The VISCACHA survey – VII. Assembly history of the Magellanic Bridge and SMC Wing from star clusters

ABSTRACT

The formation scenario of the Magellanic Bridge during an encounter between the Large and Small Magellanic Clouds ∼200 Myr ago, as proposed by N-body models, would be imprinted in the chemical enrichment and kinematics of its stars and sites of ongoing star formation along its extension. We present an analysis of 33 Bridge star clusters using photometry obtained with the SOAR 4-m telescope equipped with adaptive optics for the VISCACHA survey. We performed a membership selection and derived self-consistent ages, metallicities, distances, and reddening values via statistical isochrone fitting, as well as tidal radii and integrated masses from structure analysis. Two groups are clearly detected: 13 well-studied clusters older than the Bridge, with 0.5–6.8 Gyr and |$\hbox{[Fe/H]} < -0.6$| dex; and 15 clusters with <200 Myr and |$\rm {[Fe/H]} > -0.5$| dex, probably formed in situ. The old clusters follow the overall age and metallicity gradients of the SMC, whereas the younger ones are uniformly distributed along the Bridge. The main results are as follows: (i) we derive ages and metallicities for the first time for 9 and 18 clusters, respectively; (ii) we detect two metallicity dips in the age-metallicity relation of the Bridge at ∼200 Myr and 1.5 Gyr ago for the first time, possibly chemical signatures of the formation of the Bridge and Magellanic Stream; (iii) we estimate a minimum stellar mass for the Bridge of 3–5 × 10⁵M_⊙; (iv) we confirm that all the young Bridge clusters at |${\rm RA} < 3^h$| are metal-rich with |$\rm {[Fe/H]} \sim -0.4$| dex.

1 INTRODUCTION

The Magellanic System consists of the Large and Small Magellanic Clouds (LMC and SMC, or jointly MCs), the Magellanic Stream, Bridge, and Leading Arm. The Bridge contains mostly H i gas but also a stellar population with a mass of 1.5 × 10⁴ M_⊙ (Harris 2007). In particular, the existence of the Stream and Bridge provides evidence of tidal interactions between the MCs, and studying their metallicity content and stellar ages can enlighten our knowledge on the Bridge formation.

A few characteristics of the LMC and SMC structures are also evidence of their interaction process and the formation of the Bridge. The LMC is located at a distance of 49.6 ± 0.5 kpc (Pietrzyński et al. 2019) and presents a flat disc morphology with a single spiral arm and an asymmetric stellar bar (e.g. van der Marel 2001; Ruiz-Lara et al. 2020), as well as a warped outer stellar disc possibly connected to the Bridge (e.g. Choi et al. 2018a; Saroon & Subramanian 2022). The SMC is located at a distance of 62.4 ± 0.8 kpc (Graczyk et al. 2020) and, due to strong interactions with the LMC, presents an elongated, triaxial structure with a line-of-sight depth of up to 14 kpc in the inner regions (Subramanian & Subramaniam 2012) and ∼23 kpc in the eastern part (Nidever et al. 2013). The Bridge extent corresponds to the physical distance between the MCs (∼20 kpc), with a ∼10 kpc depth, as traced by Cepheid and RR Lyrae stars (Jacyszyn-Dobrzeniecka et al. 2016, 2017; Ripepi et al. 2017; Muraveva et al. 2018).

A mean metallicity of |$\rm {[Fe/H]}=-0.37\pm 0.15$| dex was obtained from CaII triplet (CaT) lines for 373 LMC red giants by Cole et al. (2005). The SMC has a mean metallicity of |$\rm {[Fe/H]}=-0.99\pm 0.01$| dex (Dobbie et al. 2014) and −0.9 ± 0.2 dex (Parisi et al. 2016), derived from cluster and field red giants also using CaT lines. The metallicity gradients in the inner regions of the LMC and SMC are relatively well known (e.g. Parisi et al. 2009; Choudhury, Subramaniam & Cole 2016; Parisi et al. 2016; Choudhury et al. 2018; Muñoz et al. 2023; Povick et al. in prep.), whereas the outer regions are less explored and present much higher dispersion. For the SMC, for example, Parisi et al. (2014, 2015, 2022) and Dobbie et al. (2014) detected a decreasing trend in metallicity up to a semimajor axis a ∼ 4.5° and an inversion after that, using cluster and field stars. A similar trend was obtained in Bica et al. (2020), who made a compilation of star clusters with heterogeneous ages and metallicities from the literature.

The Bridge is closely related to the MCs, sharing a common H i envelope with them particularly as a continuity of the SMC Wing (e.g. Putman et al. 2003). Metallicity measurements of the gas from absorption lines resulted to be |$\rm {[M/H]} = -1.0$| and −1.3 dex for two massive stars in different sightlines (Lehner et al. 2008), and −0.8 dex using a background quasar (Misawa et al. 2009). These values are closer to the SMC metallicity than to that of the LMC, suggesting that the Bridge may have been formed mostly from SMC material, even though radial velocity measurements indicate that LMC gas might be present as well (McGee & Newton 1986).

