Analytical and Bioanalytical Chemistry

, Volume 378, Issue 4, pp 1075–1082

Ultratrace determination of mercury in water following EN and EPA standards using atomic fluorescence spectrometry

Authors

  • Thomas Labatzke
    • Analytik Jena AG
    • Analytik Jena AG
Special Issue Paper

DOI: 10.1007/s00216-003-2416-x

Cite this article as:
Labatzke, T. & Schlemmer, G. Anal Bioanal Chem (2004) 378: 1075. doi:10.1007/s00216-003-2416-x

Abstract

Chemical vapour generation has been used in combination with atomic fluorescence spectrometry to determine mercury at ultratrace concentrations down to 0.1 ng L−1. A time-based injection of 1 mL of solution for measurement was sufficient to generate a steady-state detector response in the direct mode of measurement. The detection limit calculated from a ten-point calibration curve according to DIN 32645 was 0.26 ng L−1. Instrument noise is limited by reflected radiation from the light source rather than by the dark current of the photomultiplier. The detection limit is directly influenced by the reagent blank which was 2 ng L−1 in the experiments described. Focusing by amalgamation and subsequent thermal desorption generates a detector response which is about eight times higher in peak intensity and about twice as large in integrated intensity. The detection limit under these conditions is 0.09 ng L−1 which can be further improved by preconcentration of larger volumes of solution for measurement. The cycle time for one individual reading is about 40 s without amalgamation and 125 s with amalgamation. The linear dynamic range of the system is five orders of magnitude with a single photomultiplier gain setting. The carry-over is less than 0.3% in direct measurement mode. Reference water samples and a surface water containing approximately 5 ng L−1 were used to prove the validity of the method for real samples. Good accuracy and recoveries of 103% were calculated using the fast direct determination technique.

Keywords

Atomic fluorescence spectrometryChemical vapour generationWaterMercury

Introduction

For many years mercury has been one of the elements monitored at concentrations below 1 µg L−1 in environmental samples such as drinking water or surface water [1]. Despite rigorous control in many countries the global mercury pollution is increasing according to the United Nations Environment Program (UNEP) [2] and global world-wide pollution is higher than previously estimated. Mercury is one of the most critical heavy metals in the environment and in the biological organism. It is considered to be an endocrine-disrupting chemical and it has become obvious that even very low concentrations of total mercury, e.g. in lake water, can be accumulated by a factor of more than 10,000 in fish [3]. Most recent investigations by Frech and co-workers [4] prove that, depending on the kind of ecological system, even extremely low concentrations of mercury can be accumulated in biological organisms to levels which are no longer below threshold values for food. It is obvious that determinations of mercury at 1 ng L−1 levels and below will have to become routine in laboratories in the near future [5].

The classical method for total mercury determination at low concentrations is the so-called “chemical vapour generation” technique which was first described by Hatch and Ott [6]. Mercury, present in the sample as a variety of species is oxidized, so that the solution for measurement contains only the doubly charged cation Hg2+. In a second chemical pre-treatment step it is reduced to its metallic form by use of either stannous chloride (SnCl2), a rather mild reducing agent, or NaBH4, which is much more reactive but much less selective [7]. The neutral mercury is driven from the solution by a carrier gas (usually argon) and transported to the measurement cell. Alternatively it can be trapped on a gold/platinum gauze and, after complete stripping of all matrix components, rapidly desorbed thermally from the gauze [8]. Trapping results in focusing of mercury in the measurement cell by a factor of at least five compared with the direct mode of measurement. The increase in sensitivity can usually be transformed into five times better detection limits [9]. The time for an individual determination will be about three times longer with the amalgamation technique, because two more operating cycles, preconcentration and desorption, are required. The advantages of amalgamation are more complete analyte–matrix separation and hence an even better selectivity. The disadvantages are higher risk of contamination and carry over and longer cycle times. A further improvement in detection limit with the amalgamation technique is achievable if sample volumes beyond the optimum in direct measurement are preconcentrated and released. At the cost of larger sample volumes and longer cycle times the detection limits can theoretically be reduced in a manner which depends linearly on sample volume. However, in real life, very high preconcentration volumes require correspondingly high masses and volumes of reagents which will inevitably increase the contamination level relative to the targeted detection limit. There will be a limit of the preconcentrated volume at which the detection limit is no longer defined by the standard deviation of the spectrophotometer but by the standard deviation of the blank level.

