Examples

This page lists the main ways to run NTX, from the smallest CLI solve to the publication-figure scripts.

1. Simplest CLI Run

ntx examples/example_surface.toml --plot

This is the smallest end-to-end solve. It requires no external files and is the best first command to confirm that NTX is installed correctly. It writes a NetCDF payload and a PDF summary panel under examples/outputs/.

2. DKES-Style CLI Run

ntx examples/sample_dkes.toml

This writes a NetCDF result under examples/outputs/. Use --output examples/outputs/sample_dkes.npz when a compact NumPy archive is preferred.

3. VMEC CLI Run

ntx examples/sample_vmec.toml

This exercises the VMEC normalization path on the bundled sample wout file.

4. Open And Plot An Output File

python examples/plot_output_file.py examples/outputs/sample_dkes.nc

This reads an NTX .nc, .npz, or .h5 payload and writes:

• docs/_static/output_file_summary.png

• docs/_static/output_file_summary.pdf

The output figure contains:

• magnetic-field strength on the angular grid

• the radial-drift source on the same grid

• the solved transport coefficients

• a run-summary panel with key diagnostics

5. Python Single-Case Solve

from ntx import GridSpec, MonoenergeticCase, load_vmec_surface, solve_monoenergetic

surface = load_vmec_surface("wout.nc", psi_n=0.25)
grid = GridSpec(n_theta=9, n_zeta=11, n_xi=12)
case = MonoenergeticCase(nu_hat=1e-3, er_hat=1e-3)
result = solve_monoenergetic(surface, grid, case)

6. NEOPAX Mapping

python examples/neopax_with_ntx.py

This example:

• loads a VMEC equilibrium

• builds an NTX monoenergetic scan

• maps that scan into NEOPAX-style arrays

Use it as the minimal reference for NTX-to-NEOPAX coupling.

7. Bootstrap Current From VMEC Or Boozmn

python examples/bootstrap_current_from_vmec_or_boozmn.py

This example is the shortest NTX-only workflow:

• start from a VMEC wout file and use vmec_jax

• or, if a Boozer boozmn file already exists, use booz_xform_jax output directly

• solve a fixed-collisionality NTX radial family

• plot magnetic geometry, radial profile inputs, D11, nu_hat * D33, and a compact interior reduced bootstrap-current response

All user inputs live at the top of the file. The script prefers direct Boozer input in auto mode when a boozmn file is available and otherwise falls back to the VMEC-harmonic lane. The density and temperature derivatives are taken analytically in the example so the figure reflects the transport response rather than finite-difference edge noise.

It writes:

• docs/_static/bootstrap_current_from_vmec_or_boozmn.png

• docs/_static/bootstrap_current_from_vmec_or_boozmn.pdf

• docs/_static/bootstrap_current_from_vmec_or_boozmn.json

8. W7-X Bootstrap-Current Convergence Audit

python examples/bootstrap_current_reference_audit_w7x.py

This optional audit script rebuilds a reduced W7-X scan at several NTX resolutions, evaluates the resulting bootstrap-current profile through the imported workflow, and writes a convergence figure:

• docs/_static/bootstrap_current_reference_audit_w7x.png

• docs/_static/bootstrap_current_reference_audit_w7x.pdf

• docs/_static/bootstrap_current_reference_audit_w7x.json

9. VMEC Geometry-Family Transport Convergence

python examples/geometry_family_transport_convergence.py --preset production

This optional artifact discovers local public VMEC examples from vmec_jax, STELLOPT, and SIMSOPT checkouts, then runs a production D11/D31/D33 convergence ladder and stores D13 for the Onsager quality check. It is an NTX stress diagnostic across available geometry families, not an independent-code parity claim.

