Study of a Partial Least-squares Regression Model for Rare Earth Element Determination by Inductively Coupled Plasma Mass Spectrometry

Abstract

A partial least-squares regression (PLSR) model was developed for rare earth element (REE) determination by inductively coupled plasma mass spectrometry (ICP-MS), in order to correct interferences from REE oxides, hydroxides and isobaric spectral overlap. The total variance was explained by a 14 factor PLSR model. The square error of prediction was less than 0.005 and the oxide/hydroxide/isobaric interferences were almost completely removed. It was found that Nd, Sm, Gd, Dy and Yb played a more significant role in the model than the other REEs because they possess multiple isotopes which require repeated calibration. REE concentrations could be accurately predicted despite barium interference, even in samples with high ratios of Ba to REE, when the weights of certain isotopes in the model were set below 0.1. The PLSR model was compared with the normal calibration method (NCM) and the Gauss elimination method (GEM). The results indicated that the PLSR model was more accurate than the NCM and exhibited greater flexibility than the GEM.