EMBER: Machine-Learning Detection of Modulated Ion Acoustic Waves and Associated Core-Electron Heating in the Solar Wind with Parker Solar Probe

TL;DR

EMBER is an open-source machine learning pipeline that automatically detects modulated ion acoustic waves in Parker Solar Probe data, revealing their role in electron heating without manual inspection.

Contribution

The paper introduces EMBER, a novel ensemble detection method combining physics-based and deep learning detectors for identifying wave events in solar wind data.

Findings

EMBER recovers 93% of anomalous events at 1% false alarm rate.

Detected events show elevated core electron temperatures and Te/Ti ratios.

The method reproduces known electron heating phenomena without using electron temperature data in detection.

Abstract

Modulated ion acoustic waves (IAWs) -- including triggered ion acoustic waves (TIAWs) and frequency-dispersed ion acoustic waves (FDIAWs) -- are increasingly recognized as efficient drivers of electron heating in the solar wind through nonlinear wave-particle interactions. Identification of these events in the Parker Solar Probe (PSP) FIELDS burst-mode archive has so far relied on expert visual inspection and does not scale to the full mission. We present EMBER (Electron heating from Modulated Burst-mode Event Recognition), an open-source pipeline that converts PSP FIELDS Digital Burst Memory (DBM) voltage bursts into log-scaled Fourier spectrograms and applies a multi-detector, background-only anomaly detection suite. The suite combines physics-motivated detectors, classical outlier detectors, and deep learning detectors. The EMBER ensemble recovers 93% of the anomalous events at 1% FAR (1 false positive per 100 held-out backgrounds). Coincident SWEAP/SPAN diagnostics show that flagged intervals exhibit core perpendicular electron temperatures above the adiabatic cooling expectation and elevated Te/Ti, reproducing the preferential-heating phenomenology established by prior manual studies without any use of electron temperatures in the detection step.