Ultrafiltration separation of aquatic natural organic matter: Chemical probes for quality assurance

Abstract

Characterization of molecular size of natural organic matter (NOM) is a valuable tool when assessing its effect on the performance of water treatment systems as well as its geochemical origin. Size fractionation can be accomplished by ultrafiltration (UF). Unfortunately, membrane manufacturing generates a range of pore sizes. Many membrane manufacturers use molecular weight cutoff (MWCO) metric based on a 90% retention of given solute after specified duration of filtration. The objective of this study was to characterize the ability of different commercially available UF membranes to separate different size fractions of NOM. The UF membranes characterized were YM (regenerated cellulose, negatively charged) and PB (polyethersulfone, negatively charged) product lines by Millipore™. The probes used to represent the size, shape and charge of NOM were polymers (polyethylene glycols (PEGs), dextrans, polystyrene sulfonates (PSSs)), dyes (bromocresol green, congo red, methyl red, methyl orange) and biological molecules (vitamin B-12 and bacitracin). The results show that MWCO definition does not hold for membranes of 5 kDa and 10 kDa pore openings using most polymers and dyes. The MWCO definition holds for 1 kDa membrane for all tested probes. Under natural water conditions PSSs assume random coil configurations that are nearly identical to Suwannee fulvic acid. The results show that PSS agrees with stated MWCOs. The study demonstrates that ultrafiltration is not a simple mechanical sieving process, but that charges on the membrane and the constituent play a significant role in the rejection process. Effective probe size was increased seven- to fourteen-fold by charge interactions between the negative probes and negatively charged membrane. Uncharged molecules larger than specified MWCOs are able to pass through pores (PEGs), while small charged molecules (dyes) do not pass. For probes with low or neutral charges, shape becomes an important factor, with globular being favored over linear structure. Thus, MWCOs cannot be trusted for purposes of NOM size characterization. The study recommends the use of YM 1 K, PB 5 K and YM 10 kDa membranes for comparative-only NOM size ultrafiltration characterization within the 1–10 kDa size range.