The Bridge contains a blue, young stellar population detected by Irwin, Kunkel & Demers (1985), as well as intermediate-age (1–3 Gyr) and old stellar populations (>3 Gyr) detected later (Bagheri, Cioni & Napiwotzki 2013; Noël et al. 2013; Belokurov et al. 2017). The young population presents a higher concentration closer to the SMC Wing and another halfway between the MCs, with a strong correlation with the distribution and kinematics of the H i gas (e.g. Skowron et al. 2014; Murray et al. 2019; Zivick et al. 2019). On the other hand, the old population is uniformly distributed between the MCs, with a large number of RR Lyrae stars in the southern part (Belokurov et al. 2017). Recent proper motions from Hubble Space Telescope, Gaia, and Visible and Infrared Survey Telescope for Astronomy (VISTA) showed a flow motion of both young and red clump stars in the Wing/Bridge pointing from the SMC towards the LMC (Zivick et al. 2019; Schmidt et al. 2020; Gaia Collaboration et al. 2021b), with larger motions in regions of lower stellar density. Schmidt et al. (2020) found that the Bridge is stretching as its borders close to the SMC and LMC are moving apart relative to the proper motion of its central region.

Based on different proper motion epochs (Kallivayalil et al. 2006a; Kallivayalil, van der Marel & Alcock 2006b; Kallivayalil et al. 2013) and initial conditions, two main N-body simulations attempt to reproduce the formation of the LMC-SMC pair, providing different scenarios: the LMC captured the SMC ∼1.2 Gyr ago and they are in a bound orbit around the Milky Way (MW) since then (often referred to as bound scenario; Diaz & Bekki 2011), or it is an old interacting system formed about 5–6 Gyr ago in its first perigalactic passage, falling into the MW potential since ∼2 Gyr ago (unbound scenario; Besla et al. 2007, 2012). Although more evidence points to the unbound scenario (e.g. Conroy et al. 2021), the masses of the MW and LMC are the main sources of errors. Patel et al. (2020) computed orbits of MW and LMC satellites adopting masses of 1.0 and |$1.5\times 10^{12}\,{\rm M}_\odot$| for the MW, and 0.8, 1.8, and |$2.5\times 10^{11}\,{\rm M}_\odot$| for the LMC, showing the impact of this uncertainty on the orbits of the system members. Even so, a study on age and metallicity of Bridge objects might shed some light on its formation. Some works focused on simulating the formation and filamentary structure of the Stream (e.g. Diaz & Bekki 2012; Hammer et al. 2015), reproducing well the H i observations.

Both scenarios are able to reproduce the large-scale gas structures (Stream, Leading Arm, and gaseous Bridge), with the Bridge being formed during the most recent encounter between the MCs, with gas and stars pulled out from the SMC through tidal interactions and possibly dragged from the LMC to the Bridge (Besla et al. 2012). There is an uncertainty on the exact epoch of this encounter: 250 Myr (Diaz & Bekki 2011), 200 Myr (Gardiner & Noguchi 1996; Besla et al. 2012), 150 Myr (Zivick et al. 2018), and <250 Myr with an impact parameter of 5–10 kpc (Choi et al. 2022). This scenario of the Bridge formation would imply a gradient of increasing metallicity towards the LMC due to a minor contribution of its more metal-rich gas (Besla et al. 2012), and the presence of an older stellar population stripped from the SMC amidst a young population formed in situ. The recent encounter could also have triggered the formation of a ring-like structure in the LMC periphery (also found in the distribution of star clusters, e.g. van der Marel 2001; Bica et al. 2008; Choi et al. 2018b) and of the warp along the northeast-southwest direction.

The stellar clusters located in the Magellanic System are important tracers of the star formation history and chemodynamical evolution of the LMC-SMC pair. A catalog of 2741 stellar clusters, associations, and extended objects in the SMC and Bridge was reported by Bica et al. (2020, and references therein), containing over 400 objects in the Wing/Bridge. Most of the Wing/Bridge objects are younger than 1 Gyr (e.g. Bica et al. 2015), but older objects are found as well, reaching ∼6 Gyr (e.g. Piatti et al. 2015).

In the present work, we analysed 33 clusters located in the Wing/Bridge, that were observed with the adaptive optics system of the 4-m SOuthern Astrophysical Research (SOAR) telescope within the VIsible Soar photometry of star Clusters in tApii and Coxi HuguA (VISCACHA survey¹; Maia et al. 2019). We derived structural parameters from radial density profiles, as well as age, metallicity, distance, and reddening from decontaminated colour-magnitude diagrams (CMDs), with the aim of verifying the existence of gradients.

This paper is structured as follows. Observations and reductions are described in Section 2. Methods of analysis (radial density profile fitting, decontamination, and isochrone fitting) are detailed in Section 3. Results are presented in Section 4. The age and metallicity gradients, followed by a thorough discussion about their implications to the dynamical models, are given in Section 5. Conclusions are drawn in Section 6.

2 OBSERVATIONS AND DATA REDUCTION

The observations were carried out through 2016B-2021B semesters, using the 4.1-m SOAR (SOuthern Astrophysical Research) telescope, mounted with its ground-layer adaptive optics imager (SAMI; Tokovinin 2013; Tokovinin et al. 2016) and using a 4 K × 4 K CCD detector with 2 × 2 binning, yielding a plate scale of 0.091 arcsec pixel ⁻¹ and a field of view of |$3.1\times 3.1\, \rm {arcmin}^2$|⁠. The cluster sample was observed in V and I filters with effective exposure times of 20 and 30 min (after co-addition), respectively, reaching a signal-to-noise ratio of ∼10 and final errors of ∼0.1 mag at V ∼ 24 mag (see Maia et al. 2019, hereafter Paper I). Short exposures were also taken to sample saturated stars. Table 1 presents the log of observations for the sample clusters, containing the observation date, airmass, seeing, and the measured image quality (<1 arcsec in most cases) .

