Development of a robust technique for sampling volatile metal(loid)s in ...

Anal Bioanal Chem (2002) 374 : 1191–1198 DOI 10.1007/s00216-002-1636-9

O R I G I N A L PA P E R

Britta Planer-Friedrich · Jörg Matschullat · Broder J. Merkel · Gerhard Roewer · Peter Volke

Development of a robust technique for sampling volatile metal(loid)s in wetlands

Received: 21 March 2002 / Revised: 8 August 2002 / Accepted: 11 October 2002 / Published online: 9 November 2002 © Springer-Verlag 2002

Abstract The formation of volatile organic and inorganic metals and metalloids in aquatic environments is a known, but not very intensively investigated, process. Several techniques have been developed over the past 10 years to determine these trace components. These techniques are of limited use in wetland environments, where samples have to be taken from the soil-water interface, and require an immediate sample analysis due to thermodynamic instabilities of the volatile metal(loid)s. This paper presents an innovative sampling technique for total concentrations of volatile metal(loid)s in wetlands, based on an in situ gas-water separation via a porous PTFE membrane and stabilising the volatile metal(loid)s in a liquid sorbent (NaOCl solution). Samples may thus be collected even at remote sites, where longer storage times have to be accounted for. The sampling system was tested by means of a laboratory facility simulating the generation of arsine and dimethyl arsine under abiotic conditions as well as under field conditions. Results for sampling efficiency, reproducibility, and long-term storage are presented. Application of the sampling system in the field is shown. Keywords Volatile metal(loid)s · Volatilisation · Wetlands · Liquid sorbents · Arsenic · Hydride generation

B. Planer-Friedrich (✉) · B.J. Merkel · P. Volke Freiberg University of Mining and Technology, Department of Geology, Gustav-Zeuner-Str. 12, 09599 Freiberg, Germany e-mail: [email protected] J. Matschullat Freiberg University of Mining and Technology, Interdisciplinary Environmental Research Center, Brennhausgasse 14, 09599 Freiberg, Germany G. Roewer Freiberg University of Mining and Technology, Department of Inorganic Chemistry, Leipziger Str. 29, 09599 Freiberg, Germany

Introduction Volatile metal(loid)s in wetlands In recent decades, the study of the behaviour and contamination potential of metals and metalloids in aquatic systems, in sediments and in biota, as well as transformations between those phases, has received much attention and experienced great progress. Less attention has been paid to the transformation of metals and metalloids to the gaseous phase. Volatile metal(loid) compounds may preferably form under reducing conditions both from inorganic compounds (hydrides, e.g. from As, B, Bi, Ge, P, Pb, Sb, Se, Si, Sn, Te) and from organic compounds (volatile alkyls or carbonyls). Within the group of alkylated volatile compounds, methyls predominate, in which one H-atom is replaced by one methyl group. The transformation to volatile organic compounds in natural environments is probably microbially catalysed. Volatile methyl compounds are known from As, Bi, Cd, Ge, Hg, Pb, Sb, Se, Sn and Te. They may form both from natural sources, e.g. in volcanic exhalations, natural fires, natural gas, bog gas, geothermal gas and man-made sources such as metal(loid) processing industries, power stations, car exhausts, coal burning, waste heaps or bio reactors. So far, volatile methylated Au, Cr, Pd, Pt, Tl compounds have been produced under laboratory conditions only. Higher alkylated volatile compounds, e.g. ethyl-lead, butyl-tin, ethyl-mercury, or phenyl-mercury compounds, are supposed to derive from anthropogenic sources only. Wetlands may provide excellent conditions for the formation of volatile compounds due to high amounts of organic substances, the presence of bacteria and fungi, high bioactivity, the presence of transferable methyl groups, increased pCO2, increased temperatures, and low redox potential (anaerobic conditions due to biodegradation and oxygen consumption). Determining and investigating the behaviour of volatile metal(loid)s in wetlands is important. Over the last 20 years, wetlands have received increasing world-wide attention as passive treatment systems for highly polluted

