Trace Element Contamination: Findings of Studies on the Cleaning of Membrane Filters and Filtration Systems In Reply Refer To: July 17, 1992 Mail Stop 412 OFFICE OF WATER QUALITY TECHNICAL MEMORANDUM 92.13 Subject: Trace Element Contamination: Findings of Studies on the Cleaning of Membrane Filters and Filtration Systems PURPOSE In 1989 and 1990 Art Horowitz conducted a series of experiments with emphasis on the cleaning of filters and filtration systems. This memorandum describes the results and conclusions from those experiments. Building on the information in Office of Water Quality (OWQ) Technical Memorandum 91.10, this memo provides background for some of the decisions that the OWQ is making to provide a supportable parts-per-billion (ppb) protocol for dissolved trace elements. STUDY COMPONENTS Figure 1 provides information on the various experimental designs and identifies the three filter brands tested--MFS, Millipore, and Nuclepore. These brands were selected because Millipore and MFS represent approximately 80 percent of the Division's usage, and Nuclepore is the filter preferred by research chemists. All experiments were made on 142-mm (millimeter) diameter filters. Four experiments were conducted: 1. Cleaning of membrane filters for major and trace elements: A. Concentration of elements in successive wash aliquots of 50, 50, 100, and 200 milliters (mL), which correspond to cumulative wash volumes of 50, 100, 200, and 400 mL. B. Comparison of deionized water (DIW) versus 2 percent HNO3. C. Comparison of element concentrations in the 400-mL wash volume to National Stream-Quality Accounting Network (NASQAN) reporting limits (RLs). D. Comparison of filter brands. 2. Sample dilution effect of cleaning filters with DIW. 3. Carryover contamination in equipment: Laboratory study. 4. Carryover contamination in equipment: Field study. Table 1 provides information on the National Water Quality Laboratory's (NWQL) RLs for each experiment and for NASQAN in 1990. Reporting levels changed from experiment to experiment depending on the analytical method used; namely, inductively coupled plasma (ICP) atomic emission spectroscopy, or graphite furnace atomic absorption spectroscopy (GFAAS). CLEANING OF MEMBRANE FILTERS FOR MAJOR AND TRACE ELEMENTS Five individual filters from a single batch of each of the three brands were tested. This designation, together with the four wash volumes (50, 100, 200, and 400 mL) and the two wash solutions (DIW and 2 percent HNO3) gave a design of 120 data points (5 x 3 x 4 x 2). Results were compared by computing mean constituent concentrations and the standard deviations for the five filters for each combination of brand, volume, and wash solution. Ranks of constituent concentrations and standard deviations were also computed. The rank results are not reported in this memo, but are cited in several places. Analyses were made for 19 major and trace elements. Of these, 11 showed either: (a) no detectable concentrations throughout the entire experiment, or (b) concentrations below the study's reporting level before or in the aliquot corresponding to the 200-mL cumulative wash volume. The 200-mL volume is important because it has been the Division's "rule of thumb" in cleaning filters prior to collecting the filtrate for dissolved trace- element analysis. Elements in these two categories included silver (Ag), barium (Ba), beryllium (Be), cobalt (Co), lithium (Li), molybdenum (Mo), manganese (Mn), sodium (Na), strontium (Sr), vanadium (V), and zinc (Zn). The remaining eight elements-- which are the focus of this memo--were cadmium (Cd), copper (Cu), lead (Pb), nickel (Ni), iron (Fe), silicon (Si), magnesium (Mg), and calcium (Ca). For these elements, concentrations in the wash solutions were quite erratic, giving rise to high standard deviations (Tables 2, 3, and 4). Thus, although differences based on means alone were observable for many tested comparisons, only a few differences were statistically significant. The erratic results appear to have arisen from: (a) non-uniformity of contamination between individual filters in a batch, (b) differences in flow paths by which cleaning solutions passed through the 142-mm filters, and (c) in certain cases, contamination arising during the experiment or subsequent laboratory analyses. Comparison of 400-ML versus 200-ML Wash Volumes In this and following discussions, mean and standard deviations are described. Five observations were available for each brand of membrane filter, wash volume and wash solution. To compute means and standard deviations, "less than" values were assigned a value of one-half the reporting limit. Table 2 shows the computed means and standard deviations for the eight elements for the DIW washes (by filter brand and wash volume), whereas Table 3 shows the comparative values for the 2 percent HNO3 washes. From these tables, the following was observed: 1. Elemental concentrations generally showed a decay function, with low concentrations reached by the 200-mL cumulative wash volume. 