Gizmo-Simba: GIZMO-SIMBA Cosmological Simulations

• Gizmo-Simba is a simulation framework combining Simba’s galaxy-formation physics with GIZMO’s MFM hydrodynamics for both cosmological and cluster-scale studies.
• It recalibrates baryonic processes to match key stellar and gas observables, enhancing predictions of stellar mass fractions and AGN-driven feedback.
• The framework enables detailed analyses of oxygen absorption, WHIM properties, and cluster baryon profiles, aligning simulation outputs with observational data.

Gizmo-Simba denotes the use of the Simba galaxy-formation model within the GIZMO simulation code, and in the cluster literature it more specifically denotes the GIZMO-SIMBA realization of The Three Hundred project. In published usage, the term therefore refers both to the original cosmological Simba simulations run with GIZMO’s Meshless Finite Mass (MFM) hydrodynamics and to a cluster-zoom extension in which Simba-like baryonic physics was mildly recalibrated for lower-resolution cluster environments (1901.10203, Cui et al., 2022). It should be distinguished from the unrelated SiMBA microkinetic model-discovery tool in chemical engineering (Servia et al., 2024).

1. Definition, lineage, and nomenclature

The immediate lineage of Gizmo-Simba runs through three layers. At the numerical level sits GIZMO, introduced as a mesh-free, quasi-Lagrangian Godunov code with the MFM and MFV methods (Hopkins, 2014). At the galaxy-formation level sits Simba, introduced as a cosmological simulation suite that uses GIZMO’s MFM hydrodynamics while adding sub-resolution models for radiative cooling, star formation, stellar feedback, black-hole growth, AGN feedback, and on-the-fly dust evolution (1901.10203). At the cluster-zoom level sits GIZMO-SIMBA within The Three Hundred, which ports the Simba physical model into 324 massive cluster zooms and recalibrates selected parameters to match $z\approx 0$ cluster stellar observables (Cui et al., 2022).

This layered meaning matters because not all papers that discuss GIZMO or Simba are about the same object. Some papers use Simba for the original field/group cosmological boxes, some use GIZMO-SIMBA for the Three Hundred cluster realization, and some discuss only the GIZMO solver family. A related but separate ambiguity is orthographic: SiMBA in the chemical-kinetics literature stands for Simplest Mechanism Builder Algorithm, and is not part of the astrophysical GIZMO/Simba ecosystem (Servia et al., 2024).

2. Numerical foundation and baseline Simba physics

The numerical backbone comes from GIZMO’s mesh-free Godunov formulation. In the MFM/MFV framework, the fluid is represented by moving resolution elements with kernel-defined effective volumes $V_i$, pairwise effective face areas $\mathbf{A}_{ij}$, and a conservative finite-volume update of the form

$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$

with fluxes obtained from a Riemann problem and gradients reconstructed by a matrix least-squares estimator (Hopkins, 2014). Simba uses MFM, not MFV, so gas elements retain fixed mass while shocks and discontinuities are handled by the Godunov solver (1901.10203).

The baseline Simba suite adopts a Planck-like cosmology,

$$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$

$\sigma_8=0.82,\quad n_s=0.97,$

and its principal boxes include $100\,h^{-1}{\rm Mpc}$, $50\,h^{-1}{\rm Mpc}$, $25\,h^{-1}{\rm Mpc}$, and $12.5\,h^{-1}{\rm Mpc}$ volumes with $V_i$0 dark-matter particles and $V_i$1 gas elements in each run (1901.10203). Cooling and photoheating use Grackle-3.1 with non-equilibrium primordial chemistry, metal cooling, self-shielding via Rahmati et al., and a uniform Haardt & Madau (2012) UV background. Star formation is molecular-gas regulated, with

$V_i$2

and enrichment tracks 11 elements: H, He, C, N, O, Ne, Mg, Si, S, Ca, and Fe (1901.10203).

The black-hole model is the signature feature that differentiates Simba from its predecessor Mufasa. Black holes are seeded in galaxies once

$V_i$3

with $V_i$4 and $V_i$5, corresponding to seeding near $V_i$6 (1901.10203). Accretion is dual-mode: torque-limited accretion from cold gas and Bondi accretion from hot gas, with total growth

$V_i$7

AGN feedback is kinetic and bipolar. The radiative-mode velocity is

$V_i$8

and at $V_i$9 the model transitions toward jet mode,

$\mathbf{A}_{ij}$0

with full jet strength reached by $\mathbf{A}_{ij}$1 and a jet-mode BH threshold $\mathbf{A}_{ij}$2 (1901.10203). Simba also includes X-ray feedback in gas-poor full-jet systems and an on-the-fly dust model with production, growth, sputtering, shock destruction, and astration (1901.10203).

