Machine Learning Mapping Approach for Computing Spin\nRelaxation Dynamics

In\nthis work, a machine learning mapping approach for predicting\nthe properties of atomistic systems is reported. Within this approach,\nthe atomic orbital overlap, density, or Kohn-Sham (KS) Fock matrix\nelements obtained at a low level of theory such as extended tight-binding\nhave been used as input features to predict the electric field gradient\n(EFG) tensors at a higher level of theory such as those obtained with\nhybrid functionals. It is shown that the machine-learning-predicted\nEFG tensors can be used to compute spin relaxation rates of several\nions in aqueous solutions. From only a fraction of data used in direct\ncalculation, one can predict the quadrupolar isotropic spin relaxation\nrates with good accuracy, achieving relative errors between about\n2–8% for different ions.