Principal components analysis for the visualisation of multidimensional chemical data acquired by scanning Raman...

Raman microspectroscopy is ideally suited to surface analysis as it allows detailed chemical information to be acquired from surfaces at a relatively high spatial resolution (typically 1 microm). Using a motorised sample table or probe, it is possible to raster scan a surface to obtain spatially resolved chemical information. Visualisation of the acquired data is a problem, however, as the spectrum acquired at each point can contain several hundred individual intensity measurements. Existing visualisation methods are limited to plotting each scanned point with an intensity determined from the measured intensity at a single wavenumber, or the similarly between the point's spectrum and a reference spectrum. Such methods are wasteful as a lot of acquired information is discarded, and results are prone to misinterpretation due to background variance and instrumental noise. In this paper we introduce a new method that uses principal components analysis (PCA) to reduce the spectrum at each point to three factors that are then used to define the red, green and blue components of the corresponding point on a false colour map. To increase the effective resolution, interpolation is used to approximate the colours corresponding to points between those actually scanned. To demonstrate the technique, the internal surface of a beverage can, contaminated with a 40 microm diameter carbonised oven impurity, consisting mainly of sp2- and sp3-hybridised saturated carbon bonds, has been used as a case study.