Arsenic speciation in beverages by direct injection-ion chromatography hydride generation atomic fluorescence spectrometry

The procedure developed allows the direct speciation of arsenic in these samples with good sensitivity, selectivity, precision and accuracy. Detection limits determined using the optimized conditions were found to be between 0.16 and 2.9ng ml−1 for arsenite, dimethylarsinic acid, monomethylarsonic acid and arsenate, while standard addition studies showed that the procedure is free from matrix interferences. As no certified reference materials are available for these analytes or matrices, validation was carried out by studying spike recoveries and by comparison of results with an alternative technique.


Introduction
For many years, it was su cient for analysts to determine only the total concentration of toxic elements in samples. However, recently it has become apparent that the toxicity, mobility and bioavailability of certain elements depends heavily on their physiochemical form. This is particularly true in the case of arsenic, a well-known toxic element, which may be found in as many as 13 forms [1,2]. In general, inorganic arsenic (arsenite and arsenate) is more toxic than organic species, e.g. monomethylarsonic acid, dimethylarsinic acid, arsenobetaine , arsenocholine and tetramethylarsonium ion.
Arsenic's presence in the environment is due to both natural and anthropogenic sources. Arsenic is often found in herbicides, pesticides and insecticides (many of which environment protection agencies are now banning) [3], and it is through this route that arsenic gets into wines and other drinks. European legislation has set maximum permissible (total) arsenic concentrations of 1 and 10 ng ml ± 1 in wine and drinking water, respectively. Inorganic arsenic compounds, which are known carcinogens, are used in many manufacturing industries, e.g. glass production, wood preservation and the production of lead accumulators, and are metabolized in the body prior to excretion. Organic arsenic species, which are generally considered to be non-toxic, are often found in ® sh, seafood and mushrooms.
With the total concentration of arsenic in these samples being so low, and the number of individual forms in which it may be present, it is necessary to develop methods with suitably high sensitivity and selectivity to enable accurate determination of each individual arsenic species.
Methods for speciation have to couple the best of separation with the best detection in order to obtain the necessary sensitivity and selectivity. Atomic spectrometric methods using hydride generation have been used frequently as this particular method of sample introduction reduces many interferences and allows greater sample introduction e ciency, so allowing lower detection limits to be reached. However, most of the reported methods are based on atomic absorption which does not show su cient sensitivity for low levels of arsenic, thus meaning that pre-concentrati on steps are necessary [4± 7]. Electrothermal atomic absorption (ETAAS) has been used frequently for the determination of arsenic in several types of samples, as the technique shows good sensitivity. However, the technique su ers from serious interference e ects, making the use of chemical modi® ers necessary. Recent publications in the ® eld make it obvious that hydride generation atomic¯uorescence spectrometry and hydride generation ICP-MS are the two techniques of choice for the hydride-forming elements, as they o er the lowest limits of detection [8,9]. Although both approaches appear to o er similar detection limits, HG/ICP-MS is unsuitable for many laboratories due to high initial and running costs, together with the levels of maintenance required to keep the instrument operating. Hydride generation AFS, on the other hand, is relatively inexpensive and maintenance free, while o ering unsurpassed analytical performance in terms of linearity, sensitivity and freedom from interferences [ 10± 13]. Several publications on the determination of total arsenic levels in matrices ranging from sea water [14], wines and beers [15] to hair [16] have appeared in recent years using hydride generation atomic uorescence.
There is a large body of literature on the speciation of arsenic using ion pair chromatograph y and anion exchange chromatography [17± 22], usually using atomic absorption detection and atomic¯uorescence spectrometry. However, it is strange to ® nd that there are very few publications using hydride generation atomic¯uorescence spectrometry, and in particular, there are no publications on the speciation of arsenic as As (III) , DMA, MMA and As (V) in mineral waters and wines.
The aim of this study was to apply chromatographic separation using a strong anion exchange column, followed by hydride generation atomic¯uorescence spectrometry for the determination of As (III) , DMA, MMA and As (V) in mineral waters and wines.

HPLC system
A Spectra Physics System 1000 HPLC pump and a sixport injection valve (Part No. 7125, Rheodyne, CA, USA) were used in conjunction with a strong anion exchange column (PRPX -100, 250 £ 4.6 mm, 10 mm particle size, Hamilton) to achieve separation of the arsenic species. During the analysis of samples, a guard column (GLC 4-SAX , SGE) was used to preserve the column.

