Authors: Viacheslav Sadykov (Georgia State University), Leon Ofman (Catholic University of America, NASA Goddard Space Flight Center), Scott Boardsen (University of Maryland Baltimore, NASA Goddard Space Flight Center), Yogesh (Catholic University of America, NASA Goddard Space Flight Center), Parisa Mostafavi (Johns Hopkins Applied Physics Laboratory), Lan Jian (NASA Goddard Space Flight Center), Kristopher Klein (University of Arizona), Mihailo Martinović (University of Arizona)

Analysis of ion-kinetic instabilities in solar wind plasmas is crucial for understanding energetics and dynamics throughout the heliosphere, as evident from spacecraft observations (such as Parker Solar Probe, PSP) of complex ion velocity distribution functions (VDFs) and ubiquitous ion-scale kinetic waves. We explore machine learning (ML) and deep learning (DL) classification models to identify unstable cases of ion VDFs driving kinetic waves. Using 34 hybrid particle-in-cell simulations of kinetic protons and α-particles initialized with plasma parameters derived from the PSP wind observations, we prepare a dataset of nearly 1600 VDFs representing stable/unstable cases and associated plasma and wave properties. We compare feature-based classifiers applied to VDF moments (for instance, temperature anisotropy), such as Support Vector Machine and Random Forest, with DL convolutional neural networks (CNN) applied directly to VDFs as images in the gyrotropic velocity plane. The best-performing classifier, Random Forest, has an accuracy of 0.96±0.01, and a True Skill Score (TSS) of 0.89±0.03, with most missed predictions made near stability thresholds. We study how the variations of the temporal derivative thresholds of anisotropies and magnetic energies and sampling strategies for simulation runs affect classification. CNN-based models have the highest accuracy of 0.88±0.18 among all considered if evaluated entirely on the runs not used during the model training. The addition of the wave power spectrum as an input for the ML models leads to the improvement of instability analysis for some cases. We also discuss the limitations and potential of ML and DL for the detection of ion-driven kinetic instabilities and the estimation of the change rates of the related VDF moments using PSP’s solar wind observations near perihelia.