Chemofiltration of mercury water samples through zinc sulfide layer and determination by wavelength-dispersive X-ray fluorescence spectrometry

Abstract

Chemofiltration through zinc sulfide collected on a membrane filter was used for mercury preconcentration from water samples. Measurements of mercury radiation from the front and back sides of the sample indicate that the ion exchange proceeds on the surface and consequently mercury is collected as a thin layer of HgS on the ZnS. Therefore, filtration of mercury solutions through other sulfides, e.g., CuS, MnS and As₂S₃ will give the same radiation intensity. The loading inhomogeneity was examined using a mapping sample with 1 mm spot size and WDXRF measurements. The mercury recovery was investigated for zinc mass per unit area from 20 to 1000 μg cm⁻². Samples with mass per unit area of 200 μg cm⁻² of zinc were used because of their good recovery (ca. 100%), the fact that different flow rates between 10 mL min⁻¹ and 25 mL min⁻¹ do not influence the quantity of recovered mercury, the good precision and the ruggedness of the method. The detection limit for 1000 mL of filtered water, 100 s counting time and a molybdenum target X-ray tube operated at 50 kV and 40 mA is equal to 0.09 and 0.17 ng mL⁻¹ for Hg Lα₁ and Hg Lβ₁, respectively. The RSD characterizing the sample preparation is ca. 3.5%. The calibration was performed using standard samples in two Hg mass per unit area ranges: 0.5–20 and 0.5–3 μg cm⁻². R² and residual errors RMS of 0.9998, 0.091 μg cm⁻² and 0.9997, 0.024 μg cm⁻² were obtained for the two calibrations, respectively. The selectivity was investigated by determination of mercury in test mixtures to which selected compounds such as alkali and heavy metals, phenol and bromides had been deliberately introduced.

---

1. Introduction

Several analytical methods are applied in the determination of mercury in various types of water: inductively coupled plasma atomic emission spectrometry (ICP-AES), inductively coupled plasma mass spectrometry (ICP-MS), instrumental neutron activation analysis (INAA), electrochemical methods and, above all, cold vapor atomic absorption spectrometry (CVAAS), generally using amalgamation. Advances and difficulties in mercury determination using spectroscopic, electrochemical and radiometric methods are discussed in a review paper.¹ Spectroscopic and chromatographic techniques employed in the speciation of mercury in environmental samples are presented in ref. 2.

X-ray fluorescence spectrometry (XRF) usually requires the preconcentration of mercury prior to analysis due the low concentration of this element in various type of waters. Determination of mercury in ground and waste water by energy-dispersive X-ray fluorescence spectrometry (EDXRF) after preconcentration onto zirconium-loaded activated charcoal was carried out by Peraniemi et al.^3,4 The detection limit obtained was 58 ng mL⁻¹. Lau and Ho determined mercury among other trace elements using EDXRF after precipitation with disodium piperazino-1,4-bis(dithiocarbamate).⁵ Diethylammonium diethyldithiocarbamate-loaded polyurethane foam discs were used for the pre-concentration of phenylmercury, methylmercury and inorganic mercury in water.⁶ Mercury radiation was measured by EDXRF with an annular ¹⁰⁹Cd excitation source. Extraction of a Hg–dithizone complex with fatty acids was used for prelimary preconcentration.⁷ After the separation of solidified organic phase XRF measurement was performed. Total-reflection X-ray fluorescence spectrometry (TXRF), due to low detection limits in comparison with conventional measurement geometry, is frequently applied in the analysis of environmental samples. Nevertheless, mercury determination in water also requires prior preconcentration, e.g., using co-precipitation or cation-exchange. Hołyńska et al. described the method for mercury determination in drinking water.⁸ Trace amonts of mercury were preconcentrated by complexation using carbamates, followed by solvent extraction with methyl isobutyl ketone. The detection limit obtained was 0.06 ng mL⁻¹. Koulouridakis et al. achieved a 0.8 ng mL⁻¹ limit of quantitation using membrane complexation and TXRF.⁹ Amalgamation with a thin layer of silver¹⁰ or gold^11,12 affixed to a specular-surface quartz reflector was also used in the determination of mercury.