1192

mine and tailing effluents. Compared to active water treatment – with high energy, operation and maintenance costs over long periods and the problem of depositing the resulting sludge with an enriched contamination potential – natural and constructed wetlands are based on natural biological and chemical decontamination and immobilisation processes (biodegradation, filtration, dilution, precipitation, complexation, redox reactions, diffusion, sorption, ion exchange) on relatively large areas with long retention times to reduce concentration, load, toxicity and mobility without active intervention. The reactions are kinetically slow, but economical and sustainable, and require no maintenance. Especially due to its low costs, natural attenuation finds a lot of support. Simple “no action” can lead to various negative side effects however, e.g. the creation of more toxic metabolites and daughter products and the uncontrolled transfer of contaminants to other media. Volatilisation includes both risks. Most volatile compounds are more toxic than their corresponding dissolved species. Their transfer to the atmosphere is not a natural attenuation but an uncontrolled transfer to another environmental compartment. Since the occurrence of the volatile compounds is normally limited by their thermodynamic instability, due to spontaneous reactions with oxygen, photolytic decomposition and hydrolysis, the volatile metal(loid)s may be deposited in the immediate vicinity of the wetland. Sampling volatile metal(loid)s in wetlands Scientific investigations of monitored or enhanced natural attenuation processes must, therefore, take into account not only aquatic and solid phases but also the gaseous phase. So far, the main inhibition was the lack of an adequate sampling method from soil-water interfaces. Volatile metal(loid)s have already been trapped from point sources of well-defined gas outlets, e.g. landfill gases, cesspit gases, firedamps, natural gases [1, 2, 3] or geothermal gases [4], as well as intra-oral [5], or ambient air in urban, rural and industrial environments [6, 7, 8]. Within a wetland, however, it is difficult to trap the volatile compounds, since they derive from diffusive sources. Another problem for the sampling unit is the wet environment, which either restricts sampling to degassing from the water surface, e.g. [9] or requires a previous separation of water and gas formed in the water-saturated, organic-rich sediment. Amouroux et al. [10, 11] developed a field technique by means of purging and cryogenic trapping for pre-concentration of volatile metals and metalloids from aquatic environments The detection limits they achieved were very low (pg/L range). However two major disadvantages stand out. Sampling of water was done according to conventional techniques from a surface water column. Even though care was taken during sampling and transferring the samples to gas-tight bottles, taking the sample already disturbs the very sensitive gas-water equilibrium. Second, high-tech equipment and nearby laboratory facilities are required for the rapid analysis of sam-

ples. The focus of our research, however, is to consider that boundary conditions in wetland environments rapidly change with depth, which makes it difficult or impossible to withdraw water samples without changing the sensitive gas-water equilibrium. Consequently, degassing of water is performed in situ to obtain true concentrations of volatile compounds. Since most of the sites to be sampled for mine-water influenced wetlands are located in remote areas without the advantage of near-by laboratory facilities, stabilising the gaseous compounds is a further requirement. Long-term stability (at least 2 weeks) is difficult to achieve due to their high reactivity towards oxygen. Generally, low total volatile metal(loid) concentrations require an on-site enrichment. There are two options for sorbing and long-term stabilisation of volatile compounds: liquid or solid sorbents. Liquid sorbents alter the original speciation by dissolution and oxidation of the gases, thus only total element concentrations are obtained, e.g. inorganic arsine (AsH3), and the organic compounds monomethyl arsine (CH3)AsH2, dimethyl arsine (CH3)2AsH and trimethyl arsine (CH3)3As would all be dissolved and oxidised to arsenate. Solid sorbents, on the contrary, are supposed to preserve speciation. This is important since different volatile compounds show different chemical behaviour and different toxicities. Solid sorbents may be, e.g. porous polymers, carbon molecular sieves or graphitised carbon blacks. They are characterised by high sorption capacity due to large surfaces. Their suitability for sampling volatile metal(loid)s is handled quite controversially [12, 13, 14]. Our own investigations on efficient and reproducible long-term storage on solid sorbents are still ongoing. In the early 1990s, Klusmann developed a method for collecting reduced gases from soil, primarily for purposes of mineral, geothermal and petroleum exploration in remote areas (US Patent 4,993,874 [15]). The sampling unit consists of a container perforated in the upper part permitting selective gaseous communication with the outside environment. The volatile compounds are sorbed, dissolved, and oxidised (e.g. in Ca(OCl)2, NaOCl or H2O2) within the container. Glass wool is used inside the container to increase the surface of the oxidising solution. In wetland environments, the apparatus has to be floated on ponds, or anchored at the water-sediment interface shielded by a plastic or aluminium cone. Alternatively, even more complex, inert gas has to be released from the sampling unit while lowering the assembly in order to compensate for compression of the air in the conical space and avoid flooding of the collector. Gas sampling is only done by passive diffusion. Creating a vacuum in the collector would improve sampling efficiency. However, due to the collector perforations there is the problem of soil dust or water penetration and consequent contamination. According to personal communication with Ronald Klusmann (2001), sampling efficiency of the apparatus is low, based on laboratory calibration. The newly developed method overcomes the problems of effective gas-water separation, enables the application of a vacuum for increasing gas sampling efficiency with-

