Environ. Sci. Technol. 2002, 36, 4114-4120
On-Site Classification of Humic-Rich Hydrocolloids and Their Metal Species by Means of Online Multistage Ultrafiltration PETER BURBA* AND JOHAN VAN DEN BERGH Institute of Spectrochemistry and Applied Spectroscopy, Bunsen-Kirchhoff-Strasse 11, D-44139 Dortmund, Germany
Humic-rich hydrocolloids and their metal species (e.g., Al, Fe, Ca, Cu, Fe, Mg, Mn, Pb, Zn) in selected bogwaters of German origin (Black Forest (HO), Muensterland (VM), Arnsberger Wald (AW)) were size classified in the subparticulate and macromolecular range by on-site multistage ultrafiltration (MST-UF). For this purpose two MSTUF cascades, equipped with up to eight conventional flat membranes (0.45, 0.22, 0.1 µm (Millipore); 100, 50, 10, 5, 3, and/ or 1 kDa (Gelman PallFiltron OMEGA), were coupled and used for an online fractionation of hydrocolloidal matter immediately after water sample collection. Quantification of dissolved organic carbon (DOC) and metals was carried out off-site by conventional laboratory methods (carbon analyzer, atomic spectrometry). The size distribution of some humic-rich hydrocolloids (e.g. in HO) was exhibited to be surprisingly stable even over storage periods of 1-4 weeks. On the other side, their size distribution (e.g. in VM) considerably varied during the collection period. The natural metal loadings (e.g. Al, Fe, Mn, Zn) of hydrocolloids showed characteristic size distributions and, mostly, strong metal enrichment in subparticulate and macromolecular fractions. Further results of on-site classification of hydrocolloids, obtained by parallel single-stage ultrafiltration (P-UF) with the same membranes, were only comparable to those of MST-UF in the case of low DOC concentration.
Introduction Major fractions of organic and inorganic contaminants (e.g. pesticides, heavy metals) in aquatic environments can be bound to particulate and colloidal matter (1-4). As a consequence, the mobility and bioavailability of contaminants is strongly dependent on the nature and the behavior of their carrier hydrocolloids. Recent papers have revealed that colloids in surface and groundwaters are heterogeneously composed of natural organic matter (NOM), mainly humic substances (HS), and subparticulate inorganic constituents, particularly hydrated element oxides (5-7). Overviews on typical colloids and macromolecules in natural waters have already been presented (3, 8-10). By means of TEM (transmission electron microscopy), for instance, it could be shown that hydrocolloids in an eutrophic lake water mostly consisted of irregular networks of organic and inorganic entities (11) in the size range from some nanometers up to the micrometer level. Aggregates of larger size in surface * Corresponding author phone: ++492311392-181; ++492311392-120; e-mail:
[email protected]. 4114
9
fax:
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 19, 2002
waters can be attributed to the class of particles gradually settling to sediments or retained by filtration in porous media (11). In particular, conventional ultrafiltration techniques have successfully been employed to fractionate such hydrocolloids (3, 12, 13), despite general limitations of ultrafiltration membranes, especially their polarization and clogging effects in the presence of high NOM concentrations, their quite operational and conditional “molecular weight” cutoffs, their rather slow working time and the possible formation of macromolecular artifacts (14-17). In principle, colloids in natural waters behave like dynamic systems often subjected to fast physicochemical alterations which might be caused by complex sorption, coagulation, and precipitation processes (9). Probably, chemical reactions as well as microbial activities are other important processes for both the formation and degradation of hydrocolloids. Due to the complexity and variability of hydrocolloids, however, their characterization is still a complicated analytical task. It has to be taken into account that samples of natural waters containing colloidal matter are in principle unstable. To achieve reliable results that are representative of the natural water under study, it is mandatory to characterize hydrocolloids as quickly as possible after sampling, using appropriate on-site or even in-situ procedures, if available. In a recent overview that considers the delicate problems of reliable characterization of natural hydrocolloids it has been pointed out that an urgent demand exists for powerful field procedures enabling an improved on-site and, if required, in-situ characterization of colloidal matter (14, 18). The present study was focused on the development of new on-site classification procedures for colloidal NOM in natural waters. According to our analytical strategy, hydrocolloids and their metal species in natural waters were sizeclassified on-site, directly after water sampling, and their fractions quantified afterward off-site by conventional laboratory methods. For the on-site size-classification of hydrocolloids a mobile variant of tangential-flow multistage ultrafiltration (MST-UF), already successfully utilized for reliable online fractionations of aquatic HS (16, 17) and particulate matter (7, 19) in various water samples, was chosen. For comparison, conventional ultrafiltration through a series of parallel arranged membranes was used. In addition, humic colloids in surface waters were assessed on-site by means of mobile UV/vis spectrophotometry based on the spectral absorption coefficients a of humic matter at 254 and 436 nm, respectively. Another point of interest was to determine the size distribution of natural metal species in NOM-rich waters. For this purpose, the on-site fractionated hydrocolloids and their metal species were quantified off-site by means of conventional instrumental methods (e.g., carbon analyzer, atomic spectrometry) in the laboratory. Potential sizetransformation of hydrocolloids stored for different periods of time was a further item of this study.
