Analysis of the effect of crystal size and color distribution on fluorescence measurements of solid sugar using chemometrics

Citation
D. Baunsgaard et al., Analysis of the effect of crystal size and color distribution on fluorescence measurements of solid sugar using chemometrics, APPL SPECTR, 54(11), 2000, pp. 1684-1689
Citations number
23
Categorie Soggetti
Spectroscopy /Instrumentation/Analytical Sciences
Journal title
APPLIED SPECTROSCOPY
ISSN journal
00037028 → ACNP
Volume
54
Issue
11
Year of publication
2000
Pages
1684 - 1689
Database
ISI
SICI code
0003-7028(200011)54:11<1684:AOTEOC>2.0.ZU;2-#
Abstract
Fluorescence from sugar crystal samples has previously been used to obtain information about factory imprint and sugar quality. Solid-phase fluorescen ce has potential as a fast screening method, but the spectra are highly inf luenced by the measurement geometry and sugar crystal sample. The aim of th e present study was to examine how the fluorescence measurements are relate d to the sugar crystals for a better understanding of both. Initially, five sugar samples of varied composition were sieved into five crystal size fra ctions. Fluorescence excitation-emission landscapes of the fractions were m easured with solid transmission and reflection techniques and in solution. The transmission fluorescence was quenched at ultraviolet wavelengths, and light scatter highly influenced the reflection fluorescence. Principal comp onent analysis (PCA) showed that large crystals favored the transmission fl uorescence, whereas smaller crystals improved the reflection fluorescence m easurements. The multi-way method PARAFAC (parallel factor analysis) was us ed to resolve spectra of individual components from the fluorescence landsc apes. Transmission and solution components had similar spectral profiles at higher wavelengths, characterizing a colorant and a colorant intermediate. The resolved components of the reflection data were very influenced by sca tter. Color predictions based on a few significant wavelength variables equ aled the model results of full-spectrum models using partial least-squares regression (PLS). The variables corresponded to wavelength maxima of the re solved colorants and ultraviolet wavelengths characterizing colorant precur sors.