Speciation of Inorganic Selenium Using Flow Injection Hydride Generation Atomic Fluorescence Spectrometry Journal of Analytical Atomic Spectrometry D. W. BRYCE A. IZQUIERDO AND M. D . LUQUE D E CASTRO Department of Analytical Chemistry Fuculty of Sciences University of Cbrdoba 14004 Spain Two methods based on flow injection atomic fluorescence have been developed in order to speciate Se as Se'"-SeV'. Both methods use hydride generation of Sew with atomic fluorescence detection as the derivatization-detection step. In the first method two sample plugs are injected simultaneously in series so that the first plug passes straight to the detector to determine Se'". The second plug passes through a focused microwave device where Se"' is reduced to Sew prior to its conversion into the hydride.The Se"' content is then given as the difference between the two results. In the second method a mini-column is used to retain both the Se species; Sew and Sen are then eluted sequentially with formic and hydrochloric acids respectively. The columns can be used for preconcentration of the analytes as well as for sampling. Both methods show exceptional sensitivity [limits of detection (3s) of 0.04 pg 1-'1 and wide linear range up to 50 pg I-' with excellent linearity (9 = 0.999) and reproducibility (relative standard deviation<5%). The methods have been used in an intercomparison exercise for the Measurements and Testing Programme and have been applied to the determination of the analytes in tap water and haemodialysis samples. Keywords Flow injection; hydride generation; atomic fluorescence spectrometry; selenium speciation In the last couple of decades information on the speciation of certain elements (particularly Hg Se As and Cr) has become increasingly important as the effects of the different forms are better understood.Selenium is one such case with special importance owing to its ambivalent behaviour it is essential at low concentrations but toxic at high concentrations with a relatively small difference between these levels.'" This coupled with the fact that the concentration levels found are extremely low means that sensitive and accurate methods are needed for speciation of this element. Methods based on sophisticated expensive instruments such as isotope dilution mass spe~trometry,~ stimulated Raman scattering5 and hydride generation inductively coupled plasma mass spectrometry6 have been reported for toxic elements in the last decade.Speciation in all instances is achieved by the use of a powerful separation technique prior to the introduction of the sample into the instruments. Several flow injection (FI) methods for the speciation of Se have been developed recently including microwave-assisted hydride generation atomic absorption spectrometryY7.* and FI cathodic- and anodic-stripping v~ltamrnetry.~~'~ Atomic fluorescence has been used recently for the determi- nation of Hg and the hydride forming elernents."*l2 These methods have the advantages of excellent sensitivity with wide linear ranges are relatively free from interferences and have little memory effects. Small dedicated instruments based on atomic fluorescence which provide excellent features of the methods thus developed with accessible prices are on the market at present.PS Analyti~al'~.'~ have developed two detectors (Merlin and Excalibur) for the determination of Hg and Se Cd and As and the excellent features of these detectors have been reported in several paper^.'^-'^ The work described in the present paper had two main objectives. These were firstly to design a sensitive FI method for the speciation of Se for use in an intercomparison exercise on speciation of Se in water samples organized by the Measurements and Testing Programme (Community Bureau of Reference formerly BCR) and secondly to develop a system incorporating mini-columns to fulfil a triple aim i.e. to achieve the speciation of inorganic Se and carry out the preconcen- tration and the sampling steps in an automated continuous way.EXPERIMENTAL Apparatus An Excalibur atomic fluorescence detector (PS Analytical Orpington Kent UK) fitted with (a boosted discharge) hollow cathode lamp for Se (Photron Narre-Warren Victoria Australia) and