Reliability of the contribution profiles obtained through the SIMPLISMA Approach used as reference data in a calibration process. Application to Raman micro-analysis of dust particles
Résumé
Previous works evidenced that industrial dust samples analyzed are often aggregates of submicronic, heterogeneous particles. As a consequence, even when analyzed by Raman microspectrometry, mixtures of different compounds are observed. Such an observation motivated the use of a self-modeling mixture analysis to determine pure species present. That is the reason why the simple-to-use interactive self-modeling mixture analysis (SIMPLISMA) was used to resolve Raman imaging data obtained from the analysis of dust samples collected in the vicinity of lead and zinc smelters. Results provided the Raman spectra and respective contributions of both major and minor chemical species among which PbSO4 was always found. It was tried here to test the reliability of a given PbSO4 contribution profile for its application to data originating from other PbSO4-containing samples. This was attempted in two steps using partial least square (PLS) calibration method. Firstly, a calibration model was constructed with mean-centered data between the Raman spectra and their respective, relative contribution resolved using the SIMPLISMA approach. Secondly, the model obtained was tested in order to assess the reliability of the PbSO4 contribution profile used on different Raman imaging data of dust samples. The aim of this work is to predict the PbSO4 amount in different dusts without going through the step of SIMPLISMA resolutions. On-line monitoring of PbSO4 could thus be performed and, to a greater extent, pollutant dust concentration could be followed for a given period of time simply on the basis of PLS prediction.