Inter-laboratory comparison of ED-XRF/PIXE analytical techniques in the elemental analysis of filter-deposited...

The quantification of the elemental concentration of ambient particulate matter is a challenging task because the observed elemental loadings are not well above the detection limit for most analytical techniques. Although non-destructive nuclear techniques are widely used for the chemical characterization of ambient aerosol, only one multi-element standard reference filter material that mimics ambient aerosol composition has become recently available in the market. To ensure accuracy, reliability and comparability of instruments performance, multiple reference materials with different elemental mass loadings are necessary. In this study, an intercomparison exercise was performed to evaluate the measurement uncertainty and instruments performance using multi-element dust standard reference samples deposited on PTFE filters. The filter samples, produced by means of dust dispersion, were tested in terms of homogeneity, reproducibility and long-term stability (≈40 months). Eight laboratories participated in the exercise. The evaluation of the results reported by the participants was performed by using two sets of reference values: a) the concentrations reported by the Expert Laboratory, b) the robust average concentrations reported by all participants. Most of the reported on the certificate of analysis elements were efficiently detected in the sample loadings prepared as representative for atmospheric samples by the Expert Laboratory. The average absolute relative difference between the reported and the reference values ranged between 0.1% (Ti) and 33.7% (Cr) (CRM-2584). The participants efficiently detected most of the elements except from the elements with atomic number lower than 16 (i.e. P, Al, Mg). The average absolute percentage difference between the participants results and the assigned value as derived by the expert laboratory was 17.5 ± 18.1% (CRM-2583; Cr, Pb excluded) and 16.7 ± 16.7% (CRM-2584; Cr, P excluded). The average "relative robust standard deviation" of the results reported by all participants was 25.1% (CRM-2583) and 22.8% (CRM-2584).