Roadmap

Nucleide is pre-alpha. APIs may change without notice.

Current status

The workspace is bootstrapped with twenty-five crates, PyO3 and WASM bindings, a typed Python facade, and golden-byte fixtures. The canonical CI checks (format, clippy, workspace tests, maturin build, pytest, ruff, mypy) run on every PR.

Recently landed

• Canonical nucid representation and cross-code naming dialects (nuclei).

• Material compositions, compendium loading, and XML export (material).

• MCNP-family readers for xsdir, meshtal, SSW, PTRAC, WWINP, and MCTAL (mcnp-io).

• Serpent _res.m, _dep.m, _det.m readers (serpent-io).

• FLUKA USRBIN reader and material/compound card generation (fluka-io).

• ALARA Phase 1 interop (alara-io): deck/flux/libs/output/photon/schedule glue (solver out of scope), plus clearance / waste-classification analytics: clearance index CI = Σ Aᵢ/CLᵢ and sum-of-fractions screening over parsed inventories, caller-supplied limit tables with the EU 2013/59/Euratom Annex VII Table A vendored default, plus the Spanish CSN conditional NORM landfill tables as explicit opt-in tables (screening arithmetic, never a compliance decision).

• CRAM depletion solver and chain XML parsing (depletion).

• Multicomponent enrichment cascade solver (enrichment).

• MAGIC weight windows with OpenMC (settings.xml) and Serpent (WWINP via wwin ... wf 2) emission, plus mesh source sampling (vr-tools).

• CCCC text-subset parsers + PARTISN writer (cccc-io): ISOTXS/RTFLUX readers and deck validation (no solver).

• FISPACT-II inventory output parser (fispact-io): output-only, reusing the ALARA response frame, plus the clearance-bearing wide inventory table printed with the HAZARDS + CLEAR keywords.

• Scoped ORIGEN 2.2 TAPE5/6/9 readers (origen-io).

• R2S orchestrator landed as a scoped workflow builder (r2s): zone-to-flux linking, schedule expansion, and uniform-split photon assembly (transport and activation solving stay out of scope).

• Typed Python facade and .pyi stubs (python/nucleide/).

• MeshTally.result_array() / totals_array() zero-copy NumPy bridge (mcnp-io): owned C-order float64 (ve, group) views over one row-major flatten (plus (num_ves,) totals); to_list() stays the NumPy-free path.

• Criterion benchmarks for CRAM, cascade solving, and parser throughput (crates/*/benches/).

• Cross-code validation harness against PyNE and OpenMC (validation/).

• Activation-code I/O validation (validation/activation_vs_refs.py): ALARA/CCCC/FISPACT/ORIGEN/R2S checks on committed fixtures vs PyNE oracle probes + synthetic self-consistency.

• Documentation website with interactive WASM tutorials (website/).

• Five-dialect card emission with mass-drift report (emit).

• ARMI blueprint-key bridge to emission (emit).

• Dose factors (nuclei) and dose per gram (material).

• L3 semantic deck objects with validation (mcnp-io).

• Analytic Bateman fast path with method= selector (depletion).

• Deck-editor demo and WASM decay inventory (bindings/wasm, website).

• ARMI database-snapshot adapter (r2s).

• License-free ENDF/B-VIII.0 decay-data pack: branches, isomer masses, free-form name normalizer (nuclei).

• MeshTally.result_array() / totals_array() NumPy bridge (mcnp-io).

• Library/inventory charts in the interactive demos (website).

• Hardened guards: invalid-id fallback labels, poisoned-lock errors, cell-FILL cap, Bateman n0 validation, snapshot-mixture error.

• Rust API stability pass (all crates): every public error enum is #[non_exhaustive] (new variants are no longer breaking), crate-root error re-exports and Result aliases complete the surface, and all 25 crates carry #![warn(missing_docs)].

• Point-kinetics solver (kinetics): prescribed-reactivity PKE with inhour and prompt-jump analyses (no transport, no feedback).

• Spectroscopy toolkit (spectroscopy): smoothing, counting, calibration, X-ray, and SPE line-list readers.