1193

out contamination by soil dust or water and also shows a sampling and stabilising efficiency of about 80% in liquid sorbents. A major difference from the method of Klusmann lies in the step for collecting the gases from diffusive sources. This is spatially completely separated from the step of trapping and stabilising the volatile metal(loid)s on sorbents.

Method development and results Collection The developed gas collector cells (5 and 6 cm high, 3 and 5 cm wide) were kept small to allow for an accurate depth-orientated sampling at the soil-water interface. For the diffusive gas separation, PTFE (polytetrafluoroethylene) was chosen due to its inertness and hydrophobic characteristics. Figure 1 shows two collector cell models using a hollow porous PTFE tube. PTFE was obtained from Reichelt Chemietechnik (porosity 10 µm, side thickness 6.4 mm, Fig. 1a, left) and Bohlender (porosities 5 and 10 µm, side thickness 3 and 6 mm, Fig. 1a, right). Inserting these collector cells in the medium wetland to sample, only gas diffusion takes place via the PTFE surfaces. Water does not penetrate due to the hydrophobic properties of PTFE. The sampled gases are conducted to the trapping and stabilising unit via small PTFE capillaries.

Fig. 1a, b Collector cell models type 1 for collecting gases from diffusive sources in wet environments applying no or low (–50 mbar) vacuum; material: porous PTFE with a porosity of 5 or 10 µm and a side thickness of 3 or 6 mm

Fig. 2 Collector cell models type 2 for collecting gases from diffusive sources in wet environments applying vacuum up to –500 mbar; material: PTFE membrane with a porosity of 1 µm (M), placed in a filter of solid PTFE, protected below by a porous PTFE disk (porosity 10 µm, thickness 5 mm, left of the membrane) to the sampling media and above by a solid PTFE disk with numerous 1 mm holes for gas trespass to withstand deformation due to vacuum application (right of the membrane)

Applying vacuum to this collecting unit increases sampling efficiency. At a vacuum exceeding –50 mbar, however, while all connections were still water-proof, the bubbling point of the porous PTFE is reached and water breaks through. The different designs with porosities of 5 and 10 µm, and wall thickness of 3 and 6 mm did not show significant differences in water break-through points. A second collector cell model was developed, to apply higher vacuum for increased sampling efficiency (Fig. 2). Contrary to the first collector type, gas diffusion does not take place over the entire body of a hollow porous PTFE tube, but via a PTFE membrane (Infiltec, porosity 0.1 µm), placed in a filter of solid PTFE. The outer side is protected towards the sampling medium by a porous PTFE disk (Bohlender, porosity 10 µm, thickness 5 mm) and towards the inner side by a solid PTFE disk (Bohlender, thickness 6 mm) to withstand vacuum-induced deformation. To allow gas passage, the solid PTFE disk contains numerous 1 mm holes. This collector cell model withstands vacuum up to –500 mbar (Infiltec even cites applications with >2 bar pressure) without water leakage. Figure 3 shows a field set-up of the described equipment, with the collector cell (CC), the trapping and stabilising unit (liquid sorbent, LS), and a vacuum container (VC), the way it was already tested on four different study sites. Vacuum is applied only once and the sampling set left out in the field until the vacuum is completely diminished. The sampling capacity can then be re-calculated from the replaced vacuum volume, the amount of gas dissolved in water and the enrichment factor in liquid sorbent. For the set presented in Fig. 3, the volume of the vacuum container, all hoses, and the trapping unit is e.g. 2 L. Applying a vacuum of e.g. –500 mbar, one would sample 1 L of gas. Assuming a general gas solubility of 22.4 L/mol and considering Henry constants and respective partial pressures of the major atmospheric gases N2 (4.992×10–4 mol/kg), O2 (2.646×10–4 mol/kg), and CO2 (1.002× 10–4 mol/kg) (in sum 8.64×10–4 mol/kg), then 0.019 L gas dissolves in 1 kg water and the degassed volume of water is 1 L:0.019 L/kg=51.7 kg. The enrichment factor of trapping these gases in 100 mL liquid sorbent is therefore 51.7 L:0.1 L=517. The accuracy of the method can be further improved by taking gas samples, analysing the particular partial pressures, and considering also trace gases. Furthermore, gas diffusion has to be taken into account by means of analytical and numerical modelling helping to determine the geometry of the degassed water/soil volume.

