AN OBJECTIVE ASSESSMENT OF ANALYTICAL METHOD PRECISION - COMPARISON OF ICP-AES AND XRF FOR THE ANALYSIS OF SILICATE ROCKS

The precision of an analytical method has been evaluated objectively b y applying the method of Thompson and Howarth (1976) to the analysis i n duplicate of 55 igneous rocks covering a range of silicate matrix ty pes and analyte concentrations. Results were analysed to characterise the change in precision (S-c) of the analytical method with concentrat ion (c) according to the equation S-c=S-o+kc, where the k parameter re presents the limiting high-level precision and S-o, the precision at z ero concentration, which is related to the method detection limit (MDL ), Test materials were analysed using four analytical methods based on two analytical techniques, inductively coupled plasma-atomic emission spectrometry (ICP-AES) and X-ray fluorescence spectrometry (XRF), as operated under routine working conditions in the two participating lab oratories. The two XRF methods were major elements on fused glass disc s and trace elements on powder pellets, and the two ICP-AES methods we re major elements after a fusion decomposition technique and trace ele ments together with selected major elements, after an acid attack. Sta tistical evaluation of the data showed that significant changes in pre cision as a function of concentration (i.e. the k factor) were determi ned in 34 cases out of 78 analyte-method combinations. In cases where no significant change in precision could be detected, a grand mean pre cision, representative of the concentration range analysed was calcula ted. The S-o parameter was found to be significantly different from ze ro in 36 cases out of 72. To allow evaluation of the detection limit p erformance of all data, a maximum method detection limit (MMDL) was ca lculated, which was estimated to be on average 1.62 times greater than the MDL derived from significant values of S-o. In terms of the four methods studied, median high-level precision of the techniques used to determine major elements were found to be 0.23% relative (XRF/glass d iscs), 0.43% (ICP-AES/fusion decomposition) and 0.70% (ICP-AES/acid at tack). Typical precision values in the determination of trace elements by both techniques was 1.5%, providing elemental concentrations exten ded over a significant range. MMDL's varied from element to element bu t for XRF/powder pellet data were found to be approximately equivalent to instrumental detection limits (IDL's) calculated from background c ount rates. However, for trace elements determined by ICP-AES/acid att ack, MMDL were found to be on average three times larger than IDL's me asured from repeated analysis of an aqueous blank. As a result of an e valuation of these data, it is proposed that appropriate figures of me rit to describe the analytical performance of a technique are: (1) med ian precision in the determination of major elements; and (2) the numb er of trace elements that can be determined to MDL's of less than one- tenth the crustal abundance of the element. These factors should then be evaluated in conjunction with logistical factors including the rate at which samples can be analysed and the cost per determination. The influence of these factors on applications of the techniques studied i n pure and applied geochemistry are discussed.