Atomic fluorescence determination of mercury in fresh water ecosystems

This paper reports on an investigation into determining nanogram/l quantities of mercury in marine and fresh water matrices using a cold vapour generation of mercury, followed by fluorescence detection. Samples were prepared for analysis using a free bromine oxidation technique. A high efficiency gas-liquid separator was used to enhance the detection of mercury. For fresh water, typical method detection limits (MDL) were determined at less than 1 nanogram/l (ng/l). For near shore seawater, the MDL was 1.2 ng/l. Method spikes, which were performed at 20 ng/l, showed mean recoveries within US EPA Contract Laboratory Protocol (CLP) acceptance criteria. System blanks averaged 0.12 ng/l, and recoveries of NIST 1641c diluted to 29.4 ng/l averaged 93.4%. A number of local rivers and streams were sampled, and mercury was determined. All results to date indicate mercury levels below the US EPA chronic water quality criteria for mercury.


Introduction
Significant concerns remain regarding mercury contamination in aqueous ecosystems. The United States Environmental Protection Agency (US EPA) has advanced water quality criteria [1-] for the protection of organisms native to water environments. The chronic water quality criterion for mercury in fresh water ecosystems is 12 nanograms/litre (ng/1), and the mercury chronic criterion for salt water is 25 ng/1. These extremely low criteria present significant problems for the currently approved US EPA regulatory methods (245.1-5 and SW-846 7470-7471) in that these methods, which are based on atomic absorption techniques, generally produce instrument detection limits between 30-200 ng/1. This level of detection is insufficient to quantify mercury levels in many borderline impacted waters.
The use of atomic fluorescence for the determination of mercury was first reported by Thomson and Reynolds in 1971 [2]. Since then, several authors [3][4][5] have described enhancements to the technique that have reduced formal instrument detection limits (IDL) for the fluorescence technique to the 1-10 ng/1 range. The authors of this paper report on the use of atomic fluorescence detection coupled with a high efficiency gas-liquid separator, the combination of which has reduced the instrument detection limit for mercury to less than ng/1. A bromine-based sample digestion technique, which results in 'cleaner' sample preparation, is also described. The potassium bromide and potassium bromate digestion reagents used in this method to produce bromine are purified by heating at 300C to remove mercury. The acids used in the procedure are purified by sub-boiling distillation in a Teflon still. Using these purified reagents results in undetectable levels of mercury (<05 ng/!) in the procedural blanks. In contrast, the currently approved US EPA Methods 245"1 and SW-846 7470-7471 use relatively large amounts of reagents that are difficult to purify completely of mercury contamination.
Another innovation in this method is the use of disposable polyethylene terephthalate (PETG) bottles for mercury sample preparation. The use of disposable sample preparation bottles reduces the problems of mercury contamination when preparation bottles are reused (reuse is common in environmental laboratories that use the currently approved US EPA regulatory methods).

Method summary
The method used in this work for the determination of mercury is based on detection of mercury by atomic fluorescence at 253"7 nanometers (nm) excitation and emission wavelength. Samples are digested before analysis with a mixture of potassium bromate and bromide in hydrochloric acid [6]. This mixture, at ambient temperatures, forms free bromine, which oxidizes both organic and inorganic forms of mercury to Hg +/ [6 and 7].
A fluorescence mercury analysis system (Merlin, supplied by P. S. Analytical Ltd, Sevenoaks, Kent, UK) was used for mercury determinations. Oxidized sample, or blank solution and SnC12 reagent, are combined in the mixing valve of the instrument reduction apparatus. Hg ++ is reduced to elemental mercury, which is carried in the sample stream. The carrier mixture containing elemental mercury is then directed into a high efficiency gas-liquid separator. A stream of fine argon bubbles sweeps elemental mercury out of solution into the gaseous phase; the gas stream containing mercury vapour is then passed through a drying tube to remove water vapour; then dry vapour is swept into the instrument's fluorescence detector. A mercury vapour lamp with a characteristic wavelength of 253"7 nm is used to excite mercury atoms in the vapour. The intensity of emission at the same wavelength is measured and used to quantitate mercury. Bottles are filled to about cm from the top and cap. Yellow colour (free bromine) was to be stable until use. For 125 ml bottles, the above amounts need to be halved. The solution is decolorized with 0.15ml of 12 hydroxylamine hydrochloride (0"075 ml for 125 ml bottles). It is then stirred and discarded; the bottles are rinsed three times with a small amount ASTM type II water. The recapped bottles are then ready for use. Cleaning should be carried out either in a cleanroom or in an area with an extremely low mercury background.