Disam et al. proposed the preconcentration of trace elements such as Hg, Ag, Cu, Bi, Pb, Cd, Sn, As, Se, Te, Zn, Co and Ni using filtration through metal sulfide layers, e.g., ZnS, MnS, CuS, PbS.¹³ In this paper, chemofiltration of mercury solutions through zinc sulfide is developed. Good recovery and repeatability of the analysis results can be achieved only for homogeneously loaded zinc sulfide over the whole area of the membrane filter. Therefore, inhomogeneity of the mass per unit area of the prepared samples and depth distribution of mercury in zinc sulfide is carefully investigated.

2. Experimental

2.1. Equipment

A wavelength-dispersive X-ray spectrometer (Philips PW1410) with an LiF(200) analyzing crystal was used for the measurements. The 1 mm side-window molybdenum target X-ray tube was operated at 50 kV and 40 mA. The incidence and take-off angles were 60° and 40°, respectively. The measurements were performed in a vacuum with rotation of the sample. For the detection of Hg Lα₁ and Hg Lβ₁, a flow-proportional counter (counter gas 90% Ar + 10% CH₄) and a scintillation counter with a thallium-doped crystal of sodium iodide 1 mm thick were used. The net intensities were determined for each sample by the measurement of fluorescent radiation of the element i (peak) and the measurement of the continuum close to the peak.

To map the sample, the wavelength-dispersive X-ray spectrometer Rigaku ZSX Primus with a 30 μm end-window rhodium target X-ray tube was used. The X-ray tube was operated at 50 kV and 80 mA for mercury and 40 mA for zinc and sulfur. An LiF (200) analysing crystal was used for Hg Lα₁ and Zn Kα measurements; Ge (111) was used for S Kα. A scintillation counter was used for the measurement of Hg and Zn radiation; the flow-proportional counter (counter gas 90% Ar + 10% CH₄) was used for sulfur radiation. The measurements were performed in a vacuum. A diaphragm of 1 mm diameter was used for small-spot analysis.

The zinc sulfide was collected on a cellulose acetate filter (0.2 μm pore size, Sartorius, 47 mm diameter) using a filtration assembly (47 mm, Sigma–Aldrich). After the collection of zinc sulfide and the filtration of the analyzed solution, the filter was protected using a 0.5% (m/V) solution of polystyrene in carbon tetrachloride. The collecting surface was 10 cm² in area.

2.2. Sample preparation

The sample was prepared in three stages: precipitation of zinc sulfide, collection of zinc sulfide onto the membrane filter and filtration of analyzed sample through the zinc sulfide layer. 2 mL of 1 mg mL⁻¹ Zn, 5 mL of 10% (m/m) thioacetamide and 0.5 mL of 10% (m/m) ammonia solution were added to 50 mL of redistilled water. Then, the mixture was heated in a water bath for 20 min. The precipitated zinc sulfide was collected onto the cellulose acetate filter and rinsed with 50 mL of redistilled water. The mass per unit area of the prepared sample was 200 μg cm⁻² of zinc (ca. 300 μg cm⁻² of zinc sulfide). Then, the analyzed water sample (from 200 to 1000 mL) was filtered through the freshly prepared zinc sulfide layer. After being filtered and dried under an IR heater, the sample was protected using a 0.5% (m/V) solution of polystyrene in carbon tetrachloride.

2.3. Reference sample preparation

Various quantities of mercury (from 5 to 200 μg) in 200 mL of redistilled water were filtered through 300 μg cm⁻² zinc sulfide layer. The samples were dried under an IR heater and protected using a 0.5% (m/V) solution of polystyrene in carbon tetrachloride.

2.4. Recovery

Two series of the samples were prepared for evaluation of the recovery. In the first series 50 μg of mercury was co-precipitated with different amounts of ZnS (from 0.2 to 10.0 mg of zinc) and then the precipitates were collected on the filters. In the second series, solutions containing 50 μg of mercury were filtered through ZnS of various thicknesses (from 0.2 to 10.0 mg of zinc) initially collected on the membrane filter.

