Among spectroscopies, Nuclear Magnetic Resonance (NMR) is a very precise local probe. However, its poor sensitivity is a major drawback, as only one nucleus over 10^5 can be detected in standard conditions. In addition, the absence of fast molecular motion in solid - state NMR leads to considerable peak broadening. In such a case, it is especially difficult to have sufficient signal - to - noise ratio (SNR) for quantitative results. During this PhD, three different innovative strategies have been implemented to improve the sensitivity limit: instrumentation [1] , fast acquisition [2] and post – processing [3 – 4]. First, we used a fast acquisition technique called "Non - Uniform Sampling" (NUS), which is based on the assumption that spectra are sparse. It is thus not necessary to respect the Nyquist - Shannon theorem during multi - dimensional spectra acquisition. Using reconstruction algorithms, data can be classically analyzed. We successfully applied it to solid - state NMR, where the difficulty is that spectra are not so sparse. An experimental time decrease by a factor 4 on two - dimensional spectra has been obtained [2]. Secondly, when poor spectra are recorded despite a week of acquisition, the only solution is to apply post - processing. We used a denoising algorithm named "Singular Value Decomposion" (SVD). Our study proved that while Lorentzian peaks are correctly denoised, Gaussian peaks needed to be taken with care. An experimental time decrease by a factor 2.3 has been achieved [3]. Moreover, SVD is a computational intensive task. We optimized it using performant algorithms and graphic cards to decrease computation time by a factor 100 [4] In this study, we focused on instrumentation, in order to increase the signal strength as close as possible from the sample. For this, we used Magic Angle Coil Spinning (MACS), which is a microcoil placed inside a standard rotor, spun at Magic Angle Spinning (MAS). Such a design is especially convenient for very small sample quantities, around 100 μg. This corresponds for instance to the amount available when scratching a thin film deposited on a substrate. By electromagnetic coupling between the probe coil and the microcoil, it is possible to decrease experiment time by a factor 7. Two designs were investigated: flat coils (2D MACS) [5] and solenoids (3D MACS) [6] and low-power experiments were implemented. References: 1. Laurent G., Instrumenter et innover en chimie physique pour préparer l’avenir , Paris, France, hal - 01139041 (2015) 2. Laurent G., RMN structurale dans le bassin parisien , Orléans, France, hal - 01745319 (2018) 3. Laurent G., Woelffel W., Barret - Vivin V ., Gouillart E., Bonhomme C., Appl. Spectrosc. Rev , doi 10.1080/05704928.2018.1523183 (2019) 4. Laurent G., Gilles P. - A., Woelffel W., Barret - Vivin V., Gouillart E., Bonhomme C., Appl. Spectrosc. Rev , doi 10.1080/05704928.2018.1559851 (2019) 5. Lehmann - Horn J. A., Jacquinot J. - F., Ginefri J. C., Bonhomme C., Sakellariou D., J. Magn. Reson. , 271, 46 – 51, (2016) 6. Sakellariou D., Le Goff G., Jacquinot J. - F., Nature , 447, 7145, 694 – 697, (2007)