Chapter 4 MATERIALS & METHODS In this chapter the materidls ztsed (chemicals, organics, colloids, nzembranes and jltration eqzlipment) an described. Membrane characteristics as provided 63, the manzgactttrer are szlmmarised. Xolzltion preparation and anabtz'caI methods are also presented, includilzg the methods zlsedfor organics, aggregate, and membrane deposit characteriration. Filtration protocols are described in the relevant chapters, micrOfi/tration, nltrajltration and nanoJiltahon, respectiveb. Membrane characteristcs szlch as szlface cha~e and morphology are also presented in these chapters. Some methods which required special attention, szlch as concentration of NOM, drawings and Lydroi&namic anabsis of thejltration eqttipment, gnthesis of hematite colloids, instrument calibration (DOC and UV), and sol~tion speciation are shown in Appendzx 1, 2, 3, 4, and 5, respectiveb. Copyright © 2001 by Andrea I. Schafer 92 MATERIALS AND METHODS All chemicals used were of analytical grade from Ajax Chemicals. 1M HC1, NaOH, and NaCl solutions were used for pH and ionic strength adjustments. For some experiments, KC1 or CaC12 were used as the electrolyte. This is inlcated in the relevant results section. Dextran standard (MW 1000 Da), which was used for NF pore size comparison, was purchased from Fluka, Australia. MlliQ water was produced with a six step method; MilliRO, Super-C Carbon Cartridge, Ion Exchange Cartridge, Ion Exchange Cartridge, Organex-Q Cartridge, Milli-Pak Filter. For DOC analysis and standards, water from a regularly sterilised MlliQplus system was used. The MilliQ quality was >l8 MR/cm. Experiments were carried out in a background buffer solution that was chosen as a simple model of natural surface waters, with a monovalent and divalent cation and a background electrolyte to allow pH adjustment without changing ionic strength. The concentration of the cation calcium, was selected after the analysis of the Mooney Moonep Dam surface water. The composition of this water is shown in Appendix 1. The composition of the model system is summarised in Table 4.1. This background solution was used in all experiments, if not othenvise indicated. The species in solution as a function of solution chemistry is described in Appendix 5. Table 4.1 Backgroozmd bzlfeer solntion composition (' frtack'~. Compound Molecular Weight Concentration Concentration Purpose NaHC03 84 1 84 Buffer NaCl 58.5 20 2.935 . 103 Background electrolyte CaCh 111.5 0.5 5 6 Representative of dominant multivalent ions present Humic substances were purchased from the International Hurnic Substances Society (IHSS, USA). Suwannee fiver Stream Reference humic (HA) and fulvic acids (FA) were used. The organics are extensively characterised by IHSS (Averett et al. (1989)). As a third organic, 5000L of surface water from the Mooney Mooney Dam (Brisbane Water National Park, NSW, Australia) were concentrated using microfiltration (MF) and reverse osmosis (RO) and freeze dried. The procedure is described in Appenlx 1. Aldrich HA, a commercially available product (Sigma Aldrich, Australia) was used for comparison in some experiments. This HA is not from a aqueous source, but nevertheless frequently used in the literature. Copyright © 2001 by Andrea I. Schafer Membranes 93 Further characterisation is reported in the organics characterisation section below. An overview over some characteristics is also shown in Chapter 2. The organics were prepared as 100 mgL-l organic carbon stock solutions by mixing the dry powder with MilliQ water without increasing the pH. The solutions were stored at 4°C in the dark. The amount of powder required for 100 mL stock solution was 18.4 mg, 18.6 mg and 200 mg for HA, FA and NOM respectively. This reflects the carbon content of the organics. Hematite was selected as a model colloid in this study due to its well understood aggregation behaviour, the monodisperse, spherical nature of the colloids and the fact that the synthesis of colloids of various primary particle sizes (40 to 500 