Traditionally the analyte vapour was measured by atomic absorption spectrometry, either in a standard AA spectrometer using mercury hollow-cathode lamps or electrodeless discharge lamps or with a dedicated mercury analyser with special low-pressure mercury lamp and photomultiplier tube or photodiode or a charge-coupled device-type detector [10]. Whereas the first type of instrument results in limits of determination (LOD) of about 100 ng L−1 without amalgamation, a specific mercury AAS analyser might afford an LOD of lower than 10 ng L−1 without amalgamation. Atomic fluorescence spectrometry (AFS) is an attractive alternative to AAS. Its fundamentals were investigated and described [11] only shortly after AAS has been commercially introduced to instrumental analysis, but its application to routine element analysis in flame atomizers was limited. If measurements are performed in argon-flushed cells after analyte–matrix separation, however, AFS has distinct advantages over AAS [12]. These are predominantly spectrophotometric detection limits, which are at least an order of magnitude superior to those of AAS with specific light source and detector, and a dynamic spectrometric working range of 4 to 5 orders of magnitude. The theoretically predicted benefits were later shown experimentally [13] and today the most sensitive mercury analysers are based on atomic fluorescence. Both the US EPA and the European Community recently issued standards which reflect the required limits of determination and the new instrumental developments for the determination of mercury. In the new standards for ambient waters in US EPA method 1631 [5] the required limits of detection and determination are 0.2 ng L−1 and 0.5 ng L−1 respectively. In European Community method EN 13506 [14] the stated working range is between 1 ng L−1 and 100 ng L−1. The method detection limit in the latter method was found to be below 1 ng L−1.

In the work discussed in this paper a new specific mercury analyser based on the chemical vapour generation technique with atomic fluorescence detection was used to study fundamental factors influencing the working range of mercury determinations. Using EPA method number 1631, the factors which define analytical quality were investigated.

Experimental

Spectrometer

A “Mercur” (Analytik–Jena, Jena, Germany) atomic fluorescence spectrometer was used for all experiments. The spectrometer is based on quantitation of the mercury resonance fluorescence at 253.7 nm. The schematics of the optical system are displayed in Fig. 1. Radiation from a low-pressure, high-intensity source (GLE, Berlin, Germany) is focused through a lens on a gas cell. The fluorescence cell has an illuminated volume of 5×20×10 mm3 (w×h×d). The incident beam generates unidirectional fluorescence. It is back-reflected by a metallized cuvette wall for maximum excitation efficiency. The fluorescence in the direction of the incident light beam and opposite to this is discarded. Fluorescence perpendicular to the incident light beam is focused directly, or back-reflected by a metallized cuvette wall through a lens and a baffle on the photomultiplier. The optical system is optimised for minimum reflected light.
Fig. 1

Optical scheme of the atomic fluorescence spectrophotometer

Dynamic range

The fluorescence obtained from a line source is linearly dependent on the intensity of the incident radiation and the concentration of fluorescing atoms. From theoretical considerations the linear dynamic range should be greater than four orders of magnitude. To match the best signal-to-noise ratio with the required dynamics of photomultiplier and amplifiers, the system is equipped with an automatic gain setting which is based on measurement on the actual line source intensity and combined with sensitivity information obtained from instrument performance measurements during first installation of the system.

