(560e) Viability of Nanofiltration and Reverse Osmosis in Removing Emerging Trace Organic Contaminants

The rejection of emerging organic micropollutantsis an important issue where recycled water is used to augment drinking water supplies. The focus of this research study was to explore alternatives of an integrated membrane system involving nanofiltration (NF) and ultra-low pressure reverse osmosis (ULPRO) in place of reverse osmosis (RO) representing a more cost-effective system because of potentially lower pressure requirements and the greater selectivity for organic micropollutants. These compounds represent a broad range of physicochemical properties such as hydrophilic ionic, hydrophilic non-ionic and hydrophobic non-ionic, and are associated with potential adverse effects for human health and aquatic life. The trace organic compounds being emphasized in this research include disinfection by-products (e.g., trichloroacetic acid, chloroform, bromoform, N-nitrosodimethylamine), pesticides, endocrine disrupting compounds (e.g., 17b-estradiol, testosterone, bisphenol A), pharmaceutical residues (e.g., ibuprofen, naproxen, gemfibrozil, carbamazepine, primidone), and chlorinated flame retardants. Trace contaminants were quantified through GC/MS analysis. A key element of this investigation was to explore whether NF and ULPRO membranes can consistently meet potable water quality requirements with respect to TOC, total nitrogen, and both regulated and unregulated trace organic compounds. While meeting these water quality goals, the operating characteristics of NF/ULPRO membranes (such as flux, fouling/scaling, and cleaning frequencies) should be comparable to thin-film composite RO membranes currently employed at full-scale facilities with operating feed pressure requirements significantly lower than the pressure range currently employed at facilities using conventional RO. After pre-screening over 15 potential NF and ULPRO products, three candidate membranes were selected and tested using a 68 L/min (18 gpm) membrane pilot skid. The three selected candidate membranes were each tested for at least 1,300 hours on microfiltered feed water at two full-scale facilities. The feedwaters represented non-nitrified microfiltered effluent provided by the West Basin Water Recycling Plant in El Segundo, CA and nitrified/denitrified microfiltered effluent provided by the Scottsdale Water Campus, AZ. Operational performance and rejection of trace organics, nutrients and total organic carbon was monitored during pilot-scale testing and compared to the performance of full-scale trains operating on the same feedwater employing conventional RO membranes (i.e., TFC-HR, Koch Membranes and ESPA2, Hydranautics). During several occasions, challenge tests were conducted at pilot-scale to examine rejection of trace organics during periods of elevated feed concentrations. The study was assisted through state-of-the-art membrane characterization to describe the fouling behavior of NF/ULPRO membranes and its role on operation (e.g., flux decline) and rejection. The TMG10 and NF-90 membranes tested on microfiltered non-nitrified wastewater (WBWRP) exhibited a relative constant temperature corrected specific flux (TCSF) after initial membrane compaction and fouling of 0.03 L/m2-hr-kPa (0.12 gfd/psi). Similar to observations made during membrane selection on laboratory-scale, the NF-90 exhibited a larger flux decline as compared to the TMG10 during pilot-scale testing. For comparison, the benchmark TFC-HR and ESPA2 membranes employed at the WBWRP under similar conditions achieved a lower stabilized TCSF of approximately 0.022 L/m2-hr-kPa (0.09 gfd/psi). During these test periods, the permeate TOC and ammonia concentrations of the TMG10 (NF-90) were below 0.2 mg-C/L (<0.3 mg-C/L) and below 2 mg-N/L (<3 mg-N/L), respectively. Comparing pilot-scale testing of the TMG10 to a period of full-scale operation indicated that the TMG10 performed similarly or better than the ESPA2 and TFC-HR in terms of TOC and ammonia permeate concentrations and rejection. Rejection of regulated organics, such as THMs, HAAs and NDMA, was also found to be similar with a poor to moderate rejection of chloroform. Non-nitrified feed water concentrations of unregulated organics were variable over the testing period and ranged from below quantification for certain compounds to the low microgram per liter range for chlorinated flame retardants. Both, the TMG10 tested on pilot-scale and the TFC-HR and ESPA2 employed on full-scale were highly efficient in rejecting unregulated organics with non detectable concentrations in the permeates. The TMG10 membrane was retested along with the NF-4040 membrane on microfiltered nitrified/denitrified wastewater at SWC. Different to the performance of the TMG10 exhibited at WBWRP, the TCSF of the TMG10 at SWC never stabilized and after 800 hours of testing finally fell below what was considered an acceptable TCSF level (0.02 L/m2-hr-kPa (0.08 gfd/psi) in order to achieve any economic benefits from reduced operating pressures. The NF-4040 exhibited a relative small initial flux decline followed by a very constant TCSF of 0.08 L/m2-hr-Pa (0.33 gfd/psi) over 1,200 hours before it steadily declined until termination of the test. As expected, during the first 1,200 hours of testing the TCSF of the