Multivariate Calibration of a Photodiode Array Spectrometer for Atomic Emission Spectroscopy

Abstract

Multiple linear regression with stepwise selection of variables is used to calibrate a photodiode array spectrometer for inductively coupled plasma-atomic emission spectrometry. The spectrometer uses a preliminary wavelength selector and an echelle grating (without an order-sorting prism) to provide high-resolution information over a wide wavelength range. Chemical systems exhibiting various obstacles to quantitation, including spectral-line overlap, interference due to broadened spectral lines, and stray light from a line of high intensity, are calibrated successfully with a multivariate technique. Sample sets used for calibration are constructed with the use of simplex lattice designs, and the method of cross-validation is used to assess the predictive performance of the calibration models. Conventional univariate calibration using spectral peak areas is performed for comparison.