a specific ultraviolet (UV) filter to allow transmission of the Se atomic fluorescence spectrum in con- junction with a solar blind photomultiplier was used. A Microdigest 301 focused microwave system (Prolabo Paris France) was used together with two Gilson (Worthington OH USA) Minipuls-3 programmable pumps fitted with rate selectors and three Rheodyne 5041 injection valves (two of which were connected in order to function as a double injection valve).Teflon tubing of id of 0.5mm was used in order to construct the FI manifold. Reagents and Solutions All reagents were of analytical-reagent grade. Ultrapure water obtained from a Millipore (Milford MA USA) Milli-Q system was used throughout. Hydrochloric acid of 6 and 1 mol I-' (Panreac Barcelona Spain) 1.5% m/v NaBH solution (Aldrich Milwaukee WI USA) in 0.1 mol 1-' NaOH solution and formic acid (Merck Darmstadt Germany) were used as the carriers in the system. The Na2Se03 (Merck) and Na2Se0 (Aldrich) stock solutions of 1 g1-I were prepared in Milli-Q water. Working solutions were prepared daily by appropriate dilution in Milli-Q water. Argon 5N (SEO Barcelona Spain) and hydrogen (Carburos Metalicos Barcelona Spain) were used in order to flush the hydrides formed to the detector.Mini-column Teflon tubing of 2mm id was filled with Dowex-1 (chloride form) strongly basic anion exchanger (200-400 mesh) (Sigma St. Louis MO USA) (with cotton wool placed at each end). The exchanger was conditioned with 0.1 moll-' HCl before being thoroughly washed with water (Milli-Q) and then used for the retention-preconcentration of the analytes. Procedure The manifold used for the speciation of Se based on dual simultaneous injection is shown in Fig. l(a). The loops of the Journal of Analytical Atomic Spectrometry December 1995 Vol. 10 1059I HCI I I / GLS NaBH. - CR IV (b 1 pp Sample 2 CHOOH HCI + w Fig. 1 FI manifolds for the speciation of selenium based on (a) simultaneous dual injection of the sample and (b) sequential elution of Se'" and Se"' where HC1=6moll-'; NaBH4=1.5% (in NaOH 0.1 moll-'); D=detector; MC=minicolumn; CHOzH=2 moll-'; G = argon; IVl-.3 =injection valves 1-3; M =microwave system; CR = cooling reactor; GLS = gas-liquid separator; PP = peristaltic pump; sample 1 = Se in HCl 6 moll-'; sample 2 = Se in H20; and W = waste (for method (b) HCl=2 mol 1-' NaBH4= 1.0'3'0 G=argon+ hydrogen) double injection valve are filled with the same sample in 6 moll-' HCl and injected simultaneously into the carrier stream also of 6 mol 1-' HCl.The plug injected via valve 1 passes to the confluence point where it merges with the stream of NaBH solution where SeIV is converted into the hydride form before continuing to the gas-liquid separator where the hydride is flushed to the detector by a stream of argon giving rise to a signal corresponding to the Se" concentration in the sample.Meanwhile the plug injected via IV passes through a reactor placed in the chamber of a focused microwave device where SeV' is reduced to Se". This plug then continues through another reactor placed in an ice-bath (in order to reduce the temperature of the sample) before continuing to the confluence point where the Se all in the form of Se" forms the hydride as explained above thus giving rise to a signal corresponding to the concentration of SeIV+SeV' initially present in the sample. The concentration of Sev' is then calculated as the difference between the two results. In the second method based on the selective elution of SeIV and Se" the speciation is performed after a step where the two oxidation states are retained on a mini-column prior to selective elution.The manifold used in this method is shown in Fig. l(b). In this case three simultaneous injections are used (with valves 2 and 3 being a double injection valve thus allowing valve 1 to be operated independently for the precon- centration of the analytes). The loops of the valves IV1 IV and IV3 are filled with the sample (this time without HCl) CH02H and HCl respectively. The three valves are changed to the inject position simultaneously so that the three plugs are inserted into the carrier stream (in this case H,O). The sample plug is carried to a mini-column containing the Dowex-1 strongly