Characterizing Magnetic Reconnection Regions Using Gaussian Mixture Models on Particle Velocity Distributions

Romain Dupuis,Martin V. Goldman,David L. Newman,Jorge Amaya,Giovanni Lapenta

DOI: https://doi.org/10.3847/1538-4357/ab5524

IF: 5.521

2020-01-21

The Astrophysical Journal

Abstract:We present a method based on unsupervised machine learning to identify and characterize regions of interest using particle velocity distributions as a signature pattern. An automatic density estimation technique is applied to particle distributions provided by particle-in-cell simulations to study magnetic reconnection regions. Its application to magnetic reconnection is new. The key components of the method involve (i) a Gaussian mixture model determining the presence of a given number of subpopulations within an overall population, and (ii) a model selection technique with a Bayesian information criterion to estimate the appropriate number of subpopulations. Thus, this method automatically identifies the presence of complex distributions, such as beams or other non-Maxwellian features, and can be used as a detection algorithm able to identify reconnection regions. The approach is demonstrated for a specific double Harris sheet simulation, but it can in principle be applied to any other type of simulation data on the particle distribution function.