PROGRAMS AND PLANS--Dissolved Trace-Element Data ARE DISSOLVED TRACE ELEMENTS LOST DURING FILTRATION AND STORAGE IN A PART-PER-TRILLION PROTOCOL? In Reply Refer To: October 29, 1992 Mail Stop 412 OFFICE OF WATER QUALITY TECHNICAL MEMORANDUM 93.03 Subject: PROGRAMS AND PLANS--Dissolved Trace-Element Data ARE DISSOLVED TRACE ELEMENTS LOST DURING FILTRATION AND STORAGE IN A PART-PER-TRILLION PROTOCOL? SYNOPSIS The purpose of this memorandum is to summarize the findings of preliminary experiments conducted to assess whether analyte loss occurs during the filtration and sample storage steps of an "ultra clean" (part-per-trillion level) trace-element protocol. The experiments were conducted under the supervision of Marty Shafer and Mark Walker, Water Chemistry Program, University of Wisconsin- Madison, Wisconsin. Specific details and results of the experiments are given in two unpublished research reports dated June 1992. Some of these details, included herein, are intentionally taken verbatim from the research reports to avoid misrepresentation. The major findings of the experiments are: 1. In general, no statistically significant changes in concentrations occurred in cadmium (Cd), copper (Cu), lead (Pb), and zinc (Zn) during filtration; and aluminum (Al), cadmium (Cd), copper (Cu), lead (Pb), and zinc (Zn) during sample storage using ultra clean procedures. This suggests that analyte loss, resulting from sorption of analyte onto acid cleaned surfaces is not a problem for part-per-trillion (ppt) trace element work. 2. For Cu, systematic low-level decreases were observed for both the filtration and sample storage experiments. The decreases were not statistically significant except in one case. It is probable that a portion of the difference could be attributed to error arising from a change in analytical sensitivity. BACKGROUND Office of Water Quality (OWQ) Technical Memorandum 91.10 described our current understanding of observed discrepancies in the levels of dissolved trace-element results generated by the standard U.S. Geological Survey (USGS) protocol in comparison to "ultra clean" protocols used by academic researchers and Howard Taylor of the USGS National Research Program. Median concentrations of arsenic, boron, beryllium, cadmium, chromium, copper, lead and zinc obtained by standard USGS methods were substantially higher than levels generated by ultra clean protocols. In addition, a study in which both standard and ultra clean methods were applied to aliquots of the same samples showed that the discrepancy in trace element concentrations resulted primarily from sample collection and field processing, rather than from laboratory analysis. Memo 91.10 listed three possible reasons for the observed discrepancies: (a) contamination introduced by the standard USGS protocol; (b) variations in sample processing techniques (e.g., differences in particle removal procedures might cause differences in the amount of colloidal material incorporated and analyzed as "dissolved"); and (c) sorption loss of dissolved trace elements during the sample collection, sample handling, filtration, and sample storage steps of the ultra clean protocol. Sorption loss could occur if: (a) exhaustive acid precleaning opens active sorption sites on sampling equipment, filters, and/or sample bottles, and (b) the sorption sites are incompletely equilibrated with excess sample during collection, processing, and storage of sample aliquots. It is expected that analyte loss would be minimal during sample storage because of the low pH resulting from sample preservation. Contamination from selected surface-water samplers and from membrane filters has been described in OWQ Technical Memoranda 92.12 and 92.13, respectively. Also, the effects of filtration artifacts on iron and aluminum concentrations have been investigated by Art Horowitz (Horowitz, 1992). To date, the USGS has not investigated analyte loss from sorption of dissolved trace elements onto equipment and materials during use of ultra clean methods. Such studies have been delayed, pending the development of an "in-house" part-per-trillion (ppt) analytical capability in the National Water Quality Laboratory (NWQL). Such a capability will not be available until October 1, 1993, at the earliest. However, the OWQ has obtained unpublished research reports from the University of Wisconsin that characterize the extent of analyte loss during use of ultra clean methods. In these experiments, analyses were done by graphite furnace atomic absorption and at concentrations well above instrument detection levels. This memorandum presents the results of those reports. ANALYTE LOSS DURING FILTRATION Design/Preparation The water samples used in this experiment were taken from southern Lake Michigan and were filtered on board ship through an acid- leached 0.4 5m track-etched filter. Unacidified subsamples of the filtrates were transported to the lab on ice and refiltered to remove remaining particles and most colloids. Particles/colloids were removed by filtering the samples