Generation and control of mercury vapour

The sample, usually located in an automated sampling unit (model AS 52 S, Analytik–Jena) is transported via a peristaltic pump to the reactor (Fig. 2a) with a sampling speed of about 10 mL min−1. A group of valves selects between reaction and waste. The procedure is a time-based injection of about 1 mL sample followed by stream of carrier acid (usually dilute HCl). Carrier and reducing agent are pumped with a second, independently operated, peristaltic pump to the reactor unit. In most steps the flow speeds of reductant and carrier are 2.5 mL min−1. During sample injection the reductant flow is increased to 4 mL min−1 and carrier is replaced by sample. The second peristaltic pump also handles waste transport. In the reactor the reduction from cationic to metallic mercury takes place. Argon at a adjustable speed of 85–170 mL min−1 is used as carrier gas to separate and transport the gaseous mercury from the liquid matrix in a separator unit (Fig. 2b). The analyte gas is then passed through a bubble sensor to the membrane drying unit. The bubble sensor allows only gas to pass downstream behind the gas–liquid separator. If liquid droplets are mixed with the reaction gas a valve is activated which automatically switches the outlet from the gas–liquid separator to waste. The reaction gas, consisting predominantly of argon, metallic mercury vapour, and water vapour, is passed through membrane tubing (Perma Pure llc. NJ, USA). There the reaction gas is dried by a counterflow of argon of about 170 mL min−1. The dried analyte gas is either collected on one or two gold collector units or is passed directly to the measurement unit. The standard operating sequences available are:
Fig. 2

(a) Flow scheme for liquids. (b) Flow scheme for gases

  1. 1.

    sample injection—reaction—separation—drying—determination in the fluorescence cell;

     
  2. 2.

    sample injection—reaction—separation—drying—collection 1—thermal desorption—determination in the fluorescence cell; or

     
  3. 3.

    sample injection—reaction—separation—drying—collection 1—thermal desorption 1—collection 2—thermal desorption 2—determination in the fluorescence cell.

     
The reaction unit with all liquid and gas streams, gas–liquid separation, and two amalgamation cells is basically as described in EPA method 1631 [5]. Additional valves, however, enable individual cleaning and flushing of parts of the system, if high concentrations of mercury are erroneously passed through the system:
  1. 1.

    Sample or carrier can be directly passed into waste without contact with the reactor unit or the gas–liquid separator. This way the sampling part can be cleaned selectively.

     
  2. 2.

    A second valve behind the gas–liquid separator can be switched to “waste” to make sure that the whole reaction and separation unit can be flushed and cleaned without contaminating the gas drying unit, collectors, or measurement cell. This valve is also active if liquid bubbles are detected in the reaction gas stream coming from the gas liquid separator.

     
  3. 3.

    Finally a high stream of about 830 mL min−1 can be activated to flush the measurement cuvette rapidly, to minimize the time for one analysis. This procedure is called “fast baseline return” (FBR).

     
A typical operating procedure for the determination of mercury without amalgamation is listed in Table 1; details for the same process with amalgamation are given in Table 2.
Table 1

Function–time procedure for determination of mercury by AFS by direct introduction of generated mercury vapour into the fluorescence cell

Step

Action

Time interval (s from start)

Flow S/C/R (mL min−1)

Flow Ar (mL min−1)

Tubing

Cuvette

1

System cleaning

0–5

C2.5/R2.5

85

830

2

Auto zero

5–10

C2.5/R2.5

85

830

3

Reaction

10–16

S10/R4

85

4

Read

10–30

C2.5/R2.5

85

5

Purge

16–30

C2.5/R2.5

85

6

Read/FBR

30–35

C2.5/R2.5

170

830

7

System cleaning

30–40

C2.5/R2.5

170

830

Flow S/C/R=flow rate of sample, carrier, and reductant

FBR= fast baseline return

Argon flows can be run in parallel through tubing to waste and via cuvette to waste

Table 2

Function–time procedure for determination of mercury by AFS with amalgamation–desorption of the generated mercury vapour

Step

Action

Time interval (s from start)

Flow S/C/R (mL min−1)

Flow Ar (mL min−1)

