Machine Learning Extraction of Tensor Polarization in Spin-1 Observables

Tensor polarization enhancement using radio frequency manipulation of the spin-1 line shape can improve the figure of merit of specialized tensor polarization observables in scattering experiments, improving the overall figure of merit. This improvement is optimized only if careful manipulation and measurement are synchronized to near-real time so that the RF manipulation can be applied after each NMR sweep. This presentation discusses the measurement scheme designed to reduce uncertainties in the tensor polarization while maximizing the enhancement as well as a novel machine learning approach designed to improve inference speed, accuracy, and precision.