An inter-laboratory comparison of elemental loadings of PM2.5 samples using energy-dispersive XRF and magnetic-sector ICP-MS

We performed an inter-laboratory study comparing elemental loadings of atmospheric particulate matter (PM) samples, determined using Energy Dispersive X-ray Fluorescence (XRF) and Sector Field-Inductively Coupled Plasma-Mass Spectrometry (SF-ICP-MS) techniques. Three hundred archived samples collected on polytetrafluoroethylene (PTFE) filters from four sites of the IMPROVE network in the United States were included in the comparison. Regression analyses were utilized to compare the techniques. For elements measured well above the respective method detection limits (MDLs), agreement between methods was typically within ±20%. Major axis regression analysis between the techniques revealed very strong correlations (r2 > 0.9) for 13 out of 20 elements (Na, Mg, Al, S, K, Ca, Ti, V, Mn, Fe, Cu, Zn and Sr), strong correlations (0.8 < r2 < 0.9) for 3 elements (Cr, Ni and Pb), a weak correlation (r2 = 0.69) for one element (Zr), and very weak correlations (r2 < 0.4) for three elements (P, As and Rb). All the elements with strong or very strong correlations yielded regression slopes within 20% of unity, except Zn, indicating that absolute element recoveries of the two techniques were similar. Inter-elemental correlations for several element pairs with common environmental sources (e.g., Ni versus V from shipping emissions) were examined with both measurement platforms, and the comparisons verify that SF-ICP-MS has better detection limits than XRF for several elements. However, one of the four selected IMPROVE sites, a coastal site with high levels of sea salt, presented poor correlations between the two methods for elements Z < 20. Although independent ion chromatography (IC) confirmed the validity of the XRF measurements, without a shared Reference Method for PM elemental content, it is challenging to determine which analytical technique is more accurate in all cases. Laboratory-prepared reference materials (RM) were analyzed by both approaches in an effort to address the relative accuracy of the methods; however, the results were not consistent with the patterns observed in the ambient samples. Further inter-comparisons on real-world environmental samples, incorporating additional analytical techniques are necessary to better assess the measurements. This study highlights the need for shared reference materials that challenge all aspects of the measurement process for the atmospheric PM monitoring community.