IllustrisTNG Cosmological Simulation

• IllustrisTNG is a suite of high-resolution cosmological simulations modeling galaxy formation and evolution with state-of-the-art gravito-magnetohydrodynamics.
• It employs advanced numerical methods and subgrid physics to capture complex processes like star formation, AGN feedback, and chemical enrichment.
• The simulation results are rigorously compared with observations, providing practical insights into galaxy evolution, cluster dynamics, and large-scale cosmic structure.

The IllustrisTNG cosmological simulation suite is a set of large-scale, high-resolution, gravito-magnetohydrodynamical computations designed to model the formation and evolution of structure in the Universe, including galaxies, their gaseous haloes, and the coevolution of dark and baryonic matter. It represents a fundamental advance over previous work both in its physical modeling and in the scale and resolution of its volume. The following entry details the central features, methodology, and scientific results of the IllustrisTNG simulations, focusing on the physics, numerical techniques, and key comparisons with observations and theory.

1. Simulation Framework and Numerical Techniques

The IllustrisTNG suite replaces and extends the original Illustris project. The main simulation volumes are TNG50 (∼50 Mpc box), TNG100 (∼100 Mpc), and TNG300 (∼300 Mpc), along with specialized re-simulations such as TNG-Cluster (zoomed regions sampling rare, high-mass clusters; (Nelson et al., 2023)). The simulations are carried out using the moving-mesh code AREPO, which solves the coupled equations of gravity and (magneto)hydrodynamics via a finite-volume, quasi-Lagrangian Voronoi mesh. Gravity is computed via a hybrid Tree-Particle-Mesh scheme, and time integration uses adaptive, hierarchical timestepping. The hydrodynamics solver is supported by a least-squares fit gradient estimator and second-order Runge–Kutta integrator for mesh movement and fluid quantities (Pillepich et al., 2017).

Resolution and dynamic range are among the highest yet attempted for cosmological hydrodynamics, e.g., TNG100 evolves 2 × 1820³ resolution elements with a baryonic mass resolution of ≈1.4 × 10⁶ M_⊙ and dark matter resolution of ≈7.5 × 10⁶ M_⊙, with gravitational softening of ≈185 pc for baryons (Nelson et al., 2018, Nelson et al., 2017). Monte Carlo tracer particles supplement the Lagrangian fluid tracking.

2. Physical Modeling of Galaxy Formation

The TNG models implement a state-of-the-art suite of subgrid models for the physics of galaxy formation:

• Gas Physics: Radiative cooling includes both primordial and metal-line processes, with self-shielding corrections and coupling to a redshift-dependent UV background.
• Star Formation and Feedback: Star formation proceeds in dense gas by a stochastic model with a fixed threshold. Supernova-driven winds are implemented via a kinetic feedback prescription with isotropic wind launching, a velocity tied to the local dark matter velocity dispersion, and explicit dependence on gas metallicity and redshift (Pillepich et al., 2017). The wind energy is split between kinetic and thermal channels via a tunable fraction.
• Stellar Evolution and Chemical Enrichment: Nine tracked metal species are enriched through stellar evolution channels (Type Ia/II SNe, AGB).
• Black Hole Physics and AGN Feedback: SMBH seeding, Bondi accretion (using the maximum of Bondi and Eddington rates), and feedback in dual modes are included. At high accretion rates, feedback is thermal (quasar-mode). At low accretion rates, a kinetic feedback model ("kinetic wind") injects mechanical energy in discrete events, with a mass-dependent Eddington ratio threshold for mode-switching (Weinberger et al., 2017). Kinetic-mode feedback efficiently quenches star formation in massive galaxies.
• Magnetohydrodynamics: For the first time in large cosmological runs, ideal MHD is self-consistently included, with the initial uniform seed field (e.g., 10⁻¹⁴ G) rapidly amplified via structure formation (Marinacci et al., 2017).

Numerical improvements, such as conserved metal advection and adaptive gravitational softening, render the simulations robust to mesh geometry and spatial/temporal resolution.

3. Galaxy Populations and Evolution

IllustrisTNG self-consistently follows the formation, growth, and transformation of diverse galaxy populations:

• Mass and Morphologies: Simulated volumes contain ∼40,000 (TNG100) to ∼100,000 (TNG300) well-resolved galaxies at z=0, displaying a broad range of morphologies (early, late, and irregular types). Morphological evolution tracks observed trends: at high z (z≳5), galaxies are clumpy and irregular; toward z∼0, disks and bulges are more prevalent (Genel et al., 2014, Pillepich et al., 2019).
• Stellar Mass Functions: The GSMF is matched to observations at z=0 and broadly tracks evolution out to z∼7, with conversion efficiency peaking at halo masses of ∼10¹² M_⊙ (Vogelsberger et al., 2014, Genel et al., 2014).
• Star Formation Histories: The cosmic SFRD evolution is reasonable, with an appropriate peak at z∼2. The sSFR vs. halo mass and redshift tracks both observed and theoretical expectations, though the redshift dependence is slightly shallower than observed between z=1–3.

Dry mergers drive continued stellar mass growth on the red sequence, particularly in massive galaxies: after quenching, galaxies with M_* > 10¹¹ M_⊙ can accumulate ∼25% of their stellar mass in dry mergers (Nelson et al., 2017).

Environmental effects are robust: dense regions host a higher fraction of red, passive galaxies, and merger rates and star formation are enhanced in locally dense environments (Koncz et al., 11 Sep 2025).

