Size exclusion chromatography as a tool for natural organic matter characterisation in drinking water treatment

Abstract

Natural organic matter (NOM), ubiquitous in natural water sources, is generated by biogeochemical processes in both the water body and in the surrounding watershed, as well as from the contribution of organic compounds that enter the water as a result of human activity. NOM significantly affects the properties of the water source, including the ability to transport metals, influence the aggregation kinetics of colloidal particles, serve as a food source for microorganisms and act as a precursor in the formation of disinfection by-products (DBPs), as well as imparting a brown colour to the water. The reactivity of NOM is closely tied to its physicochemical properties, such as aromaticity, elemental composition, functional group content and molecular weight (MW) distribution. The MW distribution is an important consideration from a water treatment perspective for several reasons. For example, low MW NOM decreases the efficiency of treatment with activated carbon, and this fraction is thought to be the portion most difficult to remove using coagulation. The efficiency of membranes in the treatment of drinking water is also influenced by the MW distribution of NOM, while some studies have shown that the low MW fraction contributes disproportionately to the formation of bioavailable organic matter, therefore promoting the formation of biofilms in the distribution system. For these reasons, understanding the MW distribution of NOM is important for the treatment of natural waters for use as drinking waters. Optimisation of a high pressure size exclusion chromatography (HPSEC) method for analysis of the MW distribution of NOM in natural waters is described (Chapter 2). Several parameters influencing the performance of HPSEC are tested and an optimised set of conditions illustrated.These parameters included eluent composition, ionic strength of the sample, flow rate and injection volume. Firstly, it was found that increasing the ionic strength of the HPSEC eluent resulted in less exclusion of NOM from the stationary phase. Stationary phases used in HPSEC contain a residual negative charge that can repel the negatively charged regions of NOM, effectively reducing the accessible pore volume. By increasing the ionic strength, interactions between the stationary phase and eluent enabled a larger effective pore size for the NOM analytes. However, increasing ionic strength of the eluent also resulted in a loss of peak resolution for the NOM portion able to access the pore volume of the stationary phase. Determining the ideal eluent composition required the balancing of these two outcomes. Matching of the ionic strength of the sample with the eluent was also an important consideration. Retention times were slightly lower when the sample ionic strength was not matched with the eluent, especially for the lowest MW material, although the effect on chromatography was minimal. Flow rate had no effect on the resolution of the HPSEC chromatogram for the portion of material able to permeate the pore space of the stationary phase. Changes in the volume of sample injected had a marked effect on the elution profile of the NOM sample. Besides the obvious limitation of detection limit, only minor changes in elution profile were obtained up to an injection volume of 100 µL. Volumes above this value, however, resulted in significant peak broadening issues, as well as an undesirable effect on the low MW portion of detected DOC.In Chapter 3, high pressure size exclusion chromatography with UV254 [subscript] and on-line detection of organic carbon (HPSEC-UV254[subscript]-OCD) was used to compare the removal of different apparent MW fractions of DOC by two process streams operating in parallel at the local Wanneroo groundwater treatment plant (GWTP). One of these two process streams included alum coagulation (operating in an enhanced coagulation mode (EC) for increased DOC removal) and the other stream included a magnetic ion exchange (MIEX®) process followed by alum coagulation (MIEX®-C). The MIEX® process is based on a micro-sized, macroporous, strong base anion exchange resin with magnetic properties, which has been designed to remove NOM through ion exchange of the anionic sites in NOM. Water was sampled from five key locations within these process streams, and the DOC at each location was characterised in terms of its MW distribution. HPSEC was carried out using three different on-line detector systems, namely OCD, UV absorbance detection at 254 nm, and fluorescence detection (λex[subscript]= 282 nm; λim[subscript] = 353 nm). This approach provided significant information on the chemical nature of the DOC in the various MW fractions. The MIEX®-C process was found to outperform the EC process: these two processes removed similar amounts of high and low MW DOC, but the MIEX®-C process showed greater removal of DOC from the intermediate MW fractions. The two coagulation processes (EC and coagulation following MIEX®) showed good removal of the fractions of highest MW, while the MIEX® process alone was found to remove DOC across all MW fractions.These results seem to indicate that anionic groups, particularly susceptible to removal with MIEX® treatment, are well distributed across all MW fractions of NOM. In agreement with previous studies, MIEX®-C outperformed EC in the overall removal of DOC (MIEX®-C removed 25 % more DOC than EC). However, 70% of the additional DOC removed by MIEX®-C was comprised of a surprisingly narrow range of medium-high MW fractions. The development of a novel online organic carbon