It writes:

• docs/_static/geometry_family_transport_convergence.png

• docs/_static/geometry_family_transport_convergence.pdf

• docs/_static/geometry_family_transport_convergence.json

Owned JAX-Native NTX+NEOPAX Dataset

python examples/owned_geometry_neopax_dataset.py
python examples/owned_finite_beta_sfincs_jax_inputs.py
python examples/owned_finite_beta_sfincs_jax_resolution_audit.py
python examples/owned_finite_beta_sfincs_jax_production_ladder_audit.py
python examples/owned_finite_beta_sfincs_jax_profile_current_audit.py
python examples/owned_finite_beta_sfincs_jax_profile_current_resolution_audit.py
python examples/owned_finite_beta_bootstrap_comparison.py
python examples/owned_finite_beta_closure_localization.py
python examples/owned_finite_beta_profile_current_observable_audit.py
python examples/owned_finite_beta_current_conditioning_audit.py
python examples/owned_finite_beta_closure_quadrature_audit.py
python examples/owned_finite_beta_source_channel_audit.py
python examples/owned_finite_beta_source_response_profile_audit.py
python examples/owned_finite_beta_radial_interpolation_audit.py --rebuild-matched
python examples/owned_finite_beta_closure_quadrature_audit.py \
  --bootstrap-json docs/_static/owned_finite_beta_field_radius_matched_bootstrap_comparison.json \
  --x-values 10 18 --n-orders 10 12 14 18 \
  --output-prefix docs/_static/owned_finite_beta_field_radius_matched_closure_quadrature_audit \
  --output-dir examples/outputs/owned_finite_beta_field_radius_matched_quadrature_probe
python examples/owned_finite_beta_source_channel_audit.py \
  --bootstrap-json docs/_static/owned_finite_beta_field_radius_matched_bootstrap_comparison.json \
  --settings 10:12 10:18 18:18 \
  --output-prefix docs/_static/owned_finite_beta_field_radius_matched_source_channel_audit \
  --output-dir examples/outputs/owned_finite_beta_field_radius_matched_quadrature_probe

These optional provenance artifacts prioritize local finite-beta stellarator input/wout pairs. The NTX/NEOPAX script builds finite-beta QA surfaces through vmec_jax -> booz_xform_jax with the physical VMEC edge-flux scale passed explicitly as psi_p, writes NEOPAX-style HDF5 scan tables, stores compact profile flux/current proxies from those same tables, and compares that path with the direct VMEC-harmonic interpolation path on the same radial and collisionality grid. Optimized finite-beta QH/QI cases are retained as direct wout-harmonic stress cases until their current-profile representation is supported by the JAX geometry stack.

The SFINCS-JAX script writes owned RHSMode=3 input decks for the same finite-beta wout, rho, collisionality, electric-field, and resolution contract. Use --run-sfincs-jax for explicit local runs; completed HDF5 outputs are ingested into the JSON sidecar with the SFINCS-reported nuPrime -> nu_n bridge and a coefficient-level NTX L13/L31/L33 comparison. The committed artifact now contains a six-point same-grid finite-beta QA ladder; use it as a smoke-resolution transport-coefficient localization tool, not as a bootstrap-current parity claim. The RHSMode=1 profile-current diagnostic uses the same finite-beta VMEC wout, analytic profiles, current observable, and Redl/NTX+NEOPAX comparison contract. The committed low-resolution artifact is a direct-profile normalization and convergence diagnostic. Its high-Nxi even/odd pitch gap is accepted at 1.32e-1 under the 1.5e-1 reduced-closure stress tolerance.

