© The Authors 2026

1 Introduction

At the heart of any advancement in astrophysics lies the crucial need to answer two fundamental questions: “where” we test our theories and “how” we improve our understanding of a phenomenon. For stellar populations, globular clusters (GCs) constitute one of the best choices to fill the role of the where, providing ideal laboratories. The how is equally critical, and sometimes more challenging to establish. For decades, the stability, resolution, and depth of the Hubble Space Telescope (HST) have made this telescope the best tool for this field. Over time, HST has become instrumental not only for the detailed analysis of individual GCs, but also for the study of their characteristics as a system of Galactic tracers. To enable these broad and impactful HST-based studies on the topic, a homogeneous framework that encompasses consistent photometric catalogs and theoretical models was essential. Extensive precious Treasury and Legacy Survey programs with HST have already demonstrated that homogeneity streamlines and simplifies analyzes and comparisons of clusters. For example, almost 20 years ago, the “ACS Survey of Galactic Globular Clusters” (Sarajedini et al. 2007, GO-10775) imaged the cores of 65 GCs (and one open cluster) to construct a uniform database of deep F606W and F814W photometry (Anderson et al. 2008). Later, when the presence of the multiple stellar populations became irrefutable, the “HST UV Legacy Survey of Galactic Globular Clusters” (Piotto et al. 2015, GO-13297) was designed to photometrically shed light on the phenomenon. Similarly to the GO-10775 survey, the GO-13927 program ultraviolet and optical campaign released homogeneous photometric (Nardiello et al. 2018) and astrometric (Libralato et al. 2022) catalogs. The combination of just these two HST programs has led to hundreds of publications and a leap forward in the understanding of GCs. The Hubble Missing Globular Cluster Survey (MGCS) is meant to continue this legacy, adding new clusters to the sample.

The MGCS is an HST Treasury program (GO-17435; PI: Massari) dedicated to observing kinematically-confirmed Galactic globular clusters that lacked suitable HST imaging data (Massari et al. 2025, hereafter Paper I). The project aims at increasing our overall knowledge of GCs and the Milky Way (MW) in general, and its key objectives include accurate age determinations of the GC (that, combined with astrometric and spectroscopic information, are essential to understand GC origin and the main MW formation events; e.g., Massari et al. 2019; Kruijssen et al. 2020; Helmi 2020; Horta et al. 2021; Khoperskov & Gerhard 2022; Callingham et al. 2022; Massari et al. 2023; Chen & Gnedin 2024; Ceccarelli et al. 2025; Niederhofer et al. 2025), the search for MW bulge relics, and the inference of structural and dynamical properties of GCs. In this work, we present the astro-photometric catalogs of the MGCS and make them publicly available to the community. We summarize the data reduction and the various steps taken to ensure photometric homogeneity in Sect. 2. We then describe the artificial-star tests included in the release (Sect. 3). Finally, we present an application made possible by our data set (Sect. 4). Specifically, we link our astro-photometric catalogs to the Gaia Data Release 3 (DR3) catalog (Gaia Collaboration 2016, 2023) and compute proper motions (PMs) for the stars in common, refining and improving (in terms of precision) the absolute PMs for most clusters in our sample.

2 Data reduction

Our sample comprises 34 clusters, 27 of which were observed through the Wide Field Channel (WFC) of the Advanced Camera for Surveys (ACS) and the Ultraviolet and Visible channel (UVIS) of Wide-Field Camera 3 (WFC3), and the remaining seven systems were imaged with the Infra-Red (IR) channel of the WFC3¹. A complete log of observations is detailed in Paper I. As described in Massari et al. (2025), the GC sample of the MGCS was selected as a compromise between technical requirements (moderate distance and reddening) and scientific requirements (systems that occupy regions of the parameter space-mass, density, environment, which are rather extreme and poorly sampled).

The ACS+WFC3 campaign was originally designed to collect data in two back-to-back visits to provide images in two filters of both the core (with ACS/WFC) and the outskirts (with WFC3/UVIS) of each target. However, sporadic visit failures² compromised this setup for some targets, and we requested a repeat of the failed visits through the HST Operations Problem Report (HOPR). In order to improve the chances for the repeats to be successful, we had to change the originally planned telescope orientation that, in turn, resulted in a partial or null overlap of the parallel fields between the two visits. This is the reason why some clusters do not have two-filter photometry in the WFC3/UVIS parallel fields.

