Environ. Sci. Technol. 2001, 35, 4530-4535
DGT as an in Situ Tool for Measuring Radiocesium in Natural Waters C H R I S M U R D O C K , † M I K E K E L L Y , * ,‡ LING-YUAN CHANG,‡ WILLIAM DAVISON,‡ AND HAO ZHANG‡ Department of Environmental Science, Lancaster University, Lancaster, LA1 4YQ, U.K. and Radman Associates, Harvey House, Bollington, Macclesfield, SK10 5JR, UK.
The application of diffusive gradients in thin-films (DGT) samplers for the measurement of cesium radionuclides in solution, using an ammonium molybdophosphate (AMP) binding agent, was tested under both laboratory and field conditions. In the former they proved able to reproduce known 134Cs concentrations (60 Bq L-1) with a high degree of accuracy and precision over periods up to ∼1 d, in freshwaters over a wide range of pH and temperature, and in saline water. In field trials in a freshwater lake receiving nuclear power station discharges, mean concentrations of 137Cs (47-61 mBq L-1) were measured over periods from 5 d to 1 month. These agreed, within error, with mean concentrations determined from grab samples but rigorous field validation of long-term (month) deployments of DGT devices proved impossible using conventional sampling procedures, due to loss of 137Cs to container walls. Identified limitations of the DGT technique included probable AMP degradation over longer periods and calibration problems if large changes in temperature and concentration occurred together. Potential limitations due to biofilm growth were considered not to be significant. Despite the limitations, the technique appears to measure concentrations accurately for deployment times of 1 month or less. It has several advantages over traditional sampling methods for monitoring radionuclides in the solution/dissolved phase, including its simplicity, provision of time-averaged mean concentrations, and automatic in-situ concentration onto a medium with ideal counting geometry for gamma spectrometry.
Introduction A variety of artificial radionuclides are found in the solution/ dissolved phase in freshwaters, as a result of authorized and accidental waste discharges, nuclear accidents, and nuclear weapons testing. As encountered in this investigation, conventional sampling methods do not easily provide mean concentration values over protracted periods of time when concentrations are changing. In contrast, the new technique of diffusive gradients in thin-films (DGT) enables a direct determination of mean concentrations of ions in solution over defined time periods, and it has been shown previously to be applicable to the measurement of many trace constituents (1). This paper describes the development of the * Corresponding author phone: 44-1524-593937; fax: 44-1524593985; e-mail:
[email protected]. † Radman Associates. ‡ Lancaster University. 4530
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 22, 2001
DGT technique for the measurement of radiocesium in freshwaters and an assessment of its potential for long-term use. A DGT device consists of two thin gel layers: a pure gel layer (diffusive gel) backed by a gel incorporating a binding agent (binding gel). A membrane filter can be added to protect the outer surface of the diffusive gel. If the binding agent reacts rapidly and strongly with the ions, then subsequent diffusion across the diffusive gel will occur at a rate determined by the concentration gradient and the ion diffusion coefficient. An additional zone of diffusive transport will exist at the boundary of the solution and the DGT device, i.e. the diffusive boundary layer (DBL). In well-mixed solutions, the DBL is considered to be of negligible thickness compared to the diffusive gel (2). In practice, the solution concentration can be calculated from the measured ion uptake by the binding gel and knowledge of the diffusion coefficient and diffusive gel (+ filter) thickness.
