HST UV Globular Cluster Survey (HUGS)

• HUGS is a comprehensive five-band ultraviolet-optical survey designed to photometrically separate multiple stellar populations in 56 globular clusters and one open cluster.
• The survey employs specialized filter combinations sensitive to C, N, and O variations, enabling precise determination of helium enhancements and stellar evolutionary parameters.
• High-precision astrometry and proper-motion data from HUGS support dynamical studies, including velocity dispersion profiles, radial anisotropy measurements, and blue straggler analyses.

The Hubble Space Telescope UV Legacy Survey of Galactic Globular Clusters, commonly abbreviated HUGS, is an HST Treasury program centered on GO-13297 and designed to complement the ACS Globular Cluster Survey by adding homogeneous ultraviolet and blue imaging to existing optical photometry of Galactic star clusters. Its defining observational product is a five-band data set in F275W, F336W, F438W, F606W, and F814W, optimized to isolate and characterize multiple stellar populations in globular clusters through wavelength regions that are sensitive to abundance variations in C, N, and O [(Piotto et al., 2014); (Soto et al., 2016)]. In its combined form, incorporating GO-13297, pilot programs, and archival material, HUGS provides uniform coverage for 56 globular clusters and the super-solar-metallicity open cluster NGC 6791, together with photometric, astrometric, and later proper-motion catalogues that support stellar-population, dynamical, and exotica studies (Nardiello et al., 2018, Libralato et al., 2022).

1. Origins, scope, and scientific rationale

HUGS was conceived in response to the now-ubiquitous recognition that globular clusters host multiple stellar populations rather than a single chemically homogeneous population. The survey was explicitly designed to exploit the ultraviolet sensitivity of WFC3/UVIS to the molecular bands of OH, NH, CN, and CH, thereby enabling photometric separation of first-generation and second-generation stars from the main sequence through the subgiant and red-giant branches and onto the horizontal branch (Piotto et al., 2014).

The program complements the ACS Globular Cluster Survey (GO-10775), which had already provided homogeneous F606W and F814W photometry for 65 nearby, low-reddening globular clusters. GO-13297 added F275W, F336W, and F438W observations for the most accessible 47 of those clusters, selected so that turnoff stars were reachable with $S/N \gtrsim 50$ in F275W, and excluded the most distant or highly reddened systems and those already observed in UVIS. NGC 6791, an old open cluster with $[{\rm Fe/H}] \approx +0.4$, was included to extend the metallicity baseline from $[{\rm Fe/H}] \approx -2.3$ up to super-solar values (Soto et al., 2016).

In the formulation given in the project overview, the combined survey was intended not only to identify multiple populations and infer CNO abundance variations photometrically, but also to combine ultraviolet and optical imaging for helium measurements at the level of $\Delta Y \simeq 0.01$–$0.04$, and to provide the astrometric foundation for internal and absolute proper-motion work over baselines of 7–14 yr (Piotto et al., 2014). A plausible implication is that HUGS was conceived from the outset as both a stellar-population survey and a long-baseline astrometric resource.

2. Observational architecture and filter basis

GO-13297 was awarded 131 HST orbits to observe 47 globular clusters plus NGC 6791. In conjunction with GO-12311, GO-12605, archival WFC3 programs, and the earlier ACS survey, the result is a homogeneous five-band survey of 57 clusters in the central fields, while the later public astro-photometric release covers 56 globular clusters and one open cluster [(Piotto et al., 2014); (Nardiello et al., 2018)].

The core of the survey is the WFC3/UVIS three-filter configuration often described as the “magic trio,” combined with the ACS/WFC optical baseline:

Filter	Approximate wavelength	Principal role in HUGS
F275W	$\lambda_0 \approx 2750$ Å	OH sensitivity
F336W	$\lambda_0 \approx 3360$ Å	NH sensitivity
F438W	$\lambda_0 \approx 4330$ Å	CN + CH sensitivity
F606W	$\lambda_0 \approx 5920$ Å	Broad optical baseline
F814W	$\lambda_0 \approx 8020$ Å	Broad optical baseline

The photometric behavior of chemically distinct populations in these bands is central to the survey design. First-generation stars are C- and O-rich and N-poor, and therefore are faint in F275W and F438W and bright in F336W; second-generation stars are C- and O-poor and N-rich, and therefore are bright in F275W and F438W and faint in F336W (Soto et al., 2016). This inversion underlies the pseudo-color index

$[{\rm Fe/H}] \approx +0.4$0

which maximizes the separation between 1G and 2G sequences on the red-giant branch and main sequence (Soto et al., 2016).