2.5. Influence of the flow rate on the recovery

50 μg of mercury in 200 mL of redistilled water was filtered through a 300 μg cm⁻² zinc sulfide layer using different flow rates between 10 mL min⁻¹ and 25 mL min⁻¹.

2.6. Influence of pH on the recovery

200 mL solutions containing 50 μg of Hg were adjusted to different pH values (from pH 1 to 5) with nitric acid and filtered through a 300 μg cm⁻² zinc sulfide layer.

3. Results and discussion

The large difference between the solubility products of ZnS and HgS (6.9 × 10⁻²⁶ and 3 × 10⁻⁵⁴, respectively) enables the attainment of quantitative ion exchange during chemofiltration. Nevertheless, several conditions have to be met in practice. First of all, good recovery and repeatability of the analysis results can be achieved for homogeneously loaded zinc sulfide over the whole area of the membrane filter. The influence of inhomogenous zinc sulfide loading may critically affect the results obtained when unloaded filter areas occur if the mass of the zinc sulfide is insufficient. Therefore, the influence of zinc sulfide mass on the recovery has been carefully investigated. Two series of the co-precipitated and filtered samples with various masses per unit area of the zinc sulfide were prepared for this purpose. For comparison, results obtained for co-precipitated and filtered samples absorption effects have to be taken into consideration. In the case of co-precipitated samples, a homogeneous distribution of HgS in ZnS can be assumed. Thus, the measured radiation intensity of mercury in ZnS of various masses can be corrected using the absorption correction factor calculated from eqn. (1).where m is the ZnS mass per unit area, χ_ZnS is total mass attenuation coefficient of ZnS for mercury and incident radiation, given by eqn. (1a).where μ_ZnS(λ) and μ_ZnS(λ_i) are the mass-attenuation coefficients of ZnS for the incident and fluorescent radiation, respectively; ϕ₁, ϕ₂ are the angles of incidence and exit of primary and fluorescent radiation, respectively; λ and λ_i are the wavelengths of primary radiation and analyte element i fluorescent radiation, respectively.

Absorption correction factors were determined using a well known emission–transmission method (measurement of Hg radiation in ZnS collected on the filter without a target, with a target located at a position adjacent to the back of the filter and the target covering the filter without ZnS) and also the two-masses method.¹⁴ Both methods gave comparable results. Fig. 1a shows the radiation intensity of the mercury in zinc sulfide before and after absorption correction. The case of filtered samples is more complicated because the absorption of mercury radiation depends on the depth distribution of the HgS. Nevertheless, investigations show that the measured intensity of the mercury radiation from the front side of the sample is equal to the radiation intensity measured from the back side divided by the transmission factor (eqn. (2)) for the ZnS layer and filter.

exp(−χ_ZnSm_ZnS) exp(−χ_filterm_filter) (2)

This indicates that the ion exchange ZnS + Hg²⁺ → Zn²⁺ + HgS proceeds on the surface and consequently mercury is collected as a thin layer of HgS on the ZnS. Filtration of mercury solutions through other sulfides e.g., CuS, MnS and As₂S₃ gives the same radiation intensity because absorption effects are not observed from these sulfides. Therefore, this can be confirmation of the collection of HgS as a thin layer on the sulfides. Fig. 1b presents the radiation intensity of mercury measured from the front side of the sample and from the back side before and after absorption correction (eqn. (2)). Therefore, the recovery can be calculated from eqn. (3).The effect of a different quantity of loaded zinc sulfide on the mercury recovery is presented in Fig. 2.

Fig. 1 Measurement of Lα₁ Hg radiation in co-precipitate samples (a) and filtrate samples (b).

Fig. 2 Recovery of mercury in samples with various thicknesses of ZnS.