Methodology Gel Preparation. The diffusion and binding gels are polyacrylamide hydrogels, made using acrylamide monomer (Boehringer), a patented cross-linker derived from agarose (DGT Research, UK), ammonium persulfate initiator, and TEMED catalyst, according to the procedure of Zhang and Davison (2). The standard diffusion gel thickness used was 0.8 mm, with some use made of other thicknesses. In the binding gel, ammonium molybdophosphate (AMP) was used as the binding agent, as it has been widely used in the analysis of radiocesium in surface waters as an adsorbant in an initial concentration step (3). Earlier work had shown that although a standard cation-exchange resin (AG50W x 8, Bio-Rad) was an effective binding agent for stable Cs in simple solutions its capacity was soon exceeded in natural waters, due to its poor selectivity (4). The two gel components were cut into 2.5 cm diameter disks and assembled, together with an outer membrane filter (cellulose nitrate (Whatman), cellulose acetate (Whatman), or polysulfone (Gelman)), in a standard plastic DGT holder with a window of 2.0 cm diameter (1). This defines the effective area of the DGT device as 3.14 cm2. This assemblage is tightly held between a backing piston and the face of the frame around the window. The membrane filter, which extends the DGT diffusive layer thickness by 0.12-0.14 mm (depending on filter type), protects the gel from particles and biological films. Laboratory Experimental Procedures. Radionuclide uptake by DGT devices was investigated in a series of laboratory experiments under controlled conditions using synthetic analogues of natural waters comprising only inorganic constituents dissolved in Milli-Q water (5), with compositions corresponding to a circumneutral soft water lake (Lake Windermere (6)) and seawater (7). The DGT assembly holding the binding gel + diffusive gel + filter was immersed in a lidded polyethylene container of 4 L of synthetic waters spiked with a known mass of a certified 134Cs standard or stable Cs. The solution was constantly stirred, and its temperature was kept constant at 20 °C using a temperature controlled water bath. Periodically, DGT devices were removed, and small aliquots of solution were taken for analysis. The DGT devices were rinsed and dismantled before counting the binding layer. The effects on DGT uptake of time, pH, gel thickness, and temperature were investigated with a 134Cs concentration of 60 Bq L-1 in synthetic lake water and, for time, in synthetic seawater. For the determination of the capacity of AMP 10.1021/es0100874 CCC: $20.00
2001 American Chemical Society Published on Web 10/09/2001
binding gels, DGT devices were deployed for 24 h periods in synthetic lake water with increasing concentrations of stable Cs. The uptake rate was also determined on DGT assemblies after use in the field experiments, to identify changes in their properties, using synthetic lake water and a 134Cs spike of 189 Bq L-1. Field Experimental Procedures. The site chosen to field test the DGT technique was Llyn Trawsfynydd, Gwynedd, Wales; it being a lake receiving authorized radioactive liquid discharges from a British Nuclear Fuel plc (BNFL) MAGNOX nuclear power station. This shallow, partly artificial lake is situated at 200 m above sea level with a largely moorland catchment of ∼58 km2 in an area of high rainfall (∼2 m y-1). The lake has an area of ∼5.1 km2, a mean depth of ∼7 m, a volume of ∼36 Mm3, and an estimated mean residence time of 0.3-0.4 y (8). The depth of the original lake has been increased by low dams to enable abstraction, initially for a small hydroelectric plant and subsequently for cooling water for the nuclear power station. The lake is mesotrophic to occasionally eutrophic and circumneutral due to the effects of the discharges and liming by the fisheries management. The pH range measured during this work was 6.1-8.0. The nuclear power station is currently being decommissioned and discharges to the lake are made on average every 5 d (9). Artificial radionuclides recorded in the discharge in 1998, in order of decreasing magnitude, comprised 3H, 90Sr, and 137Cs, with a total of 6.51 GBq of 137Cs (10). A 5 month study was undertaken in 1998 using DGT sampling periods ranging from 5 days to 5 months, i.e., 5 d, 7 d, 14 d, 1 month (5 consecutive) and 2, 3, 4 and 5 month periods. The experimental site was 100 m SW of the discharge point and 10 m from the shore. DGT devices were deployed in multiple arrays, mainly positioned in a