The astro-photometric catalogs were obtained following the prescriptions extensively described in many investigations on the same topic with HST data (e.g., Bellini et al. 2017; Libralato et al. 2018; Nardiello et al. 2018; Libralato et al. 2022; Niederhofer et al. 2024, and references therein), and consist of multiple steps. In the following, we briefly describe the main steps.

First, we made use of calibrated³, unresampled images, which are best suited for astro-photometric analyzes based on point spread function (PSF) fitting techniques like the one we performed in our project. For each image, we obtained a preliminary list of the positions and fluxes of bright and isolated sources by fitting a set of effective-PSF (ePSF) models (firstpass photometry; Anderson 2022). These ePSFs were tailored to each exposure starting from a set of library ePSF models⁴ to account for variations due to telescope breathing (see, e.g., Section 3 of Bellini et al. 2017). The positions were also corrected for the effects of geometric distortion by means of the avail-able⁵ corrections for the ACS/WFC (Anderson & King 2006), WFC3/UVIS (Bellini & Bedin 2009; Bellini et al. 2011) and WFC3/IR (Anderson 2016) cameras.

Second, we built a common reference frame in which all images could be properly transformed and analyzed simultaneously (see, e.g., Bellini et al. 2017; Libralato et al. 2018, for more details). For astrometric registration, we started from the Gaia Early DR3 (EDR3) catalog Gaia Collaboration (2016, 2021). For each cluster, we projected the Gaia catalog onto a tangent plane centered on the center of the cluster taken from the literature (see Paper I). The positions of the sources in the Gaia catalog were propagated using the available PMs to the average epoch of our HST observations. This step allowed us to anchor our reference frame onto the absolute frame of the International Celestial Reference System (ICRS). Although the axis orientation (x pointing to the west and y to the north, respectively) and scale (40 mas pixel⁻¹ and 130 mas pixel⁻¹ for the GCs in the optical and IR samples, respectively) were kept consistent across all clusters in the same sample (optical or IR), the center of the cluster in the pixel-based frame was shifted to ensure non-negative coordinates. From the photometric point of view, instrumental magnitudes⁶ were zero-pointed to the longest exposure in each camera and filter.

Third, we created the so-called bright-star list. This preliminary catalog includes only very bright sources, both saturated and not, and serves a twofold purpose: (i) it marks the position of bright objects with pronounced diffuse light, diffraction spikes, and sometimes even bleeding columns (in optical detectors), which need to be masked off in the second-pass photometry stage; and (ii) it keeps a record (position and magnitude) of saturated sources that are not measured in the second-pass photometry, so that they could still be included in the final catalog.

As a fourth and last step, we ran the second-pass photometry with the code KS2 (Bellini et al. 2017; Nardiello et al. 2018; Libralato et al. 2018, 2022), which is designed to measure neighbor-subtracted stars in crowded environments like GCs, and to enhance the detectability of faint sources by using all exposures simultaneously during the detection phase. The KS2 code measures and subtracts progressively fainter objects in various stages, the detection criteria of which can be fine-tuned for any given set of observations (Bellini et al. 2017; Libralato et al. 2022). As exhaustively described in the aforementioned literature, there is no one-size-fits-all recipe to optimally run KS2. For our work, we found a set of parameters that provided a reasonable compromise between the completeness of the catalog and the reliability of the detected sources, and kept the same parameters for all clusters in our sample (for a detailed description of the various setup parameters, we refer interested readers to Bellini et al. 2017).

The final catalog from KS2 contains positions and magnitudes of detected sources, as well as a series of quality parameters that can be used to select a bona fide sample of well-measured sources. Saturated stars from the first-pass photometry and present in the bright list not measured by KS2 are also included in the final catalog⁷.

As anticipated in Paper I, the pixel-based positions were also transformed into ICRS Equatorial coordinates by means of the Gaia EDR3 catalog. Our photometry was calibrated in the Vega-magnitude flight system as follows: (i) we ran aperture photometry (with aperture radius selected on a case-by-case basis) on the drizzled images (_drc for ACS/WFC and WFC3/UVIS, and _drz for WFC3/IR data, respectively); (ii) we corrected the drizzled-based photometry to account for the finite aperture by using the official aperture corrections provided for the instruments⁸; (iii) we cross-identified bright, unsaturated stars in common between our KS2 and drizzled-based catalogs, and computed the 2.5σ-clipped average magnitude between the two photometric sets (∆m); (iv) the final calibrated magnitudes are defined as: m_Vega-mag = m_instrumental + Δm + ZP_Vega-mag, where ZP_Vega-mag are the camera-dependent official Vega-mag zero-points⁹.

