Several spectral bands or regions often are required for contrast generation to utilize the full power Raman imaging for mapping compositional or physical property variations. To achieve this, it is most efficient to use a high throughput imaging spectrograph with a two-dimensional detector operated in pushbroom mode as a line imager. Despite data from several bands, univariate data reduction techniques are inadequate when characterizing complex materials. Multivariate statistical procedures, such as factor analysis, are useful for defining weighted sums of intensities from many Raman shifts (frequencies) and this in turn expresses chemically or structurally useful contrast.

We illustrate this approach with silicate and aluminosilicate glasses. In most cases, no single band can unambiguously identify a particular glass composition. If the glass is heterogeneous, then localized compositional and structural differences may be detected by small changes in several bands, all arising from motions of assorted Si-O moieties.