Chemometrics approach to FT-IR hyperspectral imaging analysis of degradation products in artwork cross-section

Ascertain the distribution of materials and that of their degradation products in historical artifacts is crucial to understand their conservation status. Among the different analytical techniques that can be used, FT-IR imaging supplies information on the molecular composition of the material on a micrometric-scale in a nondestructive way (i.e. respecting the physical integrity of the material/object and without inducing visible damage to the object. This is possible by limiting the sampling to very small amounts.) (K.H.A. Janssens, R. van Grieken, Non-destructive microanalysis of cultural heritage materials, Elsevier, 2004). When thin sections of the material are not exploitable for transmission, and when ATR imaging mode is not suitable due to possible damages on the sample surface, FT-IR imaging is performed in reflection mode on thick polished, matrix embedded samples. Even if many efforts have been done in the optimization of the sample preparation, the material's surface quality is a critical issue that can hinder the achievement of good infrared images. Moreover, spectral artifacts due to volume and surface interactions can yield uncertain results in standard data treatment. In this paper we address a multivariate statistical analysis as an alternative and complementary approach to obtain high contrast FT-IR large images from hyperspectral data obtained by reflection mu-FTIR analysis. While applications of Principal Component Analysis (PCA) for chemical mapping is well established, no clustering unsupervised method applied to mu-FTIR data have been reported so far in the field of analytical chemistry for cultural heritage. In order to obtain certain chemical distribution of the stratigraphy materials, in this work the use of Hierarchical Cluster Analysis (HCA), validated with a supervised Principal Component based k-Nearest Neighbor (PCA-kNN) Analysis, has been successfully used for the re-construction of the mu-FTIR image, extracting useful information from the complex data set. A case study (a patina from the Arch of Septimius Severus in the Roman Forum) is presented to validate the model and to show new perspectives for FT-IR imaging in art conservation. (C) 2017 Elsevier B.V. All rights reserved.