Search

This paper introduces a new model to find the velocity-space location of energetic ions generating ion cyclotron emission (ICE) in plasmas. ICE is thought to be generated due to inverted gradients in the v ⊥ direction of the velocity distribution function or due to anisotropies, i.e., strong gradients in the pitch direction. Here, we invert synthetic ICE spectra generated from first principles PIC-hybrid computations to find the locations of these ICE-generating ions in velocity space in terms of a probability distribution function. To this end, we compute 2D ICE weight functions based on the magnetoacoustic cyclotron instability, which reveals the velocity-space sensitivity of ICE measurements. As an example, we analyze the velocity-space sensitivity of synthetic ICE measurements near the first 15 harmonics for plasma parameters typical for the Large Helical Device. Furthermore, we investigate the applicability of a least-square subset search, Tikhonov regularization, and Lasso regularization to obtain the locations in velocity space of the ions generating the ICE. [ABSTRACT FROM AUTHOR]

The relationship between simulated ion cyclotron emission (ICE) signals s and the corresponding 1D velocity distribution function f v ⊥ of the fast ions triggering the ICE is modeled using a two-layer deep neural network. The network architecture (number of layers and number of computational nodes in each layer) and hyperparameters (learning rate and number of learning iterations) are fine-tuned using a bottom-up approach based on cross-validation. Thus, the optimal mapping g s ; θ of the neural network in terms of the number of nodes, the number of layers, and the values of the hyperparameters, where θ is the learned model parameters, is determined by comparing many different configurations of the network on the same training and test set and choosing the best one based on its average test error. The training and test sets are generated by computing random ICE velocity distribution functions f and their corresponding ICE signals s by modeling the relationship as the linear matrix equation Wf = s. The simulated ICE signals are modeled as edge ICE signals at LHD. The network predictions for f based on ICE signals s are on many simulated ICE signal examples closer to the true velocity distribution function than that obtained by 0th-order Tikhonov regularization, although there might be qualitative differences in which features one technique is better at predicting than the other. Additionally, the network computations are much faster. Adapted versions of the network can be applied to future experimental ICE data to infer fast-ion velocity distribution functions. [ABSTRACT FROM AUTHOR]