nm) is possible. WGLlle silica and clays may be more abundant in surface waters, hematite appears to be a good compromise between real systems and a simple model compound. The synthesis of monodispersed, spherical hematite colloids of four primary particle sizes is described in detail in Appendix 3. The main properties of these colloids are also given in Appendx 3. Commercially available flat sheet membranes were selected. The primary selection criterium was that the membrane be made of a hydrophlic material, which adsorbs less organics than more hydrophobic polymers. For comparison, the membranes used are listed in Table 4.2 with their pore size or molecular weight cut-off (MYVCO) as specified by the manufacturer. Table 4.2 Characteristics OfMF, UF and NF membranes used in experiments. Process Supplier Type Typical Specifications Pure Water Surface Charge Operating Pore Size [pm] Flux at pH 8 Pressure Molecular Weight Fm-2h-l] [bar1 cut-off pal MF Mdhpore GVWP GVHP UF hblhpore PLHK PLTK PLGC PLCC PLBC PWC NF Fluid CA-UF Systems TFC-SR TFC-S TFC-ULP Copyright © 2001 by Andrea I. Schafer 94 MATERIALS AND METHODS This characterisation is relatively vague, as different methods are used by each manufacturer (Readman (1991), Thorsen et al. (1997)). As a more comparable parameter, the pure water fluxes as determined in the experiments are also given, as well as the membrane zeta potential at pH 8. A new membrane was used for each experiment (except for fractionation experiments). The results of surface charge measurements of the membranes as a function of pH, pure water fluxes and electronmicrographs are shown in the MF, UF, and NF chapters, respectively. 4.4.1 Microfdtration Membranes Two microfiltration membranes @hllipore, hydrophilic (GW) and hydrophobic (GVHP)) with nominal pore sizes of 0.22 pm were used. The hydrophlic membrane is a modified hydrophobic membrane. The hydrophilic membrane was chosen for most experiments because hydrophlic membranes have a reduced adsorption capacity towards hydrophobic organics (Jucker and Clark(1994)). The membrane material is a modified polyvinylidene fluoride (PVDF). The hydrophobic membrane was soaked in a 50% ethanol solution for 10 minutes to wet the pores and then rinsed with MilliQ water. All membranes were soaked in warm MdliQ water for 30 minutes prior to use to remove any organic contamination. 4.4.2 Ultrafiltration Membranes Ultrafiltration was used for fouling, rejection, and fractionation experiments. The fractionation experiments require membranes with very low adsorption characteristics to reduce loss of organics on the membranes. It was thus necessary to find low fouling membranes, whch are available in a range of membrane molecular weight cut-offs QWXCO). The fillipore "PL series" fulfil the low adsorption condtion and they are available in seven MWCOs in the range from 1 kDa to 300 kDa. The fractionation membranes selected were the PLAC, PLBC, PLCC, PLGC, PLTIC, and PLHIC with MWCOs of 1, 3, 5, 10, 30, and 100 kDa, respectively. Fouling and rejection experiments were carried out with the 10 and 100 kDa membranes. These regenerated cellulose membranes on a non-woven polypropylene substrate are described by the manufacturer as low protein-binding and hydrophlic. The MWCO (as described in Table 4.2) is determined by a range of Dextran markers. A MWCO of 10 kDa means that 90°/o of markers with a molecular weight greater than 10 kDa were retained. Prior to use, the membranes were soaked in 0.1 M NaOH for 30 minutes and flushed with 3.4 L of MilliQ water in order to remove the glycerin preservative, whch can strongly interfere with UV and DOC analysis. Alternatively, flushng the membrane with 1L fiUiQ also removed the glycerin sufficiently. 