Air pollution, especially particulate matter (PM), is of growing concern in the United States and around the world. PM with aerodynamic diameter (AED) less than 2.5um is currently one of the two indicators for PM pollutions. The concentration of PM2.5 in ambient air is measured by the EPA-approved Federal Reference Method (FRM) or Federal Equivalent Method (FEM) PM2.5 sampler. The goal of this research was to study the factors affecting the design and performance of the PM2.5 sampler. The key component of the PM2.5 sampler is the nozzle. Two sets of nozzles (40 nozzles per set) were tested in a sampler that was placed in a wind tunnel, and penetration efficiencies. It was shown that change in convergence angle of a modified nozzle can affect impactor performance. The ?Stk50 for original and modified nozzles were 0.57 and 0.49, respectively. The slope of the efficiency curve for original and modified nozzles were 1.52 and 1.36, respectively. During the wind tunnel test, the monodisperse aerosols were generated with artifacts known as multiplets and satellites. Two artifact correction methods, the Ranade method and the APS method, were compared experimentally and theoretically in this study. The two methods produced similar results in the wind tunnel tests, where the vibrating orifice aerosol generator was finely tuned to eliminate the satellites. However, in theoretical calculation, there were differences between these two methods. The APS method was able to completely correct for the effect of satellites since the APS provided data for the complete particle size distribution, which were used to identify satellites. The Ranade method was found to be sensitive to satellites, especially for the larger particles where the sampling effectiveness was close to zero. The lognormal distribution is widely used in the theoretical calculation and modeling of PM samplers; however, it was demonstrated that the error resulting from the lack of fit of the lognormal distribution was non-trivial. In this analysis, the error was as great as 22.68% when using a lognormal distribution. Ten distribution functions were applied to fit the performance curve given for FRM PM2.5 and PM10 samplers. The Kolmogorov Smirnov test and mass concentration calculation were used to demonstrate that the Dagum distribution provided the best fit among the ten functions.
