Inductively coupled plasma-atomic emission spectrometer warning diagnosis procedure using blank solution data

Abstract

Lines available while running a blank solution were used to monitor the analytical performance of an inductively coupled plasma atomic emission spectrometry (ICP-AES) system in real time. Using H and Ar lines and their signal-to-background ratios (SBRs), simple rules in the form of a prediction table were developed by inspection of the data. These rules could be used for predicting changes in radio-frequency power, carrier gas flow rates, and sample introduction rate. The performance of the prediction table was good but not excellent. Another set of rules in the form of a decision tree was developed in an automated fashion using the C4.5 induction engine. The performance of the decision tree was superior to that of the prediction table. It appears that blank spectral information can be used to predict with over 90% accuracy when an ICP-AES is breaking down. However this is not as definitive at identifying the exact fault as some more exhaustive approaches involving the use of standard solutions.