4.4.3 Nanofdtration Membranes Nanofiltration membranes were received from Fluid Systems in San Diego, USA (now Koch Membrane Systems). Thn film composite membranes were chosen due to their low fouling characteristics compared to polysulphone membranes used in other studes. The CA-UF membrane is, as the name suggests, classed as a UF membrane and the material is cellulose acetate. However, it is treated as a NF membrane here as it is often used for similar applications according to the manufacturer, and also because it exhibits some salt rejection. Membrane characteristics as given from Copyright © 2001 by Andrea I. Schafer Membranes 95 the supplier are summarised in Table 4.3. The cut-off was specified to be about 5 kDa and the material is non-ionogenic. The active layer of this membrane is about 150 nm. CA membranes have generally a 50% lower flux than TFC membranes, but are cheaper. The TFC membranes are chemically modified to render the membranes more hydrophilic, but more details were not available. All three membranes have different additives and post-treatments in the manufacturing process. The manufacturer estimates the thckness of the active layer of the TFC membranes to be 150 to 200 nm. For the TFC-SR membrane a dfferent monomer was used compared to the other TFC membranes. \%le the TFC-S and TFC-ULP membranes are made from metaphenylene diarnine with acid chloride (a benzene ring with two to three carboxglic acid groups), the TFC-SR membrane is fabricated from a mixture of cyclo-aliphatic amine with acid chloride. This means that the TFC-S and TFC-ULP have both positive and negative functional groups, whereas the TFC-SR membrane has negative functional groups only. Marker tests with 1% lactose (180 Da) solutions at pH 6-7 showed a rejection of 94.4% and 90.6% for the TFC-SR and TFC-S membranes, respectively. Rejection of the membrane is expected to be higher ('I'akigawa (1999)). Table 4.3 Membrane Infornation from Flziid Systems Corporation (now Kocb Membrane Systems), San Diego. TFC-S TFC-SR TFC-ULP CA-UF Material Test Condttions Flux pH range Rejection Design Application Design Pressure Storage Medium Pretreatmen t TFC Polyamide (PA\) on Polpsulfone (PS) base 1 g/L NaC1, 2.5 g/L AlgSO, 25OC pH 7.5 5.6 bar 14.7 L/m'h 4-1 1 95°/a hardness, 85% C1 NF or softentng of municipal water at dtralow pressure; up to 45°C 5.6 bar (560 kPa) 0.5% sodmm meta bisulfite, ALilliQ after wash wash wlth SIdhQ TFC proprietary P-\ on PS TFC P-1 on PS base base, coated with PT';\ (dye to check for damage) 1 g/L NaCl 2 g/L NaC1 2.5 g/L AlgSO, 25°C pH 7.5 7 bar 25OC pH 7.5 5.6 bar 14.7 L/m2h 14.7 L/m2h 98.5% hardness, 98.5% C1 nanofiltration or softening Industrial or municipal of municipal water at water ultralow pressure; up to ultralow pressure 1 ppm C12; up to 45 "C 5.6 bar 3.5-12.25 bar ALilliQ after wash SUiQ after wash wash with warm AlilliQ to soak in XLdliQ remove PT',I coating Cellulose Diacetate tap water 3.5 bar 16.5 L/m2h 4-6 Not specified Surface water at moderate pressure if chlorination desired (up to 1 ppm C12) 3.5 bar (560 kPa) unknown wash with ALtUIQ Copyright © 2001 by Andrea I. Schafer 96 MATERIALS AND METHODS All membranes were stored in a refrigerator (4 K) in plastic bags in the medium in which they arrived, and sealed. A few membranes of each type were cut out, pretreated and then placed in a Petri dish in the refrigerator for use in experiments. Stirred cell systems were selected for the experimental work for a number of reasons; (i) volumes are small whch is required for the use of IHSS reference material, (ii) membrane samples are small which allows the use of a new membrane for each experiment, (iii) the solution chemistry can be precisely controlled, (iv) experiments are relatively short and thus the investigation of a great number of parameters is possible, and (v) the concentration in the cell represents the concentration in a crossflow module (recovery about 70%). A comparison of mass transfer values was demonstrated in the case of NF in Chapter 7. Drawings of the filtration equipment are shown in Appendix 2. A hydrodynamic analysis is also shown in Appendix 2. 4.5.1 Microfiltration Equipment All experiments were carried out in a magnetically stirred batch cell (volume of 110 mL, membrane area 15.2 . 