Hydride generation
On-line arsine generation was obtained by use of a peristaltic pump (PS Analytical, Kent, UK) , and various mixing coils prepared in 0.5 mm I/D PTFE tubes. The volatile hydrides were separated from other reaction byproducts in a gas± liquid separator (PS Analytical) . Moisture was removed from the volatile hydrides by passing through a membrane drying tube (Perma Pure Products, Farmingdale, NJ, USA) .

D etection
An Excalibur atomic¯uorescence spectrometer (PS Analytical) , equipped with a boosted discharge hollow cathode lamp (Photron, Victoria, Australia) , was used for detection. This system includes a hydrogen di usion ame as atom cell and optical UV ® lter with a spectral band pass of 20 nm, so allowing three resonance wavelengths of arsenic to be collected.

D ata collection
The¯uorescence signal was recorded on a potentiometer chart recorder Servoscribe RE 541.20.

Validation
In order to validate results obtained with the proposed system for the speciation of arsenic, samples were also analysed for total arsenic using a Millennium Excalibur system (PS Analytical) . Data collection and treatment was by Avalon software (PS Analytical) . Detailed explanations of the instrument are given in previously published papers [10,16].

Reagents
All reagents were of analytical grade, and de-ionized water was used throughout for the preparation of solutions.
Standard solutions (1000 mg ml ± 1 ) of arsenite and arsenate were prepared by dissolving 0.1734 g of NaAsO 2 and 0.4164 g Na 2 HAsO 4 ¢7H 2 O, respectively, in de-ionized water and diluting to 100 ml. MMA and DMA solutions (1000 mg ml ± 1 ) were prepared by dissolving 0.3894 g of CH 3 AsO(ONa) 2 ¢6H 2 0 and 0.1840 g of (CH 3 ) 2 AsHO 2 in de-ionized water and diluting to 100 ml. Working solutions were made after suitable dilution in the mobile phase. This mobile phase was 10 mM K 2 HPO 4 and 10 mM KH 2 PO 4 adjusted to pH 5.7. This was prepared by dissolving 1.74 g K 2 HPO 4 and 1.36 g KH 2 PO 4 in ¹950 ml H 2 O and adjusting the pH by dropwise addition of a 50% HCl solution until pH 5.7 was obtained. This was then diluted to 1000 ml in de-ionized water and degassed by bubbling with helium for 30 min prior to use.
A solution of 1.4% m/v sodium borohydride in 0.1 M sodium hydroxide was used as the reductant and was prepared by ® rstly dissolving 4.0 g NaOH (BDH Merck) in ¹500 ml de-ionized water. Following this, 14.0 g NaBH 4 (Aldrich) was added and dissolved before ® nally diluting the solution to 1000 ml with de-ionized water. This solution was prepared fresh daily.
Hydrochloric acid, potassium iodide and ascorbic acid were all of AnalaR grade (BDH, Merck) .

Speciation of arsenic
A schematic diagram of the ion chromatography-hydrid e generation-atomi c¯uorescence system used is shown in ® gure 1. Optimization of the system is explained in detail in a previous paper by Gomez Ariza et al. [20].
A portion (200 ml) of standard or sample is introduced via the injection valve into a mobile phase of 10 mM potassium phosphate (K 2 HPO 4 /KH 2 PO 4 ) , pH 5.7,¯owing at 0.8 ml min ± 1 . From here the samples pass onto the strong anion exchange column where the four arsenic species are separated. On elution from the column, the stream is then acidi® ed by mixing with a stream of 1.5 M HCl owing at 1.5 ml min ± 1 . The reagents then pass to a gas± liquid separator where a stream of argon¯owing at 250 ml min ± 1 purges the headspace,¯ushing the volatile hydrides and the hydrogen formed in the reaction through a semi-permeable membrane (which is continuously dried with air¯owing in the opposite direction at 2.5 l min ± 1 ) and to the detector. The hydrogen gas, which is a by-product of the hydride generation reaction, is used as fuel for the hydrogen di usion¯ame, which serves to provide free arsenic atoms. These free atoms are then excited by the boosted discharge hollow cathode lamp causing them to¯uoresce, the¯uorescence being detected by the PMT and converted to a 0± 1 V output signal, recorded on a chart recorder. Arsenic species were iden-ti® ed on the basis of retention time. Table 1 summarizes the chromatographic , hydride generation and atomic uorescence conditions used throughout the study.