vertically deployed 50 × 50 × 0.7 cm plate holding 72 devices, uniformly spaced and back-to-back. The array was suspended, from a pontoon, at 2-3 m below the surface in a total water depth of 5-6 m. Other experiments, using long strings of individual devices, showed that the plate system did not adversely influence the results by increasing the boundary layer thickness. After retrieval, the DGT devices were lightly rinsed in the field in ultrapure water and then analyzed using the methods described below. Water temperature was logged at intervals of 15 s or 15 min during DGT sampling periods, depending on their length, using a TINYTAG temperature logger. The mean temperature for each deployment was used to calculate the appropriate diffusion coefficient. To validate the DGT measurements, bulked water sample were obtained by pumping at intervals into a container over a DGT sampling period, after which the samples were filtered and acidified. This approach proved to be successful only for 1 d periods, with brief pumping every half hour, because of the following problems. The inline filtering necessary to preserve the phase distribution of the Cs required frequent filter changes because of filter clogging. Without filtering, the prior acidification needed to prevent adsorption to container walls could not be done, as it would affect the particulate phase Cs concentration. Consequently, although this procedure was acceptable for 1 d sampling, it could not be extended over longer periods because, without acidification, experiment showed considerable losses of dissolved Cs occurred, e.g. up to 28% over 28 d with twice daily pumping. This is assumed to be the result of adsorption to the container walls, since the particulate phase present was shown not to be significantly involved (e.g. by algal growth and adsorption) by the constancy of the percentage of total Cs activity present in the solution phase throughout this period (44.1 ( 7.8%). Grab samples of the lake water were taken when the site was visited, and these were treated in the same way as the 1 d integrated pump samples. In addition, BNFL routinely
takes grab samples for monitoring purposes, and their results are discussed later (9). Their samples are taken 250 m from the discharge site (150 m further away than our sampling site) at 0.2 m depth. Binding Gel, Solution, and Particulate Phase Analysis. The 137Cs and/or 134Cs activity of gels and solutions was determined by gamma spectrometry, using n-type HPGe detectors. Binding gels were prepared for counting by slow drying under an infrared lamp, and, at a late stage in the drying, they were attached to the base of a counting vessel with microporous tape. Liquid samples were prepared by evaporation to dryness in a counting vessel. Efficiency standards for gels and liquid samples were prepared as above, after spiking with a known amount of a certified 134Cs or mixed gamma emitter standard. In the case of gel efficiency standards, a sufficient mass of stable Cs was also added, to saturate the binding and produce an homogeneously labeled gel. For the field experiment, binding gels were combined, in various numbers up to 15, to improve counting statistics and then treated as above. Liquid samples were filtered through Millipore filters prior to further analysis. For a limited number of samples, the particulate phase on the filter was digested in aqua regia and then dried as above. BNFL prepares their grab samples by filtration followed by concentration by passing through an AMP/silica gel cartridge, which is then analyzed by gamma spectrometry (9). The minimum detectable activity or concentration (MDA/ MDC) is based on the counting detection limit only, which itself is based on the dual criteria of a 95% probability of being above background and a 95% probability of detection. For the experiment using stable Cs to determine AMP capacity, the DGT uptake was determined by analyzing solution samples, taken before and after DGT deployment, by Zeeman furnace atomic absorption spectroscopy. Diffusion Coefficient. In the calculation of DGT measured concentrations, the value used for the diffusion coefficient of Cs in the diffusion gel was 1.9 × 10-5 cm2 s-1 at 25 °C determined by Chang et al. (4), corrected to the appropriate temperature conditions of laboratory and field experiments using the Stokes-Einstein equation (2). This predicts the unit flux of Cs activity through a 0.8 mm diffusion gel to be 0.016 Bq cm-2 d-1 per Bq L-1 solution concentration at 15 °C.