• UQ-lite sampling kernel (linalg): seeded MVN/lognormal/LHS draws over caller-supplied blocks, plus decay branch/energy/fission-yield perturbers (perturb_fission_yields closes the last named-open hook).

• License-free ENDF/B-VIII.0 fission-yield pack (nuclei): independent and cumulative yields verbatim, 36 parents, chain-XML depletion fallback.

• Tritium 1D diffusion-trapping kernel (tritium): T1–T2 mobile/trap transport with the full surface taxonomy; recombination ends closed in steady state and transient (G5/G6); multi-layer series stacks with Sieverts or Henry internal interfaces (G7/G8/G9) plus vented-sink recombination gaps (G10/G11).

• Scoped MCNP→OpenMC/Serpent/PHITS/GDML CSG translation (csg-xlate): surfaces, cells, nested universes, and rectangular LAT=1 lattices with per-item drift reports (GDML lattices expand to per-element placements).

• Gaussian KDE source resampling (vr-tools): KdeSampler fits over caller particle vectors with deterministic draw/pdf.

• Typed legacy SDEF fixed-source reader (mcnp-io): keyword and discrete SI/SP/SB subset with canonical re-emission.

• EPA FGR 15 runtime-download dosimetry pack (nuclei + nucleide.data): hash-pinned fetch of the official EPA zip; the fgr15 module parses the seven scenario tables (nothing vendored).

• Tokamak fusion neutron sources (plasma-source): ring/point D-D and D-T sources with ion-temperature-broadened spectra (Brysk 1973; Ballabio 1998), seeded sampling to particle vectors, and MCNP SDEF + Serpent source-card emission with drift reports (MCPL projection stays caller-side).

• Parametric tokamak plasma source (plasma-source): Miller-geometry flux surfaces (Fausser 2012) with caller-supplied L/H/A-mode profiles, reactivity-weighted emission (Bosch–Hale 1992), Miller-Jacobian volume gates, and marginal-histogram source cards with joint-correlation drift. Fuel is equimolar D-T, pure D-D, or a D/T mixture at the shared ion temperature (Eriksson/DRESS two-branch rate); toroidal sectors, per-species ion temperatures, the D(d,p)T proton branch, and the pinned T-T normalization also landed (sectors scale totals by rotation/2π with a PHI marginal; sampler and cards stay neutron-only).

• Damage and gas-production metrics (damage): NRT-dpa and arc-dpa displacement functions (NRT 1975; Nordlund 2018), He/H appm and He/dpa ratios by spectral folding of caller flux with caller response functions, and UQ on the folds over caller MVN blocks; the SPECTER report (ANL/FPP/TM-197, US-gov PD) is the validation oracle behind an opt-in vendored Table VII fallback for callers with no damage-data pipeline (ASTM E693/E521 stay designation-only; PKA-spectrum solving stays out).

• Neutron spectrum unfolding (unfold): SAND-II iterative spectral adjustment (McElroy et al., AFWL-TR-67-41, 1967), STAYSL-class damped least-squares (Perey, ORNL/TM-6062, 1977), GRAVEL chi-square-weighted adjustment (Matzke, PTB-N-19, 1994), and MAXED maximum-entropy adjustment (Reginatto & Goldhagen, Health Phys. 77 (1999) 579 — journal equations only, the closed UMG package never touched) of a caller-supplied guess spectrum against measured activation rates over a caller-supplied response matrix, with per-group relative-change convergence diagnostics and a hard NotConverged past the explicit iteration cap (MAXED adds a caller chi-square target to the acceptance leg); the IRDFF-II response pack (v1 foil subset + v2 extension) ships as a runtime hash-pinned download, never vendored.

• TBR and blanket power bookkeeping (blanket): raw TBR from caller tallies, multiplicative per-port coverage haircuts, blanket energy multiplication, the tritium burn rate from pinned constants, and the breeding-margin / net-surplus fuel-cycle metrics — pure arithmetic over caller transport tallies (no transport solving, no geometry optimization, no coupling into tritium).