10-4 m" at a pressure of 100 kPa (if not otherwise indcated), pressurised with nitrogen gas. A reservoir of 1.5 L volume was connected to the stirred cell. A photo of a Perspex stirred cell with reservoir, manufactured in the university workshop, is shown in Figure 4.1. Figure 4.1 Perqex stirred cell with reservoir. All stirred experiments were stirred at 270 rpm (measured with a Philips PR 9115/00 stroboscope). A balance and stop watch were used to measure permeate volume. Experiments were conducted at a temperature of 25 + 1 OC. Copyright © 2001 by Andrea I. Schafer Filtration Equipment 4.5.2 Ultrafiltration Equipment The same system as described in the MF section and shown in Figure 4.1 was used for all rejection, fouling, and fractionation ultrafiltration experiments. The balance was connected to a PC for flux data collection. 4.5.3 Nanofiltration Equipment Nanofiltration experiments were carried out in a stainless steel stirred cell with an Amicon magnetic stirrer on a magnetic heater plate (Industrial Equipment & Control, Australia). The calibration is shown in Figure 4.2. The volume of the cell was 189 mL, the inner dameter 56.6 mm (resulting in a membrane surface area of 21.2 10-"% The stirrer speed could be varied from about 200 to 2000 rpm, with a setting of 400 rpm used routinely. The stirrer speed was measured using a Phlips PR 9115/00 stroboscope. One side of the stirrer bar was labelled to avoid measuring of half rotations. Figure 4.2 Calibration of magnetic stirrer table. Figure 4.3 Stainless steel stirred cell set-zp. Copyright © 2001 by Andrea I. Schafer 98 MATERIALS AND METHODS The stirred cell was pressurised with instrument grade air. An- was used (rather than N 2), to provide C02 for the carbonate buffer. pH changes due to the high pressure air were estimated to be less significant than with N2 (see Appendix 5 for details). A photo of the set-up is shown in Figure 4.3 and a schematic in Figure 4.4. The cell was equipped with a pressure gauge mounted in the stainless steel line after the air cylinder, a stainless steel reservoir with a volume of 2 L, a pressure release valve, a fluid inlet and outlet connection, a pressure safety valve, and a refill opening on top of the reservoir. On top of the stirred cell, a fluid inlet connection, a pressure release valve and a temperature probe fitting were mounted. The temperature was measured with a PT 100 probe, connected to a Kane-May ISM 330 indicator. To control the temperature inside the cell, it was placed in a 2 L plastic beaker, through whch tap water was circulated continuouslp. The temperature was kept constant (unless otherwise indicated) at 20 "C k 1 ()C. Permeate flux was measured by weight with a Mettler-Toledo PR 2002 (0.1 to 2100g) balance, whch was connected to a PC equipped with Mettler-Toledo BalanceLink software. Figure 4.4 Stainless steel stirred cell set-up. A: stirred cell: z~olme 185 mL; B: magnetic stirrer (Amicon, dtiven bJy magnetic stirrer table); C: membrane; D: stainless steel porons support; E: reseruoir uolnme 2000 mL, F: pressurired instrtlment air inlet, G: feed inlet, presswe release and safe9 valves; H: permeate outlet (to balance and PC). 4.6.1 pH Value A Beckrnan glass electrode (Ag/AgCl) was used for solution preparations and no contamination was observed. The electrode was only used in samples after DOC analysis and was cleaned prior to use for pH adjustment. 4.6.2 Conductivity Conductivity was measured using a Lutron CD-4303 portable instrument. 4.6.3 Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) A Perkin Elmer Optima 3000 Spectrometer was used to determine the cation content of solutions. Samples and multielement standards (0, 1, 10 and 100 mgL-l) were diluted with 5% nitric acid. All vials used were cleaned with 1 M sulphuric acid. Detection limits are 3, 5,0.1, 5, and 70 pgL-' for Fe, Al, Ca, Na, and I<, respectively. Copyright © 2001 by Andrea I. Schafer Organics Characterisation 99 The particle sol and filtration samples were diluted 1:l with HC1 (36'Yo) and heated (in a closed sample vial) to dissolve the colloidal hematite. These samples were then analysed directly. 