Model card
DirectConvert
Model, not machine. A transparent model of how a broad charged-particle (alpha) spectrum might convert to electricity, distinguishing an ideal bound (L0) from a loss-aware ledger (L1).
Model levels
- DC-L0 — ideal energy-bin stage assignment (upper bound)
- DC-L1 — loss-aware seven-channel energy ledger
- DC-L2 — electrostatic trajectories (institutional-only, gated)
Equations
- EQ-DC-STAGE-BOUND: recovered ≤ q·V, with q·V ≤ E
- EQ-DC-ETA: collection efficiency integral
- EQ-DC-BALANCE: exact energy accounting
Input domain
- Normalized alpha spectrum (MeV) with a source hash
- Ordered stage potentials (MV), charge +2e
- Versioned loss set with per-channel provenance
Loss channels
- Acceptance, orbit survival, grid transparency
- Secondary/backscatter, electronics
- Optional thermal bottoming cycle (separate subsystem)
Numerical method & tolerances
- Exact ledger before rounding; conservation ≤ 1e-6·source
- DP + brute-force + differential-evolution optimizer
- Monte-Carlo / Latin-hypercube / Morris uncertainty
Verification cases
- DC-VAL-01…09 (see the L2 matrix for trajectories)
- η = n/(n+1) uniform-spectrum literature check
- Independent-codebase L0 reproduction
Unsupported physics
- Space charge / collective effects at L0–L1
- Any hardware fabrication or operating guidance
Status
- External validation: Pending
- Version: directconvert_core 0.1.0
- Reviewer state: unreviewed
Full card in source: docs/model-cards/directconvert-l1.md.