Exposure times were set by the requirement of ultraviolet precision. Typical GO-13297 coverage per cluster consisted of 4–12 exposures in F275W with 400–1800 s each, 4–6 exposures in F336W with 300–650 s, and 4–6 exposures in F438W with 40–200 s, with the design target of $[{\rm Fe/H}] \approx +0.4$1 mag at the turnoff and $[{\rm Fe/H}] \approx +0.4$2 in F275W (Soto et al., 2016). The project overview further states that most clusters received two orbits, while more distant or reddened clusters were allocated up to six orbits, and that the observations were generally split into two visits at orientations separated by $[{\rm Fe/H}] \approx +0.4$3, with post-flash to a background of $[{\rm Fe/H}] \approx +0.4$4 per pixel for CTE mitigation (Piotto et al., 2014).

HUGS also included an ACS/WFC parallel-field component. A total of 110 parallel fields were observed in the outskirts of 48 globular clusters plus NGC 6791, centered at about $[{\rm Fe/H}] \approx +0.4$5 arcmin from the cluster center and totaling about $[{\rm Fe/H}] \approx +0.4$6 square degrees of sky, with at least one exposure in both F475W and F814W per field (Simioni et al., 2018). These outer-field data extend the radial baseline from crowded cores to cluster outskirts.

3. Reduction strategy and catalogue releases

The survey was accompanied by an explicitly tiered reduction plan. The early-release reduction used a one-pass star finder on each exposure, tied measurements to the GO-10775 reference frame, averaged photometry in F275W, F336W, and F438W, and delivered stacked UVIS images and preliminary catalogues soon after acquisition (Piotto et al., 2014). The preliminary public release later published catalogues for 57 clusters, including GO-13297 targets, NGC 6791, and nine archival programs (Soto et al., 2016).

In the preliminary reduction, calibrated _flt images were corrected for CTE losses with the Anderson & Bedin algorithm and for geometric distortion with the Bellini, Anderson & Bedin solution. The one-pass photometric procedure scanned every exposure for local maxima more than $[{\rm Fe/H}] \approx +0.4$7 above the local sky in a $[{\rm Fe/H}] \approx +0.4$8 pixel box, fit a library PSF to a $[{\rm Fe/H}] \approx +0.4$9 pixel stamp to derive an instrumental magnitude and quality parameter, normalized magnitudes to a 1000 s exposure, and transformed positions and magnitudes into the ACS Globular Cluster Survey reference frame through a six-parameter linear solution. Cross-identification to the ACS catalogue used a matching radius of 1.75 pixel with iterative rejection of $[{\rm Fe/H}] \approx -2.3$0 pixel outliers (Soto et al., 2016).

Photometric calibration in the preliminary release was based on aperture photometry on drizzled images for 10 representative clusters, from which filter zeropoint offsets were derived and then applied to all clusters, with typical zeropoint uncertainties $[{\rm Fe/H}] \approx -2.3$1 mag (Soto et al., 2016). For each star and filter, the released quantities included the calibrated magnitude, the RMS scatter after $[{\rm Fe/H}] \approx -2.3$2 clipping, and the counters $[{\rm Fe/H}] \approx -2.3$3, $[{\rm Fe/H}] \approx -2.3$4, and $[{\rm Fe/H}] \approx -2.3$5 for the numbers of good measurements, detections, and possible detections (Soto et al., 2016).

The later public catalogue release replaced the preliminary Paper VIII products with a more sophisticated astro-photometric reduction. All exposures were processed in _flc format, with perturbed, exposure-specific PSFs, Gaia DR1-tied astrometry, and KS2 multi-pass crowded-field photometry. Eight finding passes were used, with discovery iterations tied first to F606W+F814W and then to F438W, F336W, and F275W. Three photometric methods were delivered: Method 1 for bright stars, Method 2 for intermediate magnitudes, and Method 3 for the faintest stars or extreme crowding (Nardiello et al., 2018).