2. In nearly two-thirds of 48 cases (8 elements x 3 filter brands x 2 wash solutions), the mean elemental concentrations were less in the 400-mL versus the 200-mL cumulative wash volume. This trend was evident for both DIW and 2 percent HNO3 for the three filter brands, and for most of the elements. Furthermore,the observation of lower concentrations in the 400- versus the 200-mL cumulative volume was confirmed by the mean of ranks. Although the mean concentrations were typically less in the 400-mL wash volume, the difference between the 200-mL and 400-mL wash volumes tended to be small. For example, in DIW washes, the difference for Cd, Pb, and Cu was - 0.2 ug/L; and for Ca, Mg, and Si, - 0.04 mg/L. 3. Mean constituent concentrations in the 400-mL cumulative wash volume were statistically different (at the +/- 1 standard deviation level) from the 200-mL volume in only 7 of the 48 cases (15 percent). However, none of these 7 cases were confirmed by an evaluation of ranks at the +/- one standard deviation level. The cited observations suggest there may be a slight, but statistically insignificant benefit for some constituents to cleaning filters with 400 mL of DIW or 2 percent HNO3, in comparison to 200 mL. Comparison of DIW Versus Two Percent HNO3 The results for the 400-mL DIW and 2 percent HNO3 washes are summarised in Table 4; the following was observed: 1. For the 8 elements, the lowest mean concentrations were obtained as follows: * Cd - DIW, MFS; 0.06 +/- 0.02 ug/L * Cu - DIW, MFS; 0.1 +/- 0.1 ug/L * Pb - 2 percent HNO3, Nuclepore; 0.05 +/- 0 ug/L, followed closely by DIW, MFS 0.06 1 0.02 ug/L * Ni - DIW, MFS; 0.2 +/- 0.2 ug/L * Fe - 2 percent HNO3, Nuclepore; 3.1 +/- 1.6 ug/L * Si - 2 percent HNO3, Nuclepore; 5 +/- 0 ug/L * Mg - 2 percent HNO3, Nuclepore; 5 +/- 0 ug/L * Ca - 2 percent HNO3, Nuclepore; 5 +/- 0 ug/L 2. For MFS and Millipore filters, the constituent concentrations were generally lower for DIW than for 2 percent HNO3. The only exceptions were Fe with MFS, and Ca and Si for both brands (concentrations are identical for DIW and 2 percent HNO3). For certain constituents--Cu and Ni--the concentrations in the DIW were much lower than in the 2 percent HNO3. 3. For Nuclepore filters, the constituent concentrations were lower using 2 percent HNO3 (in comparison to DIW) for Pb, Ca, Mg, Fe, and Si, but higher for Ni, and identical for Cd and Cu. These observations, based on means and standard deviations of concentrations, were confirmed in nearly all cases by the statistical analysis of ranks. While comparisons of concentration means in the 400-mL cumulative wash volumes suggest that DIW and 2 percent HNO3 differ somewhat in their effectiveness for cleaning membrane filters, the differences were small for most elements. Furthermore, in only three of the 24 cases--MFS-Pb (lower in DIW), Nuclepore - Pb (lower in 2 percent HNO3), and Nuclepore - Ni (lower in DIW), were the comparative concentrations (in DIW and 2 percent HNO3) statistically different (at the +/- one standard deviation level). In summary, combining the cleaning results for the three filter brands, there was no advantage at the 400-mL cumulative wash volume to using 2 percent HNO3 as opposed to DIW for the eight elements. Comparison of Elemental Concentrations in the 400-mL DIW Wash Volumes to NASQAN Reporting Limits Concentrations of the eight elements in the 400-mL DIW wash volumes were compared to the 1990 NASQAN RLs (compare results in Table 4 to RLs in Table 1). Several observations were evident: 1. For Cd and Pb, the individual, and hence the mean concentrations for the three brands of filters were at or below the NASQAN RLs. For Cu and Ni, the individual concentrations exceeded the RL in 1 of 5 and 2 of 5 cases, respectively, for Millipore; the means for Cu and Ni were at or below the RLs for all three filter brands. 2. For Ca, the individual concentrations were at or below the NASQAN RL for the three brands of filters, and the mean concentrations were all below the limit. 3. For Mg, the individual concentrations exceeded the RL in 3 of 5 cases for MFS, 1 of 5 for Millipore, and 0 of 4 for Nuclepore. The mean concentrations were below the RL for Nuclepore, at the RL for Millipore, and slightly above for MFS. 4. For Fe, the individual concentrations exceeded the RL in 4 of 5 cases for MFS, 3 of 5 for Millipore, and 2 of 4 for Nuclepore. The means exceeded the RL for all three brands. 