Combinatorial variation of structure in considerations of compound lumping in one- and two-dimensional property representations of condensable atmospheric organic compounds. 1. Lumping by 1-D volatility with fixed

Many current models that aim to predict urban and regional levels of organic particulate matter (OPM) use either the 2 product (2p) framework for secondary organic aerosol (SOA) formation, or a static 1-D volatility basis set (1-D-VBS). These approaches assume that: 1) the compounds involved in OPM condensation/evaporation can be lumped simply by volatility with no specificity regarding carbon number nC, MW, or polar functionality; 2) water uptake does not occur; and 3) the compounds are non-ionizing. This work considers the consequences for uniphasic PM caused by the first two assumptions due to effects of the condensed-phase mean molecular weight MW ¯ and activity coefficients (ζi), including when RH (relative humidity) > 0. Setting n C  = 10 for all bins, multiple chemical structures were developed for each bin of a 1-D-VBS for un-aged SOA in the α-pinene/ozone system. For each bin, a group-contribution vapor pressure ( p L o ) prediction method was used to find multiple structures such that the groups-based log p L o for n C  = 10 and variable numbers of aldehyde, ketone, hydroxyl, and carboxylic acid groups agrees, within ±0.5, with the bin volatility. The number of possible combinations with one structure taken from each bin was 17,640. The Raster-Roulette Organic Aerosol (RROA) model was used to calculate the equilibrium mass concentrations (μg m−3) of OPM (Mo) and co-condensed water (Mw) at 25 °C for each combination for ranges of RH and ΔHC (change in parent hydrocarbon concentration). UNIFAC was used to determine the needed values of ζi. Frequency distributions from RROA for Mo, Mw, and the O:C ratio were developed. For Mo levels typical of the ambient atmosphere, then for the 1-D-VBS and all bins constrained at n C  = 10, significant RH-induced enhancement of OPM condensation was observed in the distributions. The spread of the distributions was found to increase rapidly as the level of OPM decreased. The within-bin spread of ±0.5 log units in the groups-based estimates of log p L, i o was found to cause significant spread in the distributions at lower Mo values. At the chosen n C (=10), the groups-based log p L, i o values show a spread of ±2 log units in a plot of log p L, i o vs. O:C. When seeking to advance to 2-D-grid predictive modeling of atmospheric OPM, use of an O:C vs. n C grid will therefore require reliable information (or at least empirical calibration) as to the distributions of the likely structures at each gridpoint.