4.6.4 Ion Chromatography (IC) IC was used for chloride determination for NF rejection experiments. Anions could not be analysed using ICY as humic substances interfere with the analysis (Hoffmann et al. (1986)). A Millipore Waters Model 590 instrument was used with a Model 430 Conductivity detector. The eluent used was 0.68 gL -1 boric acid (H;BO3), 0.235 gL-' gluconic acid anhydride (C6H1006) and 0.3 gL-1 lithum hydroxide @OH -6 HzO). 4.7.1 Dissolved Organic Carbon (DOC) Dissolved organic carbon was analysed using a Skalar 12 instrument. The method is based on UV- persulphate oxidation and described in detail in Appendix 4. The DOC of every sample was measured as a routine analysis. For samples containing colloids, aggregates or flocs the measured value is total organic carbon (TOC). None of the samples were filtered as this would lead to loss of organics. 4.7.2 UV/VIS Spectroscopy A Varian Cary 1E UV/VIS Spectrophotometer was used to evaluate the method and for further standard analysis. Spectra of UV/VIS in the range from 190 to 500nm were obtained and correlations established with DOC analysis. The method is further described and evaluated in Appenchx 4. UV/VIS was also a routine analysis and the wavelength was used in rejection calculations. At low wavelength (190 nm region), absorption by inorganics is observed. This is strong in the case of unpurified Mooney Mooney NOM and absent in the purified IHSS samples. The ion content of all samples is shown in section 4.7.6. The W/VIS spectrum of NOM is attributed mainly to absorption of light energy by aromatic compounds and can be broken into a series of transition bands, similar to those published for benzene (Korshin et al. (1997b)). Three transition bands can be distinguished for each aromatic chromophore in NOM - the local excitation (LE) band, the benzenoid ('2) band, and the electron-transfer (ET) band. The peaks vary in their height, width, and centre location depenchng on the composition of the NOM (Kaecbng (1998)). The presence of these various peaks can be recognised in the shoulders on the spectra as shown in Figure 4.5, however detailed analysis was not considered warranted. From Figure 4.5, it can be seen that the (probably) soil-derived fidrich HA (purified with a lOOkDa UF membrane) has the largest UV/VIS absorbance, followed by IHSS and the NOM HA fraction whch are surprisingly similar. The FA fraction of Mooney Mooney NOM has a higher absorbance than the unpurified NOM, which can be explained given the NOMs relatively high content of hydrophlic acids of a very low absorbance. The IHSS FA also has a slightly lower absorbance over the complete wavelength range. Copyright © 2001 by Andrea I. Schafer 100 MATERIALS AND METHODS 5 0.15 Figure 4.5 CV Spectra of the organics zmd. all wavelengths linear with concentration IHSS FA NOM IHSS HA Aldrich HA (c100 kDa) NOM Hydrophilic Fraction NOM HA Fraction 200 250 300 350 400 450 500 2 UVIVIS Wavelength [nm] 4.7.3 Titration The NOhl sample, which was concentrated as described in Appendix 1, was titrated using a Metrohm automatic titrator. .The titrator was operated in dynamic titration mode. The samples were acidified from ambient pH to pH 2.8 with 0.1 M HNO3 and subsequently alkalised with 0.1 M NaOH to pH 10. It was assumed that at pH 2.8 all acidic functional groups will be saturated, whereas at pH 10 all carboxylic and half of the phenolic groups were dissociated. The limitations of these assumptions were discussed in Chapter 2. The titration vessel was purged with nitrogen to eliminate C0 2. From the volume and molarity of added base and the mass of titrated DOC, the content of acidic functional groups can be calculated. Carboxylic acid content was calculated from the amount of base added until the end-point was reached. Phenolic acid content was calculated as twice the difference in titrant required to change the pH of the titrate from 8 to 10, since it was assumed that at