5. For Si, the individual concentrations exceeded the RL in 4 of 5 cases for both MFS and Millipore, and in 1 of 4 for Nuclepore. The mean concentration slightly exceeded the RL for the MFS and Millipore filters, and was at the RL for Nuclepore. Thus, the mean concentrations of Mg and Si leached from certain brands, and for Fe from all three brands, were slightly above the current NASQAN RLs. However, the higher concentrations for these elements are probably inconsequential when compared to the levels that occur in environmental samples from most NASQAN and WRD project sites. Comparison of Filter Brands Taking the lowest mean concentration for each element, whether for DIW or 2 percent HNO3, the results indicate: 1. The lowest mean concentrations of Cd, Cu, and Ni were found with MFS filters. 2. The lowest mean concentrations of Pb, Fe, Si, Ca, and Mg, were found with Nuclepore filters. However, for these cases, none of the lowest mean element concentrations were statistically different (at the +/- one standard deviation level) from concentrations measured for other filter brands. The analysis of data by ranks generally confirmed the cited findings--for both the filter brand providing the lowest rank and the lack of statistically significant differences. In summary, based on the results of this study and current NASQAN RLs, it appears that a 400-mL DIW wash of MFS, Millipore, or Nuclepore filters will sufficiently preclean the filters for: (a) trace elements at the ppb level, and (b) Fe, Si, Ca, and Mg at environmentally relevant concentrations. In determining a preferred filter brand for the ppb trace element protocol, two additional factors are important: 1. For the 142-mm diameter, Nuclepore filters are comparatively limp and awkward to use, and 2. District projects have historically not used Nuclepore filters. Thus, although the results from this study indicate no difference, statistically, between Nuclepore, MFS, and Millipore filters, there is no benefit and some disadvantages to the Division switching to Nuclepore for the ppb trace element protocol. The OWQ will continue to evaluate this situation during the remainder of fiscal year (FY) 1992, and issue a directive in FY 1993 on membrane filters to be used in the new ppb protocol. The OWQ will also conduct a new set of filter cleaning studies as a basis for specifying procedures for a parts-per-trillion protocol for trace elements. SAMPLE DILUTION EFFECT CAUSED BY CLEANING FILTERS WITH DIW The second experiment was designed to determine the potential dilution effects on elemental concentrations of DIW remaining in the filter apparatus following the cleaning of membrane filters. The results are summarized in Table 5. Design Five native-water samples were collected and placed in a churn splitter. Aliquots of each sample were processed in replicate in three different ways, using separate and clean filtration systems: 1. A MFS filter was placed in the first filtration system and 200 mL of DIW was passed through it and discarded; 1 liter of sample was processed and split into replicates labeled DIW A and DIW B in Table 5. 2. A MFS filter was placed in the second filtration system and 200 mL of native water was passed through it and discarded; 1 liter of sample was processed and split into two replicates labeled Native-A and Native-B in Table 5. 3. A MFS filter was placed in the third filtration system and 200 mL of DIW was passed through it and discarded; 1 liter of sample was processed, with the first 25 mL of filtrate discarded; the remaining filtrate was split into two replicates labeled DIW + Native-A and DIW + Native-B in Table 5. Results For each of the five native water samples tested, the three processing systems produced similar results for all elements except Fe. Filtrates from the native-water cleaning system (system 2) contained about 10 percent more Fe than filtrates from the DIW cleaned system (system 1). This 10 percent difference was probably real because it was substantially higher than the analytical imprecision. The difference probably results from dilution of the iron concentrations in the native water samples by some DIW retained in the filtration apparatus. The dilution effect noted for Fe probably occurred for the other constituents, but could not be detected because the instrumental sensitivity was insufficient at the lower (in some cases, much lower) concentrations found for these elements. However, the data indicated a slight tendency for the effect for Sr, Ba, and Ca. Based on the Fe results, it appears that the use of the DIW pre- wash can cause as much