Evaluation of Computer-Controlled Scanning Electron Microscopy Applied to an Ambient Urban Aerosol Sample

Concerns about the environmental and public health effects of particulate matter (PM) have stimulated interest in analytical techniques capable of measuring the size and chemical composition of individual aerosol particles. Computer-controlled scanning electron microscopy (CCSEM) coupled with energy-dispersive X-ray analysis (EDX) allows automated analysis of particle size, chemistry, and particle classification. In combination with manual SEM and bulk analytical techniques such as X-ray fluorescence, CCSEM can be a valuable tool for characterizing individual ambient particles and determining sources of ambient PM. The goal of this study was to examine several issues related to the quality and validity of CCSEM data. These included the stability of unattended CCSEM for multihour runs, the number of particles that must be analyzed in order to yield representative results, and errors associated with CCSEM. CCSEM was applied to the analysis of a 24 h ambient particle sample collected in Baltimore, MD. The coarse-fraction sample (PM_10-2.5) was collected with a dichotomous sampler on a polycarbonate filter. A total of 2819 particles in 78 randomly selected fields of view were analyzed by CCSEM during an unattended 8 h run. Particle diameter, aspect ratio, particle location, X-ray counts for 20 elements, and digital images of each particle and its field of view were stored. The average number of particles per field (N/F), average particle diameter (D_ave), average mass loading per field (M_ave), and average particle composition were calculated for subsets of the data and compared against results for the full data set in order to assess the stability of the CCSEM analysis over time and the number of particles needed to obtain representative results. These comparisons demonstrated excellent stability of CCSEM over the 8 h run. Physical properties (represented by N/F, D_ave, and M_ave) of the sample were well characterized by analyzing approximately 360 particles. Chemical properties of the sample (average elemental composition and major chemical class abundances) converged to within a few percent of their final values after analyzing about 1000 particles. However, for many purposes several hundred particles may provide adequate characterization. Convergence of minor class abundances was limited by statistical fluctuations as the number of particles populating a class became very small. Manual review of the CCSEM data identified errors associated with CCSEM due to missed particles, overlapping particles, contrast artifacts, sizing errors, and heterogeneous particles. Most errors could be corrected or eliminated during manual off-line review of the data or avoided by maintaining a proper particle loading on the filter.