Partitioning between polyurethane foam and the gas phase: data compilation, uncertainty estimation and implications for air sampling

Abstract

Polyurethane foam (PUF) is frequently applied for sampling semi-volatile organic compounds (SVOCs) in the gas phase. Equilibrium partition coefficients (K_PUF/G) often are used to estimate the potential for breakthrough during active air sampling (AAS) and to correct for non-linear uptake during passive air sampling (PAS). K_PUF/G is either determined experimentally or estimated, in both cases incurring uncertainties that can be carried over to other parameters. Here, a dataset of 547 measured K_PUF/G values and chemical information for 281 distinct chemicals was compiled from the peer reviewed literature. Measured log K_PUF/G were compared with predicted values to identify potential bias in data generated with a particular experimental approach. An analysis of the measured data suggests that the uncertainty of unbiased log K_PUF/G values is at best 0.2 log units at 15 °C (e.g. for hexachlorobenzene and fluoranthene), but most likely much higher. This implies that inherent passive air sampling rates obtained from the loss of a depuration compound (SR) and breakthrough volumes during AAS can presently not be known with an uncertainty of less than ca. 50%. During short PAS deployment periods, the uncertainty in the effective sampling volume (V_air) derives mainly from the uncertainty in the SR, whereas the uncertainty in K_PUF/G of the target compound will become important and even the main source of uncertainty for V_air if deployments are long or target chemicals are relatively volatile. This in turn implies that the uncertainty of V_air cannot be smaller than the uncertainty of SR and K_PUF/G and therefore again is at least ca. 50%. We strongly recommend that the uncertainty of air concentrations obtained by non-linear PAS is quantified and reported and we outline a procedure on how to do that. Because the uncertainty in K_PUF/G of target and depuration chemicals generally exceeds 30%, it may often be necessary to conduct Monte Carlo simulation.