A visual inspection of the color-magnitude diagrams (CMDs) based on ACS/WFC and WFC3/UVIS data revealed the presence of photometric discrepancies between the two sets in some clusters. While differences on the order of a few hundredths of a magnitude (i.e., of the order of the calibration errors) are expected, we found cases where the photometric offset was as high as 0.1 mag. Due to the environment of some systems (e.g., patchy or high-extinction regions), it was not trivial to identify the culprit in such discrepancies (whether they were systematic errors in the calibration or genuine reddening features) and determine which photometric set (ACS, WFC3, or both) was affected by this systematic. For this reason, we performed an external cross-check on our photometry by comparing it with the synthetic photometry of the Gaia DR3 low-resolution spectra (De Angeli et al. 2023) retrieved with the GaiaXPy¹⁰ tool.

For the GCs in the optical sample, we cross-identified the same stars in our KS2 and in the Gaia synthetic-photometry catalogs. We then computed the difference between our HST-based and the Gaia synthetic photometry. The result is presented in Fig. 1. We find an overall agreement between the two sets for all filters but the WFC3/UVIS F814W, where the average zeropoint between the HST and Gaia catalogs is about −0.02 mag. This offset is on the same order as the typical uncertainty in the VEGAmag calibration (0.01-0.03 mag; see, e.g., Nardiello et al. 2018). We cannot rule out Gaia synthetic photometry as the source of this discrepancy, but this is unlikely given that we do not observe similar zero-point offsets in other filters derived from the same spectra. We also do not find a clear magnitude or color trend. To homogenize the photometry for all GCs as much as possible, for each cluster we applied to our calibrated optical photometry the corresponding zero-point coming from the Gaia comparison. For 11 GCs with less than five stars in common between the HST catalog (in at least one filter) and the Gaia synthetic-photometry catalog, we applied the average zero-point coming from the entire sample of clusters. Table 1 provides the median values (and standard errors) of the differences between our HST and Gaia photometry before and after correction. Note that this correction does not remove all spurious systematic offsets in our photometry, as Gaia sometimes struggles in fields with high reddening and/or crowding or because of the small statistics, but it is another step towards obtaining a photometrically-homogeneous sample of GCs.

The astro-photometric catalogs obtained with second-pass photometry, which contain positions, magnitudes, and quality parameters (that can be used to select the best stars for analysis and remove spurious detections) are described in Appendix A. We also make publicly available the astrometrized stacked images (one for each filter, camera, exposure time).

3 Artificial-star tests

We performed homogeneous artificial-star injection and retrieval on all available data sets. The resulting ancillary catalogs can be used to assess the level of completeness in our data and to search and correct for potential systematic errors (as in, e.g., Bedin et al. 2009; Libralato et al. 2024). The observed CMDs of the GCs in our program revealed features and sequences related to the various populations present in the images (cluster and field stars). As such, we generated lists of artificial stars that span a wide range of colors and magnitudes covering the color-magnitude space of real sources (Bellini et al. 2017; Nardiello et al. 2023; Libralato et al. 2023) instead of adopting a single fiducial line (Anderson et al. 2008).

For each GC in the ACS+WFC3 sample, we started by constructing a Hess diagram of the observed CMD. We considered all stars between instrumental magnitude −14.25 (−0.5 mag above the saturation threshold of the long exposures, to also test the completeness in the magnitude regime only covered by short exposures) and −2.5 (at about the level of the faintest sources detected by KS2). When a cell (with size −0.2 in magnitude and −0.1 in color) of the Hess diagram contained at least 5 stars, we generated a set of artificial stars with random magnitude and color within the cell boundaries. The number of artificial stars per cell quadratically increases between 50 (bright-end of the CMD) and 250 (faint-end of the CMD), to better account for incompleteness effects in the faint regime. In the central ACS/WFC field, the associated random positions were selected from either flat (like that of field objects) or Gaussian distributions (centered in the cluster’s center and with σ of 1500 pixels) with a 50-50 ratio; only a flat distribution was considered in the parallel, field-dominated WFC3/UVIS regions. In addition, the positions of the stars were chosen so that they landed in at least one HST exposure and, for the WFC3/UVIS fields, in a region covered by two filters. Due to this empirical approach, the number of entries in the artificial-star catalogs is not exactly the same for all GCs and can range from a minimum of ~93 000 (Whiting 1) to a maximum of ~393 000 (ESO-2806). A similar approach was also used for the artificial stars of GCs in the WFC3/IR sample (simulated between −243 000 and −292 000 stars).

