Effects of sampling bias on gas–particle partitioning of semi-volatile compounds

Semi-volatile compounds distribute mass between gas and particle phases in air. Gas–particle partition coefficients [Kp=(F/TSP)/A] are used to predict the fate and transport of these compounds in the atmosphere. Measurements of gas–particle partition coefficients are biased if: (1) a fraction of mass is not quantified due to mass penetration through the sampler or reactions of collected species during sampling, (2) a fraction of particle-phase mass is measured erroneously as gas-phase, or (3) a fraction of gas-phase mass is measured erroneously as particle-phase. The latter two biases can lead to serious errors in measured partition coefficients. Generalized equations were developed to predict the impact of these errors for three sampling methods: filter–adsorbent, filter–filter–adsorbent, and denuder–filter–adsorbent. Two cases were considered: (1) the level of bias is equal for compounds of varying Kp and (2) the sampling bias is compound dependent and, hence, varies with Kp. These artifacts cause deviations to the slope and intercept in log–log plots of compound partition coefficients vs. subcooled, liquid vapor pressures that are similar to the deviations seen from non-equilibrium effects, temperature and concentration changes, and the presence of non-exchangeable material in the particle.