The released catalogues contain, for each star, calibrated magnitudes in all five bands, RMS uncertainties, QFIT, RADXS, the numbers of exposures in which the source was found and well measured, a membership probability $[{\rm Fe/H}] \approx -2.3$6 where available, equatorial coordinates, and a unique identifier (Nardiello et al., 2018). Final stellar positions in the public release are accurate to a few hundredths of a WFC3/UVIS pixel, corresponding to about 1–2 mas, and the absolute coordinate readout is tied to Gaia at epoch 2015.0 (Nardiello et al., 2018). Preliminary central-field astrometry had already reached mean position uncertainties $[{\rm Fe/H}] \approx -2.3$7 pixel, or about 5 mas, with a 7–9 yr baseline sufficient for proper-motion errors $[{\rm Fe/H}] \approx -2.3$8 mas yr$[{\rm Fe/H}] \approx -2.3$9, although a systematic rotation offset of $\Delta Y \simeq 0.01$0 between WFC3/UVIS and ACS frames remained to be corrected in the final reduction (Soto et al., 2016).

The public data products are distributed through MAST as a High Level Science Product, while the preliminary release also made per-cluster ASCII tables, distortion-corrected stacks, and RGB composites available through dedicated project web pages (Soto et al., 2016, Nardiello et al., 2018).

4. Multiple-population diagnostics and chromosome maps

The survey’s defining diagnostic is the use of ultraviolet–optical color combinations to separate multiple populations photometrically across large cluster samples. In practice, diagrams such as $\Delta Y \simeq 0.01$1 versus $\Delta Y \simeq 0.01$2 cleanly separate 1G and 2G subpopulations from the main sequence up the red-giant and horizontal branches (Soto et al., 2016). The project overview emphasized that the same five-band framework also supports $\Delta Y \simeq 0.01$3 versus $\Delta Y \simeq 0.01$4, which maximizes temperature and helium sensitivity, as well as $\Delta Y \simeq 0.01$5 versus $\Delta Y \simeq 0.01$6 and $\Delta Y \simeq 0.01$7 versus $\Delta Y \simeq 0.01$8, in which the color ordering of 1G and 2G stars reverses (Piotto et al., 2014).

Paper IX systematized this information in the “chromosome map,” defined from verticalized versions of $\Delta Y \simeq 0.01$9 and $0.04$0. In most clusters, classified as Type I, the chromosome map resolves two principal sequences associated with 1G and 2G stars. In a smaller subset, classified as Type II, the 1G and/or 2G sequences are themselves split; these clusters also show split subgiant branches in optical color–magnitude diagrams and heavy-element variations, including Fe, C+N+O, and $0.04$1-process enrichment, on the anomalous red RGB (Milone et al., 2016).

The atlas of 57 clusters established several quantitative regularities. The fraction of 1G stars spans from about 8% to about 67% and anticorrelates strongly with cluster mass, while the intrinsic RGB widths in both $0.04$2 and $0.04$3, after metallicity correction, correlate strongly with mass (Milone et al., 2016). The same work states that all 57 globular clusters in the atlas host multiple populations and that the incidence and complexity of the phenomenon both increase with cluster mass (Milone et al., 2016).

HUGS photometry also enabled quantitative helium work on a cluster-by-cluster basis. Using chromosome maps and synthetic spectra, the helium-abundance analysis of 57 clusters found that in every cluster 2G stars are consistent with being helium-enhanced relative to 1G stars, with a median average enhancement $0.04$4 and a median maximum internal variation $0.04$5; $0.04$6 ranges from less than 0.01 to more than 0.10 and correlates with both cluster mass and the color extension of the horizontal branch (Milone et al., 2018). A related RGB-bump analysis for 18 clusters found an average helium enhancement $0.04$7 for 2G stars relative to 1G stars (Lagioia et al., 2018).

5. Stellar-evolution applications

Although HUGS was designed primarily for multiple-population work, its ultraviolet depth and angular resolution made it an effective survey for hot and chemically unusual stellar populations. One early application was the “UV-route” for blue straggler stars, in which the F275W images are used to define the master source list and photometry is forced in the remaining filters. Because bright cool giants contribute much less flux in the ultraviolet, this method improves completeness for hot stars in crowded cores (Raso et al., 2017).

The blue-straggler analysis of four dense clusters quantified the gain over optical-driven catalogues. In NGC 2808, the UV-guided catalogue found 215 BSSs inside the WFC3 field, whereas an optical-driven ACS selection yielded only about 69 genuine BSSs and about 235 intruders; in NGC 6541, the UV approach found 94 BSSs versus 41 in the proper-motion-cleaned optical catalogue (Raso et al., 2017). The same study measured the $0.04$8 parameter, the area between the cumulative radial distributions of BSSs and a reference population out to the half-mass radius, and showed that the four clusters fall along the previously defined dynamical sequence, with smaller central relaxation time corresponding to larger $0.04$9 and stronger BSS central segregation (Raso et al., 2017).

