Bekki (2021) used hydrodynamical simulations to reproduce a giant molecular cloud forming GCs in a high-redshift dwarf galaxy. The simulated clusters exhibit star-to-star metallicity variations of ∼0.1 dex, which are consistent with the [Fe/H] spread that we observe among 1G stars in 47 Tucanae (see also Legnardi et al. 2022; Marino et al. 2023). Such metallicity variations are due to the merging of gas clumps and self-enrichment processes. Intriguingly, the evidence of a small increase in the fraction of metal-rich stars in 47 Tucanae is consistent with the properties of most forming GCs simulated by McKenzie & Bekki (2021), where the metal-rich stars are more centrally concentrated than the metal-poor ones.

We notice that while the metallicity distribution of the central field significantly differs from those of the external fields, we detect a less pronounced metallicity gradient only when we consider the three external fields alone. In particular, the average metallicity variations among the 1G stars of these fields differ by ∼3σ only. Since the results for the central field are obtained from RGB stars, whereas the conclusions of the external regions are derived from M dwarfs, an alternative explanation is that the observed metallicity-distribution differences in the central and external fields are not due to a radial gradient but reflect different mass functions for the groups of stars with different metallicities.

Recent works have suggested that the properties of stellar populations with different light-element abundances do not significantly depend on stellar mass. This conclusion is based on the evidence that stars with different masses in various GCs, including 47 Tucanae, share the same [O/Fe] abundances and that their 1G and 2G stars have similar mass functions (Milone et al. 2014a, 2019; Dondoglio et al. 2022; Marino et al. 2024). The potential influence of stellar mass on the observed variances in the metallicity distributions of 1G stars would not alter these conclusions. It is worth noting that the distinct mass functions among groups of metal-poor and metal-rich stars within the 1G do not significantly affect the mass functions and the oxygen abundances of 1G and 2G stars⁶.

To investigate whether the 1G color extension phenomenon occurs only in clusters hosting multiple stellar populations or not, we analyzed HST multi-band photometry of two simple-population star clusters, namely the Galactic open cluster NGC 6791 and the LMC cluster NGC 1783. The main outcomes of this second analysis include:

• We derived the standard Δ_{C F275W, F336W, F438W} versus Δ_{F275W, F814W} ChM for RGB stars in NGC 6791 and NGC 1783. The ChMs of both clusters exhibit one sequence of stars alone, which resembles the 1G sequence of GCs with multiple populations. This fact confirms that both clusters are consistent with simple populations with homogeneous nitrogen abundances. The evidence that NGC 6791 hosts stars with different metallicities is confirmed by wide-field ground-based photometry in the U, B, V, and I bands.

• The F275W–F814W color extension of both NGC 6791 and NGC 1783 is much wider than the color spread due to the observational errors alone. Hence, we concluded that extended sequences in the ChM are not prerogatives of 1G stars in GCs but are also present in simple-population clusters.

• We estimated the maximum [Fe/H] variations among the stars of NGC 6791 and NGC 1783, by assuming that iron spreads are the main drivers of the color extension observed on the ChM. We found that NGC 6791 exhibits a moderate [Fe/H] variation (δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}=0.080\pm 0.011$ dex), whereas NGC 1783 shows a much lower iron spread (δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}=0.037\pm 0.008$ dex). Such metallicity variations are comparable to those observed in metal-rich GCs with similar masses.

Based on a sample of 51 ChMs studied by Legnardi et al. (2022), we noticed that the Δ_{F275W, F814W} pseudo-color distribution significantly changes from one cluster to another. In some GCs, including NGC 2808 and NGC 6723, most 1G stars have low Δ_{F275W, F814W} values. Hence, the δ[Fe/H] distribution of their 1G stars is dominated by metal-poor stars. Conversely, other clusters, such as NGC 5272 and NGC 3201, host higher fractions of metal-rich stars with high values of Δ_{F275W, F814W}. To parameterize the color distribution of 1G stars, we computed the area under the cumulative curve of the normalized Δ_{F275W, F814W} distribution, $A_{1\mathrm{G}}^N$. Intriguingly, the values of $A_{1\mathrm{G}}^N$ exhibit no significant correlation with any of the 53 analyzed parameters of the host cluster. This fact could indicate that the relative numbers of metal-rich and metal-poor 1G stars do not depend on the properties of the host GC.

