README.md

2026VBFP

Codex/ChatGPT (July 2026)

Summary

This paper automates the construction and shape optimization of stellarator island divertors within a fixed magnetic equilibrium. Two starting coordinates on the island separatrix define curved V-shaped plates subject to a maximum field-line incidence angle. The FLARE field-line diffusion model estimates surface heat loads, and the objective penalizes both peak heat flux and trajectories striking end plates. A parameter grid finds a best design with $q_{\rm peak}=2.915,\mathrm{MW/m^2}$ and 3.487% end-plate hits, below the $10,\mathrm{MW/m^2}$ engineering limit used in the study. Gaussian-process Bayesian optimization recovers a nearly identical design, $3,\mathrm{MW/m^2}$ and 3.53%, after simulating 95% fewer divertors. Repeated optimizations expose the trade-off between the two objectives, and cross-field-diffusivity scans test heat-flux-width robustness until increasing particle losses to the vessel invalidate the expected scaling. This is a proof of principle: support, cooling, manufacturing tolerances, neutral and impurity transport, equilibrium co-optimization, and validation with higher-fidelity EMC3-EIRENE modeling remain outside its reduced model.

Contributions

1. Created a two-parameter algorithm that automatically constructs curved stellarator island-divertor plates.
2. Coupled divertor geometry generation to FLARE heat-load evaluation and engineering-oriented objectives.
3. Found shapes with peak loads near $3,\mathrm{MW/m^2}$ and low end-plate interception.
4. Reduced the number of simulated candidates by 95% using Bayesian optimization.
5. Quantified objective trade-offs and sensitivity to cross-field heat diffusivity while stating reduced-model limits.