Results Laboratory Experiments. Experiments with DGT devices using 134Cs in solution in synthetic lake water and seawater showed a linear uptake of 134Cs over a period of 25 h (Figures 1a and 2a). The mean ratio of the measured to predicted uptake of 134Cs by DGT binding layers over the period was 1.00 ( 0.01 (n ) 12) from synthetic lake water and 1.03 ( 0.03 (n ) 12) from synthetic seawater (Figures 1b and 2b). These 95% confidence intervals are based on the propagated errors associated with both the measured value, derived from the binding gel analysis, and the predicted value, based on the tracer dilution in the experimental solution. These data demonstrate the accuracy of the DGT technique for measuring 134Cs in both aqueous media. Duplicate measurements (shown individually in Figures 1b and 2b and as the mean value in Figures 1a and 2a) indicate a precision of better than 4% for the uptake of 134Cs from synthetic lake water and a similar level of precision for the seawater experiment, except for one set. Other experiments using synthetic lake water indicated that DGT Cs uptake was independent of pH from 2 to 10 (Figure 3) and varied according to theory (2) with diffusive gel layer thickness (Figure 4) and temperature (via the diffusion coefficient) (Figure 5). Again a good degree of accuracy and precision was obtained, with the measuredVOL. 35, NO. 22, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4531
FIGURE 1. (a) DGT uptake of 134Cs with time in synthetic lake water (mean and 1 SD of two replicates) and (b) the measured to predicted concentration ratio of 134Cs as a function of time (for two replicates with 1 SD analytical errors).
FIGURE 2. (a) DGT uptake of 134Cs with time by AMP-DGT devices in synthetic seawater (mean and 1 SD of two replicates) and (b) the measured to predicted concentration ratio for 134Cs as a function of time (for two replicates with 1 SD analytical errors).
FIGURE 3. Measured to predicted concentration ratio for 134Cs as a function of pH (for two replicates with 1 SD analytical errors). to-predicted ratios being 1.00 ( 0.01 (n ) 16, 95% confidence) for variation in pH, 1.00 ( 0.05 (n ) 8, 95% confidence) for temperature, and 1.00 ( 0.08 (n ) 12, 95% confidence) for variation in diffusive layer thickness. The last includes an 4532
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 22, 2001
FIGURE 4. Variation DGT uptake of 134Cs with diffusive gel thickness. anomalous value for a 0.4 mm thick diffusive gel. Since this result does not agree with previous experience (4), it is considered to be due to an error in gel manufacture. In an experiment in which the 134Cs concentration was increased incrementally halfway through DGT sampling
FIGURE 5. Variation DGT uptake of
134
Cs with temperature.
TABLE 1. Effect of Changing Bulk Solution during DGT Measurement Periodsa
137Cs
Concentration
exposure times in soln A and soln B
soln A 134Cs concn/ mBq/L
soln B 134Cs concn/ mBq/L
predicted 134Cs mean concn/mBq/L
DGT measd 134Cs mean concn/mBq/L
4h+4h 5h+5h 6h+6h
74 ( 2 54 ( 1 35 ( 2
92 ( 3 74 ( 2 54 ( 1
83 ( 4 64 ( 2 45 ( 2
84 ( 2 65 ( 2 47 ( 3
a
n ) 4.
FIGURE 6. Variation of DGT uptake of stable Cs in 24 h with concentration and determination of AMP capacity. periods, very close agreement was obtained between the DGT measured value and the predicted value (Table 1). In the experiment to determine the capacity of the AMP, uptake of stable Cs was found to increase linearly with solution concentration, up to a maximum value equivalent to 0.20 mg of Cs per mg of AMP (Figure 6), which agrees with the value of 0.213 mg/mg given by Miller et al. (11). The capacity of the single DGT device used is therefore 2.3 mg of stable Cs (equivalent to 7 GBq 137Cs or 108 GBq 134Cs). This would give a theoretical time to saturation of 1790 y in a solution of 100 ng L-1 stable Cs at 10 °C (freshwater median natural concentration (12). Field Experiments. 137Cs Mean Solution Concentration. For adequate counting sensitivity, DGT sampling over the
