C-EAGLE: High-Res Galaxy Cluster Simulations

• C-EAGLE Project is a suite of 30 high-resolution cosmological zoom simulations of massive galaxy clusters, extending the EAGLE model to extreme environments.
• It employs advanced subgrid physics with AGN feedback using the ANARCHY SPH scheme in P-Gadget, accurately reproducing key X-ray, SZ, and stellar observables.
• The project provides a mock-observational framework for calibrating methods that connect dark-matter structure to observable cluster and galaxy properties.

Searching arXiv for C-EAGLE / Cluster-EAGLE papers to ground the article in the primary literature. The C-EAGLE Project (“Cluster-EAGLE”) is a suite of cosmological hydrodynamical zoom simulations of massive galaxy clusters developed to extend the EAGLE galaxy-formation model into the cluster regime, where representative periodic volumes contain too few rare, high-mass systems for detailed study. Introduced as a set of 30 cluster resimulations spanning $10^{14}<M_{200}/\mathrm{M}_{\odot}<10^{15.4}$, C-EAGLE applies the EAGLE subgrid framework at high spatial and mass resolution to rich cluster environments, enabling simultaneous analysis of resolved cluster galaxies, black holes, intracluster gas, metals, and dynamical structure (Barnes et al., 2017). Across subsequent work, the project has served both as a laboratory for testing baryonic physics in extreme halos and as a calibration set for inference frameworks that connect dark-matter structure to observable cluster and galaxy properties (Armitage et al., 2017, Armitage et al., 2018, Pearce et al., 2020, Lovell et al., 2021, Negri et al., 2022, Vurm et al., 2023).

1. Origin, design goals, and place within the EAGLE programme

C-EAGLE was created because the original periodic EAGLE volumes, although calibrated to reproduce low-redshift galaxy observables such as the galaxy stellar mass function, the galaxy size–mass relation, and the black hole mass–stellar mass relation, were too small to sample the richest cluster environments in statistically useful numbers (Barnes et al., 2017). The project therefore asks whether a galaxy-formation model tuned primarily on field-galaxy observables also yields realistic galaxy clusters when applied to much more massive and overdense systems.

The project is built around 30 cluster zoom simulations selected from a 3.2 Gpc dark-matter-only parent volume in a Planck 2013 cosmology (Barnes et al., 2017). Candidate halos were binned into 10 logarithmic mass bins from $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$ to $15.4$, and 3 halos per bin were chosen, excluding systems with a more massive neighbour within 30 Mpc or $20\,r_{200}$ (Barnes et al., 2017). A related formulation gives the parent sample as 185,150 haloes with $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$, from which the final cluster set was drawn (Armitage et al., 2017).

A central motivation was to provide cluster simulations with both high enough resolution to resolve cluster galaxies and sufficiently realistic baryonic physics to compare with observed X-ray, Sunyaev–Zel’dovich, metallicity, and galaxy-population statistics (Barnes et al., 2017). This made C-EAGLE a complement to representative-volume EAGLE rather than a replacement for it. In later work, that complementarity became explicit: periodic EAGLE volumes represent ordinary environments, whereas C-EAGLE supplies the rare overdense regimes needed for studies of environmental bias, cluster-galaxy dynamics, and cluster outskirts (Lovell et al., 2021).

2. Numerical realization and physical model

C-EAGLE uses the EAGLE AGNdT9 galaxy formation model with the ANARCHY smoothed-particle hydrodynamics implementation in modified P-Gadget-3 / P-GADGET3 (Barnes et al., 2017, Armitage et al., 2017, Negri et al., 2022). The subgrid model includes radiative cooling and photo-heating, star formation, stellar evolution and chemical enrichment, stellar feedback, black-hole seeding and growth, and AGN feedback (Barnes et al., 2017, Armitage et al., 2017, Pearce et al., 2020).

The standard resolution parameters recur across the project: gas particle mass $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$, dark-matter particle mass $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$, gravitational softening of $2.66$ comoving kpc until $z=2.8$, and $0.70$ physical kpc at lower redshift (Barnes et al., 2017, Armitage et al., 2017, Pearce et al., 2020). The AGN feedback model uses a heating temperature $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$0 (Armitage et al., 2017, Pearce et al., 2020). In one summary of the AGNdT9 calibration, the AGN parameters are given as $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$1, $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$2, and $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$3 K (Barnes et al., 2017).

