Multi-Scale Simulations of the Universe: Comparisons of

IllustrisTNG, Thesan, and EPOCH

Qian Wu

Birkbeck, University of London, London, U.K.

Keywords: Cosmological Simulations, Galaxy Formation and Evolution, Reionisation, Particle-in-cell (PIC) Plasma

Physics, Radiation Hydrodynamics.

Abstract: Contemporarily, various multi-scale simulations of the universe are proposed. This study compares three

advanced astrophysical simulation frameworks, i.e., IllustrisTNG, Thesan, and EPOCH, each designed to

model different cosmic phenomena across varying physical scales. IllustrisTNG simulates galaxy formation

and large-scale structure evolution using moving-mesh hydrodynamics and gravitational equations from early

cosmic times to the present. Thesan focuses on the Epoch of Reionization, incorporating radiation-

hydrodynamics, dust physics, and magnetohydrodynamics to model the ionization of the intergalactic medium

by early galaxies. EPOCH applies the particle-in-cell (PIC) method to simulate kinetic plasma processes under

strong electromagnetic fields, solving Maxwell’s equations and the Lorentz force law to capture phenomena

such as shock acceleration and magnetic reconnection. These simulations offer complementary strengths:

from cosmic web formation to feedback-driven ionization and relativistic plasma dynamics. By comparing

their physical models, resolution strategies, and limitations, this paper reveals the need for multi-scale, multi-

physics integration to bridge gaps between microphysical plasma processes and the evolution of large-scale

cosmic structure.

1 INTRODUCTION

Astrophysical simulations can studies galaxy

formation to plasma physics and understand the

cosmic phenomena. There are several fundamental

physics equations to explain are based on these

simulations, such as Newton gravity to model the

large-scale cosmic structures, relativistic fluid

dynamics solve the high-energy processes and

magnetohydrodynamics to simulate the stellar

evolutions.

The history of theory and observations is linked to

astrophysical simulations of stars in the mid-20th

century; there are two methods: Monte Carlo methods

and N-body and N-body simulations (Aarseth, 2003).

In 1943, Chandrasekhar used random sampling to

avoid O(N²) complexity by using random sampling

instead of computing all pairs of gravitational forces,

which makes it faster for computing large-scale

simulations, and it can be used to estimate stellar

interactions statistically (Garaldi, et al., 2022).

Subsequently, in the 1963s, N-body simulations

https://orcid.org/0009-0002-8801-6068

calculated the gravitational forces between stars using

Newton's law of gravity, which is exerted on each

particle (star) by every other particle in a system (Yeh

et al., 2023). This method developed higher-order

integration schemes to improve simulation accuracy,

but it was limited by computational limits; increasing

the number of stars would significantly slow down

the calculations. Building on these two methods, the

Smoothed Particle Hydrodynamics techniques

(1970s) and the Barnes-Hut method (1986) improved

the modelling of gas dynamics in star formation

(Monaghan, 2005), reducing the computational

complexity from O(N²) to O(N log N) respectively

(Session 2-2). Since 2000, advanced particle-in-cell

(PIC) simulations have allowed detailed studies of

kinetic plasma processes in magnetised

environments, such as magnetic reconnection and

cosmic ray acceleration. Starting with early

astrophysical environments such as stellar winds,

supernova remnants and accretion disks, PIC methods

became essential for exploring how charged particles

272

Wu, Q.

Multi-Scale Simulations of the Universe: Comparisons of IllustrisTNG, Thesan, and EPOCH.

DOI: 10.5220/0013824100004708

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Innovations in Applied Mathematics, Physics, and Astronomy (IAMPA 2025), pages 272-280

ISBN: 978-989-758-774-0

behave in strong electric and magnetic fields

(Hansen, et al., 2024).

The resulting codes, such as EPOCH, use kinetic

plasma models and solve Maxwell's equations

alongside the Lorentz force law to simulate particle

dynamics in highly magnetised regions. This allows

realistic simulations of magnetic reconnection,

plasma heating and shock-driven acceleration in

environments such as black hole jets, pulsar

magnetospheres and relativistic shocks (Hansen, et

al., 2024). Unlike traditional fluid codes, EPOCH

tracks individual particle motions rather than

averaging over large volumes, making it suitable for

small-scale, high-energy astrophysical scenarios.