These artificial star lists were used as input in new runs of KS2. Each star was added to each image one at a time, measured in the same way as the real stars, and then removed, to not bias the measurement of other artificial stars. In the following, we considered a star as recovered if the input and output positions differ by less than 0.5 pixel, and magnitudes agree to within 0.75 mag (i.e., fluxes agree within a factor of 2). Our artificial-star catalogs list input and measured positions and magnitudes, so users are free to apply different criteria that best suit their needs.

Figure 2 illustrates the inputs, outputs, and statistics from the artificial-star test of the GC ESO-452-11. Panels a and c present the input CMDs for the ACS/WFC and WFC3/UVIS fields, respectively. The black dots are the sources from the observed catalog, while gray points represent the artificial stars in input. Panels b and d show the Hess diagrams of the recovered CMD, color-coded by the completeness level as indicated in the corresponding colorbar at the top. The 50% completeness level is reached, on average, at m_f606w ~ 27.5, i.e., 7 magnitudes below the main-sequence turn-off. The Hess diagrams in panels e and g depict the average spatial completeness of the entire catalog (again, color-coded as in the side colorbars). Panels f and g display the Hess diagrams in which cells are color-coded according to the F606W magnitude at which the completeness reaches the 50% level. For this GC, the completeness is spatially uniform, with a slightly lower level towards the edges of the field, due to the dithering strategy of the observations. The completeness level also progressively drops in the proximity of the chip gap in the ACS/WFC field as a result of higher uncorrected residuals related to the charge-transfer inefficiency of the detector (Kuhn 2024). Finally, Fig. 3 presents a more classic view of the completeness of the F606W data in the ACS/WFC (black line) and WFC3/UVIS fields (red line).

We made our artificial-star tests (both input and output catalogs) available to the community. A description of these catalogs is provided in Appendix A.

Fig. 1 Comparison of our photometry with that in the Gaia synthetic-photometry catalog. In each row (=camera and filter), the first two panels from the left show the HST- Gaia magnitude difference for all stars in all our GCs as a function of magnitude and color, respectively. The red line is set to 0 as a reference, while the light-blue lines are set at the median (solid line) ± 1σ (dashed lines) magnitude difference for all GCs. The third panel from the left shows the average zero-point difference for each cluster. Finally, the last panel on the right shows the magnitude trend after the photometric correction.

Fig. 2 Overview of the artificial-star tests for the GC ESO-452-11. Panel a shows the input CMD for the ACS/WFC data. Black dots are taken from the observed catalog, while gray points are artificial stars in input. Panel b presents the Hess diagram of recovered stars, color-coded by the completeness level as in the colorbar at the top. The analogs of these plots for the WFC3/UVIS parallel field are provided in panels c and d. Panels e and f display the Hess diagrams of the spatial completeness in the ACS/WFC region. The former is color-coded according to the average completeness level, while the latter is color-coded according to the F606W magnitude at which the completeness level reaches the 50%. The x- and y-axis units are WFC3/UVIS pixels (units in the plot labels are omitted for clarity). Similarly, panels g and h show the spatial completeness in the WFC3/UVIS region.

Fig. 3 Completeness level for the GC ESO-452-11 as a function of F606W magnitude (instrumental at the top and calibrated at the bottom) for the ACS/WFC (black line) and WFC3/UVIS (red line) data. The 50% completeness level is highlighted with a gray, dashed line.

4 Scientific application: Probing globular-cluster origins with GaiaHub proper motions

For most MGCS targets, cluster members in the CMD are embedded in foreground or background field stars. Rosignoli et al. (2026, hereafter, Paper II) discuss two methods devised to facilitate the analysis of the MGCS GCs: a statistical approach purely based on photometry (applicable only to systems with both ACS and WFC3 optical data), and PM decontamination. As an example of a possible application, we show in the following the latter method with PMs in the central fields of our GCs computed using GaiaHub¹¹ (del Pino et al. 2022). GaiaHub is a tool designed to combine HST and Gaia data to compute PMs. The HST - Gaia combination has been shown to be very effective in improving the PM precision at Gaia’s faint end.

We ran GaiaHub following the same approach as in Paper II. First, we bypassed the initial step of GaiaHub in which HST-based astro-photometric catalogs (one for each image) are created from scratch and instead used the output of our first-pass photometry stage described in Sect. 2. We ran GaiaHub with the --rewind_stars option, which uses PMs (from Gaia first, and then from GaiaHub) to propagate Gaia’s positions at the epoch of the HST observations to facilitate the cross-matching of the catalogs and transform positions onto the same reference system. We opted for this setup because cluster membership is not straightforward, as field stars significantly outnumber cluster stars in many cases. Due to how we run GaiaHub, our output PMs are already in an absolute reference frame set by the Gaia DR3 PMs. For NGC 6749, we used PMs from Paper II.

