HEALTH AND SAFETY LABORATORY, BRO AD LANE, SHEFFIELD S3 7HQ , ( ) ( ) UK L.C.K. , BGI INC, 58 GUINAN ST, WALTHAM, MA 02451 R.G. , AND RUPPRECHT AND PATASHNICK CO. INC, 25 CORPORATE CIRCLE, ( ) ALBANY, NY 12203 M.M.ABSTRACT. Aerosol samplers for ambient PM2.5 are required to possess a steep aerodynamic particle size selection curve, i.e., a 'sharp cut' at 2.5 m m aerodynamic diameter. For long-term and continuous PM monitors the selector system also requires low maintenance and the ability to operate at high loadings. While a sharp cut is easier to achieve with an impactor-based selector, the other requirements are more easily met with a cyclone. Four alternative PM2.5 selectors were tested against these criteria: two were pre-existing commercial designs and two were novel cyclone prototypes. The main aim of the work was to assess the characteristics of the selectors when clean and under various loading conditions. The aerodynami c size-selection characteristics of the PM2.5 selectors were tested before and after loading with dust, under both laboratory and eld conditions. Aerosol penetration measurements were made using an Aerodynamic Particle Sizer. Many repeat tests were performed on two specimens of the Well Impactor ( ) ( ) Ninety-Six WINS , two specimens of a novel Sharp-Cut Cyclone SCC , one ( ) member of the GK cyclone family GK4.39 , and one University Research Glass-( ) ware URG cyclone. Four loadings of the WINS and SCC were made in the laboratory using a narrow-fraction alumina dust. The penetration curves were measured after each loading. Five cumulative outdoor loadings were made by setting up four PM2.5 samplers, two with the WINS and two with the SCC, in a suburban garden during the summer months. The penetration curves were measured at weekly intervals after sampling times ranging from 96 to 132 h. Three further cumulative loadings were tested in a similar experiment in a city-center underground car park. When clean, all three PM2.5 size selectors were shown to ( ) have 50%penetration D values close to 2.5 m m, although the penetration curve 50 shape differs for the three selector designs. Under loading the D value for both 50 the WINS and SCC fell, with the decrease being largest for the WINS. With high loadings the SCC D fell to 2.35 m m and the WINS D fell to 2.15 m m. The 50 50WINS deviation was large enough to potentially lead to under sampling of PM2.5. The SCC cyclone was seen to provide a sharp cut for ambient air sampling applications that is less affected than the WINS by loading. Additionally, the SCC is a dry system whereas the WINS uses an oiled substrate. While the WINS cut point is unlikely to shift to an unacceptable degree during 24 or 96 h sampling periods, it would perform less well than the SCC over extended sampling periods.