1194

Fig. 3 Field set-up of the sampling and trapping equipment showing the collector cell (CC) installed close to the surface of the water-saturated soil, the trapping and stabilising unit (liquid sorbent, LS), and a vacuum container (VC) Hydride generation Volatile metal(loid)s had to be created in the laboratory to test the collecting unit for sampling efficiency. Even though certificated gaseous standards exist for some volatile metal(loid)s (e.g. monomethyl arsine, dimethylarsine in 100 g ampoules, or as high pressure standards in helium at about 200 ppm; personal communication Dr. Dietmar Glindemann 2001), the volatile metal(loid)s had to be generated in situ to test the collecting unit’s efficiency of sampling gases from aqueous phases. This was necessary due to the high reactivity of the volatile metal(loid)s with oxygen and their good dissolution and further reaction capacity in aqueous solutions. Figure 4 shows the laboratory set-up. A high-grade stainless steel reaction vessel (RV) (250 mm high, 100 mm diameter) was constructed as an abiotic reactor replacing the natural wetland environment. PTFE coating of the steel was necessary to avoid sorption effects and precipitation due to pH-changes from extremely acid (dissolution of metals from the high grade steel) to extremely basic (precipitation as metal(loid) oxihydrates on the vessel walls). The reaction vessel is filled to about half its height with sample solution (475 mL). Chemical reagents for hydride generation are added via a syringe (SY) and a capillary connected to the bottom of the reaction vessel. The collector cell (CC) is attached to the roof of the reaction vessel. Gases created in the reaction vessel pass the PTFE membrane of the collector cell and are finally conducted out of the vessel by a PTFE capillary leading to the trapping and stabilising unit (liquid sorbent, LS). Dimethyl arsine (DMA, (CH3)2AsH) created from dimethyl arsenic sodium salt trihydrate (C2H6AsNaO2·3H2O; Fluka p.a.≥98%) was used to optimise the method. Additional experiments with inorganic arsine AsH3 (created from AsVHNa2O4·7H2O, Fluka, p.a.≥98%) under the optimised conditions, proved the method’s efficiency. Trapping of other volatile metal(loid)s should be performed in the same way. Optimum generation of other volatile metal(loid)s may require additional time-consuming adjustments. The generation of volatile As compounds (arsanes) is a common procedure that has been used for the determination of dissolved As by HG-AAS since 1973 [16], where the transformation of dissolved to volatile As increases the detection selectivity due to analyte–matrix separation. Under acidic conditions, dimethylarsenic salt (CH3)2AsNaO2·3H2O is first transformed to dimethyl

Fig. 4a, b Laboratory set-up for generation of volatile metal(loid) compounds in a PTFE-coated reaction vessel (RV), consequent sampling with the developed collector cell (CC) and transfer to liquid sorbents (LS) for stabilising; syringe (SY) is used to add reducing agent and acid for hydride generation, N2 provides reducing environment during the whole experiment, pressure is monitored by a digital manometer (MA)

arsenic acid [(CH3)2AsO(OH), DMAA]. By adding a reducing agent (sodiumborohydride, NaBH4), which also serves as a source for nascent hydrogen, (CH3)2AsVO(OH) is reduced to volatile (CH3)2AsIIIH: + → 2(CH3 )2 AsOOH + BH− 4 + H ← 2(CH 3 )2 AsH ↑ + H3 BO 3 + H2 O This reaction is extremely pH dependent. Every As species has its own pH range for optimum hydride generation [17, 18, 19] (Fig. 5a). First experiments with HG-AAS showed optimum arsane generation from DMAA at a small range between pH 0.25 and 1.5 only, with an optimum at pH 1.1, more or less in accordance with [20, 21, 22]. Outside this pH range, hydride generation decreases significantly (Fig. 5b). Pergantis et al. explain the critical effect of acid concentration from experiments with deuterium-labelled HCl and NaBH4; generated DMA, due to its high basicity, reacts rapidly with HCl to produce water-soluble arsonium species in equilibrium with gaseous DMA [23]. The amount of chemicals needed for the reaction cannot be upscaled directly from HG-AAS applications since sample volumes are only a few mL, compared to almost 0.5 L used in the reaction vessel. In addition, there is a constant purging gas stream in the HG-AAS application, transferring the created volatile compounds directly to the detection unit. In the reaction vessel, however, the

