High-Quality Reconstruction for Laplace NMR Based on Deep Learning

Laplace nuclear magnetic resonance (NMR) exploits relaxationanddiffusion phenomena to reveal information regarding molecular motionsand dynamic interactions, offering chemical resolution not accessibleby conventional Fourier NMR. Generally, the applicability of LaplaceNMR is subject to the performance of signal processing and reconstructionalgorithms involving an ill-posed inverse problem. Here, we proposea proof-of-concept of a deep-learning-based method for rapid and high-qualityspectra reconstruction from Laplace NMR experimental data. This reconstructionmethod is performed based on training on synthetic exponentially decayingdata, which avoids a vast amount of practically acquired data andmakes it readily suitable for one-dimensional relaxation and diffusionmeasurements by commercial NMR instruments.