anionic exchanger situated in the carrier channel where SeIV and Sew are retained.As the plug of CH0,H passes through the column it selectively elutes Se'". This plug of CH02H containing SeIV then continues to the first confluence point where it mixes with a stream of 6 mol 1-' HCl before passing through a reactor placed inside the micro- wave device and then through a cooling coil (this step is not actually necessary as no reaction is taking place but it keeps the manifold simple i.e. omits the need for a splitting point). From here the stream is mixed with one of NaBH solution at the second confluence point where the Se" forms the hydride before being passed from the gas-liquid separator to the detector by the stream of argon giving the signal for Se". Following this the plug of HC1 injected via valve 3 passes through the column eluting the SeV1.This stream mixes with the stream of HC1 before passing through the microwave device where the Sevl is reduced to Se". This once cooled mixes with the stream of NaBH solution forming the hydride and then giving the signal for the SeV1 present in the initial sample. Signal measurement was carried out on the basis of peak height. RESULTS AND DISCUSSION Fundamental Steps of the Atomic Fluorescence Spectrometric Determination The determination step (derivatization and monitoring) was common to both methods and was based on the formation of the hydride derivative of Se"' and atomic fluorescence of the analyte in a hydrogen flame. Organoselenium compounds such as selenomethionine and selenocystine would not give any signal as they are not reducible under the experimental conditions used. The determination can be split into three sub- steps.Firstly the hydride has to be formed which is carried out in the presence of NaBH in an acidic medium. In the second step the hydride and the hydrogen formed are swept out of the gas-liquid separator by a stream of argon into a chemically generated hydrogen diffusion flame (the flame is maintained by the excess of hydrogen produced in the reaction between NaBH and HC1). In the third step the hydride is atomized in the flame and the atoms are detected by atomic fluorescence spectrometry. These detectors and gas-liquid sep- arators have been well documented in the literat~re.".'~.~~ Optimization Determination step The univariate method was used throughout the optimization process.For the optimization of this common step the FI variable that was studied was the flow rate the chemical variables were the NaBH and HC1 concentrations. The final variable that was vital was the flow rate of the argon. As the overall flow rate of the FI system was increased so the sensitivity increased although at values above 3.5 ml min-' the hydrogen flame became large which was characterized by a high signal-to-noise ratio zs has been reported previously in the literature.' Clearly the sampling frequency was also found to increase with the increased flow rates. As a compro- mise between sensitivity signal-to-noise ratio and sampling frequency a flow rate of 2.9 ml min-' was chosen permitting good sensitivity signal-to-noise ratio and sampling frequency.The NaBH concentration was studied in the range 0.5-2.0% m/v (in 0.1 mol 1-' NaOH solution). At concentrations lower than 1.2% m/v it was difficult to maintain the hydrogen flame as insufficient H2 was formed. Above 1.2% m/v the analytical signal improved very slightly and so 1.5% m/v was chosen as the optimum concentration. The concentration of HCl had a minimal effect (less than k 2%) on the analytical signal in the range 1.5-8 moll-'. However as the reduction of SeV' is carried out optimally in 6 moll-' HC1," this concentration was chosen. Although it was necessary for the samples to be in 6 mol I-' HCl the HCl concentration in the carrier stream could be as low as 2 mol 1-' which was sufficient to maintain 1060 Journal of Analytical Atomic Spectrometry December 1995 VoE.10the flame; however this resulted in signals for the blank (as a plug of 6mol1-' HC1 was being injected into a stream of 2 moll-' HCl) thus increasing the limit of detection and so 6 moll-' was used. The flow rate of argon was also of prime importance. As the flow rate was increased