Executable compact D-D fusion-source research, with explicit energy accounting, bounded synthetic feedback and auditable measurement evidence.
Version 0.1.0, 2026-10-11. Copyright © 2026 Bryce Lovell. Evaluation-only, source-available.
IX-MicroFusion studies externally powered, small-output D-D fusion sources. It implements real reaction cross-section mathematics and useful evaluation algorithms. The current release is a computational proof of concept, not a physically demonstrated IX-MicroFusion device. Its candidate BOMs cover a complete experimental system at the subsystem level; they are not qualified construction instructions.
| Capability | Implemented behavior |
|---|---|
| D-D nuclear model | Both neutron and proton/tritium branches; referenced Bosch-Hale coefficients; explicit validity range |
| Beam accounting | Correct lab/center-of-mass conversion and D+, D₂+, D₃+ current/nuclei accounting |
| Target calculations | Conditional non-slowing-column model and executable supplied stopping-table integration |
| Energy balance | Neutron rate, generated tritium, nuclear Q-value power, beam power and their ratio |
| Source feedback research | Noisy synthetic neutron counts, bounded current/slew, loading disturbance and latched fault handling |
| Measurement evaluation | Exact conditional on/off count analysis, multiple-window correction and artifact-backed evidence importer |
| Evidence reproduction | 192 parameter cases, 1,000 assumption-sensitivity samples, seven synthetic control campaigns and deterministic replay |
| BOM coverage | Separate 25-entry OEM sealed and 32-entry open research requirements BOMs, with sources and explicit qualification holds |
Core calculations, campaigns and regression tests use the Python standard library. No API key, paid model, GPU, web server or network is needed after downloading. Optional figure generation uses matplotlib; optional independent numerical corroboration uses SciPy. Neither is required for the core toolkit.
Extract the full folder, then open a terminal in IX-MicroFusion.
py -3.13 run.py verify
py -3.13 run.py replay --compare
py -3.13 run.py bom
Start-Process .\results\RESEARCH_REPORT.htmlOn Linux/macOS, or where python selects a supported interpreter, use python run.py .... Python 3.11+ is required. The checkout entry point works without installing the package. Windows instructions include installation and individual calculations.
verify runs 87 regression test methods, audits BOM references/coverage and compares regenerated numerical summaries, claims, fixtures and CSV traces to the committed campaign. Local Linux Python 3.12.14 verification passed with no skipped methods. CI is configured for Ubuntu/Windows × Python 3.11/3.12/3.13; its remote status must be checked after upload. Full execution records are in validation.
At 80 keV per charge, 20 µA, atomic D+ and an assumed target D column of 4×10²¹ m⁻², the non-slowing-column calculation gives:
| Calculated quantity | Result |
|---|---|
| Conditional neutron rate | 581,492.602 s⁻¹ |
| Conditional generated tritium | 557,391.879 atoms s⁻¹ |
| Incident beam power | 1.6 W |
| Nuclear fusion Q-value power | 6.64628×10⁻⁷ W, about 0.665 µW |
| Nuclear Q power / beam power | 4.15392×10⁻⁷ |
These are computed conditional outputs, not measured source performance. The column is not derived from a built target. Other electrical loads add to the beam input, and nuclear Q power is not recovered electricity. The result does not support net energy or ignition.
The shipped synthetic controller improves final-window rate error after an assumed loading drop, while cooling loss, access opening, stale telemetry and invalid sensors remove the command and latch a trip. No physical power-supply response or hardware protection has been tested.
The Berkeley HFNG publication provides experimental precedent for real compact D-D fusion and a check on simple scaling. Its Table 1 local-flux improvement is 2.21154×; the nearby approximate current-only argument predicts 2.64286×. The discrepancy is 19.5031%, so IX-MicroFusion rejects that proxy against its declared 10% accuracy gate.
This rejection is retained in the generated claim ledger. The paper's separate simulations and measurements do not validate IX-MicroFusion's absolute target yield or geometry. A scientific implementation should preserve unfavorable evidence. Details, approximate-current limitations and source access status are in the physics document and source catalog.
The preferred physical evaluation baseline is an OEM beam-target D-D source with independent diagnostics and facility controls. A separate open-target research BOM supports target/source investigation. Both cover fusion core, power, thermal, fuel/vacuum, diagnostics, control, radiation protection and facility obligations.
BOM overview links to the complete machine-readable requirements. Named supplier families were checked against manufacturer sources, with weaker access explicitly recorded. Each row specifies its requirement, quantity basis and engineering hold. No fabricated prices, verified stock, exact integration dimensions, shielding thicknesses or qualified order configuration are supplied. A compact source head does not demonstrate a microwave-sized complete installation.
This repository supplies no operational hardware driver or wiring/fabrication package. Physical source construction and operation require qualified engineering and laboratory evidence. Choosing D-D avoids supplied tritium in the baseline, but the D-D proton branch generates tritium; that obligation is included in the BOM.
py -3.13 run.py experiment .\results\synthetic_evidence_example\experiment.jsonThe bundled example contains synthetic test counts, not observations. To evaluate real records, supply a separate local bundle with actual count exposures, calibration, neutron signature, classification and protocol artifacts. The importer checks paths, units, SHA-256 integrity and background statistics, then emits a reviewable receipt.
Even a structurally valid measured bundle with significant neutron-channel excess retains physical_fusion_verified: false. Hashes and operator declarations cannot authenticate a physical experiment; independent scientific review remains required.
- Generated local research report
- Full campaign summary and claim ledger
- Executed proof of concept
- Physics equations, units and limits
- Synthetic controller behavior
- Measurement and review contract
- Architecture and intended users
- Validation gates
- BOM and qualification holds
- Sources, donor audit and software inventory
- GitHub upload instructions
python run.py replay regenerates the campaign. python run.py replay --compare regenerates into a temporary directory and compares all numerical evidence. Optional plots can be regenerated with python scripts/plot_results.py after installing matplotlib. python scripts/release_check.py verifies the supplied source-tree checksum manifest.
Physical fusion by this project's hardware, qualified stopping/loading data, absolute target-yield accuracy, coupled beam/thermal/vacuum engineering, a construction-qualified BOM, a microwave-sized installation, sustained physical source operation, independently reviewed measurements, self-sustaining ignition and net energy are not demonstrated. The repository reports those missing gates explicitly rather than assigning a speculative percentage of fusion completion.
IX-MicroFusion Research & Evaluation License v1.0 permits limited internal evaluation and withholds commercialization, production, physical deployment and public derivative redistribution without Bryce Lovell's separate written permission. It is not an open-source license. Third-party software, publications, scientific facts and supplier products retain their own rights. The license does not guarantee exclusivity over independently developed underlying ideas. See licensing details.