The finite-beta bootstrap-current script runs Redl and NTX+NEOPAX on the same QA pressure/current wout, Boozer transform, analytic profile contract, radial grid, adaptive physical nu/v support, and current normalization. It uses the normalized-radius Boozer B00(rho) convention, including dB00/dr = (dB00/d rho)/a_b, so the file-backed field object and the profile closure use the same radial coordinate. It is a production-resolution reduced-closure stress audit for this QA case; the current JSON sidecar records the remaining profile-wide Redl/NTX+NEOPAX residual and a Sonine-order convergence scan instead of promoting the figure as parity. The closure-localization script then overlays these two committed sidecars. It shows that the same-grid L13/L31/L33 coefficient ladder remains below the coefficient gate, while the profile-current residual is still larger than the 1e-1 current gate; the non-promoted follow-up is therefore the reduced profile-current observable rather than the monoenergetic coefficient bridge. The profile-current observable audit decomposes the same stress radius into the no-momentum current, applied momentum correction, correction needed to match the Redl target, Pmax trend, species-current cancellation scale, and local profile/geometry drivers. The current-conditioning audit then asks a stricter question: given the observed electron/ion cancellation, how accurate must the same-grid coefficient ladder be before coefficient uncertainty can be ruled out for a 1e-1 net-current gate? For the current finite-beta QA artifact, the stress radius needs about 1.1e-3 coefficient precision, while the completed coefficient ladder is still order 2e-2. That keeps the next step focused on production-resolution same-grid coefficient/profile-current diagnostics before any closure change is promoted. The resolution audit adds the first production stress probe: increasing the same point to 35 x 43 x 48 and tightening the VMEC harmonic cutoff leaves the coefficient floor near 2.05e-2, so the remaining finite-beta current gap is not explained by those numerical knobs. The production-ladder audit then reads the six production same-grid SFINCS-JAX points across the owned finite-beta QA radii and collisionalities. All completed coefficient differences stay below 2.07e-2; the current-conditioned precision gap remains larger than the coefficient floor, so the next open item is the profile-current closure observable. The closure-quadrature audit holds the same scan, Redl observable, profiles, and normalization fixed while varying only Sonine order and velocity quadrature. After the Boozer-field normalization fix, it records no stress-radius current gate pass; the best stress point remains above 1e-1 and the full-profile maximum remains a monitored reduced-closure residual. The source-channel audit then freezes the same momentum-restoring matrix and solves one physical source channel at a time. The one-channel solves reconstruct the corrected current to roundoff. The corrected high-order stress is a mixed density/electric and temperature-gradient response with no parallel-electric drive for this profile contract. The profile source-response audit extends that decomposition from the single stress radius to the full finite-beta profile at X=18, P=18. It plots the Redl and corrected current profiles, the Redl/NTX effective source-response multiplier, the reconstruction gate, and the response trend against Redl collisionality and trapped-particle fraction. This is a diagnostic for the next physics closure, not a fitted runtime correction. The closure-target audit reads that profile-response artifact and ranks local drivers before a runtime model is proposed. The current sidecar selects the Redl geometry factor epsilon as the strongest single driver and records that the best diagnostic model is not applied to the code. The radial-interpolation audit rebuilds the same current observable on the field radii. It changes individual radii but does not improve the full-profile maximum; the matched-radius quadrature rerun still finds zero quadrature-stable current-gate passes. The matched-radius source-channel rerun reconstructs the corrected current to roundoff and shows the same pattern: X=18, P=18 remains a quadrature-stable source-response stress diagnostic rather than a finite-beta parity claim.

It writes:

• docs/_static/owned_geometry_neopax_dataset.png

• docs/_static/owned_geometry_neopax_dataset.pdf

• docs/_static/owned_geometry_neopax_dataset.json

• docs/_static/owned_geometry_neopax_database/*.h5

• docs/_static/owned_finite_beta_sfincs_jax_inputs.png

• docs/_static/owned_finite_beta_sfincs_jax_inputs.pdf

• docs/_static/owned_finite_beta_sfincs_jax_inputs.json

• docs/_static/owned_finite_beta_sfincs_jax_resolution_audit.png

• docs/_static/owned_finite_beta_sfincs_jax_resolution_audit.pdf

• docs/_static/owned_finite_beta_sfincs_jax_resolution_audit.json

• docs/_static/owned_finite_beta_sfincs_jax_production_ladder.png

• docs/_static/owned_finite_beta_sfincs_jax_production_ladder.pdf

• docs/_static/owned_finite_beta_sfincs_jax_production_ladder.json

• docs/_static/owned_finite_beta_sfincs_jax_production_ladder_audit.png

• docs/_static/owned_finite_beta_sfincs_jax_production_ladder_audit.pdf

• docs/_static/owned_finite_beta_sfincs_jax_production_ladder_audit.json

• docs/_static/owned_finite_beta_sfincs_jax_profile_current_audit.png

• docs/_static/owned_finite_beta_sfincs_jax_profile_current_audit.pdf

• docs/_static/owned_finite_beta_sfincs_jax_profile_current_audit.json

• docs/_static/owned_finite_beta_sfincs_jax_profile_current_resolution_audit.png

• docs/_static/owned_finite_beta_sfincs_jax_profile_current_resolution_audit.pdf

• docs/_static/owned_finite_beta_sfincs_jax_profile_current_resolution_audit.json