3. The Three Hundred GIZMO-SIMBA cluster realization

Within The Three Hundred, GIZMO-SIMBA is a suite of 324 zoom re-simulated cluster regions selected from MDPL2. The parent simulation is a $\mathbf{A}_{ij}$3 box with $\mathbf{A}_{ij}$4 dark-matter particles, and each selected cluster is resimulated in a region of radius $\mathbf{A}_{ij}$5 with high-resolution particle masses

$\mathbf{A}_{ij}$6

and a fixed comoving gravitational softening of $\mathbf{A}_{ij}$7 (Cui et al., 2022). Haloes are catalogued with AHF, galaxies with Caesar, and most cluster-integrated comparisons use $\mathbf{A}_{ij}$8 and quantities within $\mathbf{A}_{ij}$9 (Cui et al., 2022).

The cluster version is not a verbatim re-run of public Simba. It is a mildly re-calibrated port designed to recover three $$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$0 stellar benchmarks in one representative cluster region: total stellar mass fraction within $$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$1, the BCG stellar mass–halo mass relation, and the satellite stellar mass function (Cui et al., 2022). The principal changes are summarized below.

Ingredient	GIZMO-SIMBA setting	Public Simba reference
SF density threshold	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$2	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$3
Metallicity floor for $$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$4	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$5	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$6
Wind normalization prefactor	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$7	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$8
BH kernel maximum radius	$$\frac{d}{dt}(V_i\mathbf{U}_i) + \sum_j \tilde{\mathbf{F}}_{ij}\cdot \mathbf{A}_{ij} = 0,$$9	$$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$0
Bondi cap mass	$$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$1	$$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$2
Maximum jet speed	$$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$3	$$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$4
Softening	fixed $$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$5 comoving	variable, minimum $$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$6

The cluster-scale consequences are substantial. At $$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$7, GIZMO-SIMBA and Gadget-X yield very similar stellar mass fractions within $$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$8, while GIZMO-SIMBA predicts a steeper $$\Omega_m=0.3,\quad \Omega_\Lambda=0.7,\quad \Omega_b=0.048,\quad H_0=68\,{\rm km\,s^{-1}\,Mpc^{-1}},$$9-$\sigma_8=0.82,\quad n_s=0.97,$0 relation, especially below $\sigma_8=0.82,\quad n_s=0.97,$1 (Cui et al., 2022). The evolutionary path is different even when the present-day stellar fraction is similar: GIZMO-SIMBA forms stars earlier, has about 50% more stellar mass at $\sigma_8=0.82,\quad n_s=0.97,$2 than Gadget-X for matched $\sigma_8=0.82,\quad n_s=0.97,$3 descendant mass bins, and then quenches rapidly so that the stellar fractions converge by $\sigma_8=0.82,\quad n_s=0.97,$4 (Cui et al., 2022). Its BCGs are typically red, with

$\sigma_8=0.82,\quad n_s=0.97,$5

and their stellar ages peak around formation redshifts $\sigma_8=0.82,\quad n_s=0.97,$6, whereas Gadget-X retains a younger, bluer BCG tail (Cui et al., 2022). The satellite stellar mass function is likewise one of the explicit successes of the recalibration (Cui et al., 2022).

4. Diffuse baryons, the WHIM, and oxygen absorption in Simba

A major use of baseline Simba has been the study of diffuse baryons and the warm-hot intergalactic medium (WHIM). In the oxygen-absorber analysis based on the $\sigma_8=0.82,\quad n_s=0.97,$7 Simba runs, the central question is whether Simba’s strong AGN jets—which had previously implied that about 70% of baryons at $\sigma_8=0.82,\quad n_s=0.97,$8 reside in the WHIM—would leave an observable imprint in intergalactic O VI, O VII, and O VIII absorption (Bradley et al., 2022). The comparison is between a full-physics run (s50) and a matched no-jet/no-X-ray run (s50nojet), using synthetic spectra generated with PYGAD, ion fractions from CLOUDY, and projected CDDFs constructed with yt/Trident (Bradley et al., 2022).

