Skip to content

Astrophysics codes

seed pillar 7 · Domain codes and toolchains

Cheatsheet page

Gotchas, not tutorials. Write the entry the day you fight the tool.

The landscape

Code Method Typical use
FLASH block-structured AMR, finite volume astrophysical hydro/MHD, nuclear burning
Athena++ static/adaptive mesh, high-order Godunov MHD, accretion discs, well-documented
Gadget-4 TreePM + SPH cosmological structure formation
AREPO moving Voronoi mesh galaxy formation, Galilean-invariant
ZEUS-MP finite difference, constrained transport classic MHD; in the course archive
yt analysis and visualisation reads essentially all of the above

Common configuration concerns

Units. Every code has an internal unit system, and the conversion to CGS is set in the configuration rather than being universal. Getting it wrong produces results that are self-consistent and physically meaningless — and they often look plausible, because the dynamics are scale-free in the relevant limit.

Boundary conditions. Periodic, outflow, reflecting, user-defined. Outflow boundaries that permit inflow are a recurring source of unphysical mass injection.

Refinement criteria. AMR codes refine on a user-specified criterion — density gradient, pressure gradient, a Jeans-length condition. The criterion matters more than the maximum level: gradient-based refinement chases noise, and a Jeans-resolution criterion is what prevents artificial fragmentation.

Analysis with yt

import yt
ds = yt.load("plt00100")             # FLASH, Athena, Gadget, AMReX, ...
print(ds.field_list)                 # what is actually in the file
print(ds.derived_field_list)         # what yt can compute

slc = yt.SlicePlot(ds, "z", "density")
slc.set_zlim("density", 1e-26, 1e-22)
slc.annotate_grids()                 # see the AMR structure
slc.save()

ad = ds.all_data()
print(ad.quantities.extrema("temperature"))
print(ad.quantities.total_mass())

ds.derived_field_list is the useful one — yt computes a large number of physical quantities from the primitive fields, correctly, including unit handling.

Failure modes

Unit system misconfigured. Internal code units converted with the wrong length or mass scale. Everything remains self-consistent, so no check inside the run catches it. Verify against an analytic scale — a free-fall time, a sound crossing time — early.

\(\nabla\cdot\mathbf{B}\) not monitored. For MHD runs, track it in dimensionless form throughout. See MHD — a growing divergence produces forces along field lines that look like physics.

Refinement criterion chasing noise. Gradient-based refinement on a noisy field refines everywhere and the run stalls. Use a physically motivated criterion, and cap the level.

Artificial fragmentation from under-resolving the Jeans length. A self-gravitating simulation that does not resolve the Jeans length by several cells fragments numerically. The clumps look like physics and are grid artefacts.

Softening length not reported. N-body results depend on it, and it is frequently omitted from papers. See N-body.

Outflow boundary admitting inflow. Mass and momentum enter from outside with no physical source. Check the global mass budget.

Restart files not bit-identical to the run state. Some codes drop diagnostic or random state on restart, so a restarted run diverges from a continuous one. Test it once, deliberately, before relying on long restart chains.

Connections

Sources

  • FLASH, Athena++, Gadget-4 and yt documentation.
  • Books — Aarseth for N-body; Goedbloed & Poedts for MHD.
  • Archive: Astro/ZEUS_MP in the course archive.