• examples/outputs/owned_finite_beta_sfincs_jax_inputs/**/input.namelist

• docs/_static/owned_finite_beta_bootstrap_comparison.png

• docs/_static/owned_finite_beta_bootstrap_comparison.pdf

• docs/_static/owned_finite_beta_bootstrap_comparison.json

• docs/_static/owned_finite_beta_closure_localization.png

• docs/_static/owned_finite_beta_closure_localization.pdf

• docs/_static/owned_finite_beta_closure_localization.json

• docs/_static/owned_finite_beta_profile_current_observable_audit.png

• docs/_static/owned_finite_beta_profile_current_observable_audit.pdf

• docs/_static/owned_finite_beta_profile_current_observable_audit.json

• docs/_static/owned_finite_beta_current_conditioning_audit.png

• docs/_static/owned_finite_beta_current_conditioning_audit.pdf

• docs/_static/owned_finite_beta_current_conditioning_audit.json

• docs/_static/owned_finite_beta_closure_quadrature_audit.png

• docs/_static/owned_finite_beta_closure_quadrature_audit.pdf

• docs/_static/owned_finite_beta_closure_quadrature_audit.json

• docs/_static/owned_finite_beta_source_channel_audit.png

• docs/_static/owned_finite_beta_source_channel_audit.pdf

• docs/_static/owned_finite_beta_source_channel_audit.json

• docs/_static/owned_finite_beta_source_response_profile_audit.png

• docs/_static/owned_finite_beta_source_response_profile_audit.pdf

• docs/_static/owned_finite_beta_source_response_profile_audit.json

• docs/_static/owned_finite_beta_closure_target_audit.png

• docs/_static/owned_finite_beta_closure_target_audit.pdf

• docs/_static/owned_finite_beta_closure_target_audit.json

• docs/_static/owned_finite_beta_radial_interpolation_audit.png

• docs/_static/owned_finite_beta_radial_interpolation_audit.pdf

• docs/_static/owned_finite_beta_radial_interpolation_audit.json

• docs/_static/owned_finite_beta_field_radius_matched_bootstrap_comparison.json

• docs/_static/owned_finite_beta_field_radius_matched_closure_quadrature_audit.png

• docs/_static/owned_finite_beta_field_radius_matched_closure_quadrature_audit.pdf

• docs/_static/owned_finite_beta_field_radius_matched_closure_quadrature_audit.json

• docs/_static/owned_finite_beta_field_radius_matched_source_channel_audit.png

• docs/_static/owned_finite_beta_field_radius_matched_source_channel_audit.pdf

• docs/_static/owned_finite_beta_field_radius_matched_source_channel_audit.json

• examples/outputs/owned_finite_beta_bootstrap_comparison/*.h5

The source-channel panel overlays the Redl density and temperature target terms on the same current observable. The current artifact reconstructs the corrected NTX+NEOPAX current to roundoff and records a mixed density/electric and temperature-gradient high-order stress response.

The profile source-response panel shows that the temperature response multiplier is not a single hidden constant: it spans the profile while preserving source sign agreement, with the high-order stress largest at the inner radius.

The closure-target panel ranks geometry, trapped-particle, and collisionality drivers for the measured response. It is a model-identification artifact only: the runtime closure remains unchanged until a physics-derived term passes the fixed-field, W7-X transfer, source reconstruction, and same-grid finite-beta coefficient gates. Its JSON sidecar also cross-links the field-radius-matched source-channel and quadrature artifacts, confirming that the matched audit uses the same stress radius, reconstructs the source response, and still has no quadrature-stable current-gate pass.