The principal result is that jets strongly alter the thermal and morphological distribution of oxygen-bearing gas, but the properties of detectable absorbers shift less dramatically than the global baryon census might suggest (Bradley et al., 2022). In Simba, all three ions are commonly associated with moderate overdensities,

$\sigma_8=0.82,\quad n_s=0.97,$9

filamentary environments, and typical halo-centric distances of

$100\,h^{-1}{\rm Mpc}$0

so most absorbers arise outside haloes rather than in intrahalo gas (Bradley et al., 2022). The typical absorber temperatures at $100\,h^{-1}{\rm Mpc}$1 are roughly $100\,h^{-1}{\rm Mpc}$2–$100\,h^{-1}{\rm Mpc}$3 K for O VI, $100\,h^{-1}{\rm Mpc}$4–$100\,h^{-1}{\rm Mpc}$5 K for O VII, and $100\,h^{-1}{\rm Mpc}$6–$100\,h^{-1}{\rm Mpc}$7 K for O VIII, with O VII and O VIII shifting to hotter medians when jets are included (Bradley et al., 2022).

Observationally, Simba matches the observed O VI CDDF fairly well, underpredicts the preliminary O VII CDDF inferred from the two intergalactic Nicastro et al. systems, and makes a genuinely predictive claim for O VIII: O VIII is the ion most sensitive to jet heating, with the high-column end diverging most strongly between jet and no-jet runs by $100\,h^{-1}{\rm Mpc}$8 (Bradley et al., 2022). The paper therefore argues that current data do not rule out Simba’s widespread jet heating, and that future X-ray missions would be required to test the high-column O VII/O VIII tails decisively (Bradley et al., 2022).

5. Cluster baryon structure and subhalo compactness

When GIZMO-SIMBA is applied to cluster environments, two recurring themes emerge. First, it performs comparatively well outside cluster cores but shows central gas-structure tensions. Second, it improves several galaxy-side statistics relative to Gadget-X without fully resolving lensing-based compactness discrepancies.

In the analysis of physical baryon profiles in The Three Hundred, GIZMO-SIMBA and Gadget-X show broadly similar behavior for $100\,h^{-1}{\rm Mpc}$9, where gas-density, temperature, pressure, and entropy profiles are more nearly self-similar and closer to low-redshift observations (Li et al., 2023). In the core, however, GIZMO-SIMBA has lower central gas density, higher central temperature, and higher, flatter entropy than Gadget-X, and the authors explicitly suggest that its AGN model may be too strong for this cluster simulation (Li et al., 2023). The same study concludes that GIZMO-SIMBA appears to contain essentially no cool-core clusters, and that its high-redshift profile evolution is stronger than in Gadget-X, especially for temperature and entropy (Li et al., 2023). This suggests that the cluster recalibration, while successful on stellar benchmarks, does not simultaneously recover all ICM core observables.

The strong-lensing subhalo study reaches a related but distinct conclusion. Using 82 GIZMO-SIMBA host clusters at $50\,h^{-1}{\rm Mpc}$0 and comparable Gadget-X hosts, the authors find that the projected cumulative sub-halo mass function in GIZMO-SIMBA is consistent with the observed sub-halo functions of MACSJ0416 and MACSJ1206, while Gadget-X falls short at low masses (Srivastava et al., 2023). GIZMO-SIMBA also yields higher cumulative $50\,h^{-1}{\rm Mpc}$1 functions and higher $50\,h^{-1}{\rm Mpc}$2 at fixed $50\,h^{-1}{\rm Mpc}$3 than Gadget-X, especially below $50\,h^{-1}{\rm Mpc}$4, which the paper attributes primarily to smaller half-mass radii and denser stellar cores (Srivastava et al., 2023). Yet neither model reproduces the observed $50\,h^{-1}{\rm Mpc}$5-$50\,h^{-1}{\rm Mpc}$6 relation inferred from strong lensing: GIZMO-SIMBA is closer, but still systematically below the observed compactness relation in the low-mass regime most relevant for galaxy-galaxy strong lensing (Srivastava et al., 2023).

6. Later Simba-family extensions and derived applications

Subsequent work has extended the Simba framework in several directions without changing the basic identification of Gizmo-Simba as a GIZMO-based Simba model.