Otherwise, there is a simulation called

Millennium, which focused only on dark matter

structure formation, but did not include baryonic

physics such as gas dynamics or radiation transfer,

making it less relevant for today's multiphysics

simulations (Vogelsberger, et al., 2020). Therefore,

after all these previous simulations, modern

astrophysical simulations are developed; for example,

the IllustrisTNG and Thesan simulations are

developed. These two simulations can modify galaxy

formation and cosmic reionisation more accurately.

This article focuses on three major modern

simulations: Illus trisTNG, THesan and EPOCH.

These simulations have their unique discoveries. First,

IllustrisTNG has significant dark matter results and

simulates galaxies' evolution from the Big Bang to the

present (Nelson, et al., 2019). It shows that dark

matter halos control galaxy formation, and the cosmic

web structure in this simulation is consistent with real

observations (Nelson, et al., 2019). In addition,

IllustrisTNG explains star formation in galaxies by

showing that the rate of star birth peaked 10 billion

years ago and that supernova explosions and black

hole feedback regulate star growth (Pillepich, et al.,

2018). It also proves that supermassive black holes at

galactic centres release AGN feedback, which heats

the surrounding gas and prevents star formation,

showing how large galaxies stop forming stars

(Pillepich, et al., 2018). The IllustrisTNG simulation

also shows the strengthening of the magnetic field in

galaxies over time (Marinacci, et al., 2018)and the

baryon cycle (where gas falls into galaxies, forms

stars and is then ejected by galactic winds) (Pillepich,

et al., 2018). These processes model the movement of

gas and the enrichment of elements in the Universe.

Second, Thesan's simulation shows the process of

early stars and galaxies emitting ultraviolet radiation,

which ionizes the neutral hydrogen in the

intergalactic medium (IGM) (Kannan, et al., 2022).

Currently, it simulates how ionization bubbles form

around the first galaxies and gradually expands,

creating a dense region that ionizes earlier and

showing an inhomogeneous reionization pattern. This

process can show that low-mass galaxies are more

efficient at letting UV photons escape, while high-

mass galaxies trap more radiation and slow down the

reionization process (Kannan, et al., 2022). Thesan

also shows that reionization has a significant effect on

galaxy clustering. Galaxies that form earlier tend to

cluster more densely, forming dense clusters with

surrounding neutral voids (Garaldi, et al., 2024).

Finally, PIC and EPOCH simulations are very

important for studying plasma physics and high-

energy astrophysics. EPOCH uses the particle-in-cell

(PIC) approach to model how electrically charged

particles and fields interact with each other (Arber, et

al., 2015). This is done at very high resolutions that

show how things move and change over time and

space. In this simulation, magnetic fields are

recreated and break down, and then reform, which

creates plasmoids and speeds up particles in fast jets

(Smith, et al., 2021). EPOCH also reproduces

physical conditions found in supernova remnants,

where shock waves interact with magnetized plasma

(Arber, et al., 2015). Through solving Maxwell’s

equations and the Lorentz force law, EPOCH tracks

the behavior of electrons and ions during these violent

events, showing how energy is converted from

magnetic fields into particle motion and radiation

(Arber, et al., 2015). These kinetic-scale processes

cannot be resolved by fluid-based simulations,

making EPOCH essential for studying relativistic jets,

pulsar magnetospheres, and cosmic ray acceleration

in extreme astrophysical settings (Smith, et al., 2021).

This paper compares three astrophysical

simulations: IllustrisTNG, Thesan and EPOCH. Each

section will investigate the equations, capabilities,

and recent results of these simulations. Finally, a

comparative analysis will highlight their strengths,

limitations, and future directions for improving

astrophysical modelling.

2 DESCRIPTIONS OF

ASTROPHYSICS

SIMULATIONS

The IllustrisTNG, Thesan and EPOCH methods are

major astrophysical simulations used to model

various cosmic processes. These simulations attempt

to reproduce the complex physical interactions in

space. They use sophisticated numerical approaches

to describe how galaxies, dark matter, stars, and

Multi-Scale Simulations of the Universe: Comparisons of IllustrisTNG, Thesan, and EPOCH

273

plasma have evolved. Mass distributions, velocity

field distributions and charge distributions are the

main outputs of these simulations (Nelson, et al.,

2019). First, they produce maps showing how mass is

distributed - including where galaxies, dark matter

and stars are located in space (Springel, et al., 2005).

