• Laurent, Guillaume P.

Abstract

Solid-state Nuclear Magnetic Resonance (NMR) is suffering from an intrinsic low sensitivity, despite recent improvements. Instrumentation, fast acquisition and signal processing approaches were investigated to circumvent this drawback as far as possible. Firstly, microcoils (Magic Angle Coil Spinning, MACS) were placed into rotors and inductively coupled to the standard probe coil. A time gain of ~ 5 was obtained for microquantities with a mass m ~ 100-200 µg. Secondly, acquisition time was decreased by mean of Carr-Purcell-Meiboom-Gill (CPMG) echoes for direct acquisition. Adequate processing is required to get the best enhancement from this technique. We provided a Python software to process data either using standard spikelets or superposition methods, or with a denoising method. A time gain of ~ 3-100 was possible. Thirdly, Non-Uniform Sampling (NUS) was chosen as a way to decrease acquisition time of indirect dimensions of multi-dimensional experiments. Poisson sampling revealed to be the best choice to limit artefacts, whereas hybrid sampling proved to be efficient on spectra with both broad and narrow peaks. A time gain of ~ 4 was achieved. Fourthly, spectra were processed with Singular Value Decomposition (SVD) denoising. We highlighted an overestimation of Gaussian peaks by ~ 20 %. Automatic thresholding was implemented, giving a time gain of ~ 2.3. Finally, computation time wasdecreased by ~ 100 by combining ‘divide and conquer’ algorithm, optimised libraries, hardware instruction calls and single precision. A comparison between Central Processing Units (CPU) and Graphical Processing Units (GPU) was provided.

Topics

• impedance spectroscopy
• experiment
• Nuclear Magnetic Resonance spectroscopy
• decomposition
• spinning