Tubing

Cuvette

1

Reaction

0–6

S10/R4

170

2

Collect

6–36

C2.5/R2.5

170

3

Clean

36–40

0/0

830

4

Auto Zero

40–45

0/0

830

5

Desorb

45–55

0/0

85

6

Desorb/FBR

55–65

0/0

170

830

7

Read

45–55

0/0

85

8

Read/FBR

55–65

0/0

170

830

9

Clean

65–80

0/0

170

830

10

Cooling

65–125

0/0

170

830

Flow S/C/R=flow rate of sample, carrier, and reductant

FBR=fast baseline return

In a fluorescence experiment the following wavelength-specific intensities can fall on the detector during or before a measurement cycle:
  1. 1.

    reflected light from the excitation source;

     
  2. 2.

    mercury fluorescence as a result of contamination of argon, tubing, and reservoirs;

     
  3. 3.

    mercury fluorescence as a result of contamination of carrier, reductant, or reagents; and

     
  4. 4.

    mercury fluorescence from the measurement solution.

     

Reflected light is quantitated and a correction is made shortly before each read cycle. It is also used to control the intensity output of the mercury radiation source and correct for minimal possible drifts in source intensity. At the same time argon, cleaned by activated carbon, which is passing directly through the fluorescence cell might add infinitesimally small Hg-specific fluorescence to the baseline reading. This also is corrected in the baseline auto-zero cycle. This reading should be very small and more or less constant. It will not add to the blank reading but might influence the baseline noise level.

In a second and independent mode of operation, the argon can be passed over the reaction unit while carrier and reduction are active. This way a “constant” blank level of carrier and reagent is automatically corrected for. If the blank level is elevated, however, it will increase the baseline noise level also.

The generated analyte gas is flushed through the absorption cell with a fairly low argon flow. Although variable between 85 and 250 mL min−1 it is usually optimised for highest mercury density and optimum peak shape in the range 85 mL min−1 without amalgamation and 170mL min−1 with amalgamation. In these circumstances mercury can be removed from the measurement volume completely after about a minute only. With FBR transient Hg can be efficiently removed within less than 5 s.

Reagents and standards

Ultrapure deionised water (resistivity at least 18 MΩ) was used for all operations. All reagents were at least “analytical grade” or higher. Borosilicate containers were used to prepare and store all standards and all solutions for measurement. For preparation of blanks and standards the containers were cleaned with bromine-containing blank solutions for at least 24 h.

In all experiments we followed closely the recommendations given by EPA method 1631 [5].

The stock solution for oxidizing and stabilizing standards and samples consists of 10.8% KBr (w/v) and 15.2% KBrO3 (w/v) in concentrated HCl; 10 mL per litre of sample or standard is used to stabilise and acidify the solutions for measurement.

The carrier consists of 1.8% (w/v) dilute hydrochloric acid.

The reduction solution contains 2% (w/v) SnCl2 in 3.6% (w/v) hydrochloric acid. It is purged for at least 2 h by bubbling purified argon through the container.

The hydroxylamine hydrochloride stock is a 12% (w/v) solution in ultrapure water. The reductant is purged from Hg by bubbling purified argon through the solution after addition of 1 mL SnCl2 solution.

Standards were prepared from a 1 mg L−1 Hg stock solution and stabilized by addition of the bromine-containing solution described above. The pre-reduction solution was added less than 1 h before measurement of the sample. Prereduced samples at concentrations below 100 ng L−1 are stable for a few hours only.

A river water certified reference material CRM-ORMS-2 (National Research Council Canada, Ottawa, Canada) was used for validation of the data. The CRM was diluted 1:5. Its original mercury concentration is 30.6 ng L−1 (±2.3 ng L−1). The CRM was handled and pretreated in the same way as blanks and samples. Unfiltered surface water from the river Saale was used for recovery experiments. The water was stabilized with the bromine-containing solution described above immediately after collection.