The zoom technique yields high-resolution cluster regions embedded within lower-resolution surroundings that preserve the correct large-scale tidal field. One description states that the high-resolution region contains no low-resolution contamination within at least $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$4 at $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$5 (Armitage et al., 2017). Another notes that all 30 clusters were simulated to at least $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$6 and that 24 extend to $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$7 as the Hydrangea sample (Negri et al., 2022). Each cluster also has a corresponding dark-matter-only realization, often denoted C-EAGLE-DMO, enabling controlled baryonic-versus-collisionless comparisons (Armitage et al., 2017).

In chemical-evolution work, the simulations explicitly track 11 chemical elements,

$\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$8

with enrichment contributions from AGB stars, Type II supernovae, and Type Ia supernovae (Pearce et al., 2020). This chemical bookkeeping is integral to the project’s ability to model intracluster metallicity profiles and abundance ratios.

3. Sample architecture, observables, and methodological strategy

C-EAGLE is unusual in combining resolved internal cluster structure with observationally motivated post-processing. The introductory project paper does not merely compare intrinsic simulation quantities with observations; it constructs mock X-ray spectra, fits single-temperature plasma models, infers density, temperature, and metallicity profiles, and derives hydrostatic masses using observational-style procedures (Barnes et al., 2017). This allows direct definitions of quantities such as $\log_{10}(M_{200}/\mathrm{M}_{\odot})=14.0$9, $15.4$0, $15.4$1, $15.4$2, and $15.4$3 in forms comparable to cluster surveys (Barnes et al., 2017).

Several later studies extend this observational strategy into other domains. Galaxy luminosities have been computed from ultraviolet to infrared using E-MILES simple stellar population modelling with dust attenuation (Negri et al., 2022). Filament environments have been identified around a Coma-like C-EAGLE cluster using DisPerSE applied to the galaxy distribution, rather than directly to gas or dark matter, in order to mimic feasible observational workflows (Vurm et al., 2023). Dynamical mass-estimator studies explicitly compare idealized 3D analyses to projected, interloper-contaminated cases meant to emulate spectroscopic surveys (Armitage et al., 2018).

This strategy makes C-EAGLE not only a simulation set but also a mock-observational framework. A plausible implication is that the project’s influence derives as much from methodological calibration as from raw numerical resolution. It provides a common numerical laboratory in which X-ray, SZ, kinematic, photometric, and environmental diagnostics can be evaluated against known ground truth.

4. Global cluster properties and baryonic realism

The foundational C-EAGLE analysis finds that the simulations reproduce many observed cluster scaling relations while also exposing systematic tensions in the baryonic model (Barnes et al., 2017). The total stellar content is in broad agreement with observed relations, with roughly 2% of the total cluster mass in stars and a relatively flat stellar mass fraction across the sampled halo-mass range (Barnes et al., 2017). The black-hole population is also consistent with observed black hole mass–stellar mass relations; the project paper reports 1358 black holes within $15.4$4, about 90% of them in satellites (Barnes et al., 2017).

X-ray and SZ properties are described as being in reasonable agreement with observations. The spectroscopic X-ray temperature–mass relation has a scatter of about $15.4$5, the soft-band luminosity relation has $15.4$6, and the SZ observable $15.4$7 has $15.4$8 in the full sample, reduced to $15.4$9 for relaxed clusters (Barnes et al., 2017). The simulations also reproduce the total metal content and its radial distribution in the ICM comparatively well, with a median iron abundance of about $20\,r_{200}$0 (Barnes et al., 2017).

The principal discrepancies lie in the gas content and core thermodynamics. The clusters are too gas rich, suggesting that the AGN feedback model is not efficient enough at expelling gas from high-redshift cluster progenitors (Barnes et al., 2017). Hydrostatic mass estimates are biased low, with a median hydrostatic bias $20\,r_{200}$1, while the full mock X-ray spectroscopic pipeline yields $20\,r_{200}$2 (Barnes et al., 2017). Core temperatures are too high, with peak central temperatures about 60% higher than observed, and entropy cores can be up to a factor of 5 larger than observed (Barnes et al., 2017). The simulations also contain no cool-core clusters under the criteria used there (Barnes et al., 2017).

These results are often interpreted as evidence that the overall EAGLE energy budget and enrichment scheme are broadly viable in cluster environments, but that the efficiency and temporal structure of AGN feedback remain imperfect (Barnes et al., 2017). This suggests that C-EAGLE is both a validation and a stress test of EAGLE’s subgrid physics.