so the peak height and sampling frequency increased although above 1000 ml min-' the argon flow rate was so great that it extinguished the hydrogen flame. As a result a flow rate of 900 ml min-' was chosen as optimum. One other stage that was common to both methods was the reduction of SeV'. A method for the on-line reduction of SeV' aided by a focused microwave device had been developed previously by workers in this laboratory,20 and was adapted for this method.The HCl concentration had to be 6 moll-' (below this concentration incomplete reduction of SeV' was observed and at concentrations above 6 moll-' no increase was observed). Microwave power of 50 W and a flow rate through the microwave device of 1.0ml min-' were also necessary for the complete reduction of Se". One modification of the determination step was developed with the object of making method 1 less expensive. In this case the NaBH concentration was decreased to 1.0% m/v (i.e. a 33.3% reduction in NaBH,) the HC1 concentration was reduced to 2mol1-' (i.e. a 66.7% reduction in HCl) and a gas mixture of argon-hydrogen (540+ 3 ml min-') was used to flush the hydride to the detector. In this way the additional flow of H gas was used to maintain the hydrogen flame.The flow rate was increased to 3.9 ml min-'. These adaptations resulted in a more cost efficient rapid method which showed slightly poorer sensitivity and reproducibility but was suitable for routine analysis. This method is subsequently called 'method l B whereas the original method 1 is referred to as 'method 1A'. Speciation based on dual simultaneous injection method 1 The speciation of Se'v-Sev' in method 1 was achieved by controlling the dynamic system. The concentration of the different oxidation states is obtained by the simultaneous injection of two plugs into the system. The first plug gives the Se" concentration and the second plug after reduction of SeV' to SeIV gives the concentration of Se" + SeV' from which the Sev' concentration can be obtained by the difference in the two results.The optimum injected volume for both plugs was 600 p1. Below this value the peak height was smaller whereas above this value the peak height remained more or less constant although the peaks tended to be a lot broader with the possibility of forming double peaks which affected the repro- ducibility. The samples were prepared in 6 moll-' HCl in order to facilitate the reduction of Sev' to Se". The conditions of the reduction process have been described previously2' and were maintained for the present work. The length of the cooling reactor [CR in Fig. l(a)] was studied over the range 0-400 cm. The purpose of this reactor was two-fold firstly to cool the stream coming out of the microwave system and secondly to allow the first peak to return to the baseline prior to the start of the second peak thus improving reproducibility. A length of 200cm was found to be optimum as it allowed sufficient separation of the peaks without being too detrimental to the sampling frequency. Speciation based on selective elution of Se" and Se" method 2 In this case speciation was achieved by on-line retention of the two species on a mini-column prior to their selective elution.After a literature search Dowex-1 strongly anionic exchanger (chloride form) 200-400 mesh was used as the packing material. The column was prepared as described under Experimental. For this system an additional channel ( H20) had to be added to the manifold in which the mini-column was placed.Again an injection volume of 600 p1 for the sample was used and an overall flow rate and flow rate through the microwave system were maintained at the values given for method 1. With initial concentrations of 3 and 6mol1-' for CHOzH and HCl respectively the elution volumes of the two acids were studied between 100 and 1000p1. In order to check that elution was complete the eluent was injected three times after every injection of SeIV in order to make sure that all of the Se" had been eluted after the first injection of CH02H. This cleaning step was repeated for the study of the elution of SeV1 with HC1. For CH02H the peak height increased as the elution volume was increased up to 600 p1 above which a