Machine Learning Reconstruction of the Magnetotail Configuration to Support Simulations of the Tearing Instability and Magnetic Reconnection

Tearing instability is the main plasma mode responsible for the onset of magnetic reconnection in the Earth’s magnetotail, the key energy release process in the magnetosphere. Its free energy is provided by the mutual attraction of parallel current filaments in the tail current sheet while its dissipation is provided by the Landau resonance with electrons or ions resulting in the electron and ion tearing instabilities [1–2]. The major mystery for over forty years has been the nearly universal stability of the ion tearing mode [3], which was found unstable only for special magnetic flux distributions with accumulated flux regions and tailward gradients of the northward magnetic field component normal to the current sheet plane [4–5]. Otherwise, the tail must be squeezed to electron scales to release the magnetic energy [6]. To understand which of the scenarios works for the event of interest we apply the lazy learning algorithm of the k nearest neighbors (NN) to archives of spacebome magnetometer data from many missions over more than a quarter century [7]. NNs represent moments in the past that are closest to the event of interest in the state space formed by global indices of storm and substorm activity as well as the solar wind driving strength. A cloud of NNs in the state/input space produces a swarm of synthetic probes, of which only a few are the observations at the moment of interest. This swarm allows one to reconstruct the tail magnetic field with unprecedented details both before and after the reconnection onset [8]. We use this information to nudge the particle-in-cell (PIC) simulations of the magnetotail, as well as to compare the simulation results with the data-derived picture. The latter can also be used in global magnetohydrodynamic (MHD) simulations of the magnetosphere to supply them with the empirical resistivity maps derived from the reconstructed X-line distributions in the tail.