Development of a chemometric methodology based on FTIR spectra for paper dating

The dating of documents is one of the pending issues to be resolved in many fields. Although there are a variety of analytical methodologies focused on the inks of documents to estimate the dating of documents, the analysis of documents itself attracted little attention. A non-destructive method based on Fourier transformed infrared (FTIR) was proposed in this research to estimate the dating of documents. First, FTIR spectra of journals dated from 1940 to 1980, naturally aged and conserved in the China Agriculture University library were collected. Second, the least squares support vector machines (LS-SVM) was applied to distinguish documents of different dating, and the accuracy was 99.26%. In order to find interested wavenumber that influence the dating process of documents, sparse partial least squares (sPLS) was applied to select informative variables. The average of selected variables was 483 after 100 runs, and the selected variables were focused on the absorption peaks of inorganic components and cellulose. Splicing sPLS with LS-SVM (sPLS–LS-SVM) built model to see the effective of selected variables. Average accuracy of sPLS–LS-SVM model was 99.34%, even the best result could reach 100.00% after 100 circle times. The present work indicates that the possibility of FTIR combined with chemometrics can estimate the dating of documents accurately. Additionally, the wavenumber which influence the dating of documents are mostly focused on cellulose and inorganic components.