Measured versus Modeled Partitioning of Several Hundred Semi-Volatile Organic Compounds Using SV-TAG with in-situ Derivatization

Wednesday, 17 December 2014
Joshua Moss1, Gabriel A Isaacman1, Nathan M Kreisberg2, Weiwei Hu3, Pedro Campuzano Jost4, Douglas A Day5, Jose L Jimenez3, Eric S Edgerton6, Karsten Baumann7, Susanne V Hering2 and Allen H Goldstein1, (1)University of California Berkeley, Berkeley, CA, United States, (2)Aerosol Dynamics Inc., Berkeley, CA, United States, (3)University of Colorado at Boulder, Boulder, CO, United States, (4)University of Colorado Boulder, Boulder, CO, United States, (5)Cooperative Institute for Research in Environmental Sciences, Boulder, CO, United States, (6)Atmospheric Research & Analysis, Inc., Cary, NC, United States, (7)Atmospheric Research & Anal., Morrisville, NC, United States
Most organic aerosol is secondary, formed by oxidation of primary gas-phase chemicals whose products condense into particles. While gas-particle partitioning is known to depend in part on the volatility of a compound, the process is poorly understood and largely unconstrained by compound-specific measurements. We directly measured gas-to-particle partitioning of over 250 semi-volatile organic compounds during the SOAS campaign in rural Alabama in the summer of 2013, and compare to a partitioning model based on their calculated vapor pressures and chemical properties. All data were collected using an SVTAG (Semi-Volatile Thermal desorption Aerosol Gas chromatograph) which uses two parallel cells to collect gas- and particle-phase organic compounds with volatilities lower than tridecane. By using an activated carbon denuder to remove all gas-phase compounds from one channel, gas-particle partitioning is directly measured on an hourly timescale. Derivatization with MSTFA, a silylating agent, allows analysis of highly oxygenated compounds that have not previously been analyzed by in-situ GC methods. Internal standards are used to correct all compounds for instrument variability in order to remove instrument biases between sampling channels, and compounds lacking temporal variability are not reported so as to exclude any internal contaminant compounds. Chemical characteristics of each compound (i.e. number of derivatized hydroxyl groups) were inferred from their mass spectra using a model we built from data in an available NIST mass spectral database. Based on chromatographic retention time and chemical characteristics we model the vapor pressures of measured compounds to within the error of current structure-activity models (i.e. SIMPOL). Measured partitioning for all observed compounds is compared to partitioning predicted from their estimated vapor pressures. While current models adequately describe the partitioning of some compounds, a large fraction of observed species are found to deviate from expectation based on environmental factors and chemical characteristics. Chemical dependence is expected in part due to the possible formation of low-volatility particle phase oligomers and organosulfates, which may decompose into smaller, more volatile fragments in the desorption process.