THE MATERIALS PROJECT--CREATING A DATABASE OF NMR TENSORS FOR MACHINE LEARNING

NMR spectroscopy gives precise information about the local bonding of materials, but spectra can be difficult to interpret, unless there is a set of well-vetted reference spectra. In addition, the ability to predict NMR tensors (both spin-1/2 and quadrupolar) and connect them to three-dimensional local environments is critical for understanding more complex, long-range interactions. In this presesntation, we will discuss computational methods that have revealed structural information for 29Si solid-state NMR by generating computed reference spectra for solids. Such predictions are useful for the identification of new silicon-containing compounds, and serve as a starting point for determination of the local environments present in amorphous structures. We have used a benchmarking set to compare CASTEP and VASP density functional theory (DFT) calculations. These have been compared to the full chemical shielding anisotropy (CSA) tensor for the spin-1/2 29Si species. Through this examination, we have established a robust and systematic method for assigning the diagonalized chemical shift/shielding tensor values. Armed with confidence in this benchmarking set, over 8,000 29Si NMR shielding tensors are available via the LSDI portion of The Materials Project. In addition, a benchmarking set for 27Al is currently under development to study this quadrupolar species. These tensors will be a guide to researchers when searching for NMR assignments, as well as a platform that can assist with experimental conditions, since the appearance of spectra can be anticipated prior to measurement.Finally, we have also reported on the combination of solid-state NMR measurements with density functional theory (DFT) calculations to demonstrate how hydrogen positions can be refined in a crystalline system. A match between the calculated structure and solid-state NMR was found by testing multiple semi-local and dispersion-corrected DFT functionals and applying them to optimize atom positions, resulting in a revised ICSD crystal structure. The results show that the combination of solid-state NMR, XRD, and DFT can improve structure refinement, for crystalline materials.