Fig. 1 gives the projected distribution of the SMC and Bridge objects relative to the SMC centre. Concentric ellipses with semi-major axis from 2 to 10° ( b / a = 1/2, inclination of 45°; Piatti et al. 2005) are used to obtain the projected cluster distance to the SMC centre. The clusters observed in the VISCACHA survey are marked in red (SAMI) and blue (Goodman, backup instrument), whereas the present sample is indicated with black border circles. Fig. 2 shows the colour composite images obtained with the V and I filters for the pairs HW81+HW82 (two young clusters), L92+L93 (young versus old cluster), and L113 (old, populous cluster).

The data were processed for bias subtraction and division by skyflats in a standard way using the IRAF CCDRED package. Cosmic rays were removed from the images using the CRUTIL package. Astrometric calibration was performed with IRAF MSCCMATCH task using Gaia Early Data Release 3 catalogues (Gaia Collaboration et al. 2021a) as astrometric reference, this way achieving typical RMS residuals inferior to 0.1 arcsec on source positions. The images World Coordinate Systems were then used to register them to a common frame and co-add them into the final mosaics. PSF photometry was carried out using a modified version of the STARFINDER code (Diolaiti et al. 2000). For the photometric calibrations, we employed Stetson (2000) and MCPS (Magellanic Clouds Photometric Survey; Zaritsky et al. 2002) fields, when available. Further details are given in Paper I.

A total of 33 clusters were observed inside 31 fields. HW81 and HW82 were observed in a same field, and there is apparently a third structure named HW81w by Bica et al. (2020) that was not retrieved. HW71se also has a nearby cluster named HW71nw, which we could not detect. Six associations observed with Goodman will be analysed in a future work: NGC456 and NGC460 (2021B, 1 night); ASS67, WG5, ASS65, and ASS66 (2020B, 2 nights). The VISCACHA sample contains five clusters in the LMC western side close to the Bridge, observed in 2017B, but we will defer the LMC-Bridge connection to a future work, after the observations of the clusters in this region are completed.

In Table 2, we report the literature results on age and metallicity for the sample clusters, including the compilation of Bica et al. (2020) and subsequent work. The reported age determinations were mainly based on visual isochrone fits, and most of them fixed the metallicity and/or distance: Glatt, Grebel & Koch (2010) and Maia, Piatti & Santos (2014) assumed |$\rm {[Fe/H]}=-0.58$| dex and 60.3 kpc; Piatti et al. (2007a, b) assumed |$\rm {[Fe/H]}=-0.7$| dex and 56.8 kpc; and Piatti (2011a, b) and Piatti et al. (2011, 2015) assumed |$\rm {[Fe/H]}=-0.7$| dex and 60.3 kpc. The only papers which provided four free parameters were Bica et al. (2015), Perren, Piatti & Vázquez (2017, ASteCA code) and Paper I, where the metallicities are derived from CMDs or CaT spectroscopy (Da Costa & Hatzidimitriou 1998; Parisi et al. 2009, 2015; De Bortoli et al. 2022). In this work, we present the isochrone fits to the observed data, indicating our results for age, metallicity, distance, and reddening as free parameters, in some cases compared with available literature results.

3 METHODOLOGY

3.1 Radial density profile: King models

We have obtained the radial density profiles (RDPs) of the sample clusters by counting the number of stars in concentric rings around the centre. As a first step, the centre of each cluster is determined iteratively from the centroid of stellar coordinates inside a 50 arcsec radius aperture, starting with an initial guess and adjusting the new centre and aperture at each step. Whenever convergence is achieved (i.e. centre coordinates change is inferior than 0.5 arcsec), the aperture size is reduced to 2/3 of its current size and the process is repeated, until aperture reaches 10–20 arcsec, depending on the cluster size and concentration. At this point, the RDP is checked to ensure that the maximum central density was reached. The 4-parameter analytical model from King (1962) was fitted to the observed profiles, according to the equation:

where ρ₀ is the projected central density, r_c and r_t are the core and tidal radii, and ρ_bg is the projected density of background stars.

Different bin sizes were used to obtain the RDPs, with the smallest bin size limited by the cluster core radius (obtained iteratively). The core and tidal radii, central and background density are obtained by a χ² minimization method to find the best-fitting model, comparing the observed and predicted density of stars in each bin. The best solution is usually uniquely defined and its uncertainties depends mostly on the star counts statistics and on the stellar background fluctuations. This method was already applied to VISCACHA data in Paper I and Santos et al. (2020, Paper II).

In this work, since we are mainly interested in the cluster's tidal radius, which gives us a measure of its size, we applied no correction for completeness. As discussed in Paper II, the tidal radius and cluster centre derived from RDPs are largely unaffected by incompleteness, in contrast with the core radius. We will defer a more complete structural analysis of these clusters to a future work. Fig. 3 shows the King profile fitting over the RDP of L113.

Although both the core and tidal radii are derived in the fitting procedure, we will only list the latter in Table 3. The fit did not converge for three sample clusters (HW59, B147, and HW82), for which the decontamination was carried out with a visual radius to limit the cluster sample. As in Paper II, we obtained a tidal radius larger than the SAMI field of view (r_t ≳ 100 arcsec) for seven sample clusters, that are probably biased towards smaller values, due to the reduced cluster coverage area (relative to the cluster full size) used in their derivation (see Olivares et al. 2019). For smaller clusters, it is possible to check that our RDP fitting has sufficient spatial coverage, by extending the sample to larger radii and verifying the stability of the derived tidal radii.