They trace the gravitational clustering of matter

through time. Then, they build structure models,

including galaxy clusters and dark matter halos

(Pillepich, et al., 2018). These maps simulate the

large-scale formation of cosmic structure through the

visual representation of the cosmic web of filaments

and clusters (Springel, et al., 2005). Dark matter halos

provide the gravitational scaffolding from which

galaxies form, and galaxies cluster along filaments,

creating an area of high-density nodes. Based on the

IllustrisTNG simulations, such models provide a

good representation of how small density fluctuations

of the early Universe grew through gravity, leading to

the formation of the cosmic web as one knows it today.

Second, the velocity field distribution shows how

cosmic objects like gas, stars, and dark matter move

through time. It uses fundamental equations such as

Newton's law of gravity (gas flows in galaxies) to

show cosmic structures such as galaxy clusters

(Pillepich, et al., 2018), Navier-Stokes equations

(stars move under gravitational binding) to solve the

formation of dark matter halos (Vogelsberger, et al.,

2014), and magnetohydrodynamics equations

(magnetic forces affect plasma motion) that evolve

over billions of years (Kannan, et al., 2022). All three

main simulations solve these equations to produce

dynamic velocity maps that reveal cosmic objects'

motion patterns and interactions. Last, charge

distribution is very useful in plasma physics and tells

us about the current distribution in the space where

charged particles like electrons and ions are present

(Fonseca, et al., 2022). This directly modifies the

evolution and interaction of the electric and magnetic

fields with charged particles, which is critical for the

investigation of magnetic field interactions and

plasma instabilities (Springel, et al., 2005). Maxwell's

equations describe charge distribution, e.g. Gauss's

law relates electric field divergence to charge density,

and Ampere's law relates magnetic field curvature to

current density (Arber, et al., 2015). These are the

equations for how moving charges create magnetic

fields. This distribution model is used by the Particle-

in-Cell simulation to calculate the movement of

charged particles by electromagnetic forces, but also

keeps track of how the electric and magnetic fields

change (Fonseca, et al., 2022).

3 LARGE-SCALE

COSMOLOGICAL

MODELLING WITH

ILLUSTRISTNG

Large cosmological simulations such as IllustrisTNG,

which includes the evolution of dark matter, baryonic

matter and complex baryon-dominated processes

such as supernova feedback, galaxy mergers, galaxy

quenching and outflows (Pillepich, et al., 2018). The

original Illustris simulation has been significantly

improved, using better computational techniques and

physical laws in the new simulation (Pillepich, et al.,

2018). Firstly, IllustrisTNG uses the AREPO code,

which uses a moving mesh approach to solve the

hydrodynamical equations and follow the evolution

of gas, stars, dark matter and black holes within

cosmic volumes ranging from 50 Mpc to 300 Mpc on

a side (Marinacci, et al., 2018). This technique

involves billions of particles representing units of

mass, velocity and energy. Thus, the basic equations

implemented in IllustrisTNG consist of the equations

of gravity and hydrodynamics (and thus an accurate

calculation of the formation of cosmic structures)

(Vogelsberger, et al., 2014). First, the gravity

equations are based on Newtonian gravity and general

relativity, using Newton's Law of Gravitation

equation (Smith, et al., 2021). F in equation means the

gravitational force, G is the gravitational constant,

m1and m2 are the masses of interacting particles, r is

he distance between them. While, to calculate the

gravitational potential, the simulation uses Poisson's

equation: ∇2Φ=4πGρ. The ∇ means divergence, Φ is

he gravitational potential; ρ is the mass density. The

results of these modelling can make more accurate

gravitational clustering and the formation of dark

matter halos. Second, hydrodynamic equations are

modelled using Smoothed Particle Hydrodynamics

(SPH), where gas dynamics are treated as fluid

particles with basic physic properties like mass,

velocity, and pressure. Explaining mass conservation

can use the continuity equation:

(





) + ∇j = 0 (1)

Here, j is the flux of q (It is flowing, which

described by its flux), also means a vector field. j =

ρv, which v is the velocity field that describes the

motion of the quantity q. While, explaining pressure

gradients and gravitational forces acting on the gas

use the Navier–Stokes momentum conservation

equation:

(





) = −(





) ∇P − ∇Φ (2)

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

274

which P is the pressure. Last, among the most

important parameters are up to three types of

feedback involved in star formation: supernova

feedback (the phenomenon in which massive stars

explode and eject energy and gas into space, creating

shock waves that heat the surrounding gas) (Springel,

et al., 2005). This heating prevents the gas from

cooling down and collapsing, which slows down the

star formation rate of the galaxy. Also, there are

stellar winds - streams of charged particles blown

away from stars, driven by radiation pressure or

magnetic activity. These winds sweep away

surrounding gas and heat the interstellar medium

(ISM), reducing the gas density and slowing star

formation. Finally, there is the feedback from active

galactic nuclei, where a supermassive black hole at

the centre of a galaxy emits jets and winds that heat

gas and drive it outwards. This mechanism is

responsible for cooling the gas and thus for star

formation. In addition, IllustrisTNG has five main

functions and outputs. The first is the mapping of

these dark matter halos, which tracks how halos form

and evolve, and illustrates how they cluster along

these cosmic filaments.

Figure 1: Impact of mergers on stellar-dark matter

alignment in IllustrisTNG galaxies (Garaldi, et al., 2021).

Figure 1 compares the effect of major and minor

mergers on the ratio of stars to dark matter (DM) in

galaxies, using data from the IllustrisTNG simulation

(TNG100-1)

(Garaldi, et al., 2021)

. The figure contains

8 scatter plots, divided into Panel A (major mergers,

top two rows) and Panel B (minor mergers, bottom

two rows). Each panel shows how galaxy

morphology (q = c/a, shape-axis ratio) and kinematics

(V/σ, rotation-to-dispersion ratio) align between stars

(x-axis) and DM (y-axis). Points above the diagonal

indicate galaxies where the DM is rounder or less

rotationally supported than the stars. Colour bars

indicate different physical quantities: Panels A1 and

B1 show stellar masses from mergers (ΔM,major or

ΔM,minor / M), A2 and B2 show recent in-situ star

formation since redshift z = 1 (Δinsitu,z≤1), and

A3-B4 show how these quantities influence galaxy

motion. The results show that large mergers cause a

stronger alignment between stars and DM in both

shape and motion, while smaller mergers show

weaker trends. Galaxies with little recent star

formation (blue dots) retain tighter coupling,

suggesting that dry mergers preserve structure better

than gas-rich mergers. This data is taken from the

IllustrisTNG website.

4 RADIATION HYDRODYNAMIC

SIMULATIONS OF

REIONIZATION

The Thesan project represents an innovative,

extensive cosmological simulation focused on the

Epoch of Reionization (EQR) (Kannan, et al., 2022).

The final phase involves the formation of the first

stars and galaxies, and the ionizing influence on

neutral hydrogen in the intergalactic medium (IGM).

The purpose of this simulation is to determine that the

timing of reionization, how quickly it occurred, and

identify the galaxies that played a crucial role in this

process. There are physical processes of Thesan

project, the most important is feedback from star

formation, positioning it as one of the most thorough

and realistic simulations of the early universe

available to date (Garaldi, et al., 2024). Other

processes are such as radiation-hydrodynamics, dust

absorption, and magnetic fields (Garaldi, et al., 2022).

First of all, Thesan's radiation hydrodynamics

couples the motion of ionising photons to the heating,

ionisation and dynamics of the gas. This is important

because starlight affects the temperature and

ionisation levels of the surrounding hydrogen gas,

and thus the growth of galaxies. In contrast, Thesan

uses the M1 moment method to handle the radiative

transfer equations, allowing the calculation of both

the energy and direction of the radiation, similar to

IllustrisTNG. The radiation equations in Thesan are

fully integrated with the gas evolution, allowing the

ionised shells surrounding galaxies to be tracked in

real time without relying on the assumption of a

uniform UV background (Kannan, et al., 2022).