pH 10 only half the phenolic groups were Issociated. A solution of a concentration of 20 mgL-I as DOC NOM were titrated. The error due to the salt content of NOM is likely to be high. Table 4.4 describes the acidq and size of the three organics used and the average molecular weight as found in the literature (for IHSS organics and purified Aldrich HA) or as measured (for NOW. The reported rvlW will be verified later (see section 4.7.7) by analysis. Table 4.4 Acid$ and average molecdar weight ofthe organics ( y~cker and Clark (1984),'Beckett et al. (1987), 'Elering and Morel(1988),'ana&red by titration (lee above), 'Clark andhcker (1993), 'Children and Elimelecb (1 996)). Type of Organic Acidq [meq.gl] liverage Molecular Weight pal Carboxylic Phenolic IHSS FA 3.41 5.45 6.1 1.51 2.05 7502 IHSS HA 4.01 4.16 2.9' 2.16 1 1 00" 15002 12004 Purified Aldnch LA 3.3(j 2.56 > 50 0006 Mooney Mooney NOM 5.14 1.34 < 10004 Copyright © 2001 by Andrea I. Schafer [...]... (2.13) 0.28 (2.33) Ca [mgL-l] 0.22 (0.22) 0 (0) 62.6 (62.6) 0.61 (0. 24) 0. 24 (0.21) 48 .6 (219.9) 0. 94 (7.83) Fe [m&'] 0.11 (0.11) 0 (0) 1 .41 (1 .41 ) 0 .46 (0.18) 0.36 (0.31) 1.2 (5 .43 ) 0.15 (1.25) Na [mgL-'1 1.52 (1.52) 0.23 (0.23) 296 (296) 3.16 (1.26) 3. 54 (3.09) 244 (11 04. 1) 12.3 (102.5) I< [mgL-l] 0.55 (0.55) 0 .41 (0 .41 ) 52 .4 (52 .4) 2.16 (0.86) 1.19 (1. 04) 1 .43 (6 .47 ) 0 .47 (3.92) 4. 7.7 High Performance... should be linear (see Figure 4. 30) 6 5 4 0 - m 00 tb - 21 HA, pH 8 97 93 - 3m 0 the slope of the graph is related to the density of the aggregates; j n i r chloride jlocs with 5 m&' ar DOG 100 r n g ~ 'FeCI, = ' D - I Y Figure 4. 30 Scattering intensio over scattenkg vector; 25 m g ~ FeCI, ' % % l - ~ s r n O c l 0 -5 .0 l -4 .5 ' l ' -4 .0 log Q 4. 9.5 l -3 .5 ' l -3 .0 ' -2 .5 [-l Photon Correlation Spectroscopy... points somewhat subjective Figure 4. 28 Results of small angle static light scattering studies of hematite agregatesformed at 10 mA4 and 100 mM KCl @H stirrer speed 220 ?m) 3, l0 rnM 2 1 , 1 -5 .0 -4 .5 KC1 (high 0 ) , 1 , 1 -4 .0 -3 .5 , log Q X - [-l 1 -3 .0 , , -2 .5 -2 .0 For the complex systems used to model surface waters, the aggregation cannot be explained so easily Figure 4. 29 shows the log I versus log... IHSSHA 10 - 0 E Organics used Other organics 4- 0 + 0 Main fulvics f om a + 2 sewage treatment 0 - C 0 a , Q cn Rhine Karst 4 HS-Hydrolysates 0 I 250 500 Kleine Kinzig I I I I 750 1000 1250 1500 - Number Molecular Weight [gmol-'] Molecularity Figure 4. 15 Seine 1750 lHSSFAStd IHSSHAS~~ HunzzfZcation diagram for the natural organics as used in this s t u 4 and other organics as reported L Htlber (1 998)... Permeate 12 1.2 18 14 1. 84 10-l[' 1.08 Mooney NOhl 100 2.0 1381 2.13 10-l'' 0. 94 NOM HA Fraction 250.3 2.5 1857 1.82 IO-'" 1.10 NOhl FA Fraction 1 14. 5 1.1 1318 2.18 1 0-1 " 0.92 NOM Hydrophhc Fraction 22.1 2.2 970 2.56 1 0-' " 0.79 The results correspond very well to the SEC results except for the NOM fractions where the LC-OCD results make more sense The molecule size is in the 0.7 to 1 .4 nm size range... the 320 0-3 500 cm-' region is due H-bonded groups such as -CH, -OH, and NH The shoulder peak at 2900 cm-' is likely to be aliphatic methylene groups (fatty acids, waxes) Peaks in the 1650 to 1750 cm-l region can be attributed to the C=O of quinones, ketones, and maybe aromatic C=C vibrations, whlch are very weak Carboxylates are found in the 1600 cm -l region and also show a small peak at 144 0 cm-l, and... Reference 52.89 4. 1 43 .40 1.17 0.58 . 296 (296) 3.16 (1.26) 3. 54 (3.09) 244 (11 04. 1) 12.3 (102.5) I< [mgL-l] 0.55 (0.55) 0 .41 (0 .41 ) 52 .4 (52 .4) 2.16 (0.86) 1.19 (1. 04) 1 .43 (6 .47 ) 0 .47 (3.92) 4. 7.7 High Performance Size. 1 Organics used - Molecularity - E 0 E 10 - 5 1 0 8 - X 0 .= a . 6- S. ca r E .g 2 + a 4- E 0 fulvics f 2 a sewage 0 C 0 4 0 a, Q Figure 4. 15. (62.6) 0.61 (0. 24) 0. 24 (0.21) 48 .6 (219.9) 0. 94 (7.83) Fe [m&'] 0.11 (0.1 1) 0 (0) 1 .41 (1 .41 ) 0 .46 (0.18) 0.36 (0.31) 1.2 (5 .43 ) 0.15 (1.25) Na [mgL-'1 1.52 (1.52) 0.23