Theoretical and Machine Learning Methods in Extracting Spin-Lattice Relaxation for Ammonia

In an attempt to improve target polarization calibration measurement,

a considerable amount of time for experimental data taking is lost. To

optimize this, it is perhaps possible to determine the thermal equilibrium

polarization while the enhanced signal is not fully relaxed. To do this

using artificial intelligence, high-quality training data for the spin-lattice

relaxation (T1) is required at different temperatures and different accu-

mulated doses for a specific polarized target material such as NH3. In this

presentation, we find a theoretical T1 value for solid NH3 at 1 Kelvin in

a 5 Tesla magnetic field using basic principles of the NH3 cell geometry

and nuclear spin interactions. We construct a model of the NH3 super cell

and used known expressions for the transition rates between spin states

and the dipolar interaction tensor to find the relaxation matrix of the sys-

tem. We then discuss the calculation of the correlation time, and compute

the eigenvalues of the relaxation matrix to find the T1 time. This model

will be parameterized with real experimental data and used to generate

training data for further studies of thermal equilibrium analysis.