as a 10 percent dilution in the constituent concentrations of native-water samples. Further, this dilution effect was only marginally reduced (range of 0-25 percent reduction for four native waters) by following the 200-mL DIW wash with a 25-mL native-water wash (system 3). CARRYOVER CONTAMINATION IN EQUIPMENT: LABORATORY STUDY Design Experiment 3 was conducted in the laboratory to determine whether carryover contamination would result if a sample having low concentrations of trace elements is processed after a sample having high concentrations. The laboratory tests were designed to determine how much carryover would occur, and how much rinsing of the filtration system with DIW or 2 percent HNO3 would be required to eliminate it. The procedure was: 1. Place a clean filter in the pre-cleaned filtration system; 2. Pass one liter of laboratory solution, containing a high concentration (see concentrations below) of trace elements and other constituents through the filtration system; 3. Discard old membrane filter and replace with a new filter; 4. Pass 200 mL of either DIW or 2 percent HNO3 through the filtration system to simulate filter cleaning; 5. Rinse the filter with four successive aliquots (50 mL, 50 mL, 100 mL, and 200 mL) of DIW or 2 percent HNO3, representing cumulative wash volumes of 250 mL, 300 mL, 400 mL, and 600 mL, respectively; and 6. Replicate the procedure five times for each type of membrane filter, and for both DIW and 2 percent HNO3 washes. The initial prepared solution contained the following: Ba (5 mg/L), Fe (5 mg/L), Pb (5 mg/L), Sr (5 mg/L), Cu (5 mg/L), Zn (5 mg/L), Cd (5 mg/L), Ni (5 mg/L), Ag (5 mg/L), Ca (100 mg/L), Mg (100 mg/L), Si (100 mg/L), and Na (100 mg/L). Results The concentration of elements in the wash solutions (see step 5) generally decreased with increasing wash volumes. However, even after 600 mL of washing, the mean elemental concentrations coming from the filtration system were considerably higher than both the 1990 NASQAN RLs and the NWQL RLs for this experiment. This observation was evident for both DIW and 2 percent HNO3 washes, and for all three filter brands. The mean concentration of constituents in the 600-mL wash aliquot (in this experiment) was also much higher than the mean concentration of constituents in the 400-mL cumulative wash aliquots from the filter-cleaning experiment (Experiment 1). The laboratory carryover experiment was designed to mimic a worst case scenario of potential sample contamination. The original trace element solution contained much higher concentrations than would be encountered in most natural waters, and the levels involved probably would never be encountered at either a NASQAN or a Benchmark sampling site. CARRYOVER CONTAMINATION IN EQUIPMENT: FIELD STUDY Design This was a complex experiment incorporating a number of blanks and split samples to detect contamination and to define precision in field processing of sample aliquots and in laboratory analysis (Table 6). However, as outlined in Figure 1, the main thrust of the experiment dealt with: 1. Processing sample aliquots from a contaminated site (site 1) through two churns and filter systems, 2. Cleaning the churns and filter systems from site 1 differently, 3. Processing aliquots of a sample from an uncontaminated site (site 2) through the differently cleaned churns and filter systems (see figure 1 for details), and 4. Processing equipment blanks at each site to determine if a cleaning procedure was sufficient to eliminate detectable carryover. Results to Determine if Carryover Occurs Between Sites The following results compare the differences in concentrations for systems 1 and 2 at the uncontaminated site (site 2). 1. Considerably higher mean concentrations of Cu, Zn, Fe, and Mn occurred in system 2, suggesting a large amount of carryover. Mean concentrations of these elements at site 1 were 358, 5,350, 48,500, and 5,225 ug/L, respectively. These concentrations were, at a minimum, 2-3 orders of magnitude higher than the levels in filter system 1 (the clean system) at site 2. 