Data access

All raw data and many data products from the THESAN simulation suite are publicly available at this page through the Globus application. You can also follow a step-by-step tutorial.

Predictions for common physical quantities from THESAN-1 are summarised below in machine-readable format. All files have an header describing the content of each (white space-separated) column.

Most THESAN data are similar in format to those from the IllustrisTNG simulations. Some novel data output and products have unique formats. Both are described below.

The raw output from a single THESAN simulation are listed here. You can get more information by clicking on the name of each entry.

• Snapshots : the full simulation output at 80 different cosmic times, between z=50 and z=5.5 (produced approximately every 10 Myr).
• Group catalogs : the properties of haloes and galaxies in the simulation, at 80 different cosmic times.
• Cartesian outputs : gas properties sampled on a Cartesian grid at high time cadence.

In addition to the raw data, we provide a number of data products, produced by post processing the THESAN data. The available data products are listed here. You can get more information by clicking on the name of each entry.

• simulation.hdf5 file : file linking datasets across output chunks, automating the loading process.
• eor.txt file : file with runtime global IGM properties.
• Merger trees : evolutionary tracks of galaxies in the simulation and their properties.
• Offset files : utility files that make easy to connect snapshot, groups and merger trees.
• Hashtables : utility files that make easy to load particles based on their spatial position.
• Cross-match files : utility files that make easy to find analog haloes across the different THESAN simulations.
• Lyman-alpha catalogs : intrinsic Lyman-α properties of galaxies in the simulation.
• Lines of sight : collection of gas properties extracted along random lines of sight through the simulation.
• Filament catalogs : the filaments identified in the simulation and their properties.
• Escape fraction catalogs: the escape fraction properties of galaxies in the simulation.
• Galaxy SED catalogs : the spectral energy distributions of simulated galaxies, obtained in post-processing from the star and dust distributions.

Not all data are available for all THESAN simulations because of physical or numerical limitations. The table below shows which data are available for each run.

• 18/06/2024 - lines of sight: added lines of sight aligned with z direction (300 per snapshot) and lightcone-like lines of sight
• 21/03/2024 - SED: specified that SFR and StellarMass refer to the inner 30kpc of subhaloes
• 02/08/2023 - Renamed ThesanHR directories and placed in the main Thesan directory
• 30/07/2023 - New filaments catalogs! Files organised into subdirectories in the filaments directory.

The reduced speed of light used in Thesan (with a value of c̃ = 0.2c) inadvertently overwrote the physical one in the black hole routines. This nominally reduces the energy injection by the black hole feedback (in both quasar and radio modes) by a factor of (c/c̃)² = 25, and allows accretion up to 5x the Eddington rate. Additionally, this issue was exacerbated by an unrelated modelling choice when coupling non-equilibrium thermochemistry to gas on the equation of state (EoS). Specifically, while developing the Thesan model we encountered numerical instabilities caused by gas that was repeatedly prevented from being classified as star-forming due to the strict temperature threshold for gas to be on the EoS. At the time, the simplest solution was to allow gas to cool below the EoS effective temperature associated with hot-phase pressure support in the ISM. The temperature is eventually set back to the EoS one, but the typical temperature of gas accreting onto black holes was colder than the IllustrisTNG model expects, further boosting the Bondi rate.

Given these two issues, accretion rates and final black hole mass achieved through self-regulation should not be trusted for scientific analyses. We also caution that the above issues might give rise to other side effects within the influence of black holes. This includes the quantitative conclusions reached in Section 5.2.1 of Garaldi et al. (2024) -- Data release and JWST comparison; (`Quenched galaxies in Thesan'), as the reduced feedback energy allows black holes to grow more massive. By extrapolation, we expect that any galaxy hosting a large black hole would look different in Thesan and IllustrisTNG if run to a low enough redshift. Fortunately, although the number of ionising photons injected in the simulation is strongly suppressed, this does not affect reionization predictions as even post-processing calculations estimate per cent level contributions from high-z AGN within the simulated volume Yeh et al. (2023). Finally, we have performed new (much smaller) simulations to verify the impact of these problems. We find that they do not significantly affect the majority of simulation results, as black hole feedback and radiation remain largely negligible even once the aforementioned artificial suppression is amended.

We stress that, since the black hole activity is negligible in Thesan (as a consequence of {i} the lack of very massive galaxies and {ii} ending the simulation at z=5.5), galaxies in the Thesan simulations do resemble the ones in IllustrisTNG (see Appendix A of Garaldi et al. (2022). Again, galaxy properties are unchanged except in a small handful of cases with earlier radio mode feedback, so Thesan remains as one of the most realistic high-resolution, large-volume radiation-hydrodynamic reionization simulations to date.

• Kannan et al. (2022) -- Overview and galaxy properties;
• Garaldi et al. (2022) -- IGM and IGM--galaxy connection;
• Smith et al. (2022) -- Lyman-alpha emission and transmission;
• Garaldi et al. (2023) -- Data release and JWST comparison;
• Borrow et al. (2023) -- THESAN-HR;