NF-4040 was more than four times higher as compared to the TFC-HR. TOC permeate concentrations of the NF-4040 were below 0.5 mg-C/L. Regulated organics exhibited a similar rejection at SWC during TMG10 testing as compared to WBWRP. No unregulated organics were detected in TMG10 permeates. Chlorinated flame retardants, carbamazepine and bisphenol-A were detected in traces in the permeate of the NF-4040 immediately after start-up. Rejection, however, improved with membrane fouling and subsequent sampling campaigns revealed no detections in permeate samples. During challenge test, where 300-500 percent elevated concentrations of unregulated organics were spiked to the feedwater, only non-ionic solutes showed slightly higher detections in the combined permeate. Although the TMG10 was considered viable after testing at WBWRP, the TCSF of the TMG10 steadily declined during testing at SWC. This decline in performance for both candidate membranes and the TFC-HR was likely due to biofouling, which can build up when a sufficient level of chloramines in the feed water is not maintained. A significant degree of biofouling of the TMG10 and NF-4040 at SWC is further confirmed through results from membrane autopsy. All membranes tested during this study exhibited a very similar fouling tendency with a more severe biofouling and inorganic scaling in the tail-end elements. The nature and severeness of fouling, however, correlated closely with the feed water quality and effectiveness of fouling control measures. Initial (organic) fouling, however, seems to result in improved rejection of organic micropollutants. Foulant precipitation and cake-layer formation result in a considerable change of membrane surface characteristics with respect to membrane hydrophobicity, surface charge and surface morphology, hence potentially affecting transport mechanisms of contaminants as compared to virgin membranes. The transport of hydrophilic ionic organic contaminants was especially hindered in NF membranes as a result of improved electrostatic exclusion, pore clogging and an increased adsorption capacity of fouled polyamide membranes. The ULPRO membrane selected for pilot- and full-scale testing (TMG10) on a non-nitrified and nitrified/denitrified microfiltered feedwater performed better or similar regarding the TCSF and feed pressure than the conventional RO membranes TFC-HR and ESPA2, respectively. The two NF membranes selected (NF-90, NF-4040) confirmed the viability of NF during pilot-scale operation in rejecting TOC and regulated trace organics as well as unregulated organics. Especially, the NF-4040 membrane provided a water quality better than expected, which is likely due to the fact that fouling of the membrane improves solutes rejection. By operating at a TCSF four times higher than the conventional TFC-HR RO membrane, the NF-4040 demonstrated the cost benefits of NF membranes, which qualifies for further research. In addition, findings from rejection studies at the laboratory-, pilot- and full-scale were developed into a model framework providing the basis to reliably predict ? a priori ? rejection of organic micropollutants by RO, NF and ULPRO membranes taking into account physicochemical properties of solutes and membranes as well as key operational conditions affecting solute rejection. Past studies on modeling membrane performance have resulted in several methods and sets of equations that can be used to model the rejection of inorganic and organic solutes. However, simple yet robust solution-diffusion models do not directly apply to membranes in which pore phenomena including physical sieving and Donnan exclusion are dominant in rejecting solutes. Transport equations developed for describing the transport of electrolytes through non-porous and porous membranes are often hindered by the complexity of the calculations as well as the numerous descriptive parameters required. Although significant advances in membrane modeling have been made in order to optimize the separation of mixtures of inorganic ions and ionic organic solutes, little work has been conducted to satisfactorily predict the rejection of nonionic organic solutes. Findings of this research clearly demonstrated that a single model does not exist which is capable of describing the mass transport of organic micropollutants during high-pressure membrane treatment. The physicochemical properties of the solutes are the key factors determining rejection and if properly considered and put in context with relevant membrane properties, rejection can be qualitatively predicted. Two approaches were developed to quantitatively describe and predict the rejection of non-ionic and ionic compounds of concern (Spiegler-Kedem model and extended Nernst-Planck equation). With the exception of highly hydrophobic compounds, the Spiegler-Kedem model provided a high to moderate degree of accuracy for the rejection of non-ionic solutes by a conventional RO, an ULPRO and an NF membrane. For ionic solutes, the Spiegler-Kedem model underpredicted rejection as expected since electrostatic exclusion is not properly considered in the model. The Donnan Steric Pore model, therefore, seemed to be more suitable for modeling ionic solutes and indeed preliminary results suggest that this model describes rejection for ionic micropollutants more accurate.