The radial-interpolation panel rebuilds the same finite-beta current audit with the monoenergetic database placed on the exact profile-current field radii. It changes individual radii but does not clear the full-profile current gate, so no runtime interpolation policy is promoted.

The matched closure-quadrature panel repeats the Sonine/quadrature sweep after removing the sparse-radius interpolation layer. It still has zero quadrature-stable current-gate passes, so the non-promoted follow-up remains a quadrature-stable reduced profile-current closure.

The matched source-channel panel repeats the physical RHS decomposition on the same field-radius-matched contract. It reconstructs the corrected current to roundoff and keeps the remaining finite-beta mismatch localized to a quadrature-stable source-response stress.

10. Bootstrap Current With NEOPAX

python examples/bootstrap_current_with_neopax.py

This is the shortest user-facing NTX + NEOPAX radial-profile workflow. It:

• loads the W7-X reference equilibrium from the local NEOPAX checkout

• builds an NTX monoenergetic scan on the archived (\rho, \nu_v, E_r) axes

• maps that scan into NEOPAX monoenergetic arrays

• evaluates the bootstrap-current profile with the no-momentum closure

• optionally overlays the momentum-correction branch

It writes:

• docs/_static/bootstrap_current_with_neopax.png

• docs/_static/bootstrap_current_with_neopax.pdf

• docs/_static/bootstrap_current_with_neopax.json

10. Fixed-Field Bootstrap-Current Validation

python examples/bootstrap_current_fixed_field_validation.py

This local archive-backed benchmark compares:

• archived Fortran SFINCS

• SFINCS-JAX reruns of the archived inputs

• Redl on the same precise-QS QA/QH reference family

• NTX+NEOPAX on the same equilibria and archived profiles

It writes:

• docs/_static/bootstrap_current_fixed_field_validation.png

• docs/_static/bootstrap_current_fixed_field_validation.pdf

• docs/_static/bootstrap_current_fixed_field_validation.json

Use this script for the fixed-field validation figure that is shown in the README. The figure intentionally overlays only bootstrap-current profiles; the interior-window relative-error gates are stored in the JSON artifact and checked by the physics-gate tests. The current artifact keeps both the Redl analytic comparison and the reduced NTX+NEOPAX total-current stress comparison below the 1e-1 gate.

11. Autodiff Inverse Problem

python examples/autodiff_inverse_problem.py

This writes docs/_static/autodiff_inverse_problem.{png,pdf} and demonstrates recovery of a Boozer harmonic from synthetic transport data using JAX gradients.

12. Precise-QS Redl Versus SFINCS Audit

python examples/precise_qs_redl_sfincs_audit.py

This local archive-backed audit uses the fixed-field Landreman–Paul QA/QH reference family from the Zenodo bundle and writes:

• examples/outputs/precise_qs_redl_sfincs_audit/precise_qs_redl_sfincs_audit.png

• examples/outputs/precise_qs_redl_sfincs_audit/precise_qs_redl_sfincs_audit.pdf

• examples/outputs/precise_qs_redl_sfincs_audit/precise_qs_redl_sfincs_audit.json

It compares the archived SFINCS bootstrap-current profiles against:

• Redl with the VMEC-side trapped-fraction path

• Redl with a Boozer-side trapped-fraction path reconstructed through booz_xform_jax

Use this script when auditing the analytic fixed-field benchmark itself, not the integrated NTX+NEOPAX workflow. The plot is an overlay-only bootstrap-current comparison; the relative-error metrics remain in the JSON artifact.