We ran GaiaHub with default settings and let it converge to the final PMs. For 29 of 34 GCs, the convergence was reached before the maximum number of iterations (ten) that the code ran. For the remaining five (Bliss 1, Gran 1, Kim 3, Koposov 1, and Mercer 5), we carefully examined the output. For Gran 1, the PM catalog presented puzzling position- and magnitude-dependent systematics that our local corrections (see below) were unable to remove. For this reason, we chose not to use the PMs for this object. For Bliss 1 and Mercer 5, we could not find any suspicious systematic in the PMs (after local corrections were applied). For the remaining two GCs (Kim 3 and Koposov 1), the few stars present in their vector-point diagrams (VPDs) of the PMs did not allow us to reach a definitive conclusion.

We also searched and corrected for systematics in the PMs that correlate with the position of stars in the field using a local correction obtained from comparing our GaiaHub and Gaia DR3 PMs (i.e., by subtracting for each star the median difference between our and Gaia PMs for the N closest neighbors, with 10 ≤ N ≤ 50 chosen as a compromise between the number of available sources and the need to select a network of neighbor stars as local as possible to the target to better capture the local systematics; see Paper II). These PM systematics can arise from inaccuracies in the HST PSF models to residual, uncorrected HST geometric distortion and charge-transfer-efficiency defects. It is difficult to discern the nature of these effects a posteriori from PMs, but it is straightforward to mitigate them (Paper II, but see also Bellini et al. 2014 and Libralato et al. 2022). We emphasize that even the locally-corrected PMs might not be free of systematics. The local corrections are computed starting from the assumption that the Gaia PMs, although less precise, are still accurate at the faint-end of the Gaia catalog. However, Gaia PMs are not perfect either, and in extreme regions like those targeted by our HST observations, Gaia PMs might also have some intrinsic systematics. Therefore, any systematic in Gaia’s PMs could be transferred to our PMs. In general, we advise users to carefully check the presence of systematics in the PMs before running any kinematic analysis. Demanding applications like the assessment of the internal motions could be complicated with our PMs for some GCs, especially those systems with very few stars. The raw and corrected PMs are included (when available) in our catalogs (see Appendix A).

We used our PMs to refine the absolute PMs of most GCs of the MGCS. For Koposov 1, Muñoz 1 and Segue 3, the VPDs are poorly populated and do not show any overdensity of stars linked to the clusters (not even using the current estimates of the absolute PM of these GCs provided by Vasiliev & Baumgardt 2021 as a reference), but just a sparse PM distribution like that of field interlopers. 2MASS-GC01 and 2MASS-GC02 have instead moderately-populated VPDs, but we could not identify cluster members either. For these reasons, we did not compute the absolute PMs for these systems. Conversely, for Kim 3 and Koposov 2, the VPDs are not very populated, and we were unable to compute local corrections for their PMs, but stars with a motion close to that of the cluster predicted by Gaia (Vasiliev & Baumgardt 2021) are present. We chose to include these systems in our analysis, although we are aware of the possible presence of uncorrected systematics.

The absolute PMs of GaiaHub were computed by computing the 3σ-clipped median value (and error) of the PMs of the cluster members (selected using both PMs and their position in the CMD). The results are summarized in Appendix B, and a few examples of our astrometric analysis are shown in Fig. 4. Our estimates are within 3σ from Gaia-based values in the literature (from the online repositories of GC parameters of Andrew Pace¹² and Holger Baumgardt¹³; Pace 2025; Vasiliev & Baumgardt 2021). Our uncertainties on the absolute PMs are on average 3.7 times smaller than those in the literature.

Given the higher precision achieved, we combined our newly-computed PMs with positions, distances and line-of-sight (LOS) velocities from Pace (2025)¹⁴ to find the associations between each GC and its putative galaxy progenitor as in Massari et al. (2019) and Massari (2025). We used the AGAMA code¹⁵ (Vasiliev 2019), adopting the gravitational potential of McMillan (2017), to determine the following orbital parameters: total energy (E), vertical angular momentum (L_z), perpendicular angular momentum (L_⊥), maximum vertical height, apocenter, and circularity. We then assigned the GCs in our sample to the corresponding progenitors following the same criteria as in Massari (2025). The results are shown in Fig. 5 and are summarized as follows:

• In situ (Bulge): ESO-452-SC11, Gran 5, VVV-CL001;

• In situ (Disk): BH 140, BH 176, FSR 1716, Garro 1, Mercer 5, NGC 6749, Palomar 8, Palomar 10, Palomar 11, Patchick 126, PWM 2;

• Sagittarius : Whiting 1;

• Helmi Streams : Gran 4;

• Gaia-Enceladus : ESO-280-6, Gran 2, Gran 3, NGC 7492, RLGC-02, UKS 1, VVV-CL002;

• Elqui: Laevens 3, Kim 3;

• High energy : Bliss 1, Koposov 2, RLGC-01.

