ToF-SIMS IMAGING AND DATA MINING APPROACHES FOR ORGANIC STRUCTURED SAMPLES

The Materials Science research and development permitted the progress of technology, considering the variety of its applications in each scientific area, including nanotechnology, metallurgy, biomaterials, and cultural heritage. The material investigation and characterization are fundamental for the scientific progress and the chemical imaging can give a real contribution in this direction. As a matter of fact, the new frontiers in microelectronics, disease detection and treatment, and chemical manufacturing demand an ability to visualize and understand molecular structures, chemical composition, and interactions and reactions in materials, even below micrometer order. Considering the growing interest towards micro- and nano- structure systems, the importance of chemical imaging techniques became evident. In fact, not only do these techniques allow to achieve a high lateral resolution able to provide detailed information, but also, they provide molecular and structure information about the samples. Undoubtedly, the investigation of the materials’ structures often requires the support of computer modelling and other advanced experimental methodologies. Nowadays, several kinds of techniques are available for chemical imaging, but they often cannot provide at the same time high spatial resolution and good molecular identification. Mass spectrometry techniques represent an excellent tool for reaching this goal. Indeed, MS imaging is and will increasingly become fundamental for many aspects of Materials Science, thanks to its ability in providing high resolution and sensitivity molecular-weight-specific images. Up to the 70s and 80s of the 20th century, the use of ToF-SIMS technique was mainly confined to the characterization of inorganic materials, but more recently, especially after the introduction of instrumentation upgrades, it has spread in many other fields, including biology, disease investigation and cultural heritage, showing very promising results in 2D and 3D imaging. However, this technique still presents some problems, especially in the case of organic and hybrid samples, due to the primary ion induced damage, the resulting sputter yield decrease and the produced big data treatments. Therefore, a big challenge is related to the secondary ion signal intensity, which is often low and does not allow to obtain a good image quality. A possible strategy to overcome this issue is the use of post-analysis data mining treatments. This approach requires computing modelling and programs able to handle limited signals and extrapolating to a higher quality result. This thesis work was focused on the ToF-SIMS imaging of organic “model” samples, which allowed to experiment different instrumentation setting to achieve the best performances with all kinds of materials. More specifically, four model samples were selected, with completely different chemical composition and structure, chemical-physical properties, and application fields. The investigated samples were a Solid Lipid Nanoparticles’ solution, PET-PC blends, dyes on 19th century textiles and fingerprints. For all samples, several issues were overcome which concerned the preparation of a solid sample suitable for ultra-high vacuum and the difficulty of obtaining and interpreting the molecular signals. Some post analysis data treatments are presented for the handling and the interpretation of the acquired data. Moreover, in this thesis a probabilistic neural network (PNN) approach was proposed, capable of extracting latent chemical information from ToF-SIMS data, working on uncompressed and unbinned raw data sets. The thesis results show that, for all investigated samples, it was possible to extract molecular information, as well as to study samples’ structure and morphology, thanks to the ToF-SIMS high mass and spatial resolution and post-analysis data treatments approaches. Although in some cases a rigorous molecular identification was not possible, a complete imaging characterization was achieved for all samples. It was also demonstrated that the PNN approach can extract chemical information from single pixel ToF-SIMS spectra, which are not otherwise interpretable, thus, speeding the acquisition and the interpretation of the chemical images. In conclusion, even though ToF-SIMS is quite an old technique, widely dealt with in specific literature, there are still unexplored fields of investigation in which this technique can be essential to get information which could not be extracted by other techniques.