13. Fixed-Field Transport-Matrix Audit

python examples/fixed_field_transport_matrix_audit.py

This local audit writes:

• examples/outputs/fixed_field_transport_matrix_audit/fixed_field_transport_matrix_audit.png

• examples/outputs/fixed_field_transport_matrix_audit/fixed_field_transport_matrix_audit.pdf

• examples/outputs/fixed_field_transport_matrix_audit/fixed_field_transport_matrix_audit.json

It runs SFINCS-JAX in RHSMode=3 on the same fixed-field QA/QH reference family and compares L13, L31, and L33 against NTX candidate channels. This audit now:

• reproduces the exact SFINCS RHSMode=3 monoenergetic overwrite for nu_n

• uses archive-backed Landreman/H. Smith bridge factors for the L13/L31/L33 channels

• narrows the remaining blocker to the L33 bridge rather than a generic sign or family-selection problem

This is still the coefficient-side gate for the public NTX+NEOPAX bootstrap-current validation figure.

14. Autodiff Derivative Audit

python examples/derivative_audit.py

This writes docs/_static/derivative_audit.{png,pdf} and compares direct JAX gradients of the dense solve against centered finite differences for:

• D11 and D33 sensitivities to a Boozer harmonic amplitude

• D11 and D33 sensitivities to the radial electric field

This is the validation baseline for the current prepared implicit-adjoint derivative implementation.

15. Prepared-Derivative Benchmark

python examples/derivative_path_benchmark.py

This writes docs/_static/derivative_path_benchmark.{png,pdf} and times:

• direct reverse-mode through solve_prepared_coefficient_vector(...)

• the prepared custom-VJP path through solve_prepared_coefficient_vector_vjp(...)

on the same D33 electric-field derivative scan.

It also writes docs/_static/derivative_path_benchmark.json for manuscript tables and reproducibility notes.

16. Autodiff NEOPAX Profiles

python examples/neopax_autodiff_profiles.py

This writes docs/_static/autodiff_neopax_profiles.{png,pdf} and demonstrates a low-dimensional electric-field profile inversion on NEOPAX-style monoenergetic arrays.

17. Autodiff Profile Uncertainty

python examples/autodiff_profile_uncertainty.py

This writes docs/_static/autodiff_profile_uncertainty.{png,pdf,json} and compares linearized covariance propagation against a small Monte Carlo ensemble for the differentiable NEOPAX-style profile fit under a prescribed Gaussian parameter perturbation.

18. Robust Bootstrap-Current Optimization

python examples/bootstrap_current_robust_optimization.py

This writes docs/_static/bootstrap_current_robust_optimization.{png,pdf,json} and compares deterministic versus robust optimization of the scalar bootstrap-current response under a prescribed Gaussian control uncertainty.

19. Ambipolar Profile

python examples/ambipolar_profile.py

This writes:

• docs/_static/ambipolar_profile.png

• docs/_static/ambipolar_profile.pdf

and demonstrates:

• building a radial NTX scan from explicit in-memory surfaces

• defining two species profiles with A1(r), A3(r), and \nu_v(r)

• visualizing the residual landscape over the scanned E_r axis

• solving a smooth ambipolar E_r(r) profile with radial regularization

• evaluating the resulting reduced bootstrap-current response profile

20. Ambipolar Profile Family

python examples/ambipolar_profile_family.py

This writes:

• docs/_static/ambipolar_profile_family.png

• docs/_static/ambipolar_profile_family.pdf

and demonstrates:

• solving a small family of ambipolar closures on one NTX radial scan

• comparing the integrated residual landscapes across explicit profile controls

• evaluating a bootstrap-current objective across that family

• selecting the best control point from a scalar objective landscape

19. Science Case: Bootstrap-Current Optimization

python examples/bootstrap_current_optimization.py

This writes docs/_static/bootstrap_current_optimization.{png,pdf} and shows a differentiable geometry-control problem:

• a VMEC-derived radial surface family

• one dominant non-axisymmetric harmonic used as the control variable

• autodiff optimization of a weighted bootstrap-current response

• explicit serial-versus-multiprocess timing annotations

This is the main application/science-case figure for a methods paper centered on bootstrap-current analysis and optimization.

It also writes docs/_static/bootstrap_current_optimization.json for the manuscript table builder.