Multi-Scale Simulations of the Universe: Comparisons of IllustrisTNG, Thesan, and EPOCH

275

Secondly, Thesan also took into account the

physics of cosmic dust, which affects the absorption

and scattering of radiation. Dust particles can either

scatter or absorb ionising photons, changing the

amount of light emitted by galaxies. This

phenomenon affects the observed luminosity of

galaxies and the timing of reionisation. By taking dust

into account, Thesan can accurately represent the

physically realistic escape fractions of photons, which

depend on factors such as galaxy mass, gas density

and metallicity (Yeh et al., 2023).

Thirdly, Magnetic fields are modelled using the

magneto-hydrodynamic (MHD) equations, which

influence gas pressure, star formation and feedback

mechanisms (Kannan, et al., 2022). Thesan develops

the magnetic field B by using a revised set of MHD

equations, which includes magnetic tension and

pressure terms in the fluid momentum equation. This

inclusion helps to explain the structure of cosmic

filaments and the turbulence within galaxies (Garaldi,

et al., 2022).

Finally, Thesan takes a complete feedback

geometry from star formation feedback, with a

particular focus on supernovae and stellar winds.

When massive stars die, they explode, releasing

energy into the surrounding gas. This energy input

prevents the gas from collapsing and thus plays a role

in regulating star formation. Thesan uses models that

simulate energy injection and thermal feedback to

represent this phenomenon. It also investigates the

effect of this feedback on the escape fraction of

ultraviolet (UV) light Thesan suggests that low-mass

galaxies are likely to have higher escape fractions as

they struggle to retain gas, while high-mass galaxies

are more effective at trapping photons due to their

denser gas environments (Yeh et al., 2023).

Moreover, Thesan is built on the AREPO-RT

code, an extension of the AREPO code that adds

radiative transfer to a moving mesh system. This code

solves the equations for mass, momentum, energy,

and ionization state of gas on a mesh that moves with

the gas flow. The moving mesh method allows for

higher resolution where the gas is denser, such as in

galaxies. Thesan also solves Poisson’s equation:

∇²Φ = 4πGρ (3)

where Φ is gravitational potential and ρ is mass

density. This equation relates mass to gravity, helping

simulate gravitational clustering. Thesan covers a

volume of 200 comoving megaparsecs (Mpc), with a

resolution of up to 2×1536³ elements, which is better

than IllustrisTNG. It also improves spatial and time

resolution to better simulate reionization (Garaldi, et

al., 2024).

Figure 2: Visualization of the ionization structure and ionization history in Thesan-1 (Neyer, et al., 2024).

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

276

Compared to IllustrisTNG, Thesan offers several

improvements. First, Thesan includes full radiative

transfer instead of assuming a fixed UV background.

Second, it calculates the escape fraction of radiation

based on actual local gas conditions rather than fixed

values (Yeh et al., 2023). Third, Thesan includes dust

physics and magnetic fields, which allows it to

simulate reionization as a patchy and dynamic

process instead of a uniform one. These features

allow Thesan to show how galaxy clustering is related

to reionised regions, something that fixed background

models cannot do (Garaldi, et al., 2022). However,

Thesan has its limitations. It focuses on the early,

high-redshift universe and does not evolve galaxies to

the present day, unlike IllustrisTNG. Also, because it

fully couples radiation with gas physics, it requires

much more computing time and memory.

Nevertheless, Thesan provides new insights into how

galaxies, gas and light interacted during the epoch of

reionisation, and will be a valuable tool for

comparison with data from modern telescopes such as

JWST and ALMA (Kannan, et al., 2022).

Figure 2 shows the Thesan-1 simulation box and

zoom-in regions during the reionisation epoch. Each

row shows a different spatial scale: the middle row is

the full 200 cMpc box, and the top and bottom rows

are zoom-in regions (marked by six white boxes).