For the samples between 100 μg cm⁻² and 1000 μg cm⁻² of zinc the recovery is close to 100%: however, for the samples below 100 μg cm⁻² a decrease in recovery is observed. This results from two effects. Firstly, the investigation shows that in the applied conditions of precipitation for zinc mass below 1 mg the quantities of precipitated ZnS are smaller than amounts resulting from stoichiometry. Secondly, when the mass of the zinc sulfide is insufficient then unloaded filter areas can occur. The loading inhomogeneity was examined using mapping sample with 1 mm spot size and WDXRF measurement. Table 1 and Fig. 3 present the mapping results for Hg, S and Zn in exemplary samples, 20 μg cm⁻² and 200 μg cm⁻², of added zinc and 50 μg of collected mercury. For both samples some inhomogeneous distribution of these elements can be observed. Relative standard deviations (RSD) for Zn and S are four times higher for a 20 μg cm⁻² sample than for a 200 μg cm⁻² sample. The standard deviations for mercury in both samples are similar, whereas the mean quantity of recovered mercury in 20 μg cm⁻² of zinc is ca. 85% of mercury quantity in 200 μg cm⁻² of zinc. This results from a significantly smaller quantity of precipitated ZnS than the added quantity of zinc (one order of magnitude). For 20 μg cm⁻² sample the areas can be observed on which only 1 μg cm⁻² of S is collected. However, in these areas ca. 3.9 μg cm⁻² of Hg is collected, which is equal to 78% of recovery (100% of recovery is equal to 5 μg cm⁻² Hg). Because theoretically 1 μg cm⁻² of S can react with 6.3 μg cm⁻² of Hg, the low recovery can be explained by unloaded zinc sulfide over micro-areas which cannot be detected using a 1 mm spot size X-ray beam. Fig. 3 shows a similar distribution of Hg, S and Zn on the filter area; simultaneously it is worth emphasing that in the 200 μg cm⁻² of the zinc sample, the distribution of mercury is dependent on the S distribution to a smaller degree than in the 20 μg cm⁻² of zinc.

Fig. 3 Mapping results for Hg, S and Zn in exemplary samples 20 μg cm⁻² and 200 μg cm⁻² of added zinc and 50 μg of collected mercury.

Table 1 Mapping results using diaphragm of 1 mm diameter and WDXRF measurements

Element		20 μg cm^−2 Zn	200 μg cm^−2 Zn
Hg	Min./μg cm^−2	3.88	4.75
	Max./μg cm^−2	4.48	5.22
	Average/μg cm^−2	4.19	4.95
	SD/μg cm^−2	0.16	0.15
	RSD (%)	3.8	3.0
S	Min./μg cm^−2	1.01	93.7
	Max./μg cm^−2	1.37	100.2
	Average/μg cm^−2	1.14	97.9
	SD/μg cm^−2	0.10	1.99
	RSD (%)	8.8	2.0
Zn	Min./μg cm^−2	1.90	191
	Max./μg cm^−2	2.59	206
	Average/μg cm^−2	2.29	201
	SD/μg cm^−2	0.21	4.9
	RSD (%)	9.2	2.4

---

Samples with a mass per unit area of 200 μg cm⁻² of zinc were adopted for all experiments because of: good recovery, achievement of a good flow rate (the flow rate decereases with the increase of ZnS mass per unit area), good precision and ruggedness of the method. The experiment shows that for the 200 μg cm⁻² sample, the different flow rates between 10 mL min⁻¹ and 25 mL min⁻¹ do not influence the quantity of recovered mercury. The influence of the pH values on the mercury recovery was examined using 200 mL of solution containing 50 μg of Hg adjusted to different pH values with nitric acid. Fig. 4 shows that zinc sulfide dissolves during the filtration of a solution with a pH below 3. Nevertheless, the amount of remaining ZnS is sufficient to obtain almost 100% recovery with the experiment conditions described. Nevertheless, the analysis of solutions with pH below 3 is not recommended, expecially if the volume of the analyzed solution is higher than 200 mL.

Fig. 4 Influence of the pH on the Hg, Zn and S radiation.

