PHANGS Survey Overview

• PHANGS Survey is an international, multiwavelength initiative that maps star formation and feedback processes in nearby galaxies at high spatial resolution.
• It integrates observations from ALMA, HST, VLT/MUSE, and JWST to produce uniform catalogs and detailed datasets covering molecular clouds, star clusters, H II regions, and nebulae.
• The survey employs advanced data reduction, machine learning classification, and SED fitting techniques, providing vital benchmarks for understanding galactic structure and star formation cycles.

The Physics at High Angular Resolution in Nearby GalaxieS (PHANGS) Survey is an international, multiwavelength collaboration designed to comprehensively chart the processes of star formation, star cluster evolution, feedback, and the interstellar medium (ISM) across the disks of nearby star-forming galaxies. By integrating observational campaigns with ALMA (Atacama Large Millimeter Array), HST (Hubble Space Telescope), VLT/MUSE (Very Large Telescope/Multi Unit Spectroscopic Explorer), and JWST (James Webb Space Telescope), PHANGS has established a reference dataset of unprecedented depth and spatial resolution, enabling the paper of star-forming units (molecular clouds, H II regions, star clusters, and associations) from scales of a few parsecs up to galactic environments over $\sim$100,000 star-forming objects.

1. Survey Design and Scientific Goals

PHANGS was conceived to address fundamental questions regarding the star formation cycle—from the assembly and collapse of giant molecular clouds (GMCs) to the emergence and disruption of young stellar clusters, and the impact of feedback on the ISM—by observing physical processes at matched $\sim$10–150 pc resolution in a representative sample of $\sim$90 nearby ($D<23$ Mpc) spiral galaxies (Leroy et al., 2021).

Key scientific objectives include:

• Quantifying the efficiency and timescales of star formation across diverse galactic environments (disks, bars, spiral arms, galaxy centers) (Lee et al., 2021, Emsellem et al., 2021).
• Characterizing the demographics, lifecycles, and internal properties of GMCs and star clusters (Leroy et al., 2021, Turner et al., 2021).
• Investigating hierarchical, scale-free star formation and the dissolution of fractal structure over time (Turner et al., 2022).
• Calibrating feedback-regulated star formation models by direct measurement of H II region sizes, superbubble energetics, and pressure budgets (Chandar et al., 24 Mar 2025, Watkins et al., 2023).
• Establishing empirical constraints for theoretical models and simulations via uniform, high-resolution, multi-phase ISM datasets (Lee et al., 2022).

2. Multi-Observatory Data Collection and Processing

Each observational component of PHANGS delivers a unique dataset:

• ALMA: CO(2–1) spectral cubes at $\sim$1″ (50–150 pc) resolve the cold molecular gas, i.e., GMCs. The pipeline incorporates multi-array mosaics with total power integration and strict/broad masking, supporting accurate moment maps, GMC catalogs, and cloud property measurements (Leroy et al., 2021, Leroy et al., 2021).
• HST: Five-band (F275W–F814W) UV–optical imaging, plus narrow-band H$\alpha$, enables identification and photometric classification of star clusters, compact associations, and hierarchical stellar structures down to $\sim$2–10 pc. Candidate clusters are separated from stars by multiple concentration indices, then morphologically classified by both human inspection and deep transfer-learning CNNs (ResNet18, VGG19-BN) (Lee et al., 2021, Wei et al., 2019, Whitmore et al., 2021). Watershed algorithms segment multi-scale associations (Larson et al., 2022).
• VLT/MUSE: Integral field spectroscopy across 4800–9300 Å (0.7″ resolution) maps $\sim$20,000 nebulae per galaxy, yielding ionized gas kinematics, chemical abundances, and star formation histories at $\sim$50–100 pc (Emsellem et al., 2021, Groves et al., 2023).
• JWST: Eight-band 2–21 μm IR imaging (NIRCam+MIRI) traces embedded star formation, small dust grain properties, and PAH emission, connecting obscured phases of the star formation cycle to the exposed populations seen in HST/ALMA data (Lee et al., 2022, Kim et al., 2022).
• Ancillary Data: Accurate distances via TRGB analyses place all physical measurements on a firm absolute scale (Anand et al., 2020).