Each column shows a different property: the left

column shows the ionised hydrogen fraction 𝑥 HII,

the middle column shows the reionisation redshift 𝑧

reion, and the right column shows the effective

ionised bubble size. The colours represent physical

values - blue to yellow in 𝑥 HII mark neutral to

ionized gas, with brightness based on overdensity 𝑓

𝛿; 𝑧 region shows when each region was ionized

(higher redshift means earlier), and 𝑅 eff shows how

large each ionized region is (Neyer, et al., 2024). All

5 cMpc bars indicate the physical scale. The figure

shows that reionisation is patchy, shaped by galaxies

and gas. It supports Thesan's model of radiation

hydrodynamics and feedback. It shows how Thesan

tracks local radiation escape, bubble growth and

galaxy clustering during reionisation. This

demonstrates Thesan's strength in resolving the

structure and timing of early cosmic light.

5 ASTROPHYSICAL KINETIC

PLASMA DYNAMICS

EPOCH (Extendable PIC Open Collaboration) is a

simulation framework designed to study high-energy

plasmas in astrophysical and laboratory environments.

It differs from the other two cosmological

computations mentioned in the last two software

sections, which mainly study large-scale structures

and the process of galaxy formation. EPOCH

operates on kinetic scales, where the collective

interactions of a few particles and electromagnetic

fields govern the behaviour of matter (Arber, et al.,

2015).

The PIC approach tracks particles with respect to

phase space and solves Maxwell’s equations on a grid.

Every particle has position, velocity, charge, and

mass. These particles drift in space and treat such

translocations as inventive electric (E) and magnetic

(B) fields. The fields then act on the particles via the

Lorentz force law. This looping continues for time

steps. The main Maxwell’s equations are:

∇·E =







(4)

𝛻·𝐵 = 0 (5)

∇∙E = −





(6)

∇ ∙ B = μ₀J + μ₀ε₀





(7)

where E is the electric field, B is the magnetic field,

ρ is charge density, J is current density, ε₀ is vacuum

permittivity, and μ₀ is vacuum permeability. These

fields are propagated over a grid using a finite-

difference time-domain (FDTD) method (Smith, et al.,

2021). Moreover, the particles are accelerated via

the Lorentz force law (for particle motion):

𝐹 = 𝑞(𝐸 + 𝑣 × 𝐵) (8)

where F = force, q = particle charge, v = particle

velocity, E and B = local electric and magnetic fields.

These calculations allow EPOCH to simulate non-

linear plasma physics with the accuracy afforded by

relativity. One example is the use of EPOCH for

laser-plasma interactions, where an extremely high

intensity laser beam is fired at plasma targets. This

allows studies of the acceleration of ions and the

generation of wake fields, which are important for

energy transfer in astrophysical explosions (Arber, et

al., 2015). A further highlight of EPOCH is its

handling of magnetic reconnection, a physically

relevant process in plasma environments where

magnetic field lines break and reconnect (Smith, et al.,

2021). This converts stored magnetic energy into

kinetic and thermal energy, resulting in flares, jets

and bursts. It is predicted that reconnection events

lead to the release of gamma-ray bursts in the

magnetospheres of pulsars, where neutron stars spin

up and generate intense magnetic fields. EPOCH

models the fine structure of the fields and the

dynamics of the particle pairs produced (Smith, et al.,

2021).

Compared to traditional fluid-based simulations

such as IllustrisTNG or Thesan, which use

hydrodynamic or magnetohydrodynamic (MHD)

Multi-Scale Simulations of the Universe: Comparisons of IllustrisTNG, Thesan, and EPOCH

277

equations, EPOCH solves kinetic plasma equations at

the particle level. It does not model gravity or large-

scale cosmic structure, but addresses microphysics

that fluid codes cannot resolve. This gives it high

fidelity for studying plasma behaviour in atoropes or

extreme conditions. Currently, EPOCH has its own

limitations. It cannot model large galaxies or

cosmological evolution in time. In addition, due to the

small time steps and the need to track millions of

particles in PIC simulations, EPOCH is extremely

computationally expensive, especially in 3D runs

(Arber, et al., 2015). In conclusion, EPOCH is not

intended to replace large-scale simulations, but to

complement them. It provides a more detailed

understanding of the plasma microphysics

responsible for radiation, energy transport and

particle acceleration in many astrophysical settings,

ranging from supernova remnants and neutron stars to

active galactic nuclei (Arber, et al., 2015).

