New deterministic neutronics workflow slashes stellarator simulation time and unlocks faster tritium breeding design

A stellarator’s twisted modular coil array confines glowing plasma – the complex three-dimensional geometry that makes fast and accurate neutronics simulation essential for reactor design

(Image courtesy of Max-Planck-Institut für Plasmaphysik)

A new automated stellarator neutronics workflow can now deliver accurate stellarator blanket simulations fast enough for live design iteration, a development that directly changes how engineers handle tritium breeding, coil fast neutron flux, and plasma confinement constraints during early reactor design. Timo J. Bogaarts and Felix Warmer of Eindhoven University of Technology and the Max-Planck-Institut für Plasmaphysik published the work in March 2026, developing a deterministic model that addresses the prohibitive computational cost of conventional Monte-Carlo methods, which has long blocked rapid stellarator scaling iteration.

Why conventional stellarator neutronics fail

Stellarator reactor design depends heavily on neutronic parameters. Engineers must verify sufficient tritium breeding and confirm that fast neutron coil flux stays within safe limits before a configuration can advance. However, conventional Monte-Carlo codes require between 100 million and one billion particle samples to resolve three-dimensional neutron responses in stellarator geometry. That sampling burden makes rapid design iteration impractical, particularly during coil optimisation phases where geometry changes frequently.

To address this, Bogaarts and Warmer developed a deterministic neutronics model based on a discontinuous-Galerkin spatial discretisation combined with novel stellarator-symmetric boundary conditions. This approach solves the full six-dimensional neutron distribution directly, without stochastic sampling. Additionally, the model introduces automated variance reduction, which the authors show improves the figure-of-merit of subsequent Monte-Carlo blanket simulations by approximately two orders of magnitude.

Benchmark results and tritium breeding performance

The workflow was benchmarked against the established OpenMC Monte-Carlo code in two stellarator geometries: the HELIAS5 configuration and the novel SQuID plasma and coil geometry. In the HELIAS5 case, the deterministic code agreed with Monte-Carlo results within 0.2% for tritium breeding, within 10% for fast neutron coil flux, and within 20% for neutron power deposition. Convergence studies confirmed that significantly fewer than 168 energy groups are needed to maintain this accuracy. In one specific comparison, the deterministic code completed the simulation in 12 CPU-hours, while the equivalent OpenMC run required 1,726 CPU-hours.

The SQuID benchmark, using a non-uniform layered blanket geometry, produced similarly strong agreement. Fast neutron coil fluxes across all five coils in the half-module matched within 15%. A full SQuID simulation applying stellarator-symmetric boundary conditions returned a tritium breeding ratio of 1.52 in the fully closed layered geometry, a figure the authors note is consistent with previous studies using comparable blanket configurations, though not yet reflective of a final design including ports or divertors. Furthermore, that simulation confirmed that neutronic engineering constraints can be satisfied without filling all available blanket space. Therefore, the authors note, coil optimisation could exploit that remaining space to improve overall reactor economics.

Implications for stellarator neutronics workflow deployment

The stellarator neutronics workflow’s parametric geometry package, SBGeom, is fully differentiable and GPU-compatible, making it directly suitable for gradient-based coil and blanket optimisation. This positions the tool as a practical component in integrated stellarator design frameworks. Planned future work includes coupling to thermo-mechanical codes, adding divertor and port geometries, and embedding the deterministic model directly within optimisation loops. Together, these developments could substantially accelerate early-stage stellarator design iteration, moving blanket configurations from parametric concept toward detailed engineering assessment with far greater speed than conventional methods allow.

Stay ahead in the fusion revolution explore more breakthroughs from leading innovators in clean energy technology.