1195 duction step; pH is at about 12. The solution is acidified again (pH 1–2) by adding 40 mL of 1 N HCl within 20 min. All As that did not degas and re-dissolve in the solution is transferred again to (CH3)2AsO(OH) and reduced in a second step with 20 mL 15% NaBH4 in 5% NaOH inserted within 15 min. A final N2 flow for 20 min ensures that all volatile compounds leave the reaction vessel via the PTFE membrane. The whole experimental time is 90 min. With this two-step approach, As concentrations in the reaction vessel solution are below the determination limits (AAS 1.5 µg As/L) after the reaction. Further reduction of the amount of chemical reagents or reduction of the injection duration leads to incomplete transfer from the aqueous to the gaseous phase with 20–60% of the originally added As remaining in the reaction vessel. A pre-reduction of DMAA to organosulfur derivates of As(III) by addition of L-cysteine (C3H7NO2S) as suggested for HG-AAS, e.g. by [24, 18, 25] was not successful. The final solution becomes alkaline, and hydride generation is lower than with high acidity [25]. This leads to an incomplete transfer of dissolved to volatile DMA. Sorption and analysis

Fig. 5 a pH dependency of arsane generation from As(III), As(V), MMAA and DMAA (modified from [21]). b pH dependency of arsane generation from DMAA (data from [21, 22, 26] and own experiments) permeability of the PTFE membrane of the collector cell slows down the gaseous flow, leading to longer residence times of volatile compounds in the reaction vessel. Since the reducing agent NaBH4 is stored in alkaline solution (NaOH) for stability reasons, the pH increases significantly during the creation of the volatile compounds. Not all As is completely reduced to dimethylarsine and degassed from solution. The remainder is re-transformed to dissolved Na(CH3)2AsVOO, once the solution is alkaline again; a problem that does not occur in the HG-AAS device. A special procedure was therefore developed to generate a complete transformation of dissolved to volatile As and degassing via the PTFE membrane. Contrary to field applications, where a vacuum is applied to transfer the volatile arsenic through the PTFE membrane, a continuous N2 flow had to be kept during the whole experiment in the laboratory. The continuous N2 flow maintains reducing conditions, to exclude the oxidation of volatile As, and transfers the volatile As at a constant rate through the PTFE membrane. The N2 pressure was kept as low as possible (approximately 30 mbar differential pressure) to leave the pressure inside the reaction vessel below the gas transfer rate limited by the permeability of the PTFE membrane. The 475 mL of acidified DMAA standard solution (100 mL 1 N HCl in 1 L of DMAA solution, pH 1.1) in the reaction vessel is purged with N2 for 10 min at the beginning, to create reducing conditions. Flow from the bottom to the top allows for a complete oxygen removal. Within 20 min, 20 mL of 15% NaBH4 (p.a.≥96%, Merck) in 5% NaOH are then applied by a syringe connected to the bottom of the vessel (approximately 2 mL every 2 min). This stepwise addition of small amounts of reducing agent was chosen because a strong arsine generation at the beginning strongly enhances the pressure in the vessel. This increase is due to the limiting permeability of the PTFE membrane and may even lead to backflow via the bottom capillary. The reaction vessel is purged with N2 for another 5 min after this first re-

