Speaker
Description
Understanding and controlling the detachment and power-exhaust behavior of SPARC is essential for ensuring divertor survivability in a device characterized by narrow scrape-off layers and high parallel heat fluxes.
We apply SOLPS-NN, a neural network surrogate model trained on SPARC SOLPS-ITER simulations, to explore the SOL and divertor operational space for SPARC.
In this context, we present and focus specifically on the detachment physics and divertor operational limits that the currently available model enables us to explore.
Using targeted scans of fueling, upstream density, and divertor neutral pressure, we analyze the operational domain when SPARC enters partial and full detachment. Our approach enables us to e.g. verify the appearance of how the rollover of particle and heat fluxes emerges in diagnostic signals, but also whether divertor heat-flux limits can be satisfied while maintaining favorable upstream confinement conditions. These two-dimensional reconstructions highlight the structure and movement of the detachment front and reveal inner–outer divertor asymmetries in electron temperature reduction, neutral accumulation, and momentum-loss processes, which can be used to inform fast real time control models.
We further quantify the extent of SPARC’s viable operational space and examine its robustness by scanning pivotal parameters within the surrogate model, i.e. the cross-field transport coefficients. These controlled variations probe uncertainties in predicted behavior, explore sensitivity to heat flux width scaling and assess how modest parameter adjustments might reduce model-reality gaps.
Using the surrogate model to unlock key physics insights, we demonstrate how neural network tools can swiftly map SPARC’s divertor operational space. These methods reveal detachment margins early and steer scenario development toward better protection of plasma-facing components.