detector (OCD) for use with HPSEC for determining the MW distribution of NOM is described in Chapter 4. With UV absorbance detection, the magnitude of the signal is based on the extinction coefficient of the chromophores in the analytes being investigated; whereas the signal from an OCD is proportional to the actual organic carbon concentrations, providing significantly more information. The development of an online OCD involved the separation of analytes using HPSEC, removal of inorganic carbon species which may interfere with organic carbon determination, oxidation of the organic carbon to carbon dioxide, separation of the produced carbon dioxide from the aqueous phase and subsequent detection of the gaseous carbon dioxide. In the new instrument, following separation of components by HPSEC, the sample stream was acidified with orthophosphoric acid to a concentration of 20 mmol L-1[superscript], resulting in a pH of ≤ 2, in order to convert inorganic carbon to carbon dioxide. This acid dose was found to remove greater than 99 % of inorganic carbon once the acidified sample was passed through a hydrophobic polytetrafluoroethylene (PTFE) membrane allowing the passage of dissolved gases (under negative pressure from a vacuum pump) but restricting the flow of the mobile phase.Several factors influenced the oxidation of the organic carbon in the next step, including the dose of persulfate, the type and intensity of UV radiation and the composition of the capillary through which the sample stream passes. Through optimisation of this process, it was found that a persulfate dose of 0.84 mmol L-1[superscript] in the sample stream was required for optimum oxidation efficiency. A medium pressure UV lamp was compared to a vacuum UV lamp for its efficiency in oxidation of organic carbon to carbon dioxide. While the medium pressure lamp produced a far smaller percentage of its total radiation at the optimum wavelength for oxidation of organic compounds, the greater overall intensity of the medium pressure lamp was shown to be superior for this application. The composition of the capillary was shown to have a considerable effect on the oxidation efficiency. A quartz capillary, internal diameter 0.6 mm, was compared with a PTFE capillary, internal diameter 0.5 mm, for the oxidation of organic carbon by external UV treatment. While peak width, an important consideration in chromatographic resolution, was greater for the larger internal diameter quartz capillary, the lower UV transparency of PTFE combined with the shorter contact time, due to the reduced internal diameter of the capillary, resulted in a less efficient oxidation step using the PTFE capillary. The quartz capillary was therefore chosen for use in the UV/persulfate oxidation step for oxidation of organic carbon to carbon dioxide. Separation of the produced carbon dioxide from the sample stream was achieved by sparging with nitrogen and contacting the gas/liquid mixture with a hydrophobic PTFE membrane, restricting the passage of the liquid while allowing the nitrogen and carbon dioxide gases to pass to the detection system.The only factor influencing this separation was the flow of the nitrogen sparge gas, with a flow of 2 mL min-1[superscript] found to be optimum. Detection of produced carbon dioxide was via a Fourier transform infrared (FTIR) spectrometer with a Iightpipe accessory. The Iightpipe accessory was designed for use as a detector for gas chromatography and the small size of the detector cell was ideal for use with this application. Using the new system described, concentrations of a single peak could be determined with a detection limit of 31 ng and a determination limit of 68 ng. The development of the new OCD allowed characterisation of NOM in terms of its MW distribution and the UV and fluorescence spectral properties of each MW fraction. Further characterisation of MW fractions of NOM from a local groundwater bore was carried out by separation of the fractions by preparative HPSEC, followed by off-line analysis. Preparative HPSEC involved the injection of a pre-concentrated groundwater sample multiple times, using a large scale HPSEC column, then collecting and combining material of identical MW. This allowed each MW fraction of the sample to be further characterised as described in Chapter 5. Preparative HPSEC has only previously been applied to a small number of samples for the concentration and fractionation of NOM, where the structural features of the various MW fractions were studied. In the current research, more extensive studies of not only the chemical characteristics, but also the disinfection behaviour, of the MW fractions were conducted. Separation of the sample was conducted on a large diameter silica-based HPSEC column, with fraction collection based on semi-resolved peaks of the HPSEC chromatogram. Nine MW fractions were collected by this method.After concentration and dialysis to remove the buffer salts in the HPSEC mobile phase, each fraction was re-analysed by analytical HPSEC-UV254[subscript] and showed a single Gaussian shaped peak, indicating discrete MW fractions had successfully been collected. Analysis of the collected MW fractions indicated that 57 % of the organic carbon was in Fractions 3 and 4, with 41 % in Fractions 5-9, leaving only 2 % in Fractions 1 (highest MW) and 2. For each of the nine MW fractions, chorine demand and 7 day trihalomethane formation potential (THMFP) were measured on dilute solutions of the same DOC concentration, and solid state 13[superscript]C NMR spectra were recorded on some of the solid isolates obtained after Iyophilisation of the separate or combined dialysis