Signal Deconvolution and Generative Topographic Mapping Regression for Solid-State NMR of Multi-Component Materials.

Area of Science:

• Materials Science
• Spectroscopy
• Data Analysis

Background:

• Solid-state nuclear magnetic resonance (ssNMR) spectroscopy is crucial for understanding native structures and dynamics in solid materials.
• Analyzing multi-component solid materials using ssNMR is challenging due to broad and overlapping spectral signals.
• Effective signal deconvolution and prediction are essential for accurate ssNMR analysis.

Purpose of the Study:

• To investigate signal deconvolution methods (STFT, NTF, NMF) for complex ssNMR spectra.
• To explore generative topographic mapping regression (GTMR) for predicting NMR signals and material properties.
• To demonstrate the application of these methods in analyzing macromolecular samples.

Main Methods:

• Short-time Fourier transform (STFT) for signal deconvolution.
• Non-negative tensor/matrix factorization (NTF, NMF) for spectral separation.
• Generative topographic mapping regression (GTMR) for signal and property prediction.

Main Results:

• STFT and NTF successfully separated ssNMR spectra of cellulose degradation samples into distinct components (cellulose, proteins, lipids).
• GTMR accurately predicted cellulose degradation products (acetate, CO2) and computed NMR signals from physical properties of polylactic acid.
• ssNMR spectra of poly-ε-caprolactone were resolved into crystalline and amorphous signals using these methods.

Conclusions:

• The developed signal deconvolution and prediction methods effectively address the challenge of overlapping spectra in ssNMR.
• These techniques enhance the characterization of macromolecules and support the design of novel solid materials.
• The study showcases the utility of STFT, NTF, and GTMR in diverse applications including biopolymers and plastics.