Delivers orders of magnitude faster particle confinement optimization for commercial stellarators.

SAFEST is a fast, computer-based method that analyzes and optimizes magnetic fields in stellarators to improve energetic particle confinement, enabling efficient and flexible fusion reactor designs without needing special magnetic symmetries or expensive simulations.

Background

The development of viable fusion power relies heavily on stellarators, which are complex devices designed to confine high-temperature plasmas using three-dimensional magnetic fields. In magnetic confinement fusion, sustaining the reaction requires the effective containment of energetic particles, particularly fusion-born alpha particles. These high-energy particles must remain within the plasma long enough to transfer their heat and maintain the fusion process. Consequently, there is a critical need for advanced design methodologies capable of evaluating and optimizing particle confinement to ensure the stability and efficiency of future fusion reactors.

Despite their potential, stellarators inherently lack continuous spatial symmetry, which typically results in the poor confinement and rapid loss of energetic particles. Current approaches attempt to mitigate this by enforcing strict hidden symmetries, such as quasisymmetry or omnigeneity. However, these constraints severely restrict the available design space and limit engineering flexibility. Alternatively, designers rely on full-orbit-following simulations to evaluate confinement quality. Unfortunately, these high-fidelity simulations are computationally expensive and prohibitively slow for the rapid screening of large configuration databases. Furthermore, when designers attempt to use simpler metrics to save time, these tools frequently misrank the actual confinement quality of candidate configurations, leading to suboptimal reactor designs.

Technology

This computer-implemented framework evaluates and optimizes collisionless particle confinement in stellarator magnetic fusion devices. It analyzes magnetic field strength data along field lines to extract critical field values and well widths, computing trapped-particle well actions tested against a unified “iso-action” condition. The system tracks particle markers using bounce-averaged drift equations and manages orbit-class transitions. By recording each marker’s maximum radial excursion, it calculates a fast energetic-particle loss metric ($\Gamma_W$). This metric is integrated into automated optimization loops, utilizing Bayesian optimization and dimensionality reduction, to efficiently design stellarator magnetic configurations and coil sets.

This solution is highly differentiated because it is entirely symmetry-agnostic, overcoming the limitations of traditional designs that require strict hidden symmetries. By framing neoclassical confinement as a Whitham iso-action problem, it unifies all known good confinement classes into a single mathematical formulation. This expands the design space to include unconventional, mixed-symmetry, and piecewise omnigenous fields. Furthermore, by utilizing a bounce-adiabatic approximation rather than computationally expensive full-orbit-following simulations, the framework is orders of magnitude faster while maintaining strong correlation with high-fidelity calculations. This speed and accuracy make it uniquely suited for rapid screening of large configuration databases.

Benefits

  • High Computational Efficiency: Operates orders of magnitude faster than traditional full-orbit-following simulations, enabling rapid screening of large configuration databases, real-time analysis, and fast generation of machine-learning training data.
  • Expanded Design Space: Eliminates the requirement for strict continuous magnetic symmetries, allowing engineers to explore unconventional, highly optimized configurations such as mixed-symmetry and piecewise omnigenous fields.
  • Unified Confinement Framework: Consolidates all known “good confinement” classes—including quasisymmetry and omnigeneity—into a single, comprehensive mathematical formulation based on field-line independence.
  • Accurate Performance Metrics: Generates a reliable energetic-particle loss metric that accurately separates confined and lost particles, strongly correlating with high-fidelity guiding-center calculations.
  • Automated Design Optimization: Seamlessly integrates its fast loss metric into automated, gradient-based or derivative-free optimization loops to efficiently design stellarator magnetic configurations and coil sets.
  • Enhanced Reactor Stability: Facilitates the design of stellarators with low neoclassical transport and zero bootstrap current, which simplifies coil design and improves overall fusion reactor stability.

Applications

  • Stellarator magnetic configuration design
  • Fusion reactor coil optimization
  • Rapid fusion configuration screening
  • Real-time fusion experiment analysis
  • Machine learning data generation

Additional info

This computer-implemented framework optimizes stellarator particle confinement by modeling magnetic fields as finite-gap potentials. It extracts critical field values and well widths to calculate trapped-particle actions. By evaluating a field-line-independent iso-action condition and tracking particle bounce centers via bounce-averaged drift equations, it computes a fast energetic-particle loss metric ($\Gamma_W$). This metric is integrated into automated optimization loops to design high-confinement stellarator configurations.