The progenitors identified in our work are in agreement with those found by Massari (2025), with four noticeable exceptions. The first is VVV-CL001, which we now potentially associate with the in situ Bulge rather than the accreted low-energy group. The second is VVV-CL002, which, according to the new orbital parameters, is potentially associated with Gaia-Enceladus rather than with the MW bulge. Finally, our results highlight the presence of two new members of Elqui, namely Laevens 3 and Kim 3, which change their association from Sagittarius and high-energy groups, respectively. We note that the orbit of UKS 1 now seems more consistent with Gaia-Enceladus, compared to its previous uncertain association.

Our results are also consistent with those in Chen & Gnedin (2024), with the exception of UKS 1. Furthermore, a comparison with the analysis of Callingham et al. (2022) reveals discrepancies with our findings for approximately one third of the systems. Most clusters with different associations are in regions of the integral-of-motion space where classification is more challenging (e.g., −20 × 10⁴ < E < -15 × 10⁴ km² s⁻²) because of the overlap of the debris from several progenitors in a limited volume of the parameter space. Furthermore, we stress that although Callingham et al. (2022) used metallicity and age along with dynamics to improve the identification of the progenitor, most of these GCs with discrepant associations did not have age information in their work either. Thus, we expect that our more precise PM measurements provide an improvement significant enough to make these new associations more robust. Finally, we confirm an in-situ, Disk-like origin for Patchick 126 as in Garro et al. (2026).

For five GCs (Garro 1, Gran 2, Gran 3, Gran 4, Gran 5), we provide here for the first time an association based on the homogeneous dynamical framework described in Massari (2025). The above assignments agree with those of De Leo et al. (2026) for Gran 4 and Gran 5, whereas Gran 3 is unassociated according to that work and Gran 2 is linked to a more general retrograde substructure that is included in our definition of Gaia-Enceladus.

For Garro 1 our interpretation agrees with the in situ origin inferred by Pace et al. (2023). All updated associations are available on the CARMA website¹⁶ (Massari et al. 2023).

Fig. 4 Absolute PMs for a sample of GCs in our program. For each GC (two for each row), from left to right we present: the VPD of the Gaia DR3 PMs; the VPD of the PMs from GaiaHub; and a calibrated CMD. In the rightmost CMD, stars in red are likely cluster members according to their PM (made with an arbitrary magnitude-dependent PM selection). This Figure includes examples of a very-crowded region (Palomar 8); a well-populated, moderately-crowded field (ESO-452-11); a loose GC in a sparse field (Bliss 1); and a cluster observed with the WFC3/IR camera (UKS 1).

Fig. 5 Projections (left panel: E versus Lz; right panel: L⊥ versus Lz) of the integral-of-motion space for the GCs analyzed in our work, color-coded according to the association with their progenitor (see legend in the right panel). The gray points represent all other GCs included in the repository of Pace (2025) as a reference.

5 Conclusion

The HST MGCS is one of the last efforts, but not least, to shed light on various aspects of MW GCs using homogeneous photometry (and astrometry). This third paper in the series focuses on the creation of the official astro-photometric catalogs of the project, which have already been used in Paper I and Paper II. All data products described in the paper are made publicly available to the community. At the time of writing, there are ~170 confirmed GCs (e.g., Vasiliev & Baumgardt 2021). Thanks to the MGCS and other HST programs (some of which are still ongoing) with similar setups, there will soon be 112 GCs with homogeneous ACS/WFC F606W and F814W photometry.

The HST MGCS observing strategy secured exquisite images for 27 GCs with the ACS/WFC+WFC3/UVIS cameras and for seven with the WFC3/IR detector. The optical sample comprises data of both the core and the outskirts of each GC, while the IR sample includes only the core. We used state-of-the-art reduction techniques for HST to obtain high-precision astrometry and photometry for all targets (see Appendix C for a collection of CMDs for these systems). Upon inspection of our optical catalogs, we found the presence of photometric discrepancies between the central and outer fields of the same cluster. Although we ignored the reason for this discrepancy, we devised a novel correction that takes advantage of Gaia -based synthetic photometry. This simple application shows that Gaia can serve as a powerful auxiliary catalog for photometry, analogously to what is currently done with astrometry.