3.2 Statistical decontamination: photometric membership

Since no proper motion data are available for the faint stars of the studied clusters, the membership probability was obtained from a statistical analysis as first described in Maia, Corradi & Santos (2010). The method compares probable cluster stars within the tidal radius with those in a nearby field. The distance of each star to the cluster centre and the local overdensity of stars in a CMD grid (compared with the same region in the field CMD) are used to define a cluster membership probability and clean the cluster sample from field stars. This method was also adopted in previous VISCACHA papers: Paper I, Dias et al. (2021, Paper III), Dias et al. (2022, Paper IV), and Bica et al. (2022).

Fig. 4 illustrates the method for the Bridge cluster HW77. In this case, the stars within a radius of 0.7r_t were compared to a nearby field, for which the lower main sequence (MS) and the red clump (RC) are similar to the cluster, but with an older main sequence turnoff (MSTO). Therefore, the decontamination was very effective in removing field stars in V ∼ 22 mag, which could deviate the isochrone fitting to an older MSTO. Also note the smooth luminosity function throughout the MS of the decontaminated sample.

3.3 Statistical isochrone fitting

The SIRIUS code developed by our group (Souza et al. 2020) applies a self-consistent Bayesian isochrone fitting to analyse the CMDs of stellar clusters. Through a Markov chain Monte Carlo (MCMC) algorithm, the code is able to derive for each cluster their probability distributions for age, metallicity, absolute distance modulus, and reddening. A likelihood function sums up a chi-square in magnitude and colour for each star compared to the closest isochrone point and divided by the photometric errors, as given below:

where the index i covers all the N member stars, mag and col are the observed V and V − I, p_memb is the membership probability, and n_* is the number of neighbouring stars in the CMD. The membership probability is implemented in such a way that higher-membership stars contribute more in the fitting process. Similarly, the n_* quantity is also considered to reduce the weight of stars in the more populated CMD regions, in particular the MS in favour of the more evolved sequences.

The Padova and TRieste Stellar Evolution Code set of isochrones (PARSEC²; Bressan et al. 2012) was adopted, covering masses of 0.1 to |$350\,{\rm M}_\odot$|⁠, metallicities of |$-1.5 < \rm {[Fe/H] (dex)} < 0.0$| and ages between 1 Myr and 10 Gyr, which represent well the SMC stellar population. For clusters younger than ∼100 Myr, the pre-main sequence stage is clearly detected in the isochrones and CMDs (see Fig. 5). Our grid of isochrones has steps of 0.01 dex both in |$\mathbf {\log (\rm {age})}$| and metallicity, but linear interpolations are carried out to get intermediate values. A total to selective extinction ratio of R_V = 3.1 is adopted.

The python libraries emcee and corner (Foreman-Mackey et al. 2013; Foreman-Mackey 2016) were adapted to carry out the MCMC sampling and produce the corner plots with the posterior distributions in different projections.

Some prior distributions were employed depending on the cluster age. For clusters older than ∼1 Gyr with a well-defined RC, selected stars in this stage are used to match the respective isochrone points, thus helping constrain the distance and metallicity. For the younger clusters, the isochrone MSTO, identified by the sharp reduction in the number of stars at the top of the MS, was used instead. In these cases, a compromise was obtained between an optimal fit of the MSTO and the entire MS, e.g. for L107 (Fig. 5), HW82 and BS226. For the very young clusters with few or no giants detected, the metallicity was first estimated from a tentative fit without any prior information, followed by a new fit with a gaussian prior, in order to obtain a better convergence and circumvent the degeneracy between (m − M)₀ and |$\rm {[Fe/H]}$|⁠.

Following what was done in Paper III and Paper IV, a metallicity prior was adopted for L91, L110, HW86, and L113, centred on the value derived from CaT spectroscopy given in Table 2, in order to improve the age accuracy. A limit of ∼1 Gyr is generally assumed as the minimum age for which the CaT is well calibrated (Parisi et al. 2009), which is coherent with the limit between old clusters with red giants and young clusters, defined by Phelps, Janes & Montgomery (1994) and Janes & Phelps (1994) as those older than the Hyades (787 Myr; Mermilliod 1996).

Cluster parameters and uncertainties were derived by fitting a skewed gaussian over the marginal posterior distribution of each parameter, as shown in Fig. 6. Even though this method produces slightly larger uncertainties than those normally derived from 16th and 86th percentiles, we have found that it better matches the expected (i.e. peak) value of the parameters’ distribution. The resulting values and uncertainties are shown in Table 3.

3.4 Integrated magnitudes and masses

To derive the total masses for the clusters, we determined their integrated apparent V magnitudes (V_int) by adding up the member star fluxes. We then converted V_int to the absolute one (M_V) by using the clusters’ individual distance and extinction. Finally, the mass and its uncertainty were calculated following the calibration with age and metallicity (fixed at |$\rm {[Fe/H]}=-0.58$| dex) of simple stellar population models given by Maia et al. (2014). Mass uncertainty comes from propagation of errors in M_V, |$\mathbf {\log (\rm {age})}$|⁠, extinction and distance. The derived integrated magnitudes and masses for the clusters are shown in Table 3.

4 RESULTS

The isochrone fitting of the sample cluster CMDs provide precise age, metallicity, distance, and reddening. We also report revised coordinates and tidal radius based on the RDPs. These results are given in Table 3, together with the uncertainties.