• Equilibrium data readers (equilib): classic-netCDF VMEC wout files (CDF-1/CDF-2 only; netCDF-4/CDF-5 convert facade-side) plus the &INDATA input-text grammar, with flux-surface Jacobian helpers for volume weighting — reads data, never solves equilibria.

• Wall-load mapping in flux coordinates (equilib): closed-form per-cell accumulation of a caller birth-rate density field onto the caller wall surface sharing the same flux coordinates, plus the angle-weighted one-field-period total that conserves the discrete births (pure glue — no transport, no shadowing).

• Arbitrary-3D birth-rate lattice source (plasma-source): the caller supplies the full 3D birth distribution as a point cloud with optional field-period symmetry (base-sector cloud replicated uniformly, totals scaled by the period count), sampled and emitted through the landed mixture/spectrum/card machinery.

• Deuterium hot-tail fraction (plasma-source): the one pinned single-tail-temperature non-Maxwellian shape on a D/T fuel mixture (bulk plus hot tail reacting at their own effective temperatures, five sub-branches through strengths, sampler, cards, and drift row).

• Closed-form ECRH accessibility kernel (plasma-source): cold electron-cyclotron resonance, the weakly-relativistic thermal shift, and the O1/X1 cut-off densities located by interpolation along a caller beamline — the per-port penalty inputs for blanket bookkeeping (no ray tracing, no launcher design).

• Coil fast-fluence / lifetime bookkeeping (damage): fast-flux sums above a caller threshold, fluence as flux × time, piecewise-constant history accumulation, and weakest-link life over caller-supplied limit tables (dpa weighting reuses the folds).

• OpenMC statepoint tally bridge (nucleide.openmc, pure Python — no Rust changes): mesh/cell tally extraction through the caller-side OpenMC API into the plain arrays the damage folds accept, plus a CSV interchange for the same dict (HDF5 never read in Rust).

• Sublet S1–S7 radiological totals (alara): total activity with the IRT α/β/γ split and decay heat over caller-supplied decay energies (each with the excluding-tritium companion), committed ingestion/inhalation hazards over caller 50-year dose coefficients (each with the companion), the transport Bq/A₂ ratio with its effective A₂, the IAEA clearance-index variant screened at ≤ 1, and the slab/point gamma dose over caller groups with the loud 0.3 m clamp — all pinned to the open CCFE-PR(16)53 preprint and FISPACT-II open docs.

• He/dpa ratio uncertainty propagation (damage): per-draw ratios over the seeded joint MVN block with k-SE gates against the bias-corrected expectation and the delta-propagated spread, plus the distribution-free 68% interval gated against the Fieller-construction quantiles (holds at honestly-sized blocks); zero-dpa draws fail loudly.

• Per-species ion temperatures on single-fuel parametric configs (plasma-source): single-fuel D-T reacts at the mass-weighted T_DT, single-fuel D-D at T_D, with per-branch Ballabio lines and bit-for-bit equal-T recovery.

Upcoming priorities

• Cut releases: vX.Y.Z tags publish Python wheels to PyPI and workspace crates to crates.io.

JOSS publication milestone (~6 months of public history)

A JOSS paper draft (paper.md, paper.bib) and the cross-code validation harness (validation/) are in place. JOSS pre-review gates require the public repository to show more than six months of active, iterative development and demonstrated research use before submission, so submission waits while the project matures in the open. Until then:

• Keep commits steady and incremental; tag real releases (0.1.0 → 0.2.0 → …) with matching CHANGELOG.md sections and PyPI wheels.
• Make research use visible (e.g. Nucleide as a dependency inside the NukeHub ecosystem / NukeIDE, public examples or notebooks) — this is the research-impact evidence the submission form asks for.
• Preserve per-version validation numbers (committed validation/results/*.json archive) so the paper can show correctness across versions.
• Prepare the mandatory JOSS AI-usage disclosure (tools/models, scope, human review statement) at submission time.
• Final steps at submission: draft-PDF proofread, Zenodo DOI into CITATION.cff and the paper, then the JOSS submission form.

Out of scope

Transport solvers, Fortran discrete-ordinates ports, ENSDF evaluators, MOAB-dependent meshing, and GUIs. Nucleide complements transport codes; it does not replace them.