Data reduction pipelines are tailored for cross-facility consistency: high-precision astrometric alignment, PSF homogenization, MOMENT-based mask generation, and rigorous quality assurance are standard (Leroy et al., 2021, Emsellem et al., 2021).

3. Catalogs of Molecular Clouds, Clusters, H II Regions, and Stellar Associations

PHANGS has released homogeneous, multi-scale catalogs representing the major star formation units:

• GMCs: CPROPS-based catalogs from ALMA cubes provide positions, radii, line-widths, masses ($\gtrsim10^5~M_\odot$), and virial parameters for $\sim$12,000 clouds (Leroy et al., 2021).
• Star Clusters/Associations: HST photometry produces a neural network class-assigned catalog of $\sim$100,000 clusters and associations, with $\sim$20,000 inspected by expert classifiers. Each object is characterized by NUV, U, B, V, I magnitudes, structural parameters, and SED-derived ages, masses, and reddenings via Bayesian CIGALE fitting (Maschmann et al., 7 Mar 2024, Turner et al., 2021). Multi-scale watershed segmentation yields hierarchical association catalogs down to 8–64 pc (Larson et al., 2022).
• H II Regions/Nebulae: MUSE provides spectroscopically classified nebulae, emission line fluxes, Balmer decrements, and strong-line metallicities for $\sim$30,000 nebulae per galaxy (Groves et al., 2023). HST H$\alpha$ imaging ($\sim$2–10 pc) enables detailed measurements of region sizes and morphologies (Chandar et al., 24 Mar 2025).

Globular cluster (GC) catalogs for 17 galaxies have also been published, using color-morphology selections to identify ancient GCs and constrain the low-mass ends of the GC luminosity function (Floyd et al., 20 Mar 2024).

4. Methodologies: Automated Classification, SED Fitting, and Machine Learning

Cluster candidate identification and classification leverages a multi-stage approach:

• Initial Selection: Multiple concentration index (MCI) techniques differentiate resolved star clusters from foreground stars using PSF-convolved flux ratios (Lee et al., 2021).
• Morphological Classification: Deep transfer learning—using ImageNet-pretrained ResNet18 and VGG19-BN architectures—reaches agreement rates of 70–80% with human visual classes for symmetric clusters, and 40–70% for asymmetric or ambiguous clusters. Both single-expert and consensus training sets are used; performance is robust to input sample and network architecture changes (Wei et al., 2019, Whitmore et al., 2021).
• SED Fitting: The CIGALE package fits broad-band photometry to simple stellar population models (Bruzual & Charlot 2003 preferred), with attention to extinction (Cardelli et al. law), nebular emission, and Bayesian priors reflecting cluster mass and age function slopes. This enables estimation of cluster ages, masses, and $A_V$ values even in the presence of age-extinction degeneracy (Turner et al., 2021).
• Hierarchical Grouping: Watershed algorithms segment point-source density maps (smoothed on $8$–$64~$pc) into multi-scale associations, self-consistently tracing the hierarchical structure of star formation (Larson et al., 2022).
• Machine Learning and Multiphase ISM Analysis: Unsupervised clustering and dimensionality reduction are applied to correlate PAH band ratios, molecular/ionized gas tracers, and dust physics over $\sim$150~pc scales, identifying new regime-dependent ISM properties (Baron et al., 6 Feb 2024).