The calibration was performed using standard samples in two Hg mass per unit area ranges. Table 2 presents the calibration parameters (least-squares method) for both Hg Lα₁ and Hg Lβ₁ radiation. The correlation coefficients and residual errors RMS (root of the mean square of the sum of the differences between the measured value and the calculated values) for Hg Lβ₁ are slightly worse than for Hg Lα₁, which results from the sensitivity of these lines. For the examined Hg mass per unit area range from 0.5 to 20 μg cm⁻², a linear relationship between radiation intensity and the mass of mercury is observed and the self-absorption effect in the HgS thin layer deposited on ZnS can be neglected. This effect increases with the mass per unit area of the sample, e.g., if the mass per unit area of mercury is above 80 μg cm⁻², an error resulting from neglect of the absorption effect exceeds 1%. Therefore, collecting mercury above this mass is not recommended.

Table 2 Calibration parameters

Line	Hg mass range/μg cm^−2	Slope/counts cm^2 s^−1 μg^−1	Intercept/counts s^−1	RMS/μg cm^−2	R ^2
Lα1	0.5–20	220.2 ± 1.2	−(11.5 ± 9.4)	0.091	0.9998
	0.5–3	208.5 ± 2.6	−(3.64 ± 4.9)	0.024	0.9997
Lβ1	0.5–20	117.5 ± 0.72	0.46 ± 5.7	0.104	0.9998
	0.5–3	116.1 ± 2.2	−(1.80 ± 4.0)	0.035	0.9995

---

The detection limit was calculated from DL = (3/k)(R/t)^1/2, where k is the sensitivity in counts cm² s⁻¹ μg⁻¹, R is the background count rate in counts s⁻¹ and t is the counting time. The obtained DLs with a counting time of 40 s and for Hg Lα₁ and Hg Lβ₁ are equal to 0.013 μg cm⁻² and 0.027 μg cm⁻² (for 1000 mL of filtered water they are equal to 0.13 and 0.27 ng mL⁻¹, respectively). Extending the counting time up to 100 s achieves a significant improvement of DL: for Hg Lα₁ and Hg Lβ₁ they are equal to 0.009 μg cm⁻² and 0.017 μg cm⁻², respectively (for 1000 mL of filtered water they are equal to 0.09 and 0.17 ng mL⁻¹, respectively).

Repeatabilty was examined by preparing ten samples: 50 μg of Hg in 200 mL was collected on 200 μg cm⁻² of zinc sulfide. The mean value of mercury determined was 48.9 ± 0.9 and 48.5 ± 1.0 for Hg Lα₁ and Hg Lβ₁, respectively. The obtained relative standard deviations of 3.2 and 3.5 for these lines, respectively, can be recognized as satisfactory. Using zinc radiation (Kα or Kβ lines) as an internal standard to compensate for the inhomogeneity of the mass per unit area did not give satisfactory improvement. The selectivity was investigated by determination of mercury in test mixtures to which selected compounds such as alkali and heavy metals, phenol and bromides were deliberately introduced. Table 3 presents the concentrations of mercury determined in these mixtures and also the obtained recoveries. The results are satisfactory with the exception of mercury determination in a bromide solution using the Hg Lβ₁ line. It results from overlapping Hg Lβ₁ with Br Kα, which is adsorbed on the zinc sulfide. Therefore, use of Hg Lβ₁ is not recommended if bromide ions are present in the water sample. It is worth emphasing that accurate results can be obtained even if large quantities of heavy metals are present. These elements are collected onto the zinc sulfide to various extents, nevertheless mercury sulfide is always deposited onto the top layer of the sample due to the extremely small solubility product. Therefore, the mercury radiation intensity is independent of absorption effects from these elements. The enhancement effect from, for example, cadmium, is also not observed. The recovery of mercury has also been investigated in various natural waters: surface, waste, river and drinking water. Table 4 presents the concentrations of mercury determined and the recoveries obtained in these exemplary waters. The confidence limits are calculated for a probability level of 0.95. The determined concentrations are well in agreement with the expected values.

