Gas/particle partitioning of n-alkanes, PAHs and oxygenated PAHs in urban Denver

Abstract

In this study, a medium volume sampler equipped with quartz fiber filters (QFFs) and a polyurethane foam (PUF)/XAD-4/PUF sandwich (PXP) was used to collect semi-volatile organic compounds (SVOCs) in both gaseous and particle (PM2.5) phases. A backup QFF (bQFF) was used to evaluate possible sampling artifact of particulate organics due to vapor-phase adsorption. A series of n-alkanes (molecular weight: 170–562) and PAHs (128–300), and two oxy-PAHs (acenaphthenone, 168; fluorenone, 180) were measured. Breakthrough experiments demonstrated that the PXP could collect all gas-phase target compounds with high efficiency, even the low molecular weight (MW) species (e.g., naphthalene). Comparing species concentrations across different sampling matrices encountered at the Denver, Colorado field site, the light n-alkanes (MW < 282) and PAHs (MW < 192) were mostly distributed into the gas phase; while those heavy n-alkanes (MW > 324) and PAHs (MW > 202) were primarily in the particle phase (Average temperature, 12.5 ± 10.1 °C). Log values of measured gas/particle (G/P) partitioning coefficients (Kmp,OM) of selected SVOCs (docosane, tricosane, fluoranthene, pyrene, acenaphthenone and fluorenone) were linearly regressed to those of theoretically-based partitioning coefficients (Ktp,OM) for comparison. Prior to Kmp,OM calculation, the gas- and particle-phase concentrations of SVOCs were corrected following two different approaches based on bQFF measurements. The first approach assumed that the bQFF associated SVOCs were from the adsorption of gaseous SVOCs (positive artifact); the second approach assumed equal contributions from positive and negative (organics evaporated from top QFF and adsorbed by bQFF) artifacts. Under both corrections, significant correlations (p < 0.05) were observed between log Kmp,OM and log Ktp,OM for the six selected SVOCs, suggesting that the predicted G/P partitioning can reasonably capture the measured G/P partitioning behavior. The large deviations (1-2 orders of magnitudes) between Kmp,OM and Ktp,OM for acenaphthenone and fluorenone might be caused by the assumption of ideality (activity coefficient = 1) and the over prediction of vapor pressures (for Ktp,OM calculation). Negative correlations were observed between regression residuals of log Kmp,OM vs. log Ktp,OM and relative humidity, which might be attributed to the use of a constant activity coefficient and the possibility of phase separation.