Sorbing the created volatile metal(loid)s in liquid sorbents requires dissolution of the gases and their oxidation in a strong oxidising agent. This results in an alteration of the original speciation, but it is a fast and reliable method for a first screening of volatile metal(loid)s. Laboratory experiments started with a high-grade stainless steel cylinder as an oxidising trap (40 cm high, 4.5 cm diameter). After experiencing As losses due to precipitation and sorption, the steel cylinder was replaced by a PTFE cylinder (20 cm high, 4 cm diameter). PTFE rings (4 mm inner, 6 mm outer diameter, 2–3 mm high) within the PTFE cylinder helped to increase reaction time and sampling efficiency (from 42–45% to 76%). Glass beads used before showed significant As sorption: 60 µg As/L were detected in a cleaning solution of 1 N HCl applied overnight after 15 experiments. Only small amounts of oxidation solution are used for trapping to achieve an enrichment of the naturally low concentrations of volatile metal(loid)s. Arsine generated from a 475 mL reaction solution is trapped in 95 mL oxidising solution (enrichment factor 5). This enrichment ratio is increased in the field with the same or an even better trapping efficiency due to much lower gas flow rates. Table 1 AAS conditions for analysing the original standard solution before the experiment (samples C control), the solution in the reaction vessel (samples R rest) and the oxidising solution (samples O oxidation) Step Drying Drying

Temp (°C)

Ramp (°C/s)

Hold (s)

Time (s)

120

10

0

10.1

150

3

Dry ashing

1350 C 1400 R 1500 O

135 C 140 R 150 O

30

40.0

30 C 40 R 60 O

38.9 C 48.9 R 69.0 O

Atomisation

2400

FP (full power)

4

4.7 0.2 R

Cleaning

2600

1000

3

3.2

Run

0.0 C 0.2 O

15 µL Sample solution+8 µL modifier (Pd(NO3)2; c(Pd)=10 g/L Pd(NO3)2 in ca. 15% HNO3; dilution 1:10); wavelength 193.7 mm, gap width 60 mm C standards in 0.1 N HCl matrix, R standards in 8 mL 1 N HCl and 19 mL 1 N NaOH matrix to 50 mL, O standards in 11.1 mL NaOCl and 11.1 mL HCl matrix to 50 mL

1196 Fig. 6 As recovery in % from original input DMAA solution after volatilising DMA, collecting it by the PTFE collector cell and trapping and stabilising it in the oxidising agent NaOCl in different dilution ratios from 1:0 (pure NaOCl) to 0:1 (deionised water, control); only one experiment for 1:1, 1:200, 1:500, 0:1; 2 experiments for 1:10; 3 for 1:50; 7 for 1:0; 10 for 1:100 (bars indicate minimum and maximum values, rhombus indicate mean values)

Fig. 7 As recovery in % from original input solution (DMAA in different concentrations) after volatilising DMA, collecting it by the PTFE collector cell and trapping and stabilising it in the oxidising agent NaOCl in different dilution ratios

After each experiment, samples were taken from the original standard solution prior to the experiment (samples C=control), from the solution in the reaction vessel (samples R=rest) and the oxidising solution (samples O=oxidation) to calculate the percentage of recovered As in the oxidising solution. All samples were analysed for total As by GFAAS (graphite furnace atomic absorption spectrometry) (AAS-EA4, Zeiss, Germany) using the platform technique. A special temperature-time-program had to be developed to analyse the complex matrix of the O samples (see Table 1). Unspecific sorption peaks appeared during analysis due to the high salt concentrations in these samples. The method development aimed at a complete separation of those peaks from the As signal by varying dry ashing conditions. This led to relatively high dry ashing temperatures and a total duration of about 70 s for this step. To avoid As losses during this thermal treatment, an increased modifier concentration, 8 µL of a 1 g Pd/L stock solution (Pd(NO3)2), was added to 15 µL sample solution. Experiments with NH4NO3 to improve the separation of unspecific and specific signals showed no success. No preparation was necessary for C- and R samples, 1 mL 1 N HCl and 2.5 mL H2O were added to 1 mL O sample due to problems with high alkalinity and the applied high As concentrations.

Hydrogen peroxide (H2O2, perhydrole p.a. 30% Merck) was not suitable for rapid oxidation. 1.5 M sodiumperoxodisulfate Na2S2O8 (p.a.≥99% Fluka) caused problems due to high matrix effects during the further analysis with GFAAS. Sodium hypochlorite (NaOCl, 6–14% Cl active, Riedel-de Haen) turned out to be effective for rapid and complete oxidation. Ca(OCl)2 would be an even stronger oxidising agent than NaOCl, but problems of calcium metal(loid) oxide precipitates could occur (in 50 mL of 1.0 M Ca(OCl)2 liquid, a maximum of 12.2 mg As can dissolve at 25 °C [15]). Testing different concentrations of the oxidising solution NaOCl, a 1:100 dilution of NaOCl with deionised water proved sufficient to obtain recovery rates above 75%, with an average of approximately 80% for DMA (Figs. 6 and 7) and an average of 82% for AsH3 (Table 2). A second oxidising solution (LS 2, Fig. 4), mounted behind the first one (LS 1) generally showed As concentrations below 5% of the original solution. Since the dissolution of gases in aqueous solution depends on the gas concentration (Henry law), the oxidising efficiency in the second oxidising solution with a lower input concentration will be lower than in the first oxidising solution. Thus, a 100% recovery cannot be expected. Lower recovery rates for higher concentrated oxidising solutions (1:0, 1:1,