Table 3 Determination of Hg in test mixtures

		Lα1		Lβ1
Matrix	Added/μg L^−1	Determined/μg L^−1	Recovery (%)	Determined/μg L^−1	Recovery (%)
a Pb, Fe, Cr, Co, Cd, Cu, Se, Mn, Ni, As.
0.5 g L^−1 Ca, Mg, Na	50	49.4 ± 1.8	99 ± 4	50.4 ± 2.1	101 ± 4
0.5 g L^−1 Ca, Mg; 5 g L^−1 Na	50	49.8 ± 1.8	100 ± 4	51.1 ± 2.1	102 ± 4
30 mg L^−1 Br	50	52.2 ± 1.8	104 ± 4	81.8 ± 2.1	164 ± 4
200 mg L^−1 phenol	50	49.6 ± 1.8	99 ± 4	46.4 ± 2.1	93 ± 4
100 μg L^−1 of heavy metals^a	50	48.6 ± 1.8	97 ± 4	47.2 ± 2.1	94 ± 4
100 μg L^−1 of heavy metals	10	10.3 ± 0.8	103 ± 2	9.3 ± 1.2	93 ± 2
500 μg L^−1 of heavy metals	50	49.1 ± 1.8	98 ± 4	48.9 ± 2.1	98 ± 4
500 μg L^−1 of heavy metals	10	10.4 ± 0.8	104 ± 2	9.5 ± 1.2	95 ± 2

---

Table 4 Recovery of Hg in natural waters

Water	Added/μg L^−1	Determined/μg L^−1	Recovery, %
Surface water	10	9.9 ± 0.8	99 ± 2
Waste water	10	9.6 ± 0.8	96 ± 2
River water	10	9.2 ± 0.8	92 ± 2
Drinking water	10	9.6 ± 0.8	96 ± 4

---

Conclusion

Chemofiltration through zinc sulfide collected on a membrane filter obtains a low detection limit of mercury. This method is selective, precise and rugged. The mercury is collected as a thin layer of HgS on the ZnS. Filtration of mercury solutions through other sulfides gives the same radiation intensity because absorption effects are not observed from these sulfides. Accurate results can be achieved even if high quantities of heavy metals are present in the filtered sample because mercury sulfide is always deposited onto the top layer of the sample.

References

1. W. L. Clevenger, B. W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem., 1997, 27, 1–26.
2. V. P. Antonovich and I. V. Bezlutskaya, J. Anal. Chem., 1996, 51, 106–113.
3. S. Peraniemi, S. Hannonen, H. Mustalahti and M. Ahlgren, Fresenius’ Z. Anal. Chem., 1994, 349, 510.
4. S. Peraniemi and M. Ahlgren, Anal. Chim. Acta, 1995, 302, 89–95.
5. O. W. Lau and S. Y. Ho, Anal. Chim. Acta, 1993, 280, 269–277.
6. T. Braun, M. N. Abbas, S. Torok and Z. Szokefalvi-Nagy, Anal. Chim. Acta, 1984, 160, 277–282.
7. F. I. Lobanov, E. A. Terent’eva, I. M. Yanovskaya and I. V. Makarov, Zavod. Lab., 1983, 49, 11–12.
8. B. Holynska, B. Ostachowicz and D. Wegrzynek, Spectrochim. Acta, Part B, 1996, 51, 769–773.
9. P. E. Koulouridakis and N. G. Kallithrakas-Kontos, Anal. Chem., 2004, 76, 4315–4319.
10. E. D. Greaves, J. A. Sosa, L. Sajo-Bohus, M. Alvarez, P. Wobrauschek and Ch. Streli, Spectrochim. Acta, Part B, 1997, 52, 945–951.
11. L. Bennun and J. Gomez, Spectrochim. Acta, Part B, 1997, 52, 1195–1200.
12. L. Bennun, V. H. Gillette and E. D. Greaves, Spectrochim. Acta, Part B, 1999, 54, 1291–1301.
13. A. Disam, P. Tschöpel and G. Tölg, Fresenius’ Z. Anal. Chem., 1979, 295, 97–109.
14. R. Sitko and J. Jurczyk, X-ray Spectrom., 2003, 32, 113–118.

---