Acknowledgments

We thank the anonymous referee for various suggestions that improved the quality of the manuscript. This work has been funded by the European Union – NextGenerationEU RRF M4C2 1.1 (PRIN 2022 2022MMEB9W: “Understanding the formation of globular clusters with their multiple stellar generations”, CUP C53D23001200006), from INAF Research GTO-Grant Normal RSN2-1.05.12.05.10 – (ref. Anna F. Marino) of the “Bando INAF per il Finanziamento della Ricerca Fondamentale 2022”, and from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie Grant Agreement No. 101034319 and the European Union – NextGenerationEU (beneficiary: T. Ziliotto).

References

All Tables

All Figures

	Fig. 1. Footprints of the regions of 47 Tucanae analyzed in this paper. The central region (blue) and field A (red) include only HST data, whereas field B (cyan) and C (orange) also cover JWST images.
In the text

	Fig. 2. mF115W vs. mF115W − mF322W2 CMD of 47 Tucanae stars within field C (Marino et al. 2024). The inset shows the ΔF115W, F322W2 vs. ΔF606W, F115W ChM.
In the text

	Fig. 3. mF814W vs. mF275W − mF814W (left) CMD and mF814W vs. CF275W, F336W, F438W (right) pseudo-CMD of NGC 6791. In both panels red triangles and blue points correspond to candidate binaries and blue stragglers, respectively, whereas the error bars represent photometric uncertainties.
In the text

Fig. 4. ChMs of 47 Tucanae for the cluster center and the three external fields, A, B, and C. The ChM of the central region is based on RGB stars, whereas the ChMs of fields A, B, and C are derived for M dwarfs in the lower part of the MS. In all ChMs, 1G and 2G stars are marked with red and gray points, respectively, whereas photometric uncertainties are represented by the error bars in the bottom left corner. The top panels illustrate the kernel-density distributions of 1G stars and observational errors.

In the text

Fig. 5. Illustration of the procedure to derive the metallicity distributions of 1G stars in 47 Tucanae. Top panels: comparison between the observed ΔC F275W, F336W, F438W vs. ΔF275W, F814W ChM for 1G stars within the central field (red points in panel a) and the ChM of simulated 1G stars (panels b and c). The stars in the yellowish ChM in panel b have an iron abundance of [Fe/H] = −0.94, while those of the reddish ChM have [Fe/H] = −0.32. The ChM of panel c represents the best-fitting simulation. The number of simulated 1G stars is about five times larger than the observed stars. Bottom panels: histograms of the observed ΔF275W, F814W (ΔF606W, F814W, ΔF606W, F115W) distributions inferred from 1G stars within the cluster center (panel d), fields A, B, and C (panels e, f, and g, respectively). In each panel, the red line corresponds to the simulated 1G ΔF275W, F814W (ΔF606W, F814W, ΔF606W, F115W) distribution that provides the best match with the observed data.

In the text

Fig. 6. Histograms of the 1G metallicity distribution inferred from RGB stars within the cluster center (top left) and M dwarfs of fields A (top right), B (bottom left), and C (bottom right). In each panel the corresponding kernel-density distribution is superimposed on the histograms with the continuous black line, whereas the red dashed line represents the kernel-density distribution derived for RGB stars in the cluster core.

In the text

Fig. 7. Average metallicity variations among 1G stars, ⟨δ[Fe/H]1G⟩, as a function of the radial distance from the cluster center. The gray points correspond to the average δ[Fe/H] of 1G stars within each analyzed field, whereas the thick and thin error bars indicate the uncertainties associated with ⟨δ[Fe/H]1G⟩ and the dispersions, respectively. On the bottom axis, the distance from the cluster center is normalized to the half-light radius from Baumgardt & Hilker (2018, rhl = 2.78 arcmin). On the top axis, instead, the radial coordinate is converted to parsec by adopting a distance of 4.41 kpc (Baumgardt & Hilker 2018). The horizontal bars mark the extension of each radial interval.