4. Feedback Processes: Impact and Observational Comparisons

Baryonic feedback—especially energy injection from both supernovae and AGN—strongly influences galactic and halo-scale properties:

• Halo Mass Function and Baryonic Suppression: Baryonic feedback reduces the abundance and masses of dark matter haloes at both low and high masses by up to ∼30% relative to dark-matter-only runs. This suppression arises predominantly from supernova-driven winds in small halos and AGN feedback in cluster-scale systems (Vogelsberger et al., 2014).
• Stellar–Halo Mass Relation: The SMHM relation is steep, with scatter ∼0.12 dex at M₍*,cen₎ (30 kpc) ∝ M₍500c₎^0.49 (Pillepich et al., 2017). Total halo mass is an excellent predictor of total stellar mass content, and proposed analytic (sigmoid) models allow a ~5–10% accuracy in the cumulative stellar mass profile given a single mass estimate.
• Quenching and Color Bimodality: Low-accretion kinetic SMBH feedback is identified as the principal quenching mechanism for massive galaxies, naturally producing a sharp blue-to-red color transition at M_*∼10^10.5 M_⊙ (Nelson et al., 2017). The timescale for transition across the green valley is ∼1.6 Gyr, dropping for higher mass systems.

Observational comparisons include consistent stellar and baryonic Tully–Fisher relations, accurate disk/elliptical fractions, and agreement with galaxy luminosity functions (in SDSS bands) [(Vogelsberger et al., 2014); (Nelson et al., 2017)]. However, the model at times overpredicts the high-mass (bright) end of the GSMF and may underpredict hot gas fractions in clusters, suggesting AGN feedback could be overly efficient in massive haloes.

5. Large-Scale Structure, Clustering, and the Cosmic Web

IllustrisTNG enables precision studies of clustering and cosmic structure:

• Matter and Galaxy Clustering: Baryonic effects increase the clustering amplitude of dark matter on small scales and suppress the total matter power spectrum by ≈20% out to k∼10 h Mpc⁻¹ relative to dark-matter-only cases, due to efficient baryon removal (Springel et al., 2017).
• Two-Point and Higher-Order Statistics: The stellar and galaxy correlation functions are close to a power law at z=0–2, with slopes declining from γ∼1.8 at z=0 to γ∼1.6 at z=1. The clustering length increases with stellar mass, and bias evolves strongly with redshift.
• Intrinsic Alignments: The three-point intrinsic alignment bispectra for dark matter halos are detected with high significance, and values of the alignment amplitude A₍IA₎ inferred from both power spectra and bispectra are consistent up to k=2 h Mpc⁻¹ (Pyne et al., 2022).

Baryonic feedback imprints scale-dependent bias, influencing baryonic acoustic oscillation (BAO) peak positions by ∼5%, but this effect can be corrected in cosmological analyses.

6. The Intracluster Medium, Magnetic Fields, and Radio Signatures

The simulations evolve magnetic fields and predict radio and X-ray emission self-consistently:

• Magnetic Field Evolution: Initial comoving seed fields of ∼10⁻¹⁴ G are amplified by turbulence, shear, radiative cooling, and feedback to μG strengths in cluster cores (Marinacci et al., 2017, Nelson et al., 2023). The morphology correlates with galaxy type: discs support larger-scale ordered fields, ellipticals host more irregular fields.
• Diffuse Radio Haloes: Synchrotron emissivity is computed from models of the relativistic electron population and local magnetic field strengths; extended radio haloes are predicted with power–mass, power–SZ, and power–X-ray scaling relations matching VLA and SKA observations. Radio surface brightness profiles are typically exponential, matching observations of Coma and Perseus (Marinacci et al., 2017).
• ICM Properties and Scaling Relations: The TNG-Cluster extension leverages zoom re-simulations from a 1 Gpc parent box to sample 352 high-mass clusters. Thermodynamic, chemical, and magnetic radial profiles are broadly consistent with X-ray and SZ observations. Magnetic fields reach ∼1 μG in cluster cores, and the ICM metallicity declines with radius from ∼0.5 Z_⊙ to ∼0.1 Z_⊙. SZ scaling relations and gas fractions in clusters approach cosmic values (Nelson et al., 2023).

7. Data Access, Tools, and Ongoing Developments

IllustrisTNG data—snapshots, halo/subhalo catalogs, merger trees, mock images, and synthetic observables—are fully publicly released via an integrated web API, with data analysis exemplified in Python, IDL, and Matlab, and with cloud-based Jupyter notebooks (Nelson et al., 2018). The simulated dataset enables direct comparisons to observational data, including mock SEDs and photometry generated by post-processing (e.g., SKIRT), which enable realistic virtual surveys from FUV through the sub-millimeter (Gebek et al., 8 May 2024).

Scientific caveats are thoroughly documented: limited resolution at the low-mass end, known temperature correction issues for low-density gas, and explicit flags for non-cosmological subhalos. Convergence and cross-sample (volume, resolution) comparisons are supported. The simulation’s flexibility allows extension to more advanced CR transport models, explicit cosmic ray feedback, and further plasma physics as required by future physical and observational constraints (Ramesh et al., 26 Sep 2024).

---

In summary, the IllustrisTNG cosmological simulation suite provides a comprehensive, physically motivated model for the formation and evolution of galaxies and large-scale structure. It synthesizes the detailed impact of baryonic processes, black hole feedback, and magnetic fields on both small (galaxy) and large (cluster and cosmic web) scales, offering a versatile tool for interpreting a wide range of multiwavelength astronomical datasets and for advancing theoretical understanding of cosmic evolution.