Reagent preparation
Reagent preparation is crucial to the success of these low-level determinations to ensure that all plasticware is properly precleaned, and that all reagents are prepared in a consistent manner in a mercury-free area.
All reagents must be suitable for mercury or trace metal analysis. Trace metals grade hydrochloric and nitric acids should be purified by sub-boiling distillation in a Teflon still. ASTM type II, or better, D! water should be used for this.
2"0 lg/ml stock mercury standard A 1000 ml volumetric flask is rinsed, and then 200 ml H20 and 20ml HNO3 added; 2"0ml of a suitable 1000ppm mercury standard (SPEX Industries, Inc., Edison, NJ, or equivalent) is measured into the volumetric flask and diluted to volume with Type II water, and mixed.
Intermediate standard preparation 25 ml of 2"0 lag/ml stock is diluted to 500 ml in 2 HNO" this gives a 100 gg/1 mercury solution.
Working standards are made from a further intermediate dilution. A ml to 100ml dilution in 2 HNOa of 100 lag/1 stock gives a 1"0 lag/1 standard. This working standard is then further diluted to give Hg standards of 0"002 to 0"05 gg/1. The 1"0 gg/1 standard should be prepared daily as used. gas to the gas-liquid separator and as a sheath gas to the fluorescence detector to improve detector performance.
The drying tube (Leeman Instruments, Lowell, MA, No. 120-00067) used was approximately 15cm by l'5cm. The ends of the tube were loosely packed with quartz wool. The tube was packed before each use with granular  figure 2. The high efficiency performance results from the fine bubbles formed by the sintered plate; these bubbles are efficient in sweeping mercury from solution and result in a factor of two enhancement in sensitivity compared with the standard gas-liquid separator used with this instrument. Lower gas flow rates are used with this Argon Reagents Waste separator as opposed to the standard separator, resulting in less dilution. These reduced flows also reduce moisture carry-over into the detector. The larger volume of the improved separator is a compromise which increases the time needed to rinse out the system, system memory time, and the time required after sample introduction for the signal to approach its maximum value (rise time). Greater efficiency results from finer argon bubbles being formed by sintered plate than are formed by tip used in standard separator. This results in more contact of gas and solution and increases likelihood of mercury transfer to gaseous phase producing greater efficiency of mercury removal per volume of argon. Greater solution-volume of cell also promotes transfer. However, this increase in volume also increases time to flush the system and thus the time of analysis.