Fig. 5 shows the isochrone fits for three of the youngest sample clusters (HW81, L107, and L114), illustrating the vertical MS and the lack of giant stars, whereas the pre-main sequence is clearly sampled. In these cases, the pre-main sequence turn-on provides the strongest constraint to the fitting procedure. Nevertheless, the relative uncertainties of the derived parameters for these clusters are higher than the sample average; these uncertainties should be considered as internal errors (see Section 3.3). Fig. 6 shows the young cluster L92, which, differently than HW81 and L107, contains some giant stars that help constraining the metallicity. In this figure, we also show the resulting corner plot, which contains the posterior distribution in each parameter in the diagonal panels and the correlation between them in the other panels, as indicated by the labels. In Fig. 7, the isochrone fits of three old clusters are given, with a well-defined MS reaching V ∼ 24 mag and well-populated giant branches and a red clump that helped to constrain the metallicity and distance. The isochrone fits for the remaining clusters are presented in Appendix  A.

4.1 Clusters with first derivation of age and metallicity

This work contains the first age derivation for nine sample clusters, namely: HW78, L101, L104, B165, L107, WG1, BS198, ICA45, and OGLB33. HW78 was analysed by Piatti et al. (2015) with near-infrared data, but was marked as a possible non-cluster. Also note that L107 was presented in the van den Bergh (1981) compilation of integrated UBV photometry with an age estimation of <4 Myr (as well as <200 Myr for L114 and >1 Gyr for L113). OGLB33 was only recently catalogued by Sitek et al. (2017) as OGLE-MBR-CL-0033, without any further analysis. All the above mentioned clusters contain very few member stars, where eight of them are very young (<50 Myr old) and the other two are older and compact (B165 is 330 Myr old and OGLB33 is 4.0 Gyr old).

For these clusters and another nine of them (see Table 2), totalling eighteen clusters, metallicity was derived for the first time. For the remaining fifteen clusters, the metallicity values available in the literature are from CaT spectroscopy for four clusters (Da Costa & Hatzidimitriou 1998; Parisi et al. 2009, 2015; De Bortoli et al. 2022) and from CMD analysis for eleven of them (Bica et al. 2015; Perren et al. 2017, and Paper I).

A comparison between the derived metallicities and the mean metallicity from Bica et al. (2020) is presented in Fig. 8, showing a good agreement within the uncertainties. The largest discrepancy is observed for L114 (⁠|$\rm {[Fe/H]}=-0.47$| dex, compared to −0.10 dex from Perren et al. 2017), followed by NGC796, K55, and L113.

4.2 Comments on specific clusters

In this section, we provide further comments on clusters that have conflicting results with the literature or particularities detected in the images and/or in the CMD analysis.

HW59

The old, compact cluster HW59 contains several blue, young stars in the surrounding field. The radial density profile was very noisy and no King function was fitted to the data, and a very small radius of 17 arcsec was adopted in the decontamination procedure to define the initial cluster sample, in order to avoid the presence of field stars. An age of 6.8 ± 2.4 Gyr and a distance of ∼69 ± 8 kpc were derived, making it the oldest and farthest cluster in our sample.

HW63 and L91

There is a large difference in the literature ages for these clusters: Glatt et al. (2010) derived young ages of 0.45 ± 0.31 and 0.79 ± 0.55 Gyr for HW63 and L91, respectively, with MCPS photometry (V ≲ 21 mag). From CT₁ Washington photometry for both (V ≲ 22.5 mag), Piatti (2011a) found 5.4 ± 1.0 and 4.3 ± 1.0 Gyr, and Perren et al. (2017, ASteCA tool) found 3.5 ± 0.5 and 4.0 ± 0.6 Gyr, for HW63 and L91, respectively. In fact, Glatt et al. (2010) states that their photometric limit hampered the age derivation for clusters older than 1 Gyr, because the MSTO was not resolved. Our results confirm old ages of 2.5 ± 0.2 and 3.9 ± 0.5 Gyr (Fig. 7) for HW63 and L91, more compatible with Perren et al. (2017) and Piatti (2011a). Note that both clusters present a MSTO at V ∼ 22 mag, requiring deep observations.

HW77

Fig. 4 shows the deepest and best-seeing CMD of the sample, reaching V ∼ 24.5 mag and |$\rm {FWHM}=0.3$| arcsec. With a derived metallicity of |$\rm {[Fe/H]}=-1.02\pm 0.11$| dex, HW77 is among the most metal-poor sample clusters, very close to the average SMC metallicity of the old field stellar population (Choudhury et al. 2018), and located in the Wing.

HW81

HW81 is particularly young, resulting in 4 ± 1 Myr and a relatively high reddening of E(B − V) = 0.21 ± 0.04 mag. The CMD contains several pre-main sequence stars up to V ∼ 21 mag (Fig. 5) and the composite image shows a blurry green region (Fig. 2), possibly indicating presence of some gas and dust of its embedded phase (<5 Myr; Oliveira, Bica & Bonatto 2018). This confirms the findings by Piatti et al. (2015) that HW81 is embedded in the H ii region N88, as previously investigated using the adaptive optics system of the Very Large Telescope (Testor et al. 2010).

L113

Narloch et al. (2021) used Strömgren photometry from the SOAR optical imager (SOI), obtaining |$\rm {[Fe/H]}=-1.14 \pm 0.10$| dex and an age in the range 3.5 to 4 Gyr for this cluster. Parisi et al. (2015) derived |$\rm {[Fe/H]}=-1.03 \pm 0.04$| dex from CaT lines. Dias et al. (2010) used integrated spectra to derive metallicity in the range |$-2.1 < \rm {[M/H] (dex)} < -1.4$| and old ages. A metallicity of ∼−1.0 dex as obtained here, as well as by Narloch et al. (2021) and Parisi et al. (2015), is more compatible with the SMC metallicity therefore we discard the possibility of having a more metal-poor value.