slight decrease in peak height was observed. Similarly for the elution volume for Se" i.e.the injected volume of HC1 the peak height increased as the injected volume was increased up to a volume of 350 pl above which a slight decrease was observed. For all further experiments 600 and 350 pl were used for CH02H and HCl respectively. The lengths of the reactors between the three injection valves were as short as possible (i.e. 10 cm). This was sufficiently long to enable the Se to be retained on the column prior to the elution by CH02H and HCl and also allowed the first peak to return to the baseline before the second peak started. As far as the concentrations of CH0,H and HCl were concerned the CH20H concentration was studied from 1 to 3 moll-'. Below 2 mol 1-' the elution was incomplete (ca. 75%) whereas above 2mol 1-' 100% elution was observed.As a result 2 moll-' was chosen as the optimum concentration. For the elution of Se" 6mol1-' HCl was used providing 100% elution. Concentrations below this value were not studied as the Sev' had to be in 6moll-' HCl in order for it to be reduced to Se" and concentrations above this were not studied as they would have no advantages over 6mol1-'. Hence 6mol1-' was used for all further experiments. The flow rates of the channels containing H 2 0 and HCl were each set at 0.5 ml min-' so that when the two channels combined at the confluence point the overall flow rate i.e. the flow rate through the microwave system was maintained at its optimum value of 1 .O ml min - ' . The details of the optimization process i.e. the variables studied the ranges studied and the optimum conditions are all outlined in Table 1.Table 1 Optimization of variables Type of variable Common variables Method 1A Method 1B Method 2 Variable Overall flow rate/ml min-' Argon flow rate/ml min-' [ HCl]/mol 1 - [NaBH,] (YO m/v) [ NaOH]/mol 1 - ' Injected volume 1/p1 Injected volume 2/pl Reactor length/cm Argon flow rate/ml min-' Hydrogen flow rate/ml min-' [HCl]/mol 1-' [NaBH,] (YO m/v) Overall flow rate/ml min-' Injected volume l/pl Injected volume 2/pl Injected volume 3/pl [CHO,H]/mol 1-' [HCl]/mol 1-' H,O flow rate/ml min-' HCl flow rate/ml min-' Range studied 1-4 1 00- 1000 1-8 1-2 100-lo00 100-1000 0-350 loo- 1000 0- 10 1-6 0.5-2 1-5 100-1000 100-1000 100-1000 1-3 - 6 0.5 0.5 Optimum value 2.9 900 6 1.5 0.1 600 600 200 540 3 2 1 3.9 600 600 3 50 2 6 0.5 0.5 Journal of Analytical Atomic Spectrometry December 1995 Vol.10 1061Interferences Once the systems had been optimized a literature search was carried out in order to know which species usually interfere in the atomic fluorescence spectrometric determination of SeIV. This revealed that Co Bi Pd Ni As Pb and Cu are the most common interferents. Various concentrations of these foreign species were then added to a solution containing 2 pg 1-' Se" with the object of determining the highest tolerable ratio of foreign species to analyte with +5% of the original signal being considered as having no interference. Method 1A was used to determine the interference effects. The results obtained are as follows (foreign species to analyte ratio tolerated without interference) Cu > 500 1; Co > 1000 1; Pb > 1000 1; Bi 500 1; Ni > 1000 1; As"' 1000 > 1; Pd 250 1.These relatively high tolerance levels particularly for Cu can be explained by the high acid concentration that is used. This ensures Cu remains in solution thus preventing it from destroying the SeH which could occur if Cu was allowed to precipitate out. The method shows excellent tolerance to foreign species. This is partly due to the fact that the hydrides are separated from the carrier stream as a gas prior to detection. Only Pd interferes strongly (at a foreign species to analyte ratio of 500 1 a reduction in peak height of 65% was observed) but is not considered important as Pd is very unlikely to be present in any great amounts in the samples being investigated.Features of the Method When the optimum value of each of the variables had been found the characteristics of the methods were studied. For the calibration curves of the two methods based on dual simultaneous injection (methods 1A and lB) nine standard solutions each of Se" SeV' and SerV+SeV1 