Experimental Section Sampling Procedure. Water samples were freshly collected in acid- and base-cleaned 20-L polyethylene containers, immediately before their on-site characterization. First, the water sample under study was coarsely filtered through a clean cotton fabric. Then, a 200 mL-portion was filtered through conventional 0.45 µm-flat membranes (Millipore Durapore, diameter: 47 mm; 1-2 membranes depending on the amount of particulate matter). After their on-site characterization, all filtered samples were stored for a couple 10.1021/es010210r CCC: $22.00
2002 American Chemical Society Published on Web 08/24/2002
FIGURE 2. Multistage ultrafiltration: concentration decrease of small molecules during their wash-out, dependent on the wash volume and the stage (stages: F1, F3, F5, F7, F8; V: wash volume, VS: stage volume (8 mL, each)).
FIGURE 1. Flow scheme of online multistage ultrafiltration (Mn: flat membranes of decreasing pore size, Fn: fractions obtained, Rn: flow-through reservoirs for fraction collection). of weeks without any stabilizing agents (e.g., mineral acid, sodium azide solution, formaldehyde) at 5-10 °C in the dark. Multistage-Ultrafiltration (MST-UF). The principle of the online MST-UF devices used for the size fractionation of hydrocolloids is shown in Figure 1 (one of two cascades). The MST-UF device was developed at the Vernadsky Institute of Geochemistry and Analytical Chemistry, Moscow, based on a Russian/German cooperation (20). Further details of the applied MST-UF technique and its operating conditions have already been described elsewhere (16, 17, 19, 20). In principle, the applied field MST-UF consisted of two serially coupled cascades which were equipped with conventional flat membranes (first cascade: 0.45, 0.22 and 0.1 µm, Millipore GV type, diameter: 47 mm; second cascade: 100, 50, 10, 5, and 3 kDa, Gelman PallFiltron OMEGA series, diameter: 47 mm). By installing hydrophilized PES (poyethersulfone)based membranes (OMEGA) into the MST-UF cascade a relatively low binding of humic-rich DOM (dissolved organic matter), reduced clogging effects and enhanced permeate rates could be gained (17). Both cascades were coupled by an intermediate flow-through reservoir (100 mL) and operated by conventional multichannel peristaltic pumps (Gilson Minipuls 2) which were powered by a 60 Ah car battery. The permeate flux of the first cascade was about 5-6 mL/min, and that of the second cascade 0.5-1.2 mL/min, depending on the concentration of DOM. The tangential flow over the membranes was kept at 10-12 mL/min; the pressure applied to the first stage of both cascades was about 1.5 bar. To detect undesired concentration polarization and clogging effects of DOM onto the membranes of the MST-UF device its permeate flux was regularly controlled, as suggested in ref 15. The cell volume VS within each cascade stage was approximately 8 mL. Before operation the cascades, including their membranes, were purified by 0.1 mol/L HCl and 0.1 mol/L NaOH, respectively, and, then, filled with high-purity water (MilliporeQ system). After fractionation of the water sample (200 mL, each), the colloid fractions generated within the cascades were washed online by 80 mL of high-purity water and, then, removed from their reservoirs Rn (see Figure 1) for further characterization. In total, such a MST-UF run of a 200 mL water sample required 4.5-6 h.