SIMBA-C replaces Simba’s instantaneous-recycling enrichment treatment with the Chem5 chemodynamical model, tracks 34 elements from H $50\,h^{-1}{\rm Mpc}$7 Ge, and makes modest feedback recalibrations so that the new chemistry remains viable (Hough et al., 2023). In the cosmological-box validation, SIMBA-C improves the $50\,h^{-1}{\rm Mpc}$8 galaxy stellar mass function knee, the low-mass main sequence/green-valley structure, black-hole tracking in dwarf galaxies, and abundance-ratio diagnostics such as $50\,h^{-1}{\rm Mpc}$9 and $25\,h^{-1}{\rm Mpc}$0 (Hough et al., 2023). In group-scale IGrM studies, a related Simba-C configuration with updated stellar feedback, the Chem5 enrichment model, and recalibrated AGN feedback improves X-ray scaling relations relative to original Simba and increases Si and Fe relative to O in the hot gas (Hough et al., 2024).

Simba-EoR is a high-redshift ISM-focused extension aimed at $25\,h^{-1}{\rm Mpc}$1 galaxy formation. It removes the KMT molecular-fraction approximation, adds explicit non-equilibrium H$25\,h^{-1}{\rm Mpc}$2 chemistry, introduces a pressure-balanced two-phase ISM model activated at

$25\,h^{-1}{\rm Mpc}$3

and couples dust temperature, ISRF heating, H$25\,h^{-1}{\rm Mpc}$4 formation, cooling, and star formation self-consistently (Jones et al., 2024). The resulting high-resolution runs produce earlier star formation activity than fiducial Simba, larger dust-to-gas ratios, and a bimodal dust-temperature distribution with a cold population near $25\,h^{-1}{\rm Mpc}$5 K and a hot population near $25\,h^{-1}{\rm Mpc}$6 K, while achieving generally tighter concordance with several $25\,h^{-1}{\rm Mpc}$7 observational constraints than baseline Simba (Jones et al., 2024).

SIMBA has also been used as an interpretive framework for radio-galaxy populations. In the MIGHTEE comparison, radio galaxies in SIMBA are classified as HERGs or LERGs according to whether black-hole growth is dominated by torque-limited cold accretion or Bondi hot accretion, respectively (Thomas et al., 2024). The study finds that, in both MIGHTEE and SIMBA, HERGs and LERGs overlap strongly in stellar mass, AGN bolometric luminosity, and Eddington fraction, while SFR remains the clearest separator because SIMBA HERGs retain about three times more cold gas than LERGs (Thomas et al., 2024). A particularly notable result is the predicted and observed existence of low-$25\,h^{-1}{\rm Mpc}$8 HERGs, which challenges the classical bright-radio HERG/LERG dichotomy in the faint-radio regime (Thomas et al., 2024).

7. Scope limits, related GIZMO method papers, and common misconceptions

Several papers often associated with “Gizmo-Simba” are not Simba papers in the strict sense. The original GIZMO methods paper establishes the mesh-free hydrodynamic framework but does not contain Simba’s galaxy-formation model (Hopkins, 2014). The anisotropic-diffusion paper documents solver capabilities available in the broader GIZMO code family—passive scalar diffusion, anisotropic conduction and viscosity, cosmic-ray diffusion, non-ideal MHD, and related operators—but it does not establish that Simba production runs use those modules (Hopkins, 2016). Likewise, the 2025 GRMHD extension adds a general-relativistic magnetohydrodynamics scheme to GIZMO’s mesh-less framework and is directly relevant to horizon-scale accretion or compact-object applications, but not to ordinary non-relativistic Simba cosmological runs (Fedrigo et al., 18 Jun 2025).

A related but more specific misconception concerns self-interacting dark matter. The GIZMO–AREPO SIDM comparison tests the Gizmo SIDM implementation in an isolated dark-matter-only $25\,h^{-1}{\rm Mpc}$9 halo during the core-expansion phase, finding better than 30% agreement in density profiles during that phase and showing that the codes can distinguish $12.5\,h^{-1}{\rm Mpc}$0 in that idealized setting (Meskhidze et al., 2022). The paper is relevant to the GIZMO code family, but it is not a Simba paper: it does not run Simba, does not include hydrodynamics or baryonic physics, and does not validate SIDM in the Simba galaxy-formation stack (Meskhidze et al., 2022).

The resulting scope boundary is precise. Gizmo-Simba is fundamentally an astrophysical GIZMO + Simba construct: a numerical-hydrodynamic platform based on GIZMO’s MFM solver, populated by Simba’s galaxy-formation prescriptions, and instantiated both in cosmological boxes and in cluster zooms. Solver papers about GIZMO, DM-only module comparisons, and the unrelated SiMBA microkinetic algorithm are adjacent to that topic, but they are not themselves the thing denoted by Gizmo-Simba (Servia et al., 2024).