Hierarchical Modeling of Runaway Electron Evolution: From Fluid to Fully Kinetic Descriptions

Descriptions of runaway electrons (REs) range from simple, but efficient, fluid mod-

els to computationally intensive kinetic descriptions. We introduce a novel approach

that enables a hierarchical treatment of RE evolution in a single framework. This ap-

proach is enabled by combining an adjoint formulation of the relativistic Fokker-Planck

equation for RE evolution and the use of a physics-informed neural network (PINN).

The resulting highly customizable framework acts as a rapid surrogate that predicts REs

at various fidelity levels. At the lowest fidelity, a single moment, typically the density

moment, is predicted, consistent with existing fluid models of RE evolution. Higher

physics fidelity models can be constructed by considering the evolution of multiple mo-

ments, such as current and average energy, thus allowing for the prediction of quantities

such as the saturated energy of the RE distribution. Ongoing work includes predicting

the full RE distribution by evolving the energy distribution of Legendre polynomials,

allowing for the reconstruction of the pitch and energy dependence of the RE distribu-

tion. The resulting framework enables a trade-off between speed and physics fidelity,

allowing the user to adapt the physics fidelity to the targeted application.