Reading AREPO data (experimental)

AREPO is a moving-mesh code: gas lives in the cells of a Voronoi tessellation that moves with the flow. It writes the GADGET HDF5 snapshot layout (shared with IllustrisTNG, which is AREPO), so Mera reads it through the same frontend as GADGET — but it is auto-detected as AREPO (from the Config group AREPO writes) and reported as Code: AREPO, and its gas-cell physics is read in physical units. Everything downstream (getvar, projection, pdf, profile, …) then runs unchanged.

Usage

getinfo / getparticles auto-detect the code from the HDF5 Header:

using Mera
info = getinfo(59, "/data/TNG/halo_59")     # prints "Code: AREPO" + redshift/H0/Ω for a cosmological run
gas  = getparticles(info; families=[0])      # PartType0 gas → :x,:y,:z,:vx,:vy,:vz,:mass,:id,:family + cell fields

msum(gas, :Msol); getvar(gas, :T, :K)        # the usual particle analysis, unchanged

:family is the particle type (0 gas, 1 DM, 4 stars, 5 BH); families= restricts the load. Use getparticles_gadget directly for the families= option on huge snapshots.

Gas-cell physics

For PartType0, the cell fields present in the file are read as columns and a per-cell volume is derived; getvar adds the thermodynamic quantities. All returned in physical units.

| Machnumber | :mach | cell Mach number (dimensionless) | | (derived) | :T, :p, :cs | T = (γ-1)·u·μ·m_H/k_B; μ from :ne (neutral-primordial fallback if absent) |

getvar(gas, :rho, :g_cm3)        # physical density
getvar(gas, :T, :K)              # temperature [K] — matches the official TNG formula
                                 # (bare `getvar(gas, :T)` is code units, as for :p/:cs)
getvar(gas, :metallicity)        # metal mass fraction
getvar(gas, :bmag, :muG)         # |B| [μG]; also :pmag (:Ba), :beta, :v_alfven (:km_s), :mach_alfven/fast/slow
pdf(gas, :rho); profile(gas, :r_sphere, :T)   # PDFs / radial profiles on the gas

The temperature reproduces IllustrisTNG's documented conversion (density to machine precision, temperature to sub-percent), so the values match what TNG itself reports.

Cosmological runs

A run is treated as cosmological when Ω_Λ > 0; then Time is the scale factor a and the comoving→physical factors are applied automatically (positions ∝ a/h, density ∝ h²/a³, mass ∝ 1/h, velocities × √a). H₀, Ω_m, Ω_Λ come from the Header, so iscosmological, redshift and the cosmology utilities work. A non-cosmological AREPO run (Ω_Λ = 0, e.g. an idealised cluster) is left in code units with a = 1.

Maps — projecting a moving mesh

A Voronoi cell is neither a grid cell nor an SPH particle, so projection offers three depositions (weighting=):

• :mass (default) — deposit each cell at its point; fast, mass-conserving, but speckly.
• :sph — smear each cell over an M4 kernel sized from its volume (h = (3V/4π)^⅓); smooth and mass-conserving — the standard moving-mesh rendering, as in yt.
• :voronoi — sample each line of sight through the nearest cell (KD-tree, capped at the cell radius): sharp, genuinely cell-respecting. Intensive maps (:T, metallicity) are exact; surface density is approximate (use :sph/:mass for conserving column mass).

projection(gas, :sd, :Msol_pc2)                 # surface density (mass-conserving)
projection(gas, :T, weighting=:sph)             # smooth temperature
projection(gas, :T, weighting=:voronoi)         # sharp, cell-respecting temperature

A fully Voronoi-exact renderer (re-tessellate + analytic polyhedron–pixel integration, as in AREPO's ArepoVTK) would be more faithful still, but is rarely needed.

How it maps onto Mera's structs

Gas is a PartDataType carrying an explicit :volume column — not a HydroDataType. A Voronoi cell has an independent per-cell volume and arbitrary shape, so it cannot be placed on Mera's power-of-two AMR octree (where the volume is a function of level alone and the cells tile to boxlen³). Keeping gas as points-with-volume is honest: stored columns pass through getvar directly, mass is read (not derived), and the particle analysis works unchanged. The reader is the GADGET frontend (getinfo_gadget/getparticles_gadget) with the gas-field and a/h handling described above.

Reference readers

This frontend agrees with the origin tools for AREPO/TNG data:

• illustris_python — the official TNG loader; the field layout, unit conventions and temperature formula are matched against it.
• yt — reads AREPO as SPH-like (smoothing length from cell volume), the same approach as Mera's weighting=:sph.
• ArepoVTK — AREPO's own Voronoi ray-tracer (the exact, re-tessellating renderer).

See also

• Multi-code support — the code-blind architecture and the sibling readers.
• Reading GADGET data — the shared HDF5 frontend (GADGET/GIZMO/SWIFT).
• Other Simulation Codes — Worked Examples — the executed notebook.