twice through acid leached 0.05 5m Nuclepore filters. The 0.05 5m filtrates were split into two aliquots, one left untreated, the other spiked with an acidified mix of trace metals designed to approximately double the ambient concentration of targeted metals (Cd, Cu, Pb and Zn). The spike additions lowered the pH of the sample from 7.6 to 7.4. The mean concentration of trace elements in the unspiked and spiked filtrates, after completion of the described filtration steps, were: Nanogram/liter (ng/L) Unspiked Spiked Cd 25.5 48 Cu 780 1,460 Pb 75 142 Zn 725 1,415 These spiked and unspiked filtrates became the "reference solutions" used to evaluate analyte loss associated with an "ultra clean" filtration apparatus and membrane filter. The filtration apparatus was comprised of a segmented Teflon column to which an all Teflon 47 mm filter holder was attached. Acid leached 0.4 5m Nuclepore filters were used and filtrates were obtained by pressurizing (10-20 psig filtered nitrogen) the column. The filter sorption experiment was run one day after the laboratory filtering and spiking. The filtration apparatus was set up under a laminar flow hood, and the filter holder loaded with an acid leached filter. Approximately 200 mL of reference solution was placed in the column and the filtration initiated. The first 25 mL of filtrate was discarded, and the next two 75 mL aliquots (filtrate "A" and filtrate "B") were collected in separate trace-metal cleaned polyethylene (LPE) bottles. Filtration rates were in the range of 0.5 to 1.0 mL min-1 cm-2. The complete experiment was run in duplicate for both the spiked and unspiked reference samples. Two blank (Milli-Q water) filtrations were also performed. After filtration, all samples were acidified with 150 5L of Ultrex HNO3, lowering the pH to about 1.6. Pre-cleaning of the Nuclepore filters was done by placing filters individually into acid washed polystyrene petri dishes, and adding 1M Ultrex HNO3. Filters were soaked for a period of approximately 24 hours, after which the acid was removed and filters thoroughly rinsed with Milli-Q water. The filtration apparatus was cleaned by leaching in 20 percent HNO3 (ACS reagent) for 4 days at room temperature, followed by a leach in 2 percent HNO3 (Baker trace- metal grade) for 4 days at room temperature. The LPE sample bottles were cleaned as follows: 1. Fill with ACS reagent acetone, soak for approximately 2 hours, remove acetone, rinse 3 times with Milli-Q water; 2. Fill with 20 percent ACS reagent HCL, soak for 4 days at room temperature, remove HCL, rinse 3 times with Milli-Q water; 3. Fill with 20 percent ACS reagent HNO3, soak for 4 days at room temperature, remove HNO3, rinse 3 times with Milli-Q water; and 4. Fill with 0.5 percent Baker Ultrex HNO3, soak until needed (but at least 4 days), remove acid, rinse 3 times with Milli-Q water, dry under laminar floow hood, and bag. Results Table 1 summarizes the average of two replicated experiments. Changes in mean concentrations of trace elements through the filtration apparatus and membrane filter are shown, along with the mean percentage changes. Positive values denote an increase in concentration; negative values a decrease. The last column gives the calculated percentage changes in concentrations required to signal a statistically significant difference between the reference solutions and either filtrate A or filtrate B. Triplicate analyses and the t-test with a 95-percent confidence level were used to calculate the percentage changes shown. Levels of Cd, Cu, Pb, and Zn in the filter blanks were undetectable; therefore, no blank corrections were made to the computed differences listed in Table 1. Concentration changes during the filtration step were less than or equal to 12 ng/L for Cd and Pb (less than or equal to 16 percent difference), and less than or equal to 80 ng/L for Cu and Zn (less than or equal to 7.7 percent difference). Only the -5.5 percent difference for Cu in the spiked sample for filtrate A was statistically significant. It is noteworthy that the four observations for Cu show a negative change, suggesting that low levels of analyte loss may have occurred for this element. ANALYTE LOSS DURING SAMPLE STORAGE Design Two separate studies were completed to evaluate the effect of refrigeration storage on trace-element concentrations (that is, analyte loss to sample bottles). One study was conducted with southern Lake Michigan water; the second with samples from selected Wisconsin rivers. Details and results of the former are described in this section, whereas, only those findings related to analyte loss are given for the latter. Both studies involved: (a) analyzing samples several weeks or months after sample collection (first analysis), (b) storage of acidified samples at 4!C for 6 to 8 months, and (c) then reanalysis of the samples (second analysis). Samples of southern Lake Michigan water were filtered on board ship through an acid leached 0.4 5m track-etched filter. Subsamples were pooled soon after filtration and placed into