2. A small increase occurred in the mean concentration of Ca in System 2, indicating some carryover. The mean concentrations for Ca was moderately high at site 1, and the ratio for site 1/site 2 was about 22 fold. 3. No discernable increase occurred in the mean concentrations of Ba, Cd, Co, Li, Mg, Ni, Si, Na, and Sr in System 2. This indicates either that carryover did not occur or the extent of carryover was very small and could not be detected at the study's analytical sensitivity. Mean concentrations of these constituents at site 1 were moderately high, and the site 1/site 2 ratio ranged from <1-41 fold. 4. No comparison is possible for Be, Cr, Mo, Pb, and V because the reported concentrations were at or below the RLs at both sites. A possible hypothesis is that the higher mean concentrations of some constituents for System 2 at site 2 were due to sample- handling and analytical imprecision. However, because the higher concentrations for System 2 occurred for those constituents having very high concentrations at the first site (e.g., Cu, Zn, Fe, Mn), it seems more likely that the cause for the higher concentrations is carryover due to a lack of adequate cleaning of System 2's equipment. The churn splitter for System 2 was intentionally not cleaned after sampling the first site, whereas the filtration system was washed with 1 liter of DIW between samples. The churn splitter may have been the source of contamination, although it is equally plausible that the DIW pre-washing of System 2's filtration apparatus was inadequate to eliminate carryover effects from this equipment. Therefore, the specific equipment responsible for the carryover of Cu, Zn, Fe, and Mn can not be determined from this experiment. Results of Acid Wash Cleaning The experiment above indicates that contaminant carryover in field handling equipment can occur at the ppb level from one water sample to the next. Therefore, a test was conducted to determine whether such contaminated equipment could be cleaned in the field. The filtration apparatus was disassembled, thoroughly cleaned (along with the churn splitter) with DIW, reassembled, and rinsed (together with the churn splitter) with 1-liter of 2 percent HNO3 followed by 1-liter of DIW. This procedure was done at both sites for system 1, after the native water sample was processed through the system, but prior to processing the respective field equipment blanks. The analytical results for the two field equipment blanks shown in Table 6 confirm the efficacy of this cleaning procedure. CONCLUSIONS 1. Nuclepore filters--with intensive nitric acid cleaning--are the choice of scientists who conduct dissolved trace element work at the parts-per-trillion level. However, with proper cleaning with DIW, MFS, and Millipore filters are acceptable for trace element work at the ppb level. Nuclepore filters could be used, but accepting their relative difficulty of handling appears to be unncessary for ppb level work. The results are only representative of the tested batches of filters and could change for different batches. Thus, the OWQ will probably select a single brand of filter for Division use, identify a single supplier, purchase large batches, and quality assure each batch prior to distribution. This implies a long-term commitment to quality control because each new batch of filters will have to be tested for contamination. 2. For ppb level trace element work, a 400-mL DIW pre-wash is adequate for cleaning and pre-conditioning membrane filters. Existing procedures, which call for processing and discarding the first 200 mL of a native water sample to clean and condition filters, can lead to filter clogging and difficulties in obtaining sufficient additional filtrate for chemical analyses. These difficulties could be eliminated by using a DIW wash on the order of 400-500 mL. However, the results from this study indicate that this could produce about a 10 percent underestimation of Fe (because of its relatively high concentrations in examined environmental samples) due to dilution by DIW unavoidably retained in the filtration system. The approximately 10 percent difference between DIW and native water pre-cleaning and conditioning can probably be reduced by processing and discarding limited volumes of native water prior to the actual sample collection. However, this study did not determine the minimum volume of native water required to eliminate the difference. On the other hand, the 10 percent difference probably falls well within the errors associated with sample collection, field processing, and subsequent laboratory analysis. 3. Constituent carryover represents a potential problem when multiple samples are obtained and processed with the same sample collection and processing equipment without proper cleaning between sites. The cleaning procedure (complete DIW wash of the disassembled system and churn followed by a 1-liter, 2 percent HNO3 rinse and a 1-liter, DIW rinse through the complete system and churn) used during this study eliminated constituent carryover. David A. Rickert Chief, Office of Water Quality This memorandum does not supersede any previous Office of Water Quality Technical Memorandum. Key Words: NASQAN, trace elements, contamination, membrane filters, equipment cleaning Distribution: A, B, S, FO, PO