MIXTURE ANALYSIS USING MEMBRANE INTRODUCTION MASS-SPECTROMETRY AND MULTIVARIATE CALIBRATION

Membrane introduction mass spectrometry has been coupled with multivar iate calibration for the analysis of volatile organic compound (VOC) m ixtures, Three mixtures of increasing complexity were modeled using mu E tivariate calibration methods. Multivariate calibration models built using partial least-squares (PLS) regression were compared with univa riate calibration for the analysis of a benzene, toluene, and p-xylene (BTX) mixture and an ethylbenzene and p-xylene (EX) mixture, The univ ariate and multivariate calibration methods performed similarly for th e BTX mixture with prediction errors of <15%. For the isomer EX mixtur e, the PLS model outperformed the univariate model having prediction e rrors between 8 and 13% compared to errors of 40% for the univariate m odel, The third mixture, containing all four analytes (benzene, toluen e, ethylbenzene, p-xylene), was modeled using PLS and resulted in cali bration relative errors of 6-15% and prediction relative errors of 3-4 8%, This paper demonstrates the feasibility of using membrane introduc tion mass spectrometry with multivariate calibration for the analysis of complex VOC mixtures.