[rogeriojorge]

Problem

Setting up a SPECTRAX case means building Hermite–Fourier coefficient arrays by hand in Python. The Landau-damping example spends 30 lines on Ck_0 and Fk_0 index arithmetic, and the field must be made consistent with Gauss's law by hand. JAX-in-Cell, by contrast, describes each species physically in its TOML file.

Change

load_parameters accepts one [species.<name>] table per species:

[input_parameters]
Lx = 1.2566        # box length in c/omega_pe
t_max = 20.0
nu = 2.0

[solver_parameters]
Nx = 9
Nn = 64

[species.electrons]
charge = -1
vth = 0.1414                  # sqrt(2T/m)/c, also the Hermite width
perturbation_amplitude = 0.01 # density * (1 + A cos(2 pi m x / L))

[species.protons]
charge = 1
mass = 1836
vth = 0.0033

• The new species_initial_state builds the following from the species tables:
  • Ns, qs, ms, Omega_cs = Omega_ce / mass, alpha_s and u_s;
  • Ck_0, from each species' density in its own Maxwellian basis;
  • Fk_0, from Gauss's law i k·E = rho / Omega_ce, plus an optional uniform background field B0.
• Each species takes charge, mass, density, vth, drift and a density perturbation. The perturbation fields use JAX-in-Cell's names: perturbation_amplitude, perturbation_mode and perturbation_axis.
• The first species must have unit mass, because it sets the field units used by Ampère's law. Otherwise a clear ValueError is raised.
• Files without species tables behave exactly as before. [input_parameters] and [solver_parameters] are now optional.

The change is 56 lines in one existing file, with no new modules.

Tests

• Densities and Gauss's law: the species TOML gives the requested densities, and its initial field satisfies Gauss's law to 1e-14.
• Landau damping at kλ_D = 0.5 is built only from the TOML. It damps at −0.1547 ω_pe, against the exact −0.1533 (1%). It oscillates at 1.410 ω_pe, against 1.416.
• Units guard: making the first species non-unit mass raises.
• The full suite passes, 10 tests in 7 s.