T otal arsenic determination
In order to validate the speciation results, samples were also analysed for total arsenic.

Characteristics of the proposed method
For the purpose of this study, the proposed system was calibrated from 0 to 10 ng ml ± 1 As (III) , DMA, MMA and As (V) . Typical equations of calibration curves and correlation coe cients are given in table 3, while limits of detection and quanti® cation in various sample matrices are shown in table 4. These limits were calculated as three and 10 times the standard deviation of 10 runs of a 2 ng ml ± 1 standard, respectively. Results show that the best sensitivity was shown for As(III) followed by MMA, DMA and ® nally As(V) . Although this is partially related to the chromatography, it is mainly due to the hydride generation step where it is well known that As(III) forms a hydride more e ciently than As(V) .
A typical chromatogram, obtained using the conditions outlined in table 1, for a mixed solution of 2.5 ng ml ± 1 As (III) , DMA, MMA and As (V) is shown in ® gure 2. The retention times for each species were found to be 2.25, 3.20, 5.20 and 9.15 min for As (III) , DMA, MMA and As (V) , respectively. These retention times were used to identify arsenic species in unknown samples.
The precision of the proposed method was studied by carrying out repeated injections …nˆ10 † of a mixed 2 ng ml ± 1 standard containing As (III) , DMA, MMA and As (V) . Precision was studied not only in aqueous standards, but also in sample matrix, i.e. white wine and mineral water. Table 5 summarizes the results, which show that overall the best precision (expressed as per cent relative standard deviation, RSD) is found for As(III) , with RSDs ranging from 1.35% in white wine to 4.20% in mineral water. The best precision was found for As(III) in white wine, although this is probably due to the fact that this was due to the As(III) found in the sample (i.e. not a spiked concentration) which was actually 5.51 ng ml ± 1 instead of 2 ng ml ± 1 which was the concentration studied for all other species. The results also show that the precision for As(V) is also notably lower in the mineral water than in white wine or the aqueous standard, again probably due to the fact that the As(V) in the mineral water was present at 15.40 ng ml ± 1 as opposed to 2 ng ml ± 1 . Overall, however, the precision ranges from 1.35 to 8.75%.

Applications of the proposed method
The proposed method was applied to the determination of arsenic species in wines and mineral waters. Figure 3 shows typical chromatogram s for various samples, showing that the only species present in the samples were As(III) and As(V) . Figure 3 shows that As(III) was found in all the wine samples tested but none of the mineral water samples. As(V) , however, was found in only one of the white and one of the red wines. More surprisingly it was found at a relatively high concentration (15.41 ng ml ± 1 ) in one of the mineral water samples (French) but not in the other (Scottish) . When analysing T able 3. Performance characteristics of the proposed method for aqueous standards.   No wine or water reference materials with certi® ed arsenic species are currently available, and so in order to validate the method, two approaches were used. Firstly, samples were spiked with known concentrations of each of the arsenic species and the recoveries calculated. Secondly, samples were analysed for total arsenic using an alternative technique.
All samples were spiked with both 2.5 and 5 ng ml ± 1 of each of the four arsenic species, and some with 7.0, 10.0 T able 5. Precision for arsenic species in different samples. For analysis of total arsenic, a PSA 10.055 Millennium Excalibur system (PS Analytical) was used. This system, which has been described elsewhere [10,16], is based on hydride generation atomic¯uorescence. Samples must ® rst be acidi® ed to 25% v/v HCl, following which KI and ascorbic acid must be added in order to convert all arsenic in the sample to As (III) to facilitate hydride generation. Results for total arsenic concentrations are also given in table 6. In all cases good agreement is observed with the results using the proposed IC-HG-AFS system.

Conclusions
Atomic¯uorescence is an extremely sensitive detection system for arsenic which can be easily coupled to an ion chromatograph y system, allowing the determination of individual arsenic species. The results reported here show that the proposed method is accurate and sensitive enough to carry out arsenic speciation in wines and mineral water. In addition, the method shows high selectivity, needs no sample pre-treatment and is free from interferences.