Figure 3 shows two 2D particle-in-cell (PIC)

simulations using EPOCH. Both simulations use a 5

ps laser pulse, but the top panel has no pre-plasma,

while the bottom panel has pre-plasma. The x-axis

(μm) shows the distance along the laser direction and

the y-axis shows the transverse spatial dimension.

The colour bar on the right shows the plasma density

from 3.0 × 10 cm-³ (blue, low density) to 3.0 × 10 22

cm-³ (yellow, high density) (Peebles, et al., 2017). In

the top image, the laser enters clean plasma and

focuses only near the dense target. The electric field

is narrow and the electrons are accelerated later. In

contrast, the lower image with pre-plasma shows

relativistic self-focusing in front of the target (~100

μm), with strong filamentation and earlier electron

generation. This shows how the pre-plasma modifies

the laser-plasma interaction. The electric and

magnetic field structures become more complex and

energy is transferred more efficiently. This figure

demonstrates the EPOCH software's ability to model

microscopic plasma physics, including laser-plasma

coupling, shock formation and electron acceleration,

that cannot be resolved by fluid simulations such as

Thesan or IllustrisTNG. EPOCH captures fine-scale

electromagnetic effects critical to astrophysical

plasmas and high-energy environments (Peebles, et

al., 2017).

Figure 3: Density and electric field from EPOCH simulations with and without the pre-plasma (Peebles, et al., 2017).

6 CONCLUSIONS

This paper has presented a comparative analysis of

three representative astrophysical simulation

frameworks: IllustrisTNG, Thesan, and EPOCH.

Each simulator targets a different scale and aspect of

the universe, using distinct physical models to study

complex cosmic phenomena. IllustrisTNG models

the co-evolution of dark matter and baryonic matter

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

278

from early times to the present, solving

hydrodynamic and gravitational equations with

moving-mesh technique, and incorporating feedback

from star formation and active galactic nuclei. It

successfully recreates the cosmic web and galaxy

evolution but lacks radiative transfer and early-

universe detail. Thesan, built upon AREPO-RT,

focuses on the Epoch of Reionization and includes

full radiation-hydrodynamics, cosmic dust physics,

and magnetic fields. It models the escape fraction of

ionizing photons and patchy reionization more

realistically, revealing the clustered growth of ionized

bubbles and the impact of feedback on galaxy

environments. However, it does not evolve galaxies

past high redshift and is computationally demanding.

In contrast, EPOCH employs particle-in-cell (PIC)

methods to simulate plasma dynamics in extreme

electromagnetic environments. It solves Maxwell’s

equations and the Lorentz force law on kinetic scales,

capturing processes such as magnetic reconnection,

shock acceleration, and laser–plasma interactions.

EPOCH provides unmatched detail in local field

structures and energy transport but cannot model

gravitational clustering or large-scale cosmic

evolution due to scale and resource limits.

Current limitations include the lack of integration

between kinetic microphysics and large-scale

cosmology. IllustrisTNG and Thesan rely on fluid

approximations and cannot resolve particle-scale

interactions, while EPOCH resolves these processes

but sacrifices volume and gravitational modelling].

Additionally, all three simulations face increasing

computational challenges, particularly Thesan and

EPOCH, which require high resolution in time and

space to maintain accuracy. Memory and runtime

constraints limit the ability to simulate longer

timescales or larger cosmic volumes. Future research

could benefit from hybrid approaches combining

radiation-hydrodynamics and kinetic plasma models,

allowing simulations to bridge scales from

reionization bubbles to relativistic jets. Better GPU

parallelisation, adaptive mesh refinement and AI-

assisted parameter tuning could reduce the number of

resources needed. As telescopic observations from

JWST, SKA, LISA, etc become more accurate,

improved simulations capable of reproducing multi-

physics signatures in space and time will be required.

In summary, IllustrisTNG, Thesan and EPOCH each

contribute their own view of the Universe,

underscoring the need for multi-scale, multi-physics

approaches in the next generations of astrophysical

simulations.

REFERENCES

Aarseth, S. J., 2003. Gravitational N-body simulations:

tools and algorithms. Cambridge University Press.

Arber, T. D., Bennett, K., Brady, C. S., et al., 2015.

