HELIOS-3D is a speculative research documentation project.
It is not:
- a fabricated chip
- a validated hardware design
- a claim of demonstrated sub-Landauer computation
- a claim that hopfions have been integrated into a working inference coprocessor
It is:
- a staged research hypothesis
- a map of plausible and implausible material/fabrication paths
- a public notebook for tracking claims, blockers, and evidence
- an attempt to separate near-term demonstrator paths from long-range speculation
Sub-Landauer behavior is treated as a long-range research question, not a demonstrated capability.
What is HELIOS-3D? HELIOS-3D is a speculative research documentation project exploring whether spintronic, topological, and thermodynamic-computing mechanisms could support future low-energy inference architectures near fundamental energy-efficiency limits.
Current Baseline Path The current defensible path is planar-first and electrically read, using established multilayer spintronic stacks. Freeform 3D hopfion fabrication is treated as a later-stage research branch, not the first demonstrator.
Why Hopfions over Skyrmions?
Hopfions theoretically bypass the Skyrmion Hall Effect due to zero net topological charge in 2D projection. [INFERRED]
Long-Range Target
The project investigates whether waste-heat-assisted operation could enable ultra-low energy computation (fJ-scale) approaching theoretical efficiency limits. While the Landauer limit ([SPECULATIVE]
The Hybrid-manufactured Energy-Landscape Inference and Operation System (HELIOS-3D) is a staged spintronic research hypothesis. The current minimum credible path is planar-first and electrically read; the freeform 3D hopfion and IFE branch is explicitly a later-stage research target, not a dependency of the first demonstrator.
The strongest late-stage candidate family currently appears to be compensated ferrimagnet/altermagnet transport and readout, because it offers magnetization-free dynamics with a plausible spin-pumping or current-driven readout path.
The project is currently in Phase 0.5: Documentation + validation scaffolding. A mock topological compiler scaffold exists in compiler/ for structural testing; full firmware and physics-validated compiler work remains Phase 2.
-
Topological Compiler: A Python-based mapping layer (in
compiler/) that translates semantic embeddings into 3D magnetization tensors. -
Prototype compiler scaffolding: Initial unit tests validate coordinate mapping fidelity and integer Hopf Index synthesis (
$Q_H=1$ ) using a mock Inverse Faraday Effect transfer function. - PINN Readiness: The environment is configured for Physics-Informed Neural Network (PINN) training to automate magnetization synthesis.
To run the current topological compiler tests and verify the physics mapping:
- Clone the repository:
git clone https://github.com/myrqyry/HELIOS-3D.git cd HELIOS-3D - Install dependencies:
uv sync
- Run tests:
uv run pytest tests/test_topological_compiler.py
To maintain scientific discipline and distinguish between established physics and architectural aspirations, HELIOS-3D uses the following tagging system:
[DEMONSTRATED]: Verifiable in peer-reviewed literature for specific materials and conditions.[INFERRED]: A plausible extrapolation from established physical principles or adjacent material systems.[PROPOSED]: A specific architectural integration or implementation path suggested by the HELIOS-3D model.[SPECULATIVE]: A theoretical target, high-risk hypothesis, or unverified performance projection.
See the Claims Matrix for claim-by-claim traceability.
Modern silicon scaling faces severe constraints driven by inelastic scattering and energy-intensive data shuttling. HELIOS-3D proposes to resolve these structural crises by migrating primary information carriers away from electrical charge. Instead, it explores the use of topologically protected spin textures—specifically skyrmions and 3D hopfions.
The system is conceptualized as a dual-core architecture:
- Magnetic Convolutional Accelerator (MCA): A deterministic sensory preprocessor hypothesized to utilize Compute-in-Memory spintronics.
[PROPOSED] - Brownian Reservoir Computing (BRC) Core: A probabilistic decision-maker designed to investigate noise-driven, non-equilibrium thermodynamic processing.
[PROPOSED]
Physical realization is theorized via a hybrid fabrication pipeline: Digital Incoherent Synthesis of Holographic light fields (DISH) for macro-scaffold creation, Two-Photon Polymerization (TPP) for refinement, and Atomic Layer Deposition (ALD) for magnetic coating. Performance targets aspire to fJ-scale energy efficiency, though these remain experimentally unverified projections. [SPECULATIVE]
The repo should be read with a strict staging rule: ALTERNATIVE_MATERIALS_AND_METHODS.md is the current baseline path, while CORE_ARCHITECTURE.md and PROPOSED_FABRICATION_PATH_AND_CONTROL.md are long-range branch documents.
| Path | Focus |
|---|---|
src/content/docs/ |
Astro + MDX documentation source (Primary) |
research_specifications/ |
Formal physics and architecture modules |
simulations/ |
MuMax3 and OOMMF configuration files |
compiler/ |
Topological Compiler implementation |
analysis/ |
Data validation and spintronic analysis scripts |
The technical documentation is live at helios-3d.vercel.app (Vercel) and mirrored to myrqyry.github.io/HELIOS-3D/ (GitHub Pages).
The site is built with Astro + MDX and automatically deployed on every push to main via the workflows in .github/workflows/.
The global trajectory of computational energy consumption poses an existential challenge to AI scaling. According to the International Energy Agency (IEA), combined electricity demand from data centers, AI, and crypto is projected to reach ~600 TWh in 2026, with high-growth scenarios surpassing 1,000 TWh by 2030—equivalent to the annual consumption of Japan. [DEMONSTRATED]
Beyond electricity, the "thirst" of AI infrastructure is a secondary crisis. Research indicates that global AI demand will account for 4.2 to 6.6 billion cubic meters of water withdrawal by 2027 (roughly half of the UK's annual withdrawal). [DEMONSTRATED]
Furthermore, as grids decarbonize, embodied carbon from semiconductor fabrication is becoming the dominant environmental challenge, accounting for up to 50% of the total lifecycle footprint for state-of-the-art AI hardware. [DEMONSTRATED]
HELIOS-3D bridges this gap by testing whether transitioning computational principles to topological magnetism can allow thermal noise to assist, rather than degrade, computation, while utilizing high-density 3D scaling to reduce the physical and embodied footprint of compute infrastructure.
This site is deployed automatically by the GitHub Actions workflow at .github/workflows/deploy.yml. On every push to main, the workflow runs pnpm check && pnpm test && pnpm build and publishes dist/ to GitHub Pages.
To enable:
- In GitHub repo settings, open Pages.
- Set Source to GitHub Actions.
- Push to
main; the workflow will deploy the first build.
For a one-off local preview:
pnpm build
pnpm preview