L114

The isochrones of 140 and 160 Myr for this cluster found by Piatti et al. (2007b) and Perren et al. (2017), respectively, are plotted in Fig. 5 together with the one of 33 Myr found here. This figure shows that, in our decontaminated CMD, the literature values do not fit the brighter member giants. Curiously, our CMD shows three fainter giant stars near V ∼ 16 mag, that might have biased the previous results to older ages. In our analysis, they have been marked as field giants; furthermore the upper MS shape and bright (super)giants also appears to corroborate our results.

NGC796

We obtained a distance of 60.5 ± 3.6 kpc in very good agreement with Kalari et al. (2018, 58.9 ± 0.8 kpc) and Paper I (⁠|$60.3^{+2.7}_{-2.4}$| kpc). The latter was obtained with the same VISCACHA data, but using a different method based on synthetic isochrones. This confirms that the distance from Bica et al. (2015) was too small, which may have resulted from shallower and noisier data from the SOI@SOAR imager without adaptive optics.

BS226

This cluster was previously analysed by Bica et al. (2015), who found similar age and metallicity (see Table 2), but like in the case of NGC796, a smaller distance of ∼40 kpc was derived. The present results are more reliable as they are based on deeper photometry and adaptive optics. Our CMD presents well-defined sequences showing that the lower MS and the few giants are better fitted with a larger distance of ∼53 kpc.

5 DISCUSSION

We analysed 33 stellar clusters located from the SMC Wing until halfway across the Bridge towards the LMC, noting that farther away most of the objects are stellar associations (Bica et al. 2015, 2020). The aim of this work was to verify if there is a metallicity and/or age gradient along the Bridge, between the SMC and the LMC and to characterize the young and old cluster populations in terms of distance and space distribution. Fig. 9 shows the projected distribution of the sample clusters, overplotted on the 449 Bridge clusters and associations. The grey points are all clusters and associations reported in Bica et al. (2020), the black diamonds are the ones classified as Bridge objects; and the colour-coded circles are the sample clusters according to their age or metallicity. Most of the older clusters are located close to the SMC, whereas beyond the SMC Wing the number of old and young clusters are comparable (with a low number statistics). It suggests a stratification of the cluster groups of similar age, with more recent cluster formation predominating in Bridge regions.

Two groups of clusters are clearly detected in our results: 13 Bridge clusters more metal-poor than |$\rm {[Fe/H]}=-0.6$| dex (more consistent with the SMC metallicity) and older than 500 Myr; and 15 young and metal-rich clusters (⁠|$-0.5\lesssim \rm {[Fe/H]\:(dex)} < -0.1$|⁠), probably formed in situ after the recent encounter 150–250 Myr ago. Such high metallicities indicate that the gas used to form clusters was originated possibly from a more metal-rich component in the SMC centre (Rubele et al. 2018) or from the LMC. The ages of the remaining five clusters (200–500 Myr, namely B147, WG13, B165, K55, and K57) with metallicity between |$\rm {[Fe/H]}\sim -0.6$| and −0.4 dex indicate they formed just before the Bridge and require further analysis. According to Besla et al. (2012), their Model 2 predicts that most of the Bridge material came from the SMC with some smaller contribution from the LMC, in agreement with the observations. Based on these two groups, hereafter we make a distinction between the clusters younger and older than 300 Myr (formed in situ versus stripped), in order to check their relation with the H i gas, their spatial distribution and gradients.

Fig. 10 presents a map of neutral hydrogen (H i) column density from the HI4PI survey (HI4PI Collaboration et al. 2016) along the SMC eastern side and the Bridge. All the sample clusters are concentrated in regions with higher gas density; however, if we consider only the clusters farther from the SMC Wing (⁠|${\rm RA > 1^h20^m}$|⁠), the young ones appear to be located in regions with slightly higher density on average, compared to the older ones. A further look into the kinematics of the two groups of clusters is needed to provide a stronger conclusion about their association to the H i gas. However, since Gaia proper motions are available only for the very bright stars of these clusters and no radial velocities are available, we leave this analysis for a future work.

We converted the RA, DEC coordinates combined with the line-of-sight distances derived in this work into a Cartesian system centred at the SMC centre following the equations by van der Marel & Cioni (2001). The z = 0 plane is tangent to the sky at (x, y) = (0, 0), where z increases towards the observer, x increases towards West, y increases towards North. Fig. 11 shows the projections of the distribution of the sample clusters around the SMC, of which four are within the SMC tidal radius of 4 kpc (Paper IV). The young clusters appear to follow a homogeneous distribution up to a radius of 13 kpc from the SMC centre, whereas the old ones are gathered in some specific regions. The fact that most of the sample clusters are at distances smaller than the SMC (z > 0) and pointing to the LMC seems consistent with the recent collision scenario (e.g. Choi et al. 2022), possibly with the SMC moving away after that and leaving gas and both the young and old stellar populations in its path.

In order to check if the Wing/Bridge clusters follow similar age and metallicity trends compared to the SMC clusters, in Fig. 12, we overplotted our results over the fig. 8 from Bica et al. (2020). In this figure, the red and blue diamonds are the sample clusters that are older and younger than 300 Myr, respectively. The old sample clusters follow the overall age and metallicity radial distributions of the SMC clusters. The young clusters clearly do not follow the radial distributions from Bica et al. (2020). The clusters within a < 4° have ages of 100–300 Myr, whereas the outer Wing/Bridge clusters have ages <100 Myr. The metallicity of these young clusters also deviate from the radial distributions from Bica et al. (2020), with an offset to more metal-rich values and a mean value of |$\rm {[Fe/H]}\sim -0.4$| dex.

