Machine learning full NMR chemical shift tensors of silicon oxides with equivariant graph neural networks

The nuclear magnetic resonance (NMR) chemical shift tensor is a highly sensitive probe of the electronic structure of an atom and furthermore its local structure. Re- cently, machine learning has been applied to NMR in the prediction of isotropic chemi- cal shifts from a structure. Current machine learning models, however, often ignore the full chemical shift tensor for the easier to predict isotropic chemical shift, effectively ignoring a multitude of structural information available in the NMR chemical shift ten- sor. Here, we use an equivariant graph neural network (MatTEN) to predict full 29Si chemical shift tensors in silicate materials. The MatTEN model predicts full tensors to a mean absolute error of 1.05 ppm and is able to accurately determine the magnitude, anisotropy, and tensor orientation in a diverse set of silicon oxide local structures. When compared with other models, MatTEN outperforms the state-of-the-art ma- chine learning models by 47%. MatTEN also outperforms historic analytical models by 57% for isotropic chemical shift and 91% for anisotropy. The software is available as a simple-to-use open-source repository, allowing similar models to be created and trained with ease.