Our work also includes comprehensive artificial-star tests for each GC. The resulting ancillary catalogs serve a dual purpose: they enable a precise assessment of the completeness level and provide a robust framework for investigating and characterizing systematic errors in our data.

We combined data from HST and Gaia using GaiaHub to determine precise PMs for the MGCS targets. Although the magnitude interval (typically 4-5 magnitudes) covered by both Gaia and HST is not enough to fully exploit our HST data, the synergy between Gaia and HST makes it possible to identify cluster members in systems where Gaia alone struggles (see also Paper II). These PMs allowed us to refine the absolute PMs of the clusters, finding estimates that are, on average, 3.7 times more precise than those in the literature based on the Gaia DR3 catalog.

We integrated our HST -Gaia PMs with positions, distances, and LOS velocities derived from the literature to characterize the clusters’ orbital parameters. This allowed us to update -and in several cases, identify for the first time - associations between these GCs and their putative progenitor galaxies, ultimately contributing to the reconstruction of the MW’s assembly history.

We release astro-photometric catalogs, astrometrized stacked images, and artificial-star tests to the community. These catalogs increase the sample of GCs with exquisite photometry, enabling the scientific community to further investigate the formation, dynamics, and evolutionary history of the MW and its satellites. These comprehensive data continue the legacy of HST, and provide a bridge for future surveys with JWST and Nancy Grace Roman Space Telescope.

Data availability

Astro-photometric catalogs and stacked images are available at the CDS via https://cdsarc.cds.unistra.fr/viz-bin/cat/J/A+A/709/A140, and at our website https://www.oas.inaf.it/en/research/m2-en/mgcs-en/

Acknowledgements

DM, SC, EP and AM acknowledge financial support from PRIN-MIUR-22: CHRONOS: adjusting the clock(s) to unveil the CHRONOchemo-dynamical Structure of the Galaxy” (PI: S. Cassisi). SC acknowledges the support of a fellowship from La Caixa Foundation (ID: 100010434) with fellowship code LCF/BQ/PI23/11970031 (PI: A. Escorza) and from the Fundación Occident and the Instituto de Astrofísica de Canarias under the Visiting Researcher Programme 2022-2025 agreed between both institutions. SS acknowledges funding from the European Union under the grant ERC-2022-AdG, “StarDance: the non-canonical evolution of stars in clusters”, Grant Agreement 101093572, PI: E. Pancino. RP acknowledges the support to this study by the INAF Mini Grant 2025 (Ob.Fu.1.05.24.07.05, CUP C33C24001390005). Based on observations with the NASA/ESA HST, obtained at the Space Telescope Science Institute, which is operated by AURA, Inc., under NASA contract NAS 5-26555. Support for Program number GO-17435 was provided through grants from STScI under NASA contract NAS5-26555. 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. This job has made use of the Python package GaiaXPy, developed and maintained by members of the Gaia Data Processing and Analysis Consortium (DPAC), and in particular, Coordination Unit 5 (CU5), and the Data Processing Centre located at the Institute of Astronomy, Cambridge, UK (DPCI). This research made use of astropy, a community-developed core python package forAstronomy (Astropy Collaboration 2013, 2018).

References

Appendix A Description of the astro-photometric catalogs

The content of the astrometric and photometric catalogs released to the community is described in Tables A.1 and A.2. The astrometric catalogs of Gran 1, Koposov 1, Muñoz 1, Segue 3, 2MASS-GC01 and 2MASS-GC02 do not include PMs because we could not find any cluster member in the VPD, or PMs did not pass our tests (see Sect. 4). For Kim 3 and Koposov 2, no PM corrections were computed, so raw and corrected PMs are identical. The artificial-star-test input-output catalogs are described in Tables A.3 and A.4.

Appendix B Absolute proper motions

Table B.1 provides the absolute PMs of the GCs in the MGCS for which this measurement was feasible (enough stars available, clear understanding of the systematics).

Appendix C Color-magnitude diagrams

Figures C.1-C.4 present a collection of CMDs for all clusters in our sample.

Fig. C.1 Collection of optical CMDs for BH-140, BH-176, Bliss 1, ESO-280-6, ESO-452-11, Palomar 8, FSR 1716, Garro 1, Palomar 10, Palomar 11, Gran 1, and Gran 2. For each cluster, the left panel presents the ACS/WFC-based CMD of the central field, whereas the right panel (when present) shows the WFC3/UVIS-based CMD of the parallel field. Red points represent stars saturated (or unsaturated but missed by KS2; see discussion in Sect. 2) in either filter. All other sources are shown as black dots.