5. Cluster dynamics, tracer populations, and environmental processing

A major branch of the C-EAGLE literature concerns the fidelity with which cluster galaxies trace the underlying gravitational potential. In the velocity-bias study, C-EAGLE is used to test the relation between galaxy velocity dispersion and cluster mass (Armitage et al., 2017). The dark-matter particle velocity dispersion within $20\,r_{200}$3 follows the expected virial-like scaling, with fitted parameters at $20\,r_{200}$4 of $20\,r_{200}$5, $20\,r_{200}$6, and $20\,r_{200}$7 for the hydrodynamic runs (Armitage et al., 2017). Velocity bias is defined as

$20\,r_{200}$8

The central result is that selection by total subhalo mass yields a positive bias, typically 5–10%, whereas selection by stellar mass yields an almost unbiased estimator, with bias $20\,r_{200}$9, out to $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$0 and with little dependence on aperture from $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$1 to $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$2 (Armitage et al., 2017). This is interpreted in terms of infall time, stripping, and dynamical friction: recent infallers are dynamically hot, while satellites accreted more than $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$3 ago become progressively colder (Armitage et al., 2017).

The companion mass-estimator study evaluates Jeans, virial, and caustic methods using stellar-mass-selected galaxies with $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$4 (Armitage et al., 2018). Across these methods, cluster mass estimates are approximately unbiased on average, with scatter in the range 0.09–0.15 dex (Armitage et al., 2018). Averaging the three estimators yields an unbiased combined estimate with scatter

$M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$5

even when interlopers are included and $M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$6 is not known in advance (Armitage et al., 2018). By contrast, mock X-ray hydrostatic masses are about 30% lower than dynamical masses (Armitage et al., 2018).

Environmental processing of star formation is another C-EAGLE application. A study combining EAGLE and C-EAGLE defines a star-formation concentration index

$M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$7

and shows that low-$M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$8 galaxies, with centrally concentrated star formation, are more common in denser environments at low redshift (Wang et al., 2023). C-EAGLE cluster satellites display a particularly pronounced low-$M_{\rm 200c}>10^{14}\,\mathrm{M_\odot}$9 tail, consistent with stronger outside-in environmental quenching in rich clusters (Wang et al., 2023). The same work finds that the quenching timescale decreases with redshift, from a median $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$0 Gyr at $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$1 to $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$2 Gyr at $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$3, and that the outside-in signature weakens or disappears by $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$4–2 (Wang et al., 2023).

Taken together, these results establish C-EAGLE as a framework for studying how galaxies behave as tracers, baryonic subsystems, and environmentally processed satellites within cluster potentials.

6. Intracluster medium, metals, outskirts, and cosmic-web interfaces

C-EAGLE has been used extensively to study the thermodynamic and chemical structure of the intracluster medium. In the metallicity-evolution analysis, the project’s 30 high-resolution cluster zooms are used to test the early enrichment model (Pearce et al., 2020). The study employs 29 clusters in its main analysis because CE-27 experienced an extreme early AGN-driven event that expelled most of its gas (Pearce et al., 2020). Using mass-weighted abundances,

$$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$5

the paper finds that total metallicity, Si, and O show very little redshift evolution in cluster outskirts beyond roughly $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$6–$$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$7, out to at least $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$8 (Pearce et al., 2020). Fe is the exception, evolving at large radius because of its delayed Type Ia supernova origin (Pearce et al., 2020).

In the cluster cores, the same study reports strong redshift evolution: metallicity is higher at high redshift and decreases toward low redshift, apparently driven by accretion of low-metallicity gas and by interactions between outflowing, AGN-heated gas and surrounding material (Pearce et al., 2020). The median $$m_{\rm gas}\simeq 1.8\times 10^6\,\mathrm{M}_{\odot}$$9 Fe profile has the correct qualitative shape but is low in normalization by about a factor of 1.5 (Pearce et al., 2020). A particle-tracking analysis of cluster CE-08 indicates that only about 2% of the gas originally in the core at $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$0 remains there by $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$1, illustrating the degree of core processing (Pearce et al., 2020).