The complex concentration dynamics of dissolved macromolecules, penetrating online the membranes of a MSTUF cascade, and their successive wash out can be modeled by the material balance of a cascade of continuous agitator vessels connected together by a filtration stage, as previously outlined in ref 21. In an ideal case, the macromolecules under study either quantitatively penetrate the installed membranes (permeation coefficient PC ) 1) or are completely retained on one of them (PC ) 0), dependent on their molecular size. The retention and permeation of macromolecules in one of the UF stages can be described by the simplified eq 1
cp ) cr‚PC
(1)
where cp is the concentration of permeating molecules and cr is the concentration of retained molecules. Introducing pure wash water into a single UF stage of constant cell volume, the continuous wash-out of molecules, permeating the installed membrane on the basis of a conditional PC, can be assessed by the well-known eq 2
cr ) cro‚e -PC V/Vs
(2)
where cro is the concentration of retained molecules at the beginning of UF, V is the volume of the wash solution (L), and Vs is the cell volume of the UF stage (L). Consequently, molecules having a rather small permeation coefficient (e.g. PC < 0.25) for that UF stage need a considerably increased volume V of wash water for their extensive permeation. In the case of a multistage UF cascade, as used in the present study, the continuous washout of molecules from an UF stage n into the successive one (n + 1) can principally be assessed by a differential equation of the type (3), according to ref 21
Vs
dcrn ) Pf(cp(n-1) - cpn) dt
(3)
where Pf is the permeate flow (L/h), cp(n-1) is the concentration of molecules transported into the stage n, and cpn is the concentration of molecules washed out from stage n. The differential equations obtained for all stages of an online MST-UF unit can generally be solved by means of Laplacian transformations, in analogy to the material balance of a cascade of continuous agitator vessels as discussed in ref 21. The MST-UF unit applied consisted of eight filtration stages (F1-F8), each having a constant cell volume Vs of approximately 8 mL. Using equations of the type (3) a set of washout curves for small molecules, as represented in Figure 2 for the UF stages F1, F3, F5, F5, F7, and F8, can be assessed. They exhibit that a wash water volume of 10 Vs is practically sufficient to remove widely small molecules (PC ) 1 assumed VOL. 36, NO. 19, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4115
TABLE 1. Location and Characteristics of the Bogwaters Studied On-Site total metal contentc (µg/g DOC) label HO16 HO17 VM6 VM7 VM8 VM11 AW3 AW4 a
location Hohlohsee (HO), Black Forest, Germany Venner Moor (VM), Northrhine-Westphalia, Germany Arnsberger Wald (AW), Northrhine-Westphalia, Germany
DOCa
(mg/L)
25.1 ( 0.6 22.9 ( 0.4 70.3 ( 1.5 71.5 ( 2.7 83.3 ( 0.1 77.8 ( 0.6 14.1 ( 0.1 13.1 ( 0.3
a254 nmb
a254/a436
pH
Al
Fe
Mn
Zn
3.45 3.98 3.95 3.65 3.86 4.23 3.83 3.86
10.3 11.2 9.5 9.2 10.4 7.4 13.6 15.6
3.5 4.2 3.8 4.0 4.2 4.5 3.9 4.2
4100 3490 14660 9090 8570 8770 3475 4198
11150 9470 26100d 23170 19340 21600 62200 62520
360 260 1090d 1010 770 940 5320 8780
4420 3540 970d 1510 1010 1430 3760 7330
sr from n ) 5. b a ) absorbance A‚L‚m-1‚mg-1 of DOC. c By FAAS and ICP-OES, respectively (mean of 3 runs). d TXRF values (see ref 21).
for all membranes) from that MST-UF unit, except small remnants in F7 and F8. To achieve an operational compromise between the wash water volume required and its quite slow flow-through (0.5-1 mL/min) a relatively small wash water volume of 80 mL was used. In the case of electrolyte-rich humic colloids fractionated by MST-UF (16), however, it has to be taken into account that an operational wash-out of colloid fractions by highpurity water can systematically shift their size distribution. Additionally, humic colloids can be subject to further size alterations when the pH value and humic-substance concentration are considerably changed (16). On the other hand, bogwater samples of relatively low conductivity, as investigated in this UF study (conductivity: 35-85 µS/cm, dependent on the sample), might be washed out even by high-purity water, provided that the pH value of the UF fractions is efficiently buffered by the inherent fulvic and humic acids. Potential clogging effects of hydrocolloids during their fractionation on ultramembranes, “broad” nmw cutoffs of the ultramembranes applied and conditional CP values between >0 and 5 mg/mL DOC) a preceding DOM digestion was carried out, preferably by means of an oxidizing UV photolysis (UV-digester 705, Metrohm AG, Switzerland; conditions: 10 mL-sample, 3% H2O2, pH 2.0 adjusted by HNO3, 1 h digestion). Metal blanks 4116
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 19, 2002
of the applied MST-UF technique were assessed by processing 200 mL of high-purity water (MilliporeQ) through the MSTUF device prepurified by 0.1 mol/L NaOH (suprapur) and 0.1 mol/L HCl (suprapur), respectively. ICP-OES. Metal determinations in water samples were preferably carried out by simultaneous ICP-OES (spectrometer: Thermo Jarrell Ash IRIS AP) according to the recommendations of the manufacturer. To achieve detection limits as low as possible, sample volumes of 6-8 mL, each, were nebulized (Meinhard nebulizer, 2 mL/min flow rate), and integration times of 40 s (3 measurements) were chosen. AAS. Metal determinations at the lower µg/L-level were carried out by graphite-furnace AAS (GF-AAS, spectrometer: Perkin-Elmer Zeeman/3030) using the standard addition method for calibration and the Zeeman effect for background correction. TXRF. The accuracy of the ICP-OES and AAS determinations in humic-rich samples was controlled by energydispersive TXRF, according to refs 22 and 23. 100 ng/mL Ga spiked to the samples under study was used as internal standard, for X-ray excitation of the analytes a molybdenum tube (spectrometer: Seifert EXTRA II). The “lifetime” of the measurements was 300 s, each. Direct TXRF determinations of humic-rich aqueous solutions (