20. Profile-Control Optimization

python examples/profile_control_optimization.py

This writes:

• docs/_static/profile_control_optimization.png

• docs/_static/profile_control_optimization.pdf

and demonstrates:

• building a differentiable scalar control on top of the profile closure

• optimizing that control directly against a bootstrap-current objective

• reusing the ambipolar solve inside a JAX optimization loop

21. Profile-Basis Optimization

python examples/profile_basis_optimization.py

This writes:

• docs/_static/profile_basis_optimization.png

• docs/_static/profile_basis_optimization.pdf

and demonstrates:

• using a small radial basis to perturb the profile closure

• optimizing several control amplitudes simultaneously

• retaining a compact, publication-grade figure for a higher-dimensional optimization workflow

22. Profile Transport Loop

python examples/profile_transport_loop.py

This writes:

• docs/_static/profile_transport_loop.png

• docs/_static/profile_transport_loop.pdf

and demonstrates:

• iterating a simple self-consistent profile closure on top of the ambipolar solve

• updating A1(r) and A3(r) directly from transport mismatches

• tracking accepted-step transport-loss descent together with the ambipolar closure

• smoothing the updated force profiles radially before the next ambipolar solve

23. Primitive Profile Transport

python examples/primitive_profile_transport.py

This writes:

• docs/_static/primitive_profile_transport.png

• docs/_static/primitive_profile_transport.pdf

and demonstrates:

• reconstructing A1(r) and A3(r) from primitive density and temperature profiles

• comparing initial and final residual/current profiles for the primitive closure

• updating density and temperature instead of thermodynamic-force channels directly

• enforcing explicit density/temperature source-target closure terms in addition to the transport mismatch

• exposing the derived monoenergetic force profiles alongside the final primitive state

24. Performance Scaling

python examples/performance_scaling.py --cpu-json ... --gpu-json ...

This writes publication-style CPU/GPU scaling figures from benchmark JSON payloads.

25. Prepared-Geometry Reuse Profile

python examples/prepared_geometry_reuse_profile.py --preset paper

This writes:

• docs/_static/prepared_geometry_reuse_profile.png

• docs/_static/prepared_geometry_reuse_profile.pdf

• docs/_static/prepared_geometry_reuse_profile.json

and profiles direct repeated solves, prepared geometry reuse, and a compiled prepared solver on one fixed geometry. It is a performance artifact for optimization workflows, not a physics-parity claim.

26. Profile Force Reconstruction Audit

python examples/profile_force_reconstruction_audit.py

This writes:

• docs/_static/profile_force_reconstruction_audit.png

• docs/_static/profile_force_reconstruction_audit.pdf

• docs/_static/profile_force_reconstruction_audit.json

and validates the primitive-profile reconstruction path against the archived precise-QS QA/QH benchmark family by comparing NTX-reconstructed A1(\rho) and A3(\rho) against the exact derivatives implied by the archived density, temperature, and normalized electric-field inputs. This is a monitored benchmark-family stress test for the current primitive-profile builder, not a parity claim.

26. Validation Summary

python examples/validation_summary.py

This writes:

• docs/_static/validation_summary.png

• docs/_static/validation_summary.pdf

• docs/_static/validation_summary.json

It is the recommended core validation figure for a methods paper because it combines transport trends, Onsager closure, and Legendre convergence on the same DKES-style and VMEC benchmark surfaces. This benchmark is anchored to the monoenergetic convergence/benchmarking literature of Escoto et al. 2024 and to the low-collisionality regime discussion of Helander, Parra, and Newton 2017.

The JSON sidecar freezes the plotted curves, low-collisionality tail slopes, and convergence metrics so the benchmark can be reused in tests and manuscript artifacts without scraping the figure.

27. Full Publication Bundle

python examples/make_publication_figures.py

This regenerates the manuscript-ready figure bundle and writes a manifest to docs/_static/publication_figure_manifest.json.

For the frozen paper subsets:

python examples/make_publication_figures.py --figures main_text
python examples/make_publication_figures.py --figures supplement

To build manuscript tables and reproducibility metadata from the current validated assets, run:

python scripts/build_manuscript_artifacts.py