Choose a research question
| Laboratory | Primary question | Additional study tools |
|---|---|---|
| FusionSim | Reaction rates, density scaling and the included heating balance | Thermal radiation comparison; independent quadrature fixtures; relative-energy distribution import; time-dependent burn and ash (0-D); measured EXFOR cross sections vs the pinned baseline |
| Isotope Explorer | Q values, thresholds and event kinematics | Mass-covariance import; centre-of-mass and laboratory frame verification; cross-section contracts; measured branch and angular validation (EXFOR) |
| Virtual Reactor Lab | Mirror configuration scenarios and single-particle scales | Physical field probe; mirror notebook scenario; heating closure check; exact coil-field reference; held-out field-map comparison |
| Torus Lab | Prescribed toroidal orbit verification | Uniform-field helix phase benchmark; matched-time convergence; G-EQDSK convention inspection |
| DirectConvert | Conditional particle conversion and plant balance | Correlated planar energy–angle retarding model; separate heat reservoirs with individual Carnot checks; self-consistent collector gap (1-D PIC); steady gap theory; campaign manifest check |
| AlphaTrack | Object counts, segmentation robustness and validation | Threshold and scale sensitivity; manual touching-object partition; annotated-image comparison; commit–reveal custody holdout (four steps) |
| Reactor 3D | Composed system coupling and boundaries | Component evidence review; ledger and heating inspection; double-count gates; exact coil-field reference; held-out field-map comparison |
New in 7.5.2: evidence pipelines and research models
Each panel below is a separate study with its own settings and exports. None of them turns a calculation into an experimental validation; every result states physicalValidationEstablished: false.
| Panel (laboratory) | What it computes | What it does not establish |
|---|---|---|
| Time-dependent burn and ash (FusionSim) | 0-D p, ¹¹B and thermal-alpha densities with separate ion and electron energies, alpha heating split, electron–ion exchange, bremsstrahlung, confinement losses and fuelling, with an energy and particle ledger. Checked against an independent SciPy solver. | No measured burning p–¹¹B history exists; transport profiles and synchrotron radiation are not modelled. |
| Measured cross sections vs pinned baseline (FusionSim) | Separately selectable EXFOR measurements (IAEA-NDS, CC BY 4.0) compared with the unchanged Sikora–Weller table: ratios, covariance χ² with correlated normalisations and a fitted normalisation factor. | Consistency does not validate a reactor model, and an inconsistency does not show which source is wrong. |
| Measured branch and angular validation (Isotope Explorer) | Compares the event sampler's ground-state branch setting and isotropic angular shape with published branch fractions and Legendre coefficients at matching proton energies. | Alpha singles spectra and three-alpha correlations are not tabulated in the data used. |
| Exact coil-field reference and held-out field map (Virtual Reactor Lab, Reactor 3D) | Exact vacuum fields of coaxial loops and solenoids; the difference between the analytic mirror and a matched two-loop coil set; a held-out comparison with a vector-field map (manufactured by default). | No calibrated measured map of the actual coils is included. |
| Self-consistent collector gap and steady theory (DirectConvert) | Planar 1-D particle-in-cell gap with space charge, particle fates and an energy ledger, plus all steady cold-beam states and the critical current. | No collisions, secondary electrons, grids or multi-dimensional geometry; no collector measurement exists. |
| Collector campaign manifest check (DirectConvert) | Checks a campaign record against docs/protocols/COLLECTOR-CAMPAIGN-PROTOCOL.md. | A complete manifest is documentation, not a measurement. |
| Custody steps 1–4 (AlphaTrack) | Label commitment before predictions; frozen predictions quoting that commitment; a reveal that quotes the freeze it answers, so predictions frozen after the reveal are rejected; then verification of the whole chain and plate-grouped bootstrap scoring. | Identities, timing and label quality are declared, not authenticated. The shipped example is manufactured throughout — labels, predictions, protocol and custodian names — and its predictions are the labels moved one pixel, so its perfect score measures nothing. |
Requirement-by-requirement status: docs/v752-requirements.json. Items that need outside evidence stay open in docs/v752-work-ledger.json (WP-X1 to WP-X8).
Run, inspect, reproduce
- Pick the exercise. Choose the visual demonstration or the numerical study, and read its qualification gates before entering values.
- Calculate the main run. Enter the laboratory inputs and calculate. Current results and data carried over from a previous run are always distinguished on screen.
- Open a study tool. Expand its section and edit the labelled scientific inputs (or import documented settings JSON), then calculate. Each tool holds its own state and reports its own result.
- Export a research package. The package records inputs, results, methods, data hashes and numerical studies together. AlphaTrack additionally includes decoded image pixels — keep the original acquisition files as well.
- Transfer only compatible quantities. A centre-of-mass event, a uniform-source expectation and a device-frame particle flux are three different kinds of record. Each transfer names what it carries and what it does not.
Portable experiment records
- Export and re-importExport an experiment as JSON, then import the file or paste its contents into a compatible laboratory. Inputs can be compared side by side before they are applied.
- Validated before anything changesThe whole document is checked first: release, laboratory, control set, numeric syntax, ranges and input checksum. Only then do form values change, and the receiving laboratory recomputes the result rather than trusting stored numbers.
- Images stay separateAlphaTrack stores pixel identity, not image bytes. Load the matching original image first, then apply the experiment; save the result JSON as its own file.
- Snapshot compatibilityCurrent exports identify application release 7.5.2 separately from the retained 7.4.5 snapshot compatibility tag. The explicit importer also accepts earlier 7.4.x snapshots, preserving supplementation history where it applies. A result file is not an input snapshot and is never accepted as one.
- What a checksum meansChecksums detect change, not authenticity. A valid import confirms that the document is intact and internally consistent; it says nothing about whether the physical model or operating point is appropriate for your question.
Records, hashes and provenance
Release-owned source and result objects record source, frame, units, transformation and hash where applicable. Primary structured JSON exports contain inputs, model version, timestamp, outputs and stated limitations; standalone tabular CSV exports repeat the controlling inputs and provenance so a single row stays interpretable, while PNG accompanies them as a visualisation rather than a complete record.
Stale inputs disable the previous exports. Results that fall in a negative or invalid domain remain visible and labelled rather than being hidden, so a failed configuration is still an inspectable outcome.
The checksum manifest covers the package payload except the explicitly listed self-referential integrity files, and data normalisation is deterministic — canonical input and data hashes support replay; timestamps, run identifiers and browser image rasterization may differ.
Related pages
Method reference — the equations, published data sources and validity domains behind each laboratory.
Verification record — historical numerical baseline; see the v750 verification record below.
Installation check — confirms every required file is present, correctly typed and matched to this release.
New measured-data and reproducibility workflows
AlphaTrack includes a plate and calibration registry, locked evaluation, adjudication and bounded pixel-batch analysis. DirectConvert accepts signed I–V readings, background, calibration and frozen held-out predictions. Torus and reactor workspaces compare declared field measurements with a current numerical run. These workflows preserve evidence identity; supplied measurements and calibration declarations still require independent review.
Use Sensors or Compare in Torus for uncertainty, evidence and external-reference studies. Open Research tools in the other labs. Typed inputs retain an advanced JSON view and downloadable results. Manufactured examples exercise software only.