Pure Isotropic Proton NMR Spectra in Solids using Deep Learning.

Area of Science:

• Solid-state Nuclear Magnetic Resonance (NMR) spectroscopy
• Computational chemistry and data science

Background:

• Proton solid-state NMR spectral resolution is often limited by spin-spin dipolar interactions.
• Magic-angle spinning (MAS) is a technique used to reduce this broadening through coherent averaging.
• Current MAS techniques at accessible spinning rates cannot fully eliminate dipolar interactions.

Purpose of the Study:

• To develop a deep learning approach for obtaining pure isotropic proton NMR spectra.
• To overcome the limitations of MAS in completely removing dipolar broadening.

Main Methods:

• A novel deep learning model was developed to process NMR data.
• The model analyzes two-dimensional magic-angle spinning (MAS) NMR spectra acquired at various spinning rates.
• The approach aims to extract pure isotropic proton spectra.

Main Results:

• The deep learning model successfully determined pure isotropic proton spectra.
• Application to eight organic solids demonstrated the model's efficacy.
• Achieved high-resolution 1H solid-state NMR spectra with narrow isotropic linewidths (50-400 Hz).

Conclusions:

• Deep learning offers a powerful solution to enhance resolution in proton solid-state NMR.
• This method significantly improves the quality of NMR spectra from organic solids.
• The approach provides a pathway to more detailed structural and dynamic information from solid samples.