Sample and standard solution preparation
Procedures for standard and sample preparation should be performed in an area determined to be free from mercury contamination. Standards for low level fluorescence work are prepaired by taking 10"0, 5"0, 2"0, 1"0, 0"4, and 0 ml aliquots of the 1-0 gg/1 stock with a rinsed pipette; and then placed in 250 ml precleaned and rinsed PETG bottles. Each standard is diluted to 200 g to give 0"050, 0"020, 0"010, 0"005, and 0"002 gg/1 standards; 25 ml + HC1, followed by 5 ml of 0" N potassium bromate/ bromide solution, is added to each bottle. Then 20 ml of type I! water is added to make a total 250 ml of prepared standard. Standards bottles should be kept sealed at all times when standard or reagents are not being added, to prevent absorption of atmospheric mercury. To prepare samples, 100 ml of sample is added to tared precleaned 125 ml PETG media bottle. Any spike needed (0"50 ml of spiking solution for a 20 ng/1 spike) is then added. Similarly, the laboratory control sample (LCS) is prepared by pipetting 1"0 ml of 5"88 gg/1 1641C solution into a 250 ml precleaned bottle and diluting to 200 g (29"4 ng/1) a procedure blank is similarly prepared with 200 ml of water. To each 100 ml of solution, 12"5 ml of 4-hydrochloric acid is added; this is followed with  figure 3 for a 50 ng/1 standard, a 2 ng/1 standard, and a blank. Each standard was processed through the entire digestion procedure. Sample baseline response is measured in the earlier part of the curve. The curve approaches a maximum around 100 s. Peak height response is measured during a 48 period when the response is close to maximum. The curve then returns to baseline before another determination is initiated. The total analysis time is approximately 4"3 min. Slightly more time is required to return to baseline after higher level samples. The response of a 2 ng/1 standard is clearly distinguishable from the blank response. A calibration curve from 0-50 ng/1 is shown in figure 4. Regression linearity has generally been excellent, with all determined regressions having linear correlation coefficients at 0"9985 or greater.
Results on digested verification standard are listed in table 2. Recovery, as a percentage of theoretical, ranged from 85 to 104 for a 29"4 ng/1 standard over the study period.
Results for digestion blanks range from -0"37 to 0"47 ng/1 for all runs over the study period (see table 3). The average determined blank value + blank standard deviation is 0" 12 _ _ 0"25 ng/1.     MDLs have been determined in two sample matrices, laboratory tapwater and unfiltered, nearshore Manchester, Washington seawater. The published US EPA MDL procedure was used as guidance for MDL determination [9].
Data tbr an MDL determinations for tapwater are shown in table 4. The mean concentration determined was 0"56 0" 11 ng/1 with the MDL determined to be 0"34 ng/1. The MDL was also determined on two other nonconsecutive days to be 0"43 ng/1 and 0"60 ng/1. For fresh water, Manchester Environmental Laboratory has established a reporting limit of ng/1 for this method.
Data are shown in table 5 for a determination of an MDL for Manchester seawater. The sample was not filtered, and particulates may contribute to sampling problems and raise the MDL. Filtering of seawater was not tested. If particulates which were somewhat difficult to suspend contained mercury and were not uniformly suspended, an increase in variability could result. The mean concentration was 3"17 _ _ 0"40ng/1 for a determined MDL of 1"2 ng/1.  The MDLs presented here are at least one order of magnitude lower than the water quality criteria for fresh water and seawater. These levels were determined without the additional expense and difficulty of doing gold amalgamation concentration of mercury prior to analysis.

Results--real world samples
Analytical results from a number of water samples are listed in table 6. Results for spike recoveries on some of these samples are listed in table 7. Spike levels used were generally 20 ng/1. All results are within _. 25 except for the 2 ng/1 duplicate spike on Manchester seawater. In this case the spike level of 2 ng/1 was very close to the instrument detection limit where more noise in the spike recovery data will normally be observed. The reliability of the entire analytical and sampling protocol was further demonstrated when selected samples were re-run after several weeks' storage. There was no significant change in results for these samples (table 6). The MDL, recovery, and blank data presented in this paper demonstrate that it is possible to obtain reliable ng/1 level fresh water mercury data at levels less than a tenth of the EPA chronic water quality criteria. These data also indicate that levels of mercury in these selected Washington waters are generally well below the chronic water quality criteria for mercury.
Data from earlier papers [6 and 7-1 have demonstrated the effectiveness of bromine oxidation for releasing mercury from a number of matrices. It is significant that this new method is proving to be robust and is performed at room temperature. Also, it is possible to obtain the required MDLs needed without the additional step and expense of amalgamation concentration.  Future work will include the continued use of this method to characterize low level waters and work on exploring the use of bromine oxidation as a method of sample preparation for mercury in fresh water and marine sediments.