      ⚠ ADVERSARIAL REVIEW: FAIL (2026-05-27)

      8 named gaps: CRETIN method list fabricated — Scott 2001 §3 explicitly says 1D long-characteristics + 2D/3D discrete-ordinate integral [Kunasz-Auer]; no Monte Carlo anywhere in the paper; page invents MC to motivate what it ships; wavelength swap — "254 Å Ar He-α" conflates Scott §5.3 (Ne-like Fe X-ray laser, 254 Å) with §5.1 (Ar He-α ~4 Å, ~3 keV); FEYNMAN anchor off 6-8× ("~5-8 reabsorptions" claimed; ~45 actual from page's own algorithm); MC is gray (single mfp, no Voigt, no redistribution); imports only `Plot` from shared/ — none of the audited Humlíček/M&O-fits/redistribution kernels enter; L7.1/L7.2/L7.5 unimplemented. Do not trust the "CRETIN-class" framing or the Scott-2001 citations here. Per-layer defect list at RUBRIC.

From 1D eigenmodes to 3D ray-tracing

Layers 1–6 used 1D geometries — slab, cylinder (axisymmetric), sphere — for which the Holstein kernel has a closed-form integral and the eigenmode expansion is exact. Real experiments are rarely 1D.

Hydrodynamic instabilities in ICF implosions break spherical symmetry. Fluorescent lamps have end-caps. LightCell-class TPV emitters have inlet flow, hot-mirror boundaries on some sides only, and PV-cell coupling on others. Once symmetry breaks, the eigenmode expansion has no closed form and you must solve the transport equation directly.

CRETIN (Scott 2001) uses one of three methods, switching by regime:

Long-characteristics ray tracing in 1D — equivalent to discrete-ordinate $S_N$. Fast for axisymmetric problems.
Discrete-ordinate integral formalism in 2D and 3D — angles sampled on quadrature grid, optical-depth integration along each ray.
Monte Carlo photon-tracing when source-and-sink geometries are complicated and analytical ray integration is impractical (e.g. multi-material ICF capsules, complex mirror boundaries).

Below: a live 2D Monte Carlo photon-tracing simulation. Each press launches a fresh batch of photons from your chosen source; they fly, get absorbed-and-reemitted, and either escape or get quenched. The cumulative source function $S(\mathbf{r})$ builds up on the heatmap.

Explore

Initial photon source

Opacity $k_0 L$ 10

Quenching probability ε 0.01

Photons per batch 2000

Source function $S(x, y)$ (2D Monte Carlo)

Cumulative photon-absorption density. Boundary is the 2D vapour cell.

Escape angular distribution $I(\Omega)$

Polar plot of escape direction. Anisotropy reveals geometry.

Escape histogram by number of absorptions

$p_i$ = fraction of photons escaping after exactly $i$ absorptions (M&O Sec 4.3).

Total photons launched: 0 · Escaped: 0 · Quenched: 0 · Mean absorptions before escape: 0

FEYNMAN's worked example. Start with point source / k₀L = 10 / ε = 0.01. Launch 2000 photons. Watch the heatmap fill — the source function spreads radially but stays roughly centered. Mean reabsorptions should be ~5–8. Now switch to boundary excitation at the same opacity. The source function is now wall-localized; very few photons make it to the cell centre. The escape histogram skews toward $i = 0$ — photons born near the boundary often escape on first emission. This is why Scott 2001 §5.1 reports that 1D and 2D ICF simulations are inadequate for capsule-asymmetry analysis: the source function inherits the excitation geometry, and lower-dimensional models smear out the very feature being diagnosed.

The Scott 2001 ICF case

Scott 2001 §5.1 reports CRETIN simulations of an imploded Nova capsule, with argon dopant in the central DH fuel and titanium in the inner shell. Both serve as spectroscopic indicators — argon for fuel temperature at peak compression, titanium for shell distortion. The published images (Scott 2001 Figs 1–2) show the 254 Å Ar He-α line viewed from four orthogonal directions; in a symmetric implosion they should be identical, but the 3D simulation reveals shell distortions that print themselves into the line emission profile.

The key result of that section: 1D simulations agreed with neither the line ratios nor the directional asymmetry. 2D simulations got the line ratios but missed the directional asymmetry. Only 3D simulations reproduced both. The 3D NLTE solver isn't a luxury for ICF — it's the only tool that can extract the asymmetry from the spectra.

What this layer does NOT do

2D only; the live MC sim is in $(x, y)$ with cylindrical-cell boundary. True 3D ICF capsule rendering requires GPU; left for a Rust→WASM build (Layer 7v2).
Single-line, CFR-only — no PFR, no multi-level. To couple to Layer 3/4/6 physics, the MC photon kernel would need to be replaced by the full $R_{II}$ redistribution function and the multi-level CRM.
No discrete-ordinate $S_N$ quadrature comparison; deferred to a separate page.
The "ICF capsule" interactive is conceptual — the live MC is 2D circular, not the 3D Nova geometry.

FEYNMAN ✓ live MC + worked case · Victor ✓ button + sliders + live heatmap · Kay ⏳ MC photon kernel is layer-local; should be factored to shared/mc.js for Layer 8 · Scott ✓ Scott 2001 §5.1 cited · Tri Dao spirit ⏳ no GPU; CPU JS only at this stage