FIGURE 7. Comparison of the 5 d 137Cs mean activity concentration measured by DGT and those obtained from grab water samples, Llyn Trawsfynydd. shortest deployment period used of 5 d required a relatively large effective sampling area of 47.1 cm2, achieved by bulking 15 DGT binding gels for analysis. The result obtained for the 5 d mean solution concentration of 137Cs was 60.1 ( 3.2 mBq L-1 (Figure 7). The quoted standard deviation is based on the analytical uncertainty only, giving an analytical RSD of ∼5%. In comparison, the 5 d water sample (comprised of samples filtered after each day, for which adsorption losses were experimentally determined as being negligibly small) gave a 137Cs mean concentration of 51.3 ( 2.7 mBq L-1 (Figure 7). However, this is not a true integrated sample for the period, being based on grab samples every 0.5 h. Furthermore, the individual 1 d samples showed considerable variation in mean concentration over the 5 days (factor of 1.8), with a mean value of 53.8 ( 11.8 mBq L-1. For the five successive 1 month DGT deployment periods, the effective gel sampling area ranged from 3.1 to 12.6 cm2 (1-4 binding gels). The DGT measured 137Cs mean solution concentrations for individual months ranged from 47 to 60.9 mBq L-1, with a mean analytical RSD of 4.9%. Replication of DGT measurements for 1 month and 4 month deployments showed good precision, with RSD values of 2.5% (n ) 4) and 1.8% (n ) 3), respectively. As explained in the Methodology, time integrated water samples to validate the DGT results could not be obtained over 1 month periods and only a general comparison can be made with the results from individual grab samples and those taken by BNFL (mean of four grab samples taken during a month, but not from precisely the same location as ours) (Figure 8). The DGT results for the 7 and 14 d periods during the first month are also shown on this figure. Not unexpectedly, the grab sample results show a wider variation than obtained by DGT, although the overall agreement is good. The DGT 137Cs mean solution concentration for the whole 5 months of 55.2 ( 7.1 mBq L-1 is statistically identical to the value of 58.8 ( 22.8 mBq L-1 obtained by BNFL, with considerable overlap of uncertainties at the 1 σ level. This value can also be compared to the approximate annual mean concentration calculated from the annual discharge and the residence time estimates of 54-72 mBq L-1. Effect of Time on DGT Performance. The 137Cs mean concentrations obtained from the 2, 3, 4, and 5 month DGT deployments were compared with the cumulative record of the 1 month deployments (Figure 9). The longer deployments increasingly underestimate the cumulative uptake, with the DGT uptake over 5 month’s continuous deployment being VOL. 35, NO. 22, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4533
Discussion and Conclusions
FIGURE 8. Comparison of the 1 month 137Cs mean activity concentration measured by DGT and those obtained from grab water samples, Llyn Trawsfynydd.
FIGURE 9. Comparison of a cumulative plot of the 137Cs uptake from 1 month’s DGT deployment (squares) with the uptake from successive 2, 3, 4, and 5 month deployments (diamonds), Llyn Trawsfynydd. 81% of the expected value. This result was confirmed retrospectively by a laboratory experiment in which DGT devices previously deployed at Trawsfynydd for 1 month and 5 months were exposed to a solution of 134Cs of known activity. (Although 134Cs is present in the discharge, its concentration in the lake water is below the DGT detection limit.) This experiment showed that the 134Cs uptake rate at the end of 1 and 5 month’s deployment, compared to a nondeployed control, was 87% and 29% of the control value, respectively. A 13% reduction of the uptake rate after 1 month corresponds to a mean reduction in uptake over the whole month of 6.5% if, as a first approximation, it is assumed that the effect is linear with time. The measured 137Cs solution concentrations given above have not been corrected for this source of error. Temperature Effect. The monthly mean temperatures, based on measurements every 15 min., varied from 6.2 to 15.7 °C over the 5 months, which corresponds to a 27% increase in the diffusion coefficient. The maximum monthly temperature standard deviation was 14.3%, which translates to a 9% uncertainty in mean 137Cs solution concentration for the month, assuming that the concentration did not vary over that period. 134Cs Solution Concentration. For 134Cs, the DGT concentration measurements from the 5 d and 5 × 1 month sampling programs were all below the analytical detection limit, i.e., < 4.4 to < 11 mBq L-1. These are compatible with the values obtained by BNFL from their bulked grab samples (
