Spin Echo Simulations and Supervised Machine Learning

Using a numerical simulation of a phenomenological model of nuclear spin-spin interaction mediated by collective electronic spin excitations, we study the effects of such an interaction on nuclear magnetic resonance spin echo experiments. These local interactions can result in asymmetric spin echoes and unique time domain signals that are related to, and dependent on, the various parameters of the interaction. Simulating a large number of these echoes for different interaction parameters enables the use of supervised machine learning to analyze and “reverse engineer” the parameters of the interaction, demonstrating that the approach can be used to extract information about the form of the local interaction from the time domain spin echo signal.