Dartmouth Events

Abstract: One of the central difficulties in interpreting energetic flaring events in the Universe (e.g., from black hole magnetospheres and pulsar wind nebulae) is our limited understanding of the particle acceleration which powers them. While collective plasma processes such as magnetic reconnection, shocks, and turbulence have been identified as primary engines of particle acceleration, their underlying microphysics—and thus radiative signatures—are poorly understood. A promising path forward is to cast the acceleration of individual particles in terms of the local mechanisms which accelerate them, e.g. magnetic X-points or Fermi reflections. This casting is only possible, however, if such mechanisms can be characterized merely by the “experience” they impart to particles. In this talk, I will argue that particle-in-cell (PIC) studies on particle injection in relativistic magnetic reconnection have supported this assumption directly, via reproducing key theoretical predictions, including the work done by each mechanism and their time-dependent dominance. I will conclude by discussing how unsupervised machine learning approaches, such as a vector-quantized variational autoencoder (VQ-VAE), may be applied to PIC tracer-particle trajectories to extract the underlying acceleration mechanisms, akin to discovering discrete phonetic units from raw audio. This opens the possibility for a descriptive bridge between the particle distribution function and the underlying microphysics to be constructed.

Hosted by Assistant Professor Jens Mahlmann

Join Zoom meeting

https://dartmouth.zoom.us/s/99960465228

Email Rowan.m.kowalsky@dartmouth.edu for passcode