The project also reaches beyond the virialized ICM into the cluster–cosmic-web interface. A proof-of-method analysis of the most massive C-EAGLE cluster, CE-29, applies DisPerSE to galaxies around a Coma-like halo with $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$2 (Vurm et al., 2023). Filaments identified this way account for about 50% of the hot WHIM with $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$3 in the cluster vicinity (Vurm et al., 2023). The filament gas remains in approximate free-fall down to $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$4, then slows as ambient pressure rises, with density increasing from $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$5 at large radius to $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$6 near the cluster boundary, and temperature rising from $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$7–$$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$8 K to $$m_{\rm DM}\simeq 9.7\times 10^6\,\mathrm{M}_{\odot}$$9–$2.66$0 K (Vurm et al., 2023). In one filament, denoted F9, an accretion shock at $2.66$1 is consistent with a Mach number $2.66$2 (Vurm et al., 2023).

These studies broaden the scope of C-EAGLE from virialized cluster structure to enrichment histories and cluster feeding channels. A plausible implication is that the project’s spatial reach, especially in the Hydrangea subset, is as important as its internal resolution for understanding how clusters exchange matter with their surroundings.

7. Galaxy populations, statistical emulation, and later extensions

C-EAGLE has also been used to characterize cluster galaxy populations directly. A luminosity study computes AB magnitudes for simulated galaxies from ultraviolet to infrared using E-MILES stellar population synthesis and a dust model based on the surface density of heavy elements in the gas phase (Negri et al., 2022). At $2.66$3, the $2.66$4 colour–stellar mass diagram shows a defined red sequence reaching $2.66$5, about 0.05 dex redder than EAGLE at high masses, alongside a blue cloud when field galaxies are included (Negri et al., 2022). The cluster inner regions are dominated by red-sequence galaxies, although some blue galaxies persist (Negri et al., 2022).

The same study models cluster luminosity functions with single and double Schechter forms and finds that the knee luminosity brightens toward redder bands and with cluster mass (Negri et al., 2022). The faint-end slope is comparatively stable, typically $2.66$6 to $2.66$7, with only a modest optical upturn in some mass bins (Negri et al., 2022). The simulated luminosity functions reproduce, within observational errors, the spectroscopic luminosity functions of Hercules and Abell 85, including the faint-end upturn in the latter (Negri et al., 2022). This establishes C-EAGLE as a generator of photometric cluster observables as well as intrinsic stellar-mass distributions.

A distinct line of work uses C-EAGLE as training data for machine learning. In a halo-to-galaxy mapping framework, periodic EAGLE simulations are combined with C-EAGLE cluster zooms to train a tree based machine learning method that predicts baryonic galaxy properties from dark-matter halo properties (Lovell et al., 2021). The logic is that EAGLE supplies representative average-density environments, while C-EAGLE supplies rare overdense environments, allowing the model to “learn the bias of galaxy evolution in differing environments” (Lovell et al., 2021). This learned mapping is then applied to the P-Millennium dark-matter-only simulation of volume $2.66$8, yielding predictions for key baryonic distribution functions and clustering statistics at a tiny fraction of the cost of full hydrodynamics (Lovell et al., 2021). In this role, C-EAGLE functions less as an end in itself and more as an essential environmental calibration set.

More recent work extends C-EAGLE into dark-matter phenomenology. A 2026 study resimulates representative C-EAGLE clusters in both CDM and SIDM with $2.66$9 and compares the morphology of dark matter to baryonic tracers using a Weighted Overlap Coefficient (Yoo et al., 5 Apr 2026). That analysis finds that BCG+ICL is the best morphological tracer of dark matter overall, while gas becomes relatively more dark-matter-like in SIDM than in CDM (Yoo et al., 5 Apr 2026). This suggests that the C-EAGLE framework is adaptable to controlled experiments on dark-sector microphysics in addition to baryonic astrophysics.

A recurrent misconception is that C-EAGLE is simply “EAGLE at higher mass.” The literature indicates a more specific role: it is a cluster-focused zoom programme designed to preserve EAGLE’s subgrid philosophy while exposing it to the rarest, densest environments and the observational systematics unique to clusters (Barnes et al., 2017, Lovell et al., 2021). Another plausible misunderstanding is that the project is only about intracluster gas. In fact, its published applications span cluster galaxies, kinematic mass estimators, environmental quenching, luminosity functions, metal enrichment, cluster outskirts, cosmic-web interfaces, and machine-learning calibration (Armitage et al., 2017, Armitage et al., 2018, Pearce et al., 2020, Negri et al., 2022, Vurm et al., 2023).

In aggregate, the C-EAGLE Project occupies a distinctive position in computational astrophysics: it is a high-resolution cluster laboratory, an observational calibration platform, and a bridge between detailed baryonic simulations and scalable inference models for large cosmological volumes.