This work introduces a novel method to multivariate analysis applied to fused hyperspectral datasets in the field of Cultural Heritage (CH). Hyperspectral Imaging is a well-established approach for the non-invasive examination of artworks, offering insights into their composition and conservation status. In CH field, a combination of hyperspectral techniques is usually employed to reach a comprehensive understanding of the artwork. To deal with hyperspectral data, multivariate statistical methods are essential due to the complexity of the data. The process involves factorizing the data matrix to highlight components and reduce dimensionality, with techniques such as Non-negative Matrix Factorization (NMF) gaining prominence. To maximize the synergies between multimodal datasets, the fusion of hyperspectral datasets can be coupled with multivariate analysis, with potential applications in CH. In this work, I will show examples of this approach with different combinations of datasets, including reflectance and transmittance spectral imaging, Fluorescence Lifetime Imaging and Time-Gated Hyperspectral Imaging, and Raman and fluorescence spectroscopy micro-mapping.

This work reports the analysis of the time-resolved photoluminescence behaviour on the nanosecond and microsecond time scale of fourteen historical and contemporary titanium white pigments. The pigments were produced with different production methods and post-production treatments, giving rise to a remarkable variability of titanium dioxide powders and, in some cases, to the formation of a complex surface of the crystal agglomerates. The pigments have been further characterized by Raman spectroscopy, scanning transmission electron microscopy coupled with energy dispersive X-ray spectroscopy and inductively coupled plasma atomic emission spectrometry. Our study provides a clear view of the main features of the photoluminescence (PL) emission of anatase- and rutile-based pigments. For both the polymorphs of titanium dioxide the room-temperature photoluminescence emission is complex and involves different relaxation paths, related to shallow levels close to the conduction bands and mid-gap trap states. The PL behaviour appears to be little affected by post-production treatments such as organic and inorganic coatings. Instead, the presence of niobium impurities in the TiO₂ crystal lattice, as residues of the sulphate synthesis process, induce a remarkable quenching of the visible emission of anatase-based pigments. We confirm that rutile-based and anatase-based pigments are significantly different in terms of photoluminescence behaviour. This clear distinction is a valuable point for non-invasive pigment identification by in-situ photoluminescence spectroscopy. In particular, while many organic binding media emit in the visible region, the near-infrared emission of rutile is specific and can likely be used to identify the pigment in more complex materials as paints. This research paves the way to future studies of the photo-physical properties of titanium white pigments, which is imperative to understand the risk of degradation induced by the well-known photocatalytic activity of this widely used 20^th century pigment.