HUGS also provided a uniform ultraviolet framework for horizontal-branch morphology. Using 53 clusters with GO-13297 and archival WFC3 data, the survey showed that the Grundahl jump and Momany jump occur at nearly universal effective temperatures, $\lambda_0 \approx 2750$0 K and $\lambda_0 \approx 2750$1 K, with alignment across clusters within $\lambda_0 \approx 2750$2 mag (Brown et al., 2016). Two metal-rich clusters, NGC 6388 and NGC 6441, are exceptions, with the G-jump shifted to about 13,500–14,000 K; the survey interpretation is that these shifts are likely associated with large helium enhancements in the blue horizontal-branch stars (Brown et al., 2016). The same work increased the number of Galactic clusters known to host blue-hook stars from 6 to 23 (Brown et al., 2016).

Relative-age studies are another downstream use of the HUGS database. A five-cluster analysis divided stars into POPa and POPb and used the optical turnoff color difference to infer relative ages, concluding that the populations are coeval within 220 Myr in NGC 104, 214 Myr in NGC 6121, 336 Myr in NGC 6352, 474 Myr in NGC 6362, and 634 Myr in NGC 6723 (Lucertini et al., 2020). This suggests that, at least for those systems, the formation timescale separating the principal populations is constrained to at most a few $\lambda_0 \approx 2750$3 Myr within the quoted uncertainties.

6. Astrometry, internal kinematics, and legacy use

The later HUGS program extended from photometric tagging to internal kinematics through large, homogeneous proper-motion catalogues. By combining GO-13297 images with archival HST material, Paper XXIII produced proper motions for 56 globular clusters and one open cluster, describing the resulting data set as the most complete and homogeneous collection of proper motions of stars in the cores of stellar clusters to date (Libralato et al., 2022).

The kinematic reductions used a two-step first-pass/second-pass pipeline with library ePSFs, geometric-distortion correction, Gaia DR2-tied master frames, and KS2 crowded-field measurements at multiple epochs. For each star, the released astrometric catalogues include celestial and master-frame positions, corrected and raw proper motions, uncertainties, reduced $\lambda_0 \approx 2750$4, the number of exposures used, the time baseline, and quality-of-fit parameters (Libralato et al., 2022). For well-measured bright stars, the typical proper-motion precision is 7–60 $\lambda_0 \approx 2750$5as yr$\lambda_0 \approx 2750$6 before systematics correction and 13–120 $\lambda_0 \approx 2750$7as yr$\lambda_0 \approx 2750$8 afterward, with a median one-dimensional proper-motion error of about 25 $\lambda_0 \approx 2750$9as yr$\lambda_0 \approx 3360$0 for bright unsaturated stars (Libralato et al., 2022).

These proper motions support field-star decontamination, membership assignment, velocity-dispersion profiles, anisotropy measurements, and dynamical distances. The survey-wide kinematic analysis found that $\lambda_0 \approx 3360$1 peaks at the center and declines outward, that more concentrated clusters have steeper dispersion profiles, and that dynamically young clusters with $\lambda_0 \approx 3360$2 exhibit mild radial anisotropy beyond about $\lambda_0 \approx 3360$3, whereas dynamically old clusters with $\lambda_0 \approx 3360$4 are isotropic within the HST fields (Libralato et al., 2022). The derived dynamical distances are stated to agree with Gaia-based and other literature values at about the $\lambda_0 \approx 3360$5 level (Libralato et al., 2022).

The HUGS catalogues have also been adopted directly in multi-wavelength cluster work. In M4, for example, HUGS five-band photometry, differential-reddening-corrected magnitudes, and proper-motion membership probabilities were used to identify optical counterparts to Chandra X-ray sources and VLA radio sources; that study found 24 new confident optical counterparts to Chandra sources, for a total of 40, and used HUGS color–magnitude and color–color diagrams to identify active binaries, cataclysmic variables, possible millisecond-pulsar candidates, and H$\lambda_0 \approx 3360$6 excesses (Lugger et al., 2023). This illustrates the survey’s role as a reusable astro-photometric infrastructure rather than only a source of cluster-specific CMDs.

Across its successive releases, HUGS therefore evolved from a UV imaging program optimized for multiple-population separation into a broader five-band, astrometric, and kinematic framework for Galactic cluster research. Its enduring significance lies in the combination of homogeneous ultraviolet sensitivity, crowded-field photometric reduction, long temporal baselines, and public delivery of catalogues, stacked images, proper motions, membership probabilities, and outer-field atlases (Soto et al., 2016, Nardiello et al., 2018, Libralato et al., 2022).