Ab Initio Quality NMR Parameters in Solid-State Materials Using a High-Dimensional Neural-Network Representation.

Area of Science:

• Solid-state chemistry and materials science
• Computational chemistry and spectroscopy

Background:

• Nuclear magnetic resonance (NMR) spectroscopy is crucial for analyzing local atomic order in solids.
• Interpreting NMR spectra, especially for amorphous materials, is challenging due to spectral complexity.
• Traditional methods combining molecular dynamics and ab initio calculations are limited by system size and statistical sampling.

Purpose of the Study:

• To develop an efficient and accurate method for predicting NMR parameters in large solid-state systems.
• To overcome the computational constraints of existing simulation approaches for amorphous materials.
• To enable quantitative description of local environments through improved statistical sampling.

Main Methods:

• Utilized a high-dimensional neural-network representation of NMR parameters.
• Calculated NMR parameters using an ab initio formalism.
• Applied the neural-network NMR (NN-NMR) method to (17)O and (29)Si in crystalline silica polymorphs and silica glasses.

Main Results:

• Demonstrated an efficient and accurate prediction of NMR parameters for very large systems.
• Successfully applied the NN-NMR method to analyze quadrupolar coupling and chemical shift parameters in various silica structures.
• Validated the potential of the NN-NMR approach for diverse solid-state materials.

Conclusions:

• The NN-NMR method offers a computationally efficient way to predict NMR parameters for large systems.
• This approach significantly enhances the interpretation of NMR spectra for complex and amorphous materials.
• The NN-NMR method is general and applicable to predicting NMR properties across a wide range of materials.