In the text

	Fig. 8. Reproduction of the ChMs of Fig. 4. The subsamples of 1GA and 1GB stars are highlighted in orange and cyan, respectively.
In the text

Fig. 9. Velocity dispersion profile along the radial (top panel) and tangential (middle panel) direction, together with the corresponding anisotropy profile (bottom panel). In all the panels, orange and cyan points indicate 1GA and 1GB stars, respectively, while the gray dashed lines correspond to the core and the half-light radius. On the bottom axis, the distance from the cluster center is normalized to the half-light radius from Baumgardt & Hilker (2018, rhl = 2.78 arcmin). On the top axis, instead, the radial coordinate is converted to parsec by adopting a distance of 4.41 kpc (Baumgardt & Hilker 2018). The horizontal bars mark the extension of each radial interval.

In the text

Fig. 10. ΔC F275W, F336W, F438W vs. ΔF275W, F814W ChMs for NGC 6791 (left) and NGC 1783 (right). In both panels the orange points in the bottom left corner indicate the distribution expected for a simple population derived from ASs, while the orange ellipses encircle the 68.27% of simulated points. The ΔF275W, F814W kernel-density distributions of observed and simulated stars are shown in black and orange, respectively, in the top panel of each figure. The insets show a comparison between the ChMs of the two simple-population clusters and that of the metal-rich GC NGC 6624 (Milone et al. 2017).

In the text

Fig. 11. Results from the ground-based photometry of NGC 6791. Panel a: ΔC U, B, I vs. ΔB, I ChMs of NGC 6791 (red points) and the metal-rich GC NGC 5927 (gray triangles, Jang et al. 2022). Panel b: reproduction of the ground-based ChM of NGC 6791; the two subsamples of 1GA and 1GB stars are shown as orange and cyan points, respectively. Panels c and d: I vs. B–I and I vs. U–V CMDs of RGB stars in NGC 6791, where the fiducial lines for 1GA and 1GB are plotted in orange and cyan, respectively. Panel e: Δ(X − I) (where X = U, B, and V) calculated at Iref = 15.0, against the central wavelength of the X filter. The colors inferred from the best-fitting isochrones are overplotted with blue star symbols (see text for details).

In the text

	Fig. 12. Maximum iron variations of 1G stars, δ[Fe/H]${}_{1\mathrm{G}}^{\mathrm{MAX}}$, as a function of the logarithm of the cluster mass (left) and metallicity (right). Gray dots correspond to the 55 Galactic GCs analyzed by Legnardi et al. (2022), whereas NGC 6791 and NGC 1783 are marked with red and blue triangles, respectively.
In the text

Fig. 13. Comparison between the color distribution of 1G stars in NGC 5272, NGC 6637, and NGC 6723. Top panels: ChMs of NGC 5272 (a1), NGC 6637 (a2), and NGC 6723 (a3) where the 1G stars are colored red. Middle panels: zoomed-in image of the ChMs shown in the top panels for 1G stars of NGC 5272 (b1), NGC 6637 (b2), and NGC 6723 (b3). Here, the ΔF275W, F814W pseudo-color is normalized to the width of the 1G sequence. Bottom panels: cumulative (c) and kernel-density (d) distributions of the ${\mathrm{\Delta }}_{F275W,F814W}^N$ pseudo-color for 1G stars of NGC 5272 (dark blue), NGC 6637 (purple), and NGC 6723 (orange).

In the text

	Fig. 14. Histograms of the $A_{1\mathrm{G}}^N$ parameter for all of the 51 GCs (left panel) and the selected 37 GCs with wide 1G sequences and well-separated groups of 1G and 2G stars in the ChM (right panel) (see text for details).
In the text