Results and discussion

Photometric noise and detection limits

The noise in AAS measurements near the detection limit is mainly source-flicker noise. The relationship between concentration and instrument response is logarithmic. In resonance fluorescence, the process which is used in all commercial AFS systems, the fluorescence is often detected perpendicular to the incident light beam. From theoretical considerations [15] Winefordner et al. showed that the fluorescence radiance, BF, is linearly related to the number of atoms in the cell, n0, and linearly related to the effective source intensity, I0. Winefordner [12] in his paper concluded that in non-flame atomizers AFS should afford significantly better detection limits than AAS and related the effect to the type of noise limiting AAS (source flicker) and AFS (shot noise from the detector). The absolute and relative detection limits obtained are clearly significantly lower than those obtained by AAS and are approximately 0.5 ng L−1. They are, however, still not defined by the dark-current noise of the photomultiplier but are a combined effect of the flicker of radiation reflected from the line source and the increased background photon intensity on the detector. The criterion for the optimisation of the optical system is to collect as much of the fluorescence radiation as possible to increase sensitivity and at the same time minimize reflected radiation from the source to minimize noise. As an example 250 peak readings of 20 ms data acquisition packages with the photomultiplier set to 369 V were taken. In the first experiment the excitation source was switched off, in the second it was operated under normal measurement conditions (Fig. 3). The ordinates in these examples are arbitrary intensity units but both ordinates have, in essence, the same resolution. The fluorescence intensity in Fig 3b is close to zero, because only purified argon is flushed through the cell. The standard deviation (n=250) of peak intensities was 0.43 counts in Fig 3a and 1.60 counts in Fig 3b, clearly indicating that the noise level is influenced by reflected light from the source. If fluorescence higher the background light level is generated by analyte mercury or by contamination, the absolute noise level increases with the magnitude of the fluorescence intensity, because of the direct proportionality of light flux and fluorescence intensity. In other words, the noise level of the lamp is amplified by the process of fluorescence. This effect is shown in Fig. 4 where the standard deviations of calibrated intensity units at 369 V photomultiplier voltage for different mercury concentrations are displayed. The blank solution in this case contains 2 ng L−1 Hg. From these considerations it becomes obvious that, apart from properties which are photometer constants, for example radiation source intensity and reflected intensity, the detection limit will inevitably depend on the blank level under measurement conditions. For detection limit targets below 1 ng L−1 the total blank level of all sources of contamination should be below 10 ng L−1.
Fig. 3

individual peak values from 250 readings at 369 V photomultiplier gain: (a) radiation source off (dark current noise of photomultiplier); (b) radiation source on (combined shot noise and flicker noise of reflected source radiation)

Fig. 4

Noise level of peak intensity (calibrated intensity units, n=250) for blank (~2 ng L−1 Hg, 10 ng L−1 Hg, 40 ng L−1 Hg, 100 ng L−1 Hg, 600 ng L−1)

If mercury is preconcentrated on a metallic collector, usually a gold alloy gauze, the selectivity of the method is further increased. All gaseous matrix compounds and gaseous reaction products usually pass the gold net without being trapped on the cold gauze. The mercury-trapping efficiency on the net is usually close to 100%. After a flushing period, during which all tubing is cleaned of gaseous contaminants, the gold net is heated rapidly and mercury is released within less than a second into a cleaned stream of argon gas and transported to the fluorescence cell. This way the mercury peak is focused by a factor of approximately 7–8 compared with direct introduction of the analyte vapour. If integrated intensity is evaluated, both signals should be more similar in size if the measurement is taken until the signal returns to baseline. In Fig. 5 mercury intensity after direct sample introduction (a) and after preconcentration (b) is shown. For peak evaluation the fast baseline return gas flow can be selected. Under these conditions, the peak was 7.7 times higher after preconcentration whereas the integrated intensity was barely twice as large in preconcentration mode. If the contamination level is under perfect control, the higher peak intensity can be converted proportionally into lower detection limits. Under laboratory conditions, however, even with amalgamation the LOD obtainable is probably hardly better than that required by the EPA, i.e. 0.2 ng L−1. A further increase in sensitivity by collection of mercury from higher volumes of sample might as well lead to better detection limits. With the system described in this paper a preconcentration time of up to 40 s, representing approximately 6 mL of the solution for measurement would be possible. Again, in real laboratory life, the signal-to-noise ratio of the fluorescence signal is good enough to determine sub ng L−1 mercury concentrations from about 1 mL sample.
Fig. 5

