Use the outer 1D1V kinetic oracle for validation

beach-kinetic-response is an offline research tool that fixes a matching-plane input and tests whether the outer collisionless 1D1V Vlasov—Poisson system reaches a steady or statistically stationary response. It does not evolve the outer system with the BEACH runtime. Only a certified, complete Cartesian response subset can be converted for the existing response_backend="table" path.

A successful process exit does not establish physical validity. Check each query’s classification, grid convergence, far boundary, and conservation diagnostics before conversion.

1. Run the reference sweep

Run this command from the repository root. The output directory must be new or empty.

beach-kinetic-response \
  examples/outer_kinetic_reference.toml \
  examples/outer_kinetic_k1_k6.csv \
  outputs/outer-kinetic-atlas

The example is a diagnostic sweep containing unique Zhao roots, multiple roots, and no-root regions. It is not a production fixture whose every point is expected to certify. The command writes:

Output	Purpose
kinetic_response_raw.csv	Per-query response, classification, stationarity, far-boundary, and conservation data
kinetic_response_manifest.json	Grid, velocity ranges, CFL, tolerances, commit, source hash, and raw-CSV hash
kinetic_response_profiles/	Potential, field, charge-density profiles, and time histories

2. Interpret classifications

classification	Table use
steady	Eligible after grid and domain convergence
stationary_average	Eligible only after checking autocorrelation and SEM and explicitly allowing it in the converter
unresolved_transient	Reject; revisit run time or averaging windows
secular	Reject; do not replace it with an arbitrary-time average
far_boundary_not_converged	Reject; validate the outer length and far-boundary physics first
numerical_failure	Reject; inspect failure_reason, the velocity domain, and conservation

The solver imposes $E(H)=D_H/\epsilon_0$ and $\Phi(L)=0$. It uses E(L)=0 as a convergence diagnostic rather than a third boundary constraint. Version 1 accepts only zero outward ambient-population inputs. In the raw CSV, far_field_v_m and far_charge_imbalance are maximum absolute values over the final averaging window. Use the corresponding NPZ for signed values and profiles.

ambient_electron.number_density_m3 and ambient_ion.number_density_m3 are amplitudes of the complete drifting Maxwellians feeding the far boundary. Only their $v<0$ half spaces are imposed there, so these values need not equal the computed far-field densities. Version 1 does not shoot the electron-source amplitude to E(L)=0. For a Zhao overlap test, use the ambient electron source density solved by that Zhao root. Where Zhao has no solution, scan the source density as an independent convergence axis. Do not interpret a far-boundary failure caused by an inconsistent fixed amplitude as insufficient outer length.

3. Convert a certified subset

The converter checks the raw-CSV hash, every selected classification, duplicates, and the complete Cartesian product. It does not fill missing points and accepts only steady rows by default.

beach-kinetic-table \
  outputs/outer-kinetic-atlas/kinetic_response_raw.csv \
  outputs/matching_plane_response.csv \
  --range displacement_c_m2=-1e-11:1e-11

By default it finds the manifest beside the raw CSV. Use --manifest when it is elsewhere. Add --allow-stationary-average only when the study explicitly accepts statistically stationary rows.

The resulting CSV is consumed by the existing matching-plane table backend.

4. Compare before production use

Vary at least these controls independently:

• nz and each species’ nv;
• both velocity limits;
• z_length_m;
• cfl; and
• max_time_s and the averaging window.

At points with a unique Zhao root, compare $\Phi_H$, ambient inward fluxes, and PE return and escape fluxes. Then run the same BEACH case with zhao_online and the kinetic-derived table and compare surface charge, gap potential, and absorbed currents. Finally vary the matching-plane height at three or more points. If the principal observables do not converge, classify the setup as no_matching_overlap.

This oracle is collisionless, unmagnetized, and one-dimensional, and reconstructs PEs as a half-Maxwellian from flux and mean normal energy. If a strong PE beam prevents far-field quasineutrality, do not pass certification by extending the domain alone. Test whether collisions, magnetic fields, geometric dilution, or an energy-resolved VDF require a different outer model.