Struct

`space::irbem::DriftShellOptions`

cheatah-space v0.1.0-alpha — Biome Standard 0.6.5-alpha

How finely a drift shell is resolved, and how hard the root-find works.

The two resolutions carry IRBEM's meaning exactly, so from_irbem can translate its options(3) and options(4) and a differential comparison can be made at MATCHED settings — which docs/ERROR_BUDGET.md §2(a) shows is the only setting at which a 0.01 L* target means anything, IRBEM's own default-resolution error being 0.010–0.017 at L ≈ 6.

Functions

fn static DriftShellOptions from_irbem(int t_resolution, int r_resolution) #

The settings IRBEM's options(3) and options(4) select.

dθ = π/(720·(t_resolution+1)) and Nder = 25·(r_resolution+1), both from its published option table. Used by the differential suite so the two libraries are compared at the same resolution rather than across resolutions — the distinction ERROR_BUDGET.md §2(a) insists on.

Parameters

`t_resolution`	IRBEM's `options(3)`, 0–9; the colatitude resolution.
`r_resolution`	IRBEM's `options(4)`, 0–9; the azimuth count.

Returns

the options, with everything else left at its default.

Complexity

O(1).

Allocation

none.

Unit testIrbemDriftShell.IrbemOptionsTranslateToTheDocumentedResolutions

Constants & variables

var int azimuths #

How many azimuths the shell is sampled at — IRBEM's Nder.

The default 25 is its options(4) = 0.

var double colatitude_step_deg #

The colatitude step of the polar-cap quadrature, degrees — IRBEM's dθ.

The default 0.25° is its options(3) = 0, i.e. π/720.

var int bracket_trials #

How many radii the bracketing sweep tries, spread over [bracket_low, bracket_high] × L_m.

These all go into ONE batch, which is what makes the bracketing phase ntime × Nder × trials wide rather than a serial scan.

var double bracket_low #

The low end of the bracketing sweep, as a fraction of the dipole seed L_m.

var double bracket_high #

The high end of the bracketing sweep, as a fraction of the dipole seed L_m.

Measured over L = 2…8 with IGRF-14 the true root sits within ±7 % of L_m; the sweep is widened to −18 %/+32 % so a disturbed shell still brackets.

var int refine_iterations #

How many Brent iterations follow the sweep.

Bounded on purpose: IRBEM's θ march has no bound, and an unbounded refinement inside a lock-step batch would make every shell wait for the worst one.

var double radius_tolerance #

The radius the root-find is converged to, Earth radii.

5 × 10⁻⁴ Re is ~1/20 of a quadrature cell in the footpoint colatitude it feeds, so the root-find is not the error term.

var double inaccessible_scale #

The scale of the analytic continuation below the shell, Earth radii.

Positive; it sets how fast an out-of-range bracket closes and cannot move the root. See the file brief.

var TraceOptions trace #

Tracing options for the invariant I — handed straight to trace_invariant_batch.

var TraceOptions footpoint #

Tracing options for the footpoint walk.

Longer than trace — it runs from the magnetic equator all the way to r = 1 rather than stopping at the mirror points — and finer, because its colatitude is what the flux quadrature's boundary is. MEASURED at IRBEM's default resolution, worst deviation from the converged oracle over 12 points: steps_per_l = 25 → 0.0121, 50 → 0.0101, 100 → 0.0100, 200 → 0.0099. The knee is at 50 and the default sits one doubling past it, which costs ~10 % of a host-lane L* and nothing measurable on the device.