Time resolved intensity of 10 ng L−1 Hg: (a) direct reading, peak intensity=0.0011, integrated intensity=0.012 (b) amalgamation, peak intensity=0.0087, integrated intensity=0.023

The detection limits measured for this system have been obtained under conditions specified by the CEN and US EPA norms and following the rules defined by German DIN standard 32645 [16]. Plots were constructed each containing ten equidistant points in the range between 1 and 10 ng L−1 (Fig. 6). Each point on the curve including the blank was determined three times. All data were introduced into a linear regression calculation. Under these conditions the limits of detection and determination obtained were 0.26 and 0.90 ng L−1, respectively. The contamination level of the blank solution was 2 ng L−1.
Fig. 6

Ten-point calibration curves according to DIN/EN. Each point represents an average from three individual readings: (a, b) 1 mL of the solution for measurement with (circles) and without (triangles) amalgamation; (c) amalgamation of 3 mL of the solution for measurement (squares)

With amalgamation on one gold net the limits of detection and determination were 0.090 and 0.325 ng L−1, respectively, using the 10-point calibration plot and 1 mL of sample per injection. After preconcentration of 3 mL of sample, the respective values were 0.065 ng L−1 for detection limit and 0.234 ng L−1 for quantitative determination limit.

Linear dynamic range, working range, carry-over, and relative standard deviation

The system described makes use of a conventional photomultiplier. Detector, electronic amplification, and analogue-to-digital conversion make a dynamic range of five orders of magnitude possible. Best photometric detection limits are obtained, however, when the photomultiplier voltage is optimised for small concentrations. The instrument is therefore equipped with an algorithm relating the desired working range to the optimum multiplier setting. The absolute correlation between gain and detector response is calibrated for each individual instrument on installation, using the reflected light from the source which is practically constant from instrument to instrument. As an example a fixed setting of 280 V was used to produce a calibration plot from 1 ng L−1 to 100 µg L−1(Fig. 7). Each point is the average from three replicate determinations. The plot is linear over five orders of magnitude with a correlation coefficient of 0.9974 which is well within the ISO definition of linearity [17]. To make use of the wide dynamic range it is essential to control carry-over from sample to sample. The analyte vapour generated is driven through the system by an argon gas stream of about 85 mL min−1. Higher carrier gas velocities will result in loss of peak sensitivity. Metallic mercury vapour, on the other hand, moves unidirectionally as a result of its natural diffusion. If samples with strongly varying concentrations are determined it would, under these conditions, require several minutes to bring the blank down to a tolerable level again. To minimize the time for determinations all components are minimized in volume and the system is flushed by the purge gas stream of 830 mL min−1 mentioned above to rapidly remove the mercury vapour from the measurement cell after the peak maximum has been quantitated. The carry-over from a sample of 1 µg L−1 Hg was measured by a triplicate determination of the blank solution immediately after the high standard under standard measurement conditions without amalgamation. The raw values are listed in Table 3; the corresponding carry over concentrations for the first three blanks were 0.0044, 0.0024, and 0.0019 µg L−1. The calculated mean carry-over was 0.0029 µg L−1 or 0.3% of the high standard. The measurement conditions in these experiments were in accordance with the manufacturer’s recommendations, i.e. 1000 µL sample volume, 369 V multiplier gain and a total measurement time of 50 s for 1 cycle.
Fig. 7