paired 125mL polyethylene (LPE) and Teflon (TEF) bottles. (Both bottle types were pre-cleaned using the "ultra clean" procedure described previously for the filtration experiment). Subsamples were then acidified at a rate of 2 mL of concentrated Ultrex HNO3 per liter resulting in a pH of 1.4 to 1.5. All samples were kept refrigerated for the entire cruise and transported to the lab on ice. In the lab, samples were held at 4!C until first analyzed 2- 3 weeks after sample collection. Considerable time elapsed before the first analysis was completed. Therefore, analyte loss could have occurred to sample bottles between the time of sample collection and the first analysis. After the first analysis, samples were stored for an additional 7.5 months at 4!C, and then re-analyzed for Cd, Cu, Pb, and Zn. Two Milli-Q bottle blanks were also analyzed and, except for Zn (in LPE bottles only), the desorption of trace elements from sample bottles was not significant. For Zn, the average increase in the blank for the LPE bottles over 7.5 months of refrigeration storage equated to approximately 10 percent of the sample's ambient concentration (about a 65-90 ng/L increase). Results A summary of the results from the study of Lake Michigan water is given in Table 2. With the exception of Zn in the LPE bottles, none of the trace-element concentrations were statistically different (for both bottle types) between the two analyses. The 83 ng/L increase of Zn in the LPE sample bottles is similar in magnitude to the increase in the bottle blank and presumedly was a result of contamination, either during storage or during the second analysis. In summary, there were no statistically significant losses of Cd, Cu, Pb, and Zn to either polyethylene or Teflon bottles during storage at pH 1.5 and 4!C for a period of 7.5 months. As noted above, the amount of analyte loss, if any, that occurred to samples between the time of collection and first analysis remains to be quantified. Similar results were found for Al, Cd, Pb, and Zn in the storage loss experiment conducted on Wisconsin river waters. However, although not statistically significant, a decrease in Cu concentration occurred in all four samples during storage in pre- cleaned and refrigerated Teflon bottles (polyethylene bottles were not evaluated) for 5-8 months. The decreases ranged from 35-131 ng/L--5 to 15 percent of the sample's first analysis concentration. The consistently lower copper concentrations during the second analysis might reflect some sorption loss to the Teflon sample bottles. However, some of the difference was attributed by the University of Wisconsin researchers to an analytical "...downward sensitivity drift observed over the course of the storage check (period)". SUMMARY AND FUTURE RESEARCH The experiments conducted at the University of Wisconsin provide data for selected trace elements on the extent of potential analyte loss associated with ultra clean methods. In general, changes in trace element concentrations from filtration and sample storage were not statistically significant. Also, no systematic loss occurred: (a) for Cd, Pb, and Zn during filtration (that is, on the filter holder and membrane filter); nor (b) for Al, Cd, Pb, and Zn onto bottle walls during 7.5 months of storage at about pH 1.5. However, systematic, low-level decreases in Cu concentrations were observed for the filtration experiment and for one of the sample storage experiments (Wisconsin River water), with decreases ranging from 30-131 ng/L. The University of Wisconsin preliminary test results suggest that analyte loss is not a problem for trace element work at the ppt level. The University of Wisconsin is currently completing additional analyte loss experiments. Once the NWQL establishes ppt analytical capability, the OWQ will collaborate with the University of Wisconsin to conduct analyte loss tests on sample collection equipment, filtration equipment, and storage bottles for those trace elements to be analyzed at the ppt level by the NWQL. REFERENCES Horowitz, A.J., Elrick, K.A., and Colberg, M.R., 1992, The effect of membrane filtration artifacts on dissolved trace-element concentrations: Water Research, v. 26, no. 6, p. 753-763. Shafer, Martin, 1992, An evaluation of analyte loss resulting from the storage and filtration of Lake Michigan waters: Water Chemistry Program, Water Science and Engineering Laboratory, University of Wisconsin-Madison, 12 p. [Unpublished report on file in the Office of Water Quality, Reston, Virginia] Walker, Mark, 1992, An evaluation of trace element loss resulting from the storage of Wisconsin surface waters: Water Chemistry Program, Water Science and Engineering Laboratory, University of Wisconsin-Madison, 12 p. [Unpublished report on file in the Office of Water Quality, Reston, Virginia] David A. Rickert Chief, Office of Water Quality Key words: Trace elements, analyte loss, sorption to field equipment This memorandum refers to Office of Water Quality Technical Memorandums 91.10, 92.12, and 92.13. Distribution: A, B, S, FO, PO