A machine learning approach to structure and energy in magnetic reconnection

Magnetic reconnection is a fundamentally important process in space plasmas due to the release and repartition of the magnetic energy stored within the reconnecting field. As this energy transfer significantly impacts magnetospheric dynamics, understanding the partition of this energy and how this varies across the reconnection site can provide further insight into other magnetospheric processes. Although in situ spacecraft data provide direct measurements of relevant plasma properties, it can be difficult to establish the location of spacecraft relative to the reconnection site. This frustrates efforts to evaluate the way in which energy fluxes change with distance from the central reconnection X-line. Under certain circumstances, reconstruction techniques can be used to estimate the spacecraft trajectory through individual events, but these may rely on simplifying assumptions limiting their use.

This motivates new approaches to determining where a spacecraft is relative to the reconnection structure. By utilising forefront machine learning techniques, we can more accurately study individual regions associated with the reconnection process and thus understand how they individually contribute to repartitioning the overall energy budget. In this context, we present these new applications of machine learning techniques to identify the regions in both simulation and spacecraft data.

Firstly, we present the results of a robust method which utilises k-means clustering to identify different regions encountered within the overall reconnection X-line structure. This uses plasma fluid and field variables output by a 2.5-D PIC simulation with a geometry comparable to that of reconnection in Earth’s magnetotail. We then translate this model for use in spacecraft data by implementing an approach based on a recurrent neural network to account for the temporal context of the observations. We demonstrate the use of this model on MMS observations of reconnection in the Earth’s magnetotail, examining the properties of the plasma energy flux in different regions. We conclude by discussing how this approach may find use in other contexts where reconnection is observed in space plasmas.