Linear dynamic range from 1 ng L−1 to 100 µg L−1 at 280 V multiplier gain

Table 3

Carry-over from 1 µg L−1 mercury on consecutive blank readings

Solution for measurement

Peak intensity (counts)

Peak intensity, reagent blank corrected

Concentration (µg L−1)

Concentration reagent blank corrected (µg L−1)

Reagent blank

0.00020

0.0022

High standard

0.08915

0.08895

1.002

1.000

Blank 1

0.00059

0.00039

0.0066

0.0044

Blank 2

0.00041

0.00021

0.0046

0.0024

Blank 3

0.00037

0.00017

0.0042

0.0019

The relative standard deviation at elevated concentrations of approximately ten times the limit of determination is mainly a result of the reproducibility of sample introduction and chemical reaction . The RSD (n=11) for a 10 ng L−1 standard was 0.7% without amalgamation and 0.7% with amalgamation.

Determination of mercury in water samples

To prove the data obtained from acidified standards recovery experiments were performed with a typical surface water from the river Saale, Thüringen, Germany. The water was used unfiltered and undiluted. It was collected in a precleaned borosilicate glass container and stabilized with the bromine-containing solution immediately after collection. For validation of accuracy we used the certified reference water ORMS-2 from NRCC and diluted it 1:5 to reduce the concentration to below 10 ng L−1. Both samples were determined directly, i.e. without amalgamation. For a fluorescence experiment these are the hardest conditions. The value calculated back for the undiluted reference water was 31.3 ng L−1 (±0.6 ng L−1), in full agreement to the reference value of 30.6 ng L−1 (±2.3 ng L−1). Mercury in the collected water of river Saale was found to be 3.7 ng L−1 (±0.1 ng L−1). Recovery of 5 ng L−1 Hg from river Saale water was 103±1%. The relative standard deviations both for diluted reference water and for river Saale water were between 2 and 3%. The individual data sets are listed in Table 4.
Table 4

Determination of CRM-ORMS-2 and water from the river Saale. Sets 1 to 9 are within-batch determinations with three replicates per samples. The average values and standard deviations stated in the text are averages from individual runs (between-batch average) and standard deviations

Sample

Set #

Determined (ng L−1)

s.d. (ng L−1)

r.s.d.%

Recovery (%)

CRM

1

31.9

0.97

3.0

 

2

31.4

0.43

1.4

 

3

30.7

0.84

2.7

 

Average CRM

 

31.3

0.60

  

Saale

1

3.81

0.11

2.9

 

2

3.62

0.15

4.1

 

3

3.71

0.07

1.9

 

Average Saale

 

3.71

0.10

  

Saale+5 ng L−1

1

9.03

0.02

0.2

104

2

8.99

0.25

2.8

103

3

8.92

0.20

2.2

102

Avg Saale+5 ng L−1

 

8.98

0.06

 

103

Conclusions

Although blank control of mercury in containers, tubing, gases, and all reagents and solutions used is of central importance if the element has to be determined at concentrations below 10 ng L−1, these conditions can be obtained in a routine laboratory without excessive equipment or operations. It has been found that method detection limits below 0.5 ng L−1 are possible if the blank level can be limited to about 2 ng L−1 Hg. The spectrometric detection limits of optimised atomic fluorescence spectrometers can reach the currently most stringent detection limit requirements even without amalgamation. Compared with atomic absorption spectrometry they are lower by about one order of magnitude. They can be improved by another order of magnitude if the amalgamation technique is applied. A second inherent advantage of atomic fluorescence over atomic absorption spectrometry is the linear dynamic working range of five orders of magnitude. The wide dynamic range can be used only if carry-over from sample to sample is minimised. Values of less than 0.5% carry-over are possible even for the critical element mercury by optimisation of liquid and gas velocities and pathways. Quenching has to be avoided or minimised by careful gas–liquid separation, optimised drying of the mercury cold vapour, and, in complex samples, by the amalgamation technique.

Copyright information

© Springer-Verlag 2004