were prepared in 6mol1-' HC1 with concentrations ranging from 0.15 to 50 pg 1-' of each species and were injected in triplicate into the FI systems. The methods both showed exceptionally good linearity with coefficients of correlation r2 better than 0.999 in all cases although method 1A had a slightly better sensi- tivity. In order to study the reproducibility of the results for Se" from the first and second peaks and SeV' from the second Table 2 Features of the proposed methods peak 11 solutions of both low (0.5 pg 1-') and high (25 pg 1-') Se" and SeV' concentrations were injected in triplicate.The reproducibility expressed as YO RSD was excellent in all cases although slightly better for method 1A. The limits of detection were expressed as 3s for the low level reproducibility study for method 1A and 3s for the blank signal (injected 22 times) for method 1B. Again method 1A showed slightly better character- istics with limits of detection of 0.04 and 0.06 pg 1-' and 0.11 and 0.13 pgl-' for Se" and SeV' in methods 1A and lB respectively. Sampling frequencies of 40 and 50 h-' were observed for methods 1A and lB respectively. These tests were all repeated for the method based on the selective elution of SeIV and SeV' (method 2) with the exception of those which were unnecessary (i.e.calibration and reproduc- ibility of Se" in the second peak). A sampling frequency of 15 h-l was observed for this method. The method showed a limit of detection of 0.07 pg 1-'. In addition to these tests method 2 was also studied to check its suitability for the preconcentration of Se" and Sevl. A solution of 0.025 pg 1-' of SeIV and SeV' was preconcentrated ten-fold prior to the elution step. This was carried out by filling the loop of valve 1 with sample and injecting it ten times before the loops of the double injection valve were filled with CHOzH and HCl and injected to elute the preconcentrated analytes. The reproduc- ibility of the ten-fold preconcentration process expressed as YO RSD of 11 repetitions was excellent showing the potential of the mini-columns.All the equations coefficients of correlation and reproducibilities are shown in Table 2. Applications Method 1A was used in an intercomparison exercise organized by the Measurements and Testing Programme (formerly the BCR) on the speciation of Se in water samples. Two solutions (five bottles each) of 2-9 and 20-70 pg 1-' of SeIV and SeV' were analysed. A separate batch of the same samples was also determined by FI cathodic-stripping voltammetry2' showing good agreement between the two methods. The results obtained are shown in Table 3. Following this method 2 was applied to the determination Method Analyte 1A Se" 1st peak Se" 2nd peak Sevl 2nd peak 1B SeIV 1st peak SeIV 2nd peak SeV' 2nd peak 2 SeIV Sev' Equation y = 5.70~ + 1.29 y = 5 .0 1 ~ + 0.93 ~ = 4 . 8 9 ~ + 1.34 ~ ~ 4 . 7 7 ~ - 1.97 y= 4 . 2 9 ~ - 2.39 JI = 4.27~ - 2.1 1 y = 3.00~ + 0.75 y= 1.60~-0.37 Linear range/ Pug 1-' 0.1-50 0.1-50 0.1-50 0.25-50 0.25-50 0.25-50 0.15-50 0.15-50 Regression coefficient ( r 2 ) 0.9995 0.9988 0.9995 0.9990 0.9984 0.9988 0.9993 0.9999 Limits of detection (3s)/ Pg I-' 0.04t 0.06 - 0 . l l t 0.13$ 0.07 0.06 Reproducibility RSD (%); n = l l * 5.44 (0.5) 2.98 (25) 4.76 (0.5) 4.03 (25) 5.26 (0.5) 4.93 (25) - - - 5.70 (0.025)g 6.75 (0.025)$ * At the concentration levels given in parentheses (pg 1-l). For SeIV. $ For Sevl. 0 For a sample preconcentrated by a factor of ten repeated 11 times. Table 3 Intercomparison study; all results are in p g 1-1 Method 1A FI Solution Se" Se'" + SeV' SeV' SeIV SeIV + SeV1 SeV' Low concentration 5.2 k 0.2 12.4 & 0.9 7.3 0.9 5.2 f 0.7 11.9 k 2.6 6.7 _+ 2.4 High concentration 35.3 f 1.3 78.8 k 3.7 43.5 f 3.7 37.8 f 4.2 80.8 14.8 43.1 +_ 7.4 1062 Journal of Analytical Atomic Spectrometry December 1995 Vol.10Table 4 Results (pg 1-') for analyses of spiked tap water samples Se" added Sevl added SeIV found SeV1 found 0.5 0.5 0.51 0.52 2 2 2.06 1.93 10 10 9.50 9.78 0.1 0.1 0.09" 0.104" 0.05 0.05 0.045+ 0.055-f 0.015 0.0 15 0.014+ 0.014t * After preconcentration by a factor of five. After preconcentration by a factor of ten. Table 5 Results (pug I-') for analyses of haemodialysis samples after