5. Science Results: Star Formation, Feedback, and ISM Structure

PHANGS data provide empirical quantification of star formation physics:

• GMC-Star Cluster Link and Timescales: Cross-matching HST clusters and ALMA GMCs in 11 galaxies demonstrates that star clusters remain associated with their parent molecular clouds for only $4$–$6$ Myr. Beyond this age, feedback (photoionization, winds, early SNe) disperses the cloud, implying a short overlap phase in the gas-to-stars transition (Turner et al., 2022). JWST 21 μm imaging extends these timescale measurements to more distant systems, finding embedded star formation phases of $\sim$5.1~Myr, consistent with $\sim$20% of the total cloud lifetimes (Kim et al., 2022).
• Hierarchical Star Formation: Spatial autocorrelation and cross-correlation functions (Landy–Szalay estimator) reveal strongly clustered, fractal distributions for young clusters and GMCs that dissolve over time, supporting the hierarchical star formation paradigm (Turner et al., 2022).
• Cluster Mass and Age Functions: Cluster mass functions follow a power law $dN/dM\propto M^{\beta}$ with $\beta\approx -2$ over 1–200 Myr, and age distributions $dN/d\tau\propto \tau^{\gamma}$ show that global trends are robust to classifier or method used (Turner et al., 2021, Whitmore et al., 2021).
• Feedback Signatures in ISM: Superbubble studies identify 325 candidate molecular superbubbles across 18 galaxies, with a fiducial sample of 88 showing clear expansion in CO. The energetics and ages are best matched by impulsive supernova blast wave models with $\sim$10% coupling efficiency, and as they expand, superbubbles sweep up surrounding molecular gas, potentially triggering subsequent star formation (Watkins et al., 2023).
• Pressure and Dynamics: Star formation rate surface density $\Sigma_{\rm SFR}$ and turbulent velocity dispersion $\sigma$ are tightly correlated, but this is secondary to dependencies on cloud mass and surface density (Larson relations). The virial parameter stays constant, indicating clouds remain self-gravitating even at high $\Sigma_{\rm SFR}$. Star formation efficiency per unit cloud pressure is enhanced in strong spiral arms, and on global scales, $\Sigma_{\rm SFR}\propto (Σ_{\rm gas}R)^{1.0\pm0.3}$ ($R$ is a generic dynamical rate: epicyclic, orbital, or shear), showing star formation tracks gravo-dynamical processes as well as feedback (Elmegreen, 19 Mar 2024).
• Dynamical Resonances and Secular Evolution: Gravitational torque analysis of barred galaxies finds corotation radii close to the bar's end (mean $\mathcal{R} = R_{\rm CR}/R_{\rm bar} = 1.12$), indicating predominantly fast bars, with implications for gas inflow and central star formation (Ruiz-García et al., 17 Oct 2024).

6. Data Products, Accessibility, and Community Impact

All high–level science products—ALMA cubes, HST images and catalogs, MUSE datacubes, JWST imaging, and derived property tables—are publicly released through MAST, CADC, and institutional archives (Leroy et al., 2021, Chandar et al., 24 Mar 2025). Data are science-ready, flux-calibrated, astrometrically aligned, and accompanied by detailed documentation and software tools for community use.

PHANGS public catalogs, moment maps, and cross-matched multiwavelength tables have been adopted as benchmarks for model comparison, provide training/validation data for machine learning in astrophysics, and underpin follow-up programs including resolved studies with JWST and next-generation facilities.

7. Future Prospects and Open Questions

As the PHANGS dataset matures, ongoing and planned research will:

• Expand SED fitting and cluster catalogs to the full sample of 38 HST galaxies, improving cluster property statistics and cross-environment tests (Maschmann et al., 7 Mar 2024).
• Integrate JWST-identified embedded clusters and mid-IR dust diagnostics to trace even earlier phases of star formation.
• Further probe ISM energetics, chemical abundances, and enrichment via spectroscopic mapping of nebulae and H II regions, including radial and azimuthal metallicity gradients (Groves et al., 2023).
• Refine hierarchical structure analysis using machine learning to dissect PAH–gas–star correlations with scale, radiation field, and environment (Baron et al., 6 Feb 2024).
• Test models of feedback, cloud dispersal, and cluster survival in galaxy simulations with empirical timescales and pressure measurements.
• Address the nature of faint GC populations and their connection to disks versus halos, constraining dynamics, disruption, and star formation histories (Floyd et al., 20 Mar 2024).

PHANGS exemplifies the power of panchromatic, multi-scale, multi-phase surveys for advancing the empirical basis of star formation and feedback physics and provides a canonical dataset for the broader extragalactic research community.