retentates. The larger MW Fractions 3 and 4 were found to contain a greater proportion of aromatic and carbonyl carbon, and the lower MW Fractions 5 and 6 and Fractions 7-9 contained greater proportions of aliphatic and O-aliphatic carbon, by this technique. Chlorine demand experiments on each individual fraction with a normalised DOC concentration indicated that the largest MW fraction (Fraction 1) had the lowest chlorine demand. It was concluded that material in this fraction may be associated with inorganic colloids and unavailable for reaction with chlorine. Fraction 3 had the highest chlorine demand, just over two times more than the next highest chlorine demand (Fraction 4) and approximately three times the chlorine demand of Fraction 2. The organic material in Fraction 2 was postulated to contain a mixture of the reactive material present in Fraction 3 and the colloidal associated material present in Fraction 1.NMR analysis indicated that the difference between Fraction 3 and Fraction 4 was a reduction in reactive aromatic carbon and hence the lower chlorine demand in the latter fraction. Fractions 5-8 had similar chlorine demands, lower than Fraction 4, while Fraction 9 had a very low chlorine demand similar to that of Fraction 1. For Fractions 5-9, the lower aromatic carbon content most likely resulted in the lower chlorine demand. The 7 day THMFP experiments showed some clear trends, with Fraction 1 and Fraction 2 producing the least amounts of THMs but having the greatest incorporation of bromine. Fractions 3 and 4 produced the greatest concentration of THMs with the lowest bromine incorporation, perhaps as they contained fast reacting THM precursors and the higher chlorine concentrations resulted in greater amounts of chlorinated THMs. Fraction 5 and Fraction 6 produced similar levels of THMs over 7 days to Fractions 7-9 (approximately 75% of the amount formed by Fractions 3 and 4), however, Fractions 7-9 formed these THMs more quickly than Fractions 5 and 6, with slightly greater amounts of bromine incorporation. It was thought that the increased speed of formation was due to the smaller MW of these fractions and a simpler reaction pathway from starting material to formation of THMs, as well as some structural differences. This research marks the first report of significantly resolved MW fractions being isolated and their behaviour in the presence of a disinfectant being determined. While the high MW fractions had the greatest chlorine demands and THMFPs, these fractions are also the easiest to remove during coagulation water treatment processes, as shown in Chapter 3. The lowest MW material formed significant amounts of THMs, and also formed THMs more quickly than other MW fractions.This has important implications from a water treatment perspective, as the lowest MW material is also the most difficult to remove during conventional treatment processes. Solid samples of NOM were isolated from water samples taken from four points at the Wanneroo GWTP using ultrafiltration and subsequent Iyophilisation of the retained fractions, as described in Chapter 6. The sampling points were following aeration (Raw), following treatment by MIEX®, following treatment by MIEX®-C and following treatment by EC. Elemental analysis, FTIR spectroscopy, solid state 13[superscript]C NMR spectroscopy and HPSEC-UV254[subscript]-0CD analysis were used to compare the four isolates. Treatment with MIEX®-C was found to remove the greatest amount of NOM. Additionally, treatment with MIEX®-C was able to remove the largest MW range of NOM, with the remaining material being depleted in aromatic species and having a greater proportion of aliphatic and O-aliphatic carbon. EC treatment completely removed the NOM components above 5000 Da, but NOM below this was not well removed. NOM remaining after the EC train had a lower aromatic content and more aliphatic oxygenated organic matter than the RW. The remaining organic matter after MIEX® treatment contained less aromatic material compared to the RW, but had a greater aromatic content than either of the EC or MIEX®-C samples. HPSEC was a significant analytical technique used throughout this research. Initial optimisation of an HPSEC method was an important development which allowed improved resolution of various MW fractions. The application of this technique and comparison of three detection systems for the study of DOC removal showed, for the first time, the performance of MIEX® treatment at a full scale groundwater treatment facility.The use of various HPSEC detection systems allowed significant characterisation of the MW fractions, more information than had previously been gathered from such a sample set. This work demonstrated the need for OCD when applying HPSEC to the study of NOM. As such, a system was constructed that built on previously developed systems, with the use of a small detector cell enabling detection limits capable of measuring even the most dilute natural and treated water samples. To study the individual MW fractions in detail, preparative HPSEC was applied and, for the first time, the disinfection behaviour of various MW fractions was examined. Interestingly, the lowest MW fractions, acknowledged to be the most recalcitrant to conventional water treatment processes, produced significant quantities of THMs. Also the formation kinetics of THMs from the low MW fractions indicated that THMs were formed as quickly as, or perhaps even at faster rates than from the larger MW fractions. Finally, structural characterisation of NOM at four stages of the Wanneroo GWTP indicated MIEX®-C treatment was superior to EC, of significant interest for the water industry.