imas: exclude fast-ion charge from the impurity split

[d-burg]

effective_impurity_charge treated ne - ni as impurity charge. With a beam population that difference also contains the fast ions, so nz was inflated and Z_imp drifted off the real impurity charge.

The fix, in two parts

1. Subtract the fast-ion charge and renormalize Zeff. Only ne_th = ne − Σ_s Z_s n_s^fast is neutralized by thermal ions. The IDS Zeff is the thermal-species numerator over the full ne, so the inversion must also use Zeff · ne / ne_th; the subtraction alone recovers only half the bias (adversarial review, verified numerically: raw 1.95 → half-applied 3.19 → true 6.00 for C6 at 25 % fast fraction). The inversion lives in physics.impurity_charge_with_fast_ions, wired into the IMAS reader. The Zeff consumed by the bootstrap and the forward solve deliberately stays the full-ne one — the renormalization is local to the Z_imp / p_imp derivation, so the wider blast radius that concern implied never materializes.

2. Run every impurity consumer on the thermal ne. Baseline gains z_fast (fixed across draws, like p_fast), plumbed through generate_bouquet into perturb_kinetic_equilibrium, so run.py's forward-solve assemblies, the per-draw impurity pressure, the baseline reference assembly, and the Zeff-primary channel all use ne − z_fast with the same thermal-derived Z_imp the reader stores. Previously the reader derived Z_imp/p_imp on thermal ne while the solves and draws used the full ne — a σ=0 pressure skew of e·z_fast·ti/Z_imp between reader and solver. The Zeff-primary ni derivation uses the exact thermal inverse ni = (Z·ne_th − Zeff·ne)/(Z−1) and its draw clip generalizes to [ne_th/ne, Z_imp·ne_th/ne]; both reduce exactly to the old forms when z_fast is absent, so the reconstruction path is bit-untouched.

Motivation

On a DIII-D discharge with 6.3 MW NBI the fast fraction reaches 28 % through the L→H transition, and the pressure-completeness gap tracked it closely with the first (subtraction-only) form of the fix:

fast fraction	gap before	gap after (subtraction only)
10.9%	3.4%	0.27%
28.0%	12.6%	0.56%
21.3%	7.6%	0.48%
6.4%	1.5%	0.19%
1.7%	0.4%	0.09%

The 2% guard was rejecting three of those slices outright. The residual gaps in that table are the signature of the half-applied Z_imp (p_imp still ~1.9× inflated); the renormalization is expected to close most of them.

No behaviour change when no ion carries density_fast.

Tests

tests/test_fast_ion_dilution.py (10 tests): exact recovery of the true impurity charge through the shipped helper, the half-applied form pinned as a regression marker, the fast-ion-free reduction, the pathological z_fast ≥ ne degradation, the imas.py wiring, exact ni/nz recovery through the thermal inversion, the Baseline.z_fast field, and source-level pins for all four consumer sites. Fast suite: 399 passed.

🤖 Generated with Claude Code

[Copilot]

🟢 Approval recommended

The functional changes are coherent end-to-end (reader → baseline → draw/solve paths) and are backed by targeted regression tests; remaining feedback is limited to documentation/test robustness.

Pull request overview

This PR corrects the IMAS impurity inversion so fast-ion charge is not mistakenly attributed to impurities, and propagates the resulting “thermal-electron” density (ne_th = ne - z_fast) through downstream pressure construction paths.

Changes:

• Add physics.impurity_charge_with_fast_ions() and extend main_ion_density_from_zeff(..., z_fast=...) to support fast-ion dilution while keeping IMAS Zeff normalized to full ne.
• Wire IMAS baseline reading to compute z_fast = Σ Z_s n_s^fast, compute (Z_imp, ne_th) via the helper, and use ne_th for impurity pressure.
• Thread z_fast through bouquet generation / equilibrium perturbation / forward-solve assembly and add regression tests for the corrected behavior and wiring.

File summaries

File	Description
tests/test_fast_ion_dilution.py	Adds regression tests covering corrected impurity inversion and wiring across IMAS + solver paths.
bouquet/physics.py	Introduces impurity_charge_with_fast_ions and extends Zeff→ni inversion to account for z_fast.
bouquet/io/imas.py	Computes z_fast from IMAS ions and uses the helper to derive Z_imp and thermal ne_th for impurity pressure.
bouquet/baseline.py	Adds Baseline.z_fast to carry fast-ion charge density on the kinetic grid.
bouquet/TokaMaker_interface.py	Threads z_fast into perturb/draw plumbing, Zeff-primary ni derivation bounds, and impurity pressure calculation.
bouquet/run.py	Ensures forward-solve and sigma=0 consistency pressure assembly uses thermal ne when z_fast is present.

Review details

• Files reviewed: 6/6 changed files
• Comments generated: 3
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.