Solar Wind Reconnection Exhausts in the Inner Heliosphere Observed by Helios and Detected via Machine Learning

Reconnecting current sheets in the solar wind play an important role in the dynamics of the heliosphere and offer an opportunity to study magnetic reconnection exhausts under a wide variety of inflow and magnetic shear conditions. However, progress in understanding reconnection can be frustrated by the difficulty of finding events in long time-series data. Here we describe a new method to detect magnetic reconnection events in the solar wind based on machine learning, and apply it to Helios data in the inner heliosphere. The method searches for known solar wind reconnection exhaust features, and parameters in the algorithm are optimized to maximize the Matthews Correlation Coefficient using a training set of events and non-events. Applied to the whole Helios data set, the trained algorithm generated a candidate set of events that were subsequently verified by hand, resulting in a database of 88 events. This approach offers a significant reduction in construction time for event databases compared to purely manual approaches. The database contains events covering a range of heliospheric distances from ∼0.3 to ∼1 au, and a wide variety of magnetic shear angles, but is limited by the relatively coarse time resolution of the Helios data. Analysis of these events suggests that proton heating by reconnection in the inner heliosphere depends on the available magnetic energy in a manner consistent with observations in other regimes such as at the Earth's magnetopause, suggesting this may be a universal feature of reconnection.