Comparison of size-resolved PM elements measured using aluminum foil and Teflon impaction substrates: Implications for ultrafine particle source apportionment and future sampling networks in California

Abstract

Measurement networks for ultrafine particulate matter (PM0.1) have been limited by the high costs for equipment, supplies, and labor associated with the need to collect PM0.1 samples on multiple substrates for full chemical analysis. Here we explore whether a single cascade impactor loaded with aluminum foil substrates is sufficient for PM0.1 source apportionment calculations in order to reduce those costs. An extraction method previously designed to measure elements on Teflon substrates was modified to accommodate features of aluminum foil substrates. Regression analysis between co-located aluminum foil and Teflon substrates in the particle diameter range 0.1–1.8 μm showed good agreement (R > 0.7) for 18 elements. Regression in the diameter range 0.1–0.18 μm (quasi-ultrafine particulate matter) was used to characterize the uncertainty introduced by the aluminum foil extraction method for the elements Li, K, V, Br, Rb, Mo, Cd, Sn, Sb, and Ba. This uncertainty was used to generate 30 simulated aluminum foil PM0.1 datasets at each of three sites, followed by source apportionment analysis using Positive Matrix Factorization (PMF). At two of the three sites, the PM0.1 source contributions calculated using aluminum foil substrates alone were almost identical to the PMF results from combined aluminum foil and Teflon substrates. The PM0.1 source contributions calculated using aluminum foil substrates at the third site were closer to the results from a previous Chemical Mass Balance (CMB) study than to the PMF results from the combined aluminum foil and Teflon substrates, possibly because the CMB study also relied exclusively on samples collected using aluminum foil substrates. The success of the PM0.1 source apportionment approach using aluminum foil substrates in a single cascade impactor provides a viable method for reducing costs in PM0.1 sampling networks by 40–47%. Similar results may be achievable at locations outside of California.