1197 Table 2 As recovery in % from original AsVHNa2O4.7H2O input solution after volatilising AsH3, collecting it by the PTFE collector cell and trapping and stabilising it in the oxidising agent NaOCl (dilution 1:100), testing different concentrations from 16–95 µg As/L and reproducibility at 75 µg As/L

Table 3 Long term stability of the 1:100 diluted solution, comparing analysis 1–2 days after the experiment to analysis after 6 weeks of storage in a refrigerator. Mean losses were 5% in glass and 8% in PE bottles

As concentrations (µg/L)

Recovery (%)

Mean standard deviation

As concentrations (µg/L)

Recovery (%)

Mean standard deviation

16 30 45 72 95

76 82 84 82 87

82%±4.0

75 75 75 75 75

79 80 83 82 88

82%±3.5

As concentrations (µg/L)

As losses after 6 weeks (%)

In glass bottles 1–2 days after the experiment

In PE bottles 1–2 days after the experiment

In glass bottles 6 weeks after experiment

In PE bottles 6 weeks after experiment

In glass bottles

In PE bottles

284 136 373 22 70

278 132 373 21 72

251 135 382 20 69

255 129 394 19 63

12 0 –3 8 1

8 2 –6 10 13

1:10) than for the 1:100 solutions are probably not related to insufficient trapping but to problems with AAS analysis due to the high matrix effects. Dilutions of 1:200 and 1:500 lead to incomplete trapping; the second oxidising solution (pure NaOCl) mounted behind the first one showed higher As concentrations (up to 25%). A blind test with deionised water in the first oxidising solution showed As concentrations of less than 1% in the first solution and 79% in the second solution (pure NaOCl). Acidifying pure NaOCl from pH 11.8 to pH 4.0 (1:1 NaOCl to 1 N HCl) led to a poor recovery of 9%, similar to strongly alkaline oxidising solutions (pH 12.3, 1:1 NaOCl to 1 N NaOH) with 58% recovery. Experimental reproducibility is very satisfactory (Figs. 6 and 7, Table 2). Six experiments performed by two different people (DMA standard solution 72 µg As/L, NaOCl dilution 1:100) showed recoveries between 75% and 82% with a mean of 79% and a standard deviation of 2.6% for DMA. Five reproducibility experiments with an As(V) standard solution (75 µg As/L) showed similar recoveries between 79% and 88%, with a mean of 82% and a standard deviation of 3.5% for AsH3. The long term stability of the 1:100 diluted solution is shown in Table 3. Maximum losses during 6 weeks of storage in a refrigerator were 12% in glass bottles and 13% in PE bottles, compared to results obtained 1–2 days after the experiment. Mean losses were 5% in glass and 8% in PE bottles. Even though the losses in glass bottles were slightly lower than in PE bottles, PE bottles are recommended for further analysis in order to avoid sorption effects of other metal(loid)s on the glass walls, due to the high pH value (pH 11.1–11.3).

Conclusion Element budgets and mass balances are most often lacking for the processes determining volatile metal(loid)s behaviour. These aspects are of great relevance, however, particularly at the interface of pedospheric and hydrospheric processes with the atmosphere. To overcome current problems, this works suggests a new methodological approach to sample volatile metal(loid)s in the field. Laboratory experiments with inorganic As and dimethyl As

compounds confirmed the possibility of sampling volatile metal(loid)s from wetland environments with a specially designed PTFE collector cell via a PTFE membrane withstanding a vacuum of more than –500 mbar. Trapped gases are dissolved and oxidised in 1:100 diluted NaOCl with a sampling efficiency of about 80%. The oxidising solutions can be stored in the refrigerator for 6 weeks with an average loss of only 5–8%, both in PE and glass bottles. Due to the small size of the equipment, the low amount of chemicals needed (