Contemporary particle-in-cell approach to laser-plasma

modelling. Plasma Physics and Controlled Fusion,

57(11), 113001.

Fonseca, R. A., Silva, L. O., Tsung, F. S., et al., 2002.

OSIRIS: A three-dimensional, fully relativistic particle

in cell code for modeling plasma based accelerators.

Computational Science—ICCS 2002: International

Conference Amsterdam, The Netherlands, April 21–24,

2002 Proceedings, Part III 2, 342-351.

Garaldi, E., Kannan, R., Smith, A., Springel, V., Pakmor,

R., Vogelsberger, M., Hernquist, L., 2022. The

THESAN project: properties of the intergalactic

medium and its connection to reionization-era galaxies.

Monthly Notices of the Royal Astronomical Society,

512(4), 4909-4933.

Garaldi, E., Kannan, R., Smith, A., et al., 2024. The thesan

project: public data release of radiation-hydrodynamic

simulations matching reionization-era JWST

observations. Monthly Notices of the Royal

Astronomical Society, 530(4), 3765-3786.

Hansen, E. C., Garcia-Rubio, F., Adams, M. B., et al., 2024.

Feasibility and performance of the staged Z-pinch: A

one-dimensional study with FLASH and MACH2.

Physics of Plasmas, 31(4).

Kannan, R., Garaldi, E., Smith, A. E., Pakmor, R., Springel,

V., Vogelsberger, M., Hernquist, L., 2022. Introducing

the thesan project: radiation-magnetohydrodynamic

simulations of the epoch of reionization. Monthly

Notices of the Royal Astronomical Society, 511(3),

4005-4030.

Marinacci, F., Vogelsberger, M., Pakmor, R., et al., 2018.

First results from the IllustrisTNG simulations: radio

haloes and magnetic fields. Monthly Notices of the

Royal Astronomical Society, 480(4), 5113-5139.

Monaghan, J. J., 2005. Smoothed particle hydrodynamics.

Reports on progress in physics, 68(8), 1703.

Nelson, D., Springel, V., Pillepich, A., et al., 2019. The

IllustrisTNG simulations: public data release.

Computational Astrophysics and Cosmology, 6, 1-29.

Neyer, M., Smith, A., Kannan, R., et al., 2024. The

THESAN project: connecting ionized bubble sizes to

their local environments during the Epoch of

Reionization. Monthly Notices of the Royal

Astronomical Society, 531(3), 2943-2957.

Peebles, J., Wei, M. S., Arefiev, A. V., et al., 2017.

Investigation of laser pulse length and pre-plasma scale

length impact on hot electron generation on OMEGA-

EP. New Journal of Physics, 19(2), 023008.

Pillepich, A., Springel, V., Nelson, D., 2018. Simulating

galaxy formation with the IllustrisTNG model. Monthly

Notices of the Royal Astronomical Society, 473(3),

4077-4106.

Smith, J. R., Orban, C., Rahman, N., McHugh, B., Oropeza,

R., Chowdhury, E. A., 2021. A particle-in-cell code

Multi-Scale Simulations of the Universe: Comparisons of IllustrisTNG, Thesan, and EPOCH

279

comparison for ion acceleration: EPOCH, LSP, and

WarpX. Physics of Plasmas, 28(7).

Springel, V., White, S. D., Jenkins, A., et al., 2005.

Simulations of the formation, evolution and clustering

of galaxies and quasars. Nature, 435(7042), 629-636.

Vogelsberger, M., Genel, S., Springel, V., et al., 2014.

Introducing the Illustris Project: simulating the

coevolution of dark and visible matter in the Universe.

Monthly Notices of the Royal Astronomical Society,

444(2), 1518-1547.

Vogelsberger, M., Marinacci, F., Torrey, P., Puchwein, E.,

2020. Cosmological simulations of galaxy formation.

Nature Reviews Physics, 2(1), 42-66.

Yeh, J. Y., Smith, A., Kannan, R., et al., 2023. The thesan

project: ionizing escape fractions of reionization-era

galaxies. Monthly Notices of the Royal Astronomical

Society, 520(2), 2757-2780.

IAMPA 2025 - The International Conference on Innovations in Applied Mathematics, Physics, and Astronomy

280