In order to determine new discriminating criteria for the
organic substances, multivariate analyses were carried out on
the reference materials as described in paragraph 2.3.The results
of PCA calculation using the CH spectral region data points as
variables with a 2 cm−1 step are presented Fig. 4. It illustrates the
distribution of the 37 reference materials on the PC1 and PC2 score
plot. The glues can form a cluster thanks to the PC1 (negative
contribution); the oils are all on one side of the PC1 (between
10 and 20). However, the resins are much dispersed and seem to
be hardly separated with PC1 but quite sorted with PC2. Most of
the triterpenic resins have a negative contribution on PC2 and the
diterpenic are from 0 to 15 on PC2. However, this is insufficient to
consider that the different materials are well sorted or clustered.
Indeed, some samples are quite dispersed: for example pine resins
(colophony and Venice turpentine) and others are clustered but
among other materials; for instance frankincense samples are
“mixed” with the diterpenic resins. Moreover some of the refer-
ence samples are processed: for example, two samples of coloph-
ony are in a raw form and one is a colophony varnish, same for the
oils, which explain the oils dispersion. Material transformation or
alteration is consequently an obstacle to agood clustering when
the variables for PCA are spectra data points.