Semi-empirical approach to the calculation of instrumental detection limits in inductively coupled plasma atomic emission spectrometry

Abstract

The instrumental LOD in ICP-AES can be calculated from the sensitivity and the two principal noises that contribute to the standard deviation of the blank,σ_B. The only inputs required are a single measurement of a known concentration of the analyte, a single measurement of the blank signal and the RSD of the signal from a relatively high-concentration Mn solution at the 257.61 nm line. The sensitivity is calculated from the mean signal for the known concentration of analyte and the RSD of the Mn 257.61 nm signal gives an estimate of the source flicker noise. If desired, the Mn solution can be matrix matched to suit the specific analysis, but this may not be necessary. The RSD of the analyte solution can be used to estimate the source flicker noise; however, the RSD of a Mn solution is preferred because it has been shown experimentally to be very similar to the flicker noise component in the blank for a range of matrices. Once the source flicker noise has been estimated, it can be presumed constant unless some parameter likely to affect it has been changed. With constant flicker noise, only a single measurement of the analyte solution is required, to measure the sensitivity. A single measurement of the blank then gives the information needed for the calculation of the shot noise contribution to σ_B. The quadratic sum of the source flicker noise and the shot noise provides an acceptably accurate estimate of σ_B at any specified integration time. This, with the measured sensitivity and blank signal, allows the calculation of the RSDB and the limit of detection. The method allows a considerable saving of time compared with the usual method of repetitively measuring blanks and can be applied inversely to the diagnosis of spectrometer noise characteristics and for the optimisation of ICP instruments. The degree of improvement obtained by such optimisation can, of course, be small if the system is already close to the optimum.