Evaluation of MidIR fibre optic reflectance: detection limit, reproducibility and binary mixture discrimination.

MidIR fibre optic reflectance (MidIR-FORS) is a promising analytical technique in the field of science conservation, especially because it is non-destructive. Another advantage of MidIR-FORS is that the obtained information is representative, as a large amount of spectral data can be collected. Although the technique has a high potential and is almost routinely applied, its quality parameters have not been thoroughly studied in the specific application of analysis of artistic materials. The objective of this study is to evaluate the instrumental capabilities of MidIR-FORS for the analysis of artwork materials in terms of detection limit, reproducibility, and mixture characterisation. The study has been focused on oil easel painting and several paints of known composition have been analysed. Paint layers include blue pigments not only because of their important role along art history, but also because their physical and spectroscopic characteristics allow a better evaluation of the MidIR-FORS capabilities. The results of the analysis indicate that MidIR-FORS supplies a signal affected by different factors, such as the optical, morphological and physical properties of the surface, in addition to the composition of materials analysed. Consequently, the detection limits established are relatively high for artistic objects (Prussian blue - PB 2.1-6.5%; Phthalocyanine blue - Pht 6.3-10.2%; synthetic Ultramarine blue - UM 12.1%) and may therefore lead to an incomplete description of the artwork. Reproducibility of the technique over time and across surface has been determined. The results show that the major sources of dispersion are the heterogeneity of the pigments distribution, physical features, and band shape distortions. The total dispersion is around 4% for the most intense bands (oil) and increases up to 26% when weak or overlapped bands are considered (PB, Pht, UM). The application of different pre-treatments (cutoff of fibres absorption, Savizky-Golay smoothing algorithm, polynomial baseline offset, Standard Normal Variate algorithm - SNV) to the raw spectra allows improving these results to maximum values of 15%. Finally, the capabilities of PCA and MidIR-FORS to discriminate between binary mixtures were tested. The results demonstrate that it is possible to differentiate mixtures depending on the range of concentration of their components, within specific limits of detection.