Fig. C.2 As in Fig. C.1 but for Gran 3, Gran 4, Gran 5, Kim 3, Koposov 1, Koposov 2, Laevens 3, Munoz 1, NGC 6749, NGC 7492, Patchick 126, and PWM 2.

Fig. C.3 As in Figs. C.1 and C.2 but for RLGC-01, Segue 3 and Whiting 1.

Fig. C.4 Collection of IR CMDs for 2MASS-GC01, 2MASS-GC02, Mercer 5, RLGC-02, UKS 1, VVV-CL001, and VVV-CL002. Colors have the same meaning as in the other Figures in this Appendix.

All Tables

All Figures

Fig. 1 Comparison of our photometry with that in the Gaia synthetic-photometry catalog. In each row (=camera and filter), the first two panels from the left show the HST- Gaia magnitude difference for all stars in all our GCs as a function of magnitude and color, respectively. The red line is set to 0 as a reference, while the light-blue lines are set at the median (solid line) ± 1σ (dashed lines) magnitude difference for all GCs. The third panel from the left shows the average zero-point difference for each cluster. Finally, the last panel on the right shows the magnitude trend after the photometric correction.

In the text

Fig. 2 Overview of the artificial-star tests for the GC ESO-452-11. Panel a shows the input CMD for the ACS/WFC data. Black dots are taken from the observed catalog, while gray points are artificial stars in input. Panel b presents the Hess diagram of recovered stars, color-coded by the completeness level as in the colorbar at the top. The analogs of these plots for the WFC3/UVIS parallel field are provided in panels c and d. Panels e and f display the Hess diagrams of the spatial completeness in the ACS/WFC region. The former is color-coded according to the average completeness level, while the latter is color-coded according to the F606W magnitude at which the completeness level reaches the 50%. The x- and y-axis units are WFC3/UVIS pixels (units in the plot labels are omitted for clarity). Similarly, panels g and h show the spatial completeness in the WFC3/UVIS region.

In the text

	Fig. 3 Completeness level for the GC ESO-452-11 as a function of F606W magnitude (instrumental at the top and calibrated at the bottom) for the ACS/WFC (black line) and WFC3/UVIS (red line) data. The 50% completeness level is highlighted with a gray, dashed line.
In the text

Fig. 4 Absolute PMs for a sample of GCs in our program. For each GC (two for each row), from left to right we present: the VPD of the Gaia DR3 PMs; the VPD of the PMs from GaiaHub; and a calibrated CMD. In the rightmost CMD, stars in red are likely cluster members according to their PM (made with an arbitrary magnitude-dependent PM selection). This Figure includes examples of a very-crowded region (Palomar 8); a well-populated, moderately-crowded field (ESO-452-11); a loose GC in a sparse field (Bliss 1); and a cluster observed with the WFC3/IR camera (UKS 1).

In the text

	Fig. 5 Projections (left panel: E versus Lz; right panel: L⊥ versus Lz) of the integral-of-motion space for the GCs analyzed in our work, color-coded according to the association with their progenitor (see legend in the right panel). The gray points represent all other GCs included in the repository of Pace (2025) as a reference.
In the text

Fig. C.1 Collection of optical CMDs for BH-140, BH-176, Bliss 1, ESO-280-6, ESO-452-11, Palomar 8, FSR 1716, Garro 1, Palomar 10, Palomar 11, Gran 1, and Gran 2. For each cluster, the left panel presents the ACS/WFC-based CMD of the central field, whereas the right panel (when present) shows the WFC3/UVIS-based CMD of the parallel field. Red points represent stars saturated (or unsaturated but missed by KS2; see discussion in Sect. 2) in either filter. All other sources are shown as black dots.

In the text

	Fig. C.2 As in Fig. C.1 but for Gran 3, Gran 4, Gran 5, Kim 3, Koposov 1, Koposov 2, Laevens 3, Munoz 1, NGC 6749, NGC 7492, Patchick 126, and PWM 2.
In the text

	Fig. C.3 As in Figs. C.1 and C.2 but for RLGC-01, Segue 3 and Whiting 1.
In the text

	Fig. C.4 Collection of IR CMDs for 2MASS-GC01, 2MASS-GC02, Mercer 5, RLGC-02, UKS 1, VVV-CL001, and VVV-CL002. Colors have the same meaning as in the other Figures in this Appendix.
In the text