NEO_JAX¶
NEO_JAX is a standalone neoclassical transport code for stellarators.
It computes effective ripple and related trapped-particle diagnostics from Boozer-coordinate geometry, provides a clean Python API, integrates directly with Geometry Inputs and Pipelines, and uses JAX for just-in-time compilation, vectorized surface evaluation, and automatic differentiation.
What The Documentation Covers¶
Governing equations for \(\epsilon_{\mathrm{eff}}^{3/2}\), trapped-particle classes, field-line integrals, and reference scalings.
Fourier reconstruction, spline interpolation, Newton refinement, RK4 integration, rational-surface handling, and workload controls.
boozmn inputs, in-memory Boozer objects, VMEC→Boozer→NEO pipelines,
and the exact data model consumed by the solver.
What stays on device, which paths are JIT-friendly, and how NEO_JAX is used in optimization and design studies.
NeoConfig, control-file fields, runtime environment variables, CLI
switches, diagnostics, and result containers.
Regression structure, CI coverage, performance guardrails, and comparisons against established reference cases where appropriate.