The peak metallicity of the SMC main body for the younger stellar populations is about ∼−0.5 dex (e.g. Rubele et al. 2018). We interpret our results as evidence that the gas that was stripped out from the SMC to form the Magellanic Bridge came from the innermost region of the SMC that was enriched by stellar evolution. This metal-rich gas was the material used to form the star clusters along the whole extension of the gaseous Wing and beginning of the Bridge (⁠|${\rm RA} < 3^{\rm h}$|⁠), forming clusters with similar metallicity around |$\rm {[Fe/H]} = -0.4$| dex. This result is in agreement with the analysis of B-type supergiants (Lee et al. 2005). The more distant gaseous bridge (⁠|${\rm RA} > 3^{\rm h}$|⁠) seems to be more metal-poor (e.g. Ramachandran, Oskinova & Hamann 2021), which is beyond the scope of this work, nevertheless it is an evidence that the formation of the whole extension of the Bridge has a complex history. In this sense, we speculate that the clusters older than the formation of the Bridge are more metal-poor and were dragged from the SMC along with the gas during the formation of the Bridge.

In Fig. 13, we show the age-metallicity relation (AMR) including literature and present results. Based on the study by Parisi et al. (2022, in particular the AMR in their figure 16) in the present AMR plot, we selected three popular chemical evolution models, namely the classical Pagel & Tautvaisiene (1998, PT98), the closed box model from Da Costa & Hatzidimitriou (1998, DH98), and the two merger scenarios by Tsujimoto & Bekki (2009, TB09). The latter includes mergers with a mass ratio of 1:1 and 1:4 that should have occurred ∼7.5 Gyr ago. In this figure, black circles and blue pentagons have spectroscopic metallicities from CaT (Da Costa & Hatzidimitriou 1998; Parisi et al. 2009, 2015; De Bortoli et al. 2022; Parisi et al. 2022), whereas the blue pentagons are from the VISCACHA collaboration (Paper I, Paper III, and Paper IV). The three sets of chemical evolution models, despite their significant differences in terms of assumptions, reproduce well the data.

When comparing Fig. 13 with the AMR from Parisi et al. (2022), it is clear that our data help to better visualize the AMR, in particular including several young objects below 1 Gyr (which do not have red giants, and consequently no measurable CaT lines) and a group of 4 clusters with ∼1 Gyr and |$\rm {[Fe/H]}\sim -1.0$| dex. Additionally, Parisi et al. (2022) isolate in their figure 18 an AMR for three Wing/Bridge clusters (L110, HW86, and L113), and we complement the sample with ten more clusters older than 1 Gyr.

A few important remarks can be deduced from the present AMR plot: (i) a large dip is identified by five metal-poor clusters with ages around 1–1.5 Gyr, with the metallicity decreasing from |$\rm {[Fe/H]}=-0.6$| to ∼−1.0 dex, followed by a rapid chemical enrichment; (ii) a smaller dip is observed to start around 200 Myr ago, with the metallicity decreasing from |$\rm {[Fe/H]}=-0.3$| to ∼−0.6 dex; (iii) the sample clusters older than 1.5 Gyr are slightly more metal-rich than the literature and VISCACHA points. Three clusters at 200 and 300 Myr (B147, WG13, and B165) present some noise in detecting the exact epoch of the metallicity dip, which is consistent with the uncertainties faced by the models in reproducing the recent encounter.

As far as we know, both metallicity dips are detected for the first time in the present work. Comparing them to the dynamical history of the System, both make sense given the recent encounters between the MCs that formed the Stream and Bridge, 2 Gyr and 150–250 Myr ago respectively (Besla et al. 2010; D’Onghia & Fox 2016; Choi et al. 2022). The three clusters that mark the prominent dip at 1.5 Gyr, namely HW77, BS187 and BS198, are the same that deviate from the overall age and metallicity gradients of the SMC. The same applies to two metal-poor clusters from the literature sample: IC1708 (Northern Bridge; Dias et al. 2021) and K9 (West Halo; Parisi et al. 2015, 2022). Additionally, the young Bridge clusters BS233 and BS235 were identified to have |$\rm {[Fe/H]}=-1.3$| and −1.0 dex by Bica et al. (2015, 2020) from CMDs.

Harris (2007) suggested that the star formation in the Bridge commenced about 200–300 Myr ago and continued over an extended interval, until about 40 Myr ago. They determine the Bridge star formation history over a wide area and, by applying a Kroupa mass function and a 10 Gyr stellar population, they calculate an upper mass limit for the Bridge of 1.5 × 10⁴ M_⊙. In the present analysis of around one third of the Bridge clusters, we obtain a total mass of 10⁵M_⊙. Extrapolating this value to the ∼100 clusters and ∼300 associations in the Wing/Bridge (Bica et al. 2020), a conservative estimate of 3–5 × 10⁵M_⊙ appears to be more reasonable for the Bridge stellar mass.

Zivick et al. (2018) ran models to explain kinematic information and concluded that the Bridge was formed about 150–250 Myr ago. As a consequence, the young stellar population would have been formed in situ in the Bridge therefore representing the nearest extragalactic stellar population formed from tidal debris. The fact the Wing/Bridge older clusters are more metal-rich than the overall SMC cluster distribution could indicate that they were dragged from the SMC main body that is more metal-rich than the SMC outskirts (Glatt et al. 2010; Rubele et al. 2018).

6 CONCLUSIONS

We have derived ages, metallicities, distances, and masses for 33 star clusters from deep observations obtained with the adaptive optics of the SAMI imager, at the SOAR telescope. These clusters are among the ∼100 star clusters located in the Bridge between the SMC and LMC. The sample includes a few clusters in the SMC Wing and along the Bridge (⁠|${\rm RA} < 3^{\rm h}$|⁠) but not close to the LMC, because between the LMC and halfway the Bridge, there are essentially no clusters but only sparse stellar associations. Our results include metallicities for 18 clusters and ages for 9 clusters derived for the first time. Based on the masses of our sample clusters, we estimate a minimum stellar mass of the Bridge to be |$3\!-\!5\times 10^5\,{\rm M}_\odot$|⁠, more than one magnitude higher than previous estimates.

Interestingly, we have found 13 metal-poor clusters (⁠|$\hbox{[Fe/H]} < -0.6$| dex) with ages between 500 Myr and 6.8 Gyr. They were probably formed from SMC gas and stripped from the SMC, as predicted by Besla et al. (2012, Model 2). We showed that these clusters follow strictly the SMC age and metallicity gradients, as well as age-metallicity relation given by the chemical evolution models.

We have also found 15 young (≲200 Myr) and metal-rich (⁠|$-0.5 < \text{[Fe/H](dex)} < -0.1$|⁠) clusters, which probably formed in situ after the Bridge formation. Given the more metal-rich regime, the gas that originated this population could have been extracted from the more metal-rich component in the SMC centre (Rubele et al. 2018). This population does not follow the SMC age and metallicity gradients, is roughly constant around |$\rm {[Fe/H]} = -0.4$| dex, which indicates that they were formed in situ from the metal-rich gas that constitute the Wing and Bridge closer to the SMC (⁠|${\rm RA} < 3^h$|⁠). An extraction of gas from the LMC is less likely because the Bridge region closer to the LMC is actually more metal-poor.

A particularly interesting group are the intermediate-age (∼1 Gyr) and metal-poor (⁠|$\hbox{[Fe/H]} < -0.8$| dex) Bridge clusters, namely BS187, HW77, and BS226, together with BS233 and BS235 (Bica et al. 2015), and another two from Parisi et al. (2022) and De Bortoli et al. (2022): IC1708 (Northern Bridge) and K9 (West Halo). These clusters mark a metallicity dip on the age-metallicity relation around 1–1.5 Gyr, decreasing the metallicity from −0.6 to −1.0 dex. It is possible that this episode, followed by a rapid chemical enrichment, happened due to the infall of metal-poor gas, which is consistent with the formation epoch of the Magellanic Stream. These clusters are good candidates to be observed with high-resolution spectroscopy with the future giant telescopes.

Another metallicity dip seems to be present at around 200 Myr when metallicity dropped from −0.25 to −0.55 dex approximately, suggesting a second infall of metal-poor gas. This was the time when the Magellanic Bridge itself was formed (Zivick et al. 2018). In summary, the formation of the Magellanic Stream and Magellanic Bridge left marks in the chemical evolution of the SMC and Wing/Bridge clusters. Dedicated chemical evolution models shall enlighten the explanation of these metallicity dips.

ACKNOWLEDGEMENTS

We thank an anonymous referee for the remarks that improved this paper. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001. R.A.P.O. acknowledges FAPESP (proc. 2018/22181-0). F.F.S.M. acknowledges financial support from Conselho Nacional de Desenvolvimento Científico e Tecnológico – CNPq (proc. 404482/2021-0) and from FAPERJ (proc. E-26/201.386/2022 and E-26/211.475/2021). B.B. and E.B. acknowledges partial financial support from FAPESP, CNPq, and CAPES. B.D. acknowledges support by ANID-FONDECYT iniciación grant No. 11221366. J.F.C.S. acknowledges support by FAPEMIG (proc. OET-00020-22). S.O.S. acknowledges FAPESP (proc. 2018/22044-3), the support of Deutsche Forschungsgemeinschaft (DFG, project number 428473034), and the DGAPA–PAPIIT grant IA103122. L.O.K. acknowledges partial financial support by CNPq (proc. 313843/2021-0) and UESC (proc. 073.6766.2019.0013905–48). D.M. gratefully acknowledges support by the ANID BASAL projects ACE210002 and FB210003 and by FONDECYT Project No. 1220724. A.P.-V. acknowledges the DGAPA-PAPIIT grant IA103122.

Based on observations obtained at the Southern Astrophysical Research (SOAR) telescope (projects SO2016B-018, SO2017B-014, CN2018B-012, SO2019B-019, SO2020B-019, SO2021B-017), which is a joint project of the Ministério da Ciência, Tecnologia, e Inovação (MCTI) da República Federativa do Brasil, the U.S. National Optical Astronomy Observatory (NOAO), the University of North Carolina at Chapel Hill (UNC), and Michigan State University (MSU).

This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.

DATA AVAILABILITY

The data underlying this article are available in the NOIRLab Astro Data Archive (https://astroarchive.noirlab.edu/) or upon request to the authors.

Footnotes

References

APPENDIX A: CMDS WITH THE ISOCHRONE FITS

Figs A1, A2, and A3 present the decontaminated V versus V − I CMDs of the remaining Bridge clusters. The CMDs are colour-coded by membership probability and include the best-fitting isochrone, as well as previous literature results when available.