preconcentration by a factor of ten Sample A B Se" found 0.32 0.08 SeV1 found 0.09 0.06 of Se" and SeV1 in tap water.Even after preconcentration by a factor of 20 no Se was found in the water. Thus Se" and Sev' at concentrations of 0.015 0.05 0.1 0.5 2 and 10 pg 1-1 were added to the water and the recoveries checked (preconcen- trating the first three solutions on the column prior to elution). Finally two samples used in the haemodialysis procedure in a local hospital were analysed. The maximum legal Se (total) concentration in such samples is 90 pg 1-'. These samples along with the a tap water sample spiked with 0.05 pg1-' were analysed after a preconcentration factor of ten. All the results obtained for both methods are shown in Tables 4 and 5. CONCLUSION Two extremely sensitive methods and one slightly less so (suitable for routine analysis) with exceptionally good linearity and linear ranges reproducibility and sampling frequencies have been developed and applied to real samples.The develop- ment of mini-columns for the retention and specific elution of Se" and SeV' allows the speciation of Se to be carried out giving an SeV1 concentration that is independent of the Se" concentration which is therefore more accurate than when the concentration of SeV1 is calculated as a difference. The methods developed show increased automation in com- parison with other more expensive complex methods reported recently for speciation of Se. The methods reported here follow the trend of increased automation with steps such as precon- centration elution reduction speciation and detection all being carried out on-line.In addition the methods compare favourably in terms of automation limits of detection linear range sampling frequency simplicity and cost (particularly method lB which shows 33.3 and 66.7% saving in the amounts of NaBH and HC1 used respectively as well as a 25% saving in time) with other methods reported recently. The mini-columns developed should be easily applicable to use in field sampling studies after investigations of resin capacity and the effect of humic/fluvic acids and pH of the sample have been carried out. PS Analytical is thanked for the loan of an Excalibur atomic fluorescence selenium detector. One of the authors (D.W.B.) would like to thank the Measurements and Testing Programme (formerly BCR) for a doctorate grant covering the expenses incurred during his stay in Spain.REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Schroder H. A. Frost D. V. and Balassa J. J. Chronic Dis. 1970 23 227. Robinson W. O. J . Assoc. Off. Agric. Chem. 1933 16 423. Underwood E. J. Trace Elements in Human and Animal Nutrition Academic Press New York 4th edn. 1977. Tanzer D. and Heuman K. Anal. Chem. 1991 63 1984. Leong M. B. D'Silva A. P. and Fassel V. A. Anal. Chem. 1986 58 2594. Buckley W. T. Budac J. J. Godfrey D. V. and Koening K. M. Anal. Chem. 1992 64 724. Cobo-Fernandez M. G. Palacios M. A. Chakraborti D. Quevauviller Ph. and Camara C. Fresenius' J. Anal. Chem. 1995 351 438. Pitts L. Worsfold P. J. and Hill S . J. Analyst 1994 119 2785. Bryce D. W. Izquierdo A. and Luque de Castro M . D. Fresenius' J. Anal. Chem. 1995 351 433. Bryce D. W. Izquierdo A. and Luque de Castro M. D. Anal. Chim. Acta. 1995 308 96. Stockwell P. B. and Corns W. T. Int. Lab. 1994 19 33. Ebdon L. Goodall P. Hill S. J. Stockwell P. B. and Thompson K. C. J. Anal. At. Spectrom. 1993 8 723. Stockwell P. B. Thompson K. C. Henson A. Temmerman E. and Vandecasteele C. Int. Lab. 1989 14 45. Corns W. T. Stockwell P. B. Ebdon L. and Hill S . J. J. Anal. At. Spectrom. 1993 8 71. Stockwell P. B. and Corns W. T. J. Autom. Chem. 1993 15,79. Corns W. T. Ebdon L. Hill S. J. and Stockwell P. B. J. Autam. Chem. 1991 13 267. Stockwell P. B. and Corns W. T. Hydrocarbon Asia 1993 October 36. Corns W. T. Ebdon L. Hill S. J. and Stockwell P. B. Analyst 1992 117 717. Ebdon L. Corns W. T. Stockwell P. B. and Stockwell P . M. J. Autom. Chem. 1989 11 247. Stockwell P. B. and Corns W. T. Analyst 1994 119 1641. Bryce D. W. Izquierdo A. and Luque de Castro M . D. Analyst 1995 120 2171. Paper 5/